SYSTEM SIMULATION MODELING ” . OF A BEEF CATTLE ENTERPRISE : ‘TO INVESTIGATE MANAGEMENT ‘ 7 ‘ DECISION MAKING STRATEGIES ' ; Dissertation for'thiei Degree'of Ph.‘-D,;ffi* ‘ ’MICHIGANVTSTATE} UNIVERSITY? i - V; 7 {TMICHAEL RAYMOND IASKE , f : _§.‘:j*1:9767 "j V I l i I _——————-i —i, ‘ IIIIIIIIIJIIIIIIIIIIIIILIIZIIIIIIIIIIII L 5:33am}? Raiin’é‘sifmt Evita” E'nitszzsity Jan-1g? -;. ,rv-I-‘N'Tm- This is to certify that the thesis entitled , ". '- A System Simulation Modeling of a Beef Cattle Enterprise to Investigate Management Decision Making Strategies presented by Michael Raymond Jaske ‘ has been accepted towards fulfillment I of the requirements for Ph. D. degree in Systems Science 4&5 M a ) 7’1 Unfit/EELS My professor \ Date Ma 21, 1976 0—7 639 Li?! H 0mm 1: of he ( Tue ular cape ted 0X9 ABSTRACT KIA SYSTEMS SIMULATION MODELING OF A BEEF CATTLE ENTERPRISE To INVESTIGATE MANAGEMENT DECISION MAKING STRATEGIES By Michael Raymond Jaske Management decision making can be greatly improved by provision of better information about alternative choices to the decision maker. The complexity of agricultural systems, and livestock ones in partic— ular, has resulted in considerable neglect in early model building efforts to supply better information. Simulation models have the capability of surmounting many of the problems of modeling complica— ted dynamic system behavior. This thesis develops a simulation model of a general beef cattle enterprise to allow investigation of alter— native management decision making strategies. The model simulates the behavior of the major physical and financial variables of the enterprise through time. Although the model is intended for general use, attention has been concentrated on modeling of the land extensive cow/calf range operation. The model consists of five major components and several secondary components. MajOr components are (1) cattle demography, (2) forage growth, (3) feed stock accounting, (4) nutrient impacts on growth and reproduction, and (5) management decision making. The financial component is the most important component of the secondary group. Land allocation among alternative uses and crop production are not modeled and are arbitrarily specified by the user. A large FORTRAN computer program nd subrout A high alien of pt descriptor the nutrie Inetive ch; as a femct tennenatm end cost an profit The needs of Planning Wants o Tenant, Standarh h the “I IIII IIntro “I bl Michael Raymond Jaske and subroutines accomplish this system simulation. A highly disaggregated herd population model allows full explor— ation of population dynamics. Using TDN as the basic feed quality descriptor, and net energy for growth and maintenance where possible, the nutrient impact component predicts rates of weight gain and repro— ductive dynamics as a result of feed intake. Forage growth is modeled as a function of the exogenous weather variables of solar radiation, temperature, and rainfall. The financial component determines revenue and cost as a result of physical events and these determine the effect on profit, cash flow, depreciation, taxation, and debt levels. The management component has been developed to provide for the needs of managers in ongoing operating decisions as well as investment planning. Detailed control variables allow realistic simulation of events of interest to managers. The interactive structure of the com— ponent, with straight forward decisions made endogenously following standard economic criteria and more complex decisions made exogenOusly by the model user, is implemented in a batch mode to allow creation of files storing the model's values at decision points where user control is required. A decision tree of alternative courses of action can be developed by cataloging the files created at each encounter of a decision point. Three examples of management strategies are evaluated to demon— strate the capability of the model to assist decision makers. These are (1) early versus late calf weaning, (2) the rate of cattle herd development in investment projects, and (3) general profit maximiza— tion. These examples are discussed in terms of financial criteria. have be denonst state I which I Michael Raymond Jaske The thesis concludes with a summary of the accomplishments that have been realized, conclusions about the worth of the model and the demonstration examples presented, and a discussion of the current state of validation of the model along with improvements and extensions which appear to be helpful. rfi—iw SYSTEM SIMULATION MODELING OF A BEEF CATTLE ENTERPRISE TO INVESTIGATE MANAGEMENT DECISION MAKING STRATEGIES By Michael Raymond Jaske A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Electrical Engineering and Systems Science 1976 ACKNOWLEDGEMENTS This thesis has been completed with the help of many people. My major advisor, Dr. Thomas J. Manetsch, has been most helpful in reviewing various drafts, but most importantly he gave me the oppor- tunity to find my own direction and to develop as a professional. Dr. R. J. Deans and his student Margaret Schuette are responsible for initiating this project and the cooperative effort among us has been very beneficial. Dr. G. E. Rossmiller has provided advice on economic matters and potential uses of the simulation model in economic development. Other members of my committee——Dr. G. L. Park, Dr. R. A. Schlueter, and Dr. J. S. Frame—-have provided me with advice and encouragement. My wife, Marsha, has been most patient and supportive during the years of my graduate education. The financial support of the Department of Electrical Engin— eering and Systems Science, and the Korean Agricultural Sector Simulation Project, funded by AID contract csd-2975, is grate- fully acknowledged. This thesis could not have been completed without this support. ii j LIST OF IIST 0F Chapter TABLE OF CONTENTS LIST OF TABLES LIST OF FIGURES Chapter I. II. III. IV. INTRODUCTION GENERAL CONCEPTS 11.1 Systems Concepts 11.2 Modeling 11.3 Simulation 11.4 Use of Simulation Models THE PROBLEM STATEMENT 111.1 General Review of the Problem 111.2 Previous Approaches and Review of the Literature 111.3 Rationale for Selecting Simulation 111.4 Detailed Problem Statement GENERAL DESCRIPTION OF THE MODEL 1V.l Overall Model Organization IV.2 Specific Components 1V.3 Simulation Model Operation IV.4 Summary DETAILED DESCRIPTION OF THE MODEL V 1 Cattle Demography Component V 2 Forage Growth Component V.3 Feed Stock Component V 4 Nutrient Impact Component V 5 Management Decision Making Component Exogenous Decision Making Endogenous Decision Making 6 Secondary Model Components 7 Simulation Model Calling Structure .8 Summary 9 Glossary of Variables l7 19 25 25 27 31 32 39 39. 70 75 79 79' 97 112 115 122 127 144 160 195 197 198 Ill. VIII. IPPEII BIBLI VI. MODEL TESTING AND VALIDATION V1.1 Approaches to Validation V1.2 Validation Tests Logical Consistency Tracking Historical Data Expert "Eyeballing" Prediction V1.3 Summary VII. MANAGEMENT STRATEGIES V11.l Strategic Decision Making V11.2 Illustration of Strategy Investigations V11.3 Summary VIII. SUMMARY AND CONCLUSIONS VIII.1 SUmmary V111.2 Conclusions V111.3 Extensions to the Model V111.4 Implementation APPENDIX SYSTEMS SCIENCE REFERENCES BIBLIOGRAPHY iv 208 208 212 212 242 245 256 256 258 258 259 288 290 290 294 296 303 304 305 Table 5 Table 5 Table 5 Table I Table Table Table Table Table Tabl Tab: TaT Table 4.1 Table 5.1 Table 5.2 Table 5.3 Table 6.1 Table 6.2 Table 6.3 Table 6.4 Table 6.5 Table 6.6 Table 6.7 Table 6.8 Table 6.9 Table 6.10 Table 6.11 Table 7.1 LIST OF TABLES Age and Sex Description of the Cattle Herd Cohorts Definition of Crop and Cattle Expense Variables Definition of Resource Variables Parameter Values and Definitions for the Production Cost Element of the Financial Component Total Herd Population at Selected Points in Time Versus DT Step Size Comparison of Computed(Corrected) and Theoretical Herd Deaths Over Time Feed Stock Allocations Feed Stock Values Compared Over Time Sample Run of Feed Stock Use and Desired Allocation Selected Financial Variables Over a Simulated Time Horizon of One Year Comparison of Calculated and Simulated Values of Financial Variables Forage Digestibility Factors Investigated to Obtain an Improved Digestibility Range Weather Effects on Forage Growth and Forage Digestibility Sensitivity Testing of the Parameters of the Forage Growth Model Effect of Grazing and Harvesting Policies on Forage Growth and Obtainability Results of Sample Runs Investigating the Effects of Early Versus Late Weaning of Calves 46 172 174 185 215 217 220 222 223 226 234 237 248 266 Table 1.2 Table 7.1 Table I. Table I. Table 7 Table Table Table Table Table Change in Net Worth Over Time Breeding Method Cost Comparison Financial Effects of Weaned Calf Retention Policies Financial Effects of Alternative Feeding Plans Mature Breeding Cow Conditioning Effects from Alternative Feeding Plans vi Page 275 279 280 282 286 Figure 2. Figure 2 Figure 2 Figure T Figure Figure Flu Figure 2.1 Figure 2.2 Figure 2.3 Figure 2.4 Figure 4.1 Figure 4.2 Figure 4.3 Figure 4.4 Figure 4.5 Figure 4.6 Figure 4.7 Figure 4.8 Figure 4.9 Figure 4.10 Figure 4.11 Figure 4.12 Figure 4.13 Figure 5.1 LIST OF FIGURES Page General system diagram 9 System diagram illustrating multiple components 9 Block diagram symbols for mathematical operations 16 Relation between simulation step size and simulation error 18 General system diagram for a beef cattle enterprise 40 Graphical illustration of the response Y(t), to an input, X(t), for a distributed delay process 49 Graphical illustration of the response, Y(t), to an input, X(t), for a discrete delay process 50 The accumulated calving fraction versus time 58 Decision mechanism for slaughter cohort sales 63 Decision mechanism for forage harvesting 64 Decision mechanism for feed stock sales and purchases to obtain wintering nutrient requirement 65 Decision mechanism for herd feeding allocations 66 Blowup of the general system diagram to show functional details of the management decision making component 67 Flow diagram of the management decision making component in relation to the simulation model 71 "Decision Tree" of alternative strategies using the simulation model's restart capability 74 Enterprise system diagram illustrating physical and information flows between components 77 Calling sequence of the simulation model components to simulate the beef cattle enterprise through time 78 Decomposition of a kth order distributed delay into a series of k first order delays 80 Figure 5 Figure 5 Figure .‘ Figure Figure Figure Figur Figur Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Figure Kth order distributed delay with losses 81 Block diagram of subroutine HDMOG4 87 Determination of the annualized birth rate by numerical approximation of the deriviative of the accumulated births curve 90 Determination of BFRAC. values from the breeding history of each cohort subpopulation 92 Determination of the photosynthetically active quantity of plant material per unit area 99 Plot of points determining the temperature quality index, TPF(t), from the average temperature 99 Plot of points determining the soil moisture quality index, SMF(t), from the soil moisture 100 Plot of points determining the soil nutrient quality index, SNFn(t), from the soil nutrient 100 value Relationship between the grazing waste factor, AWFn(t), and the stocking rate, STOCKLn(t) 103 The intermediate digestibility, BASEDGn(t), as a function of pseudo—age, PINDEXn(t) 106 The growth quality factor,FRQUAL(t), as a function of the integral of the temperature index over the growth season 107 Plot of forage density digestibility factor against per animal density 103 Block diagram of the forage growth component 111 Block diagram of the feed stock component 114 Block diagram of the nutrient impact component 121 Flow chart of subroutine INQUIR of the management component 125 Flow chart of subroutine RESPNS of the management component 128 Figure 5.1 Figure 5. Figure 5. Figure 5 Figure Figure Figure 5.19 Figure 5.20 Figure 5.21 Figure 5.22 Figure 5.23 Figure 5.24 Figure 5.25 Figure 6.1 Figure 6.2 Figure 6.3 Figure 6.4 Figure 6.5 Figure 6.6 Figure 6.7 Figure 6.8 Figure 6.9 Page Subpopulation values from the female calf cohort prior to weaning action by management 133 Subpopulation values from the replacement heifer cohort following weaning action by management 133 Graph of culling rates for each subpopulation of the mature cow cohort using accumulated calving rates as a guide 139 Printed message at the spring growth onset decision point requesting values for specified control variables 145 Yearly pattern of crop harvests as exogenously specified 161 Linear interpolation between exogenously specified annual rainfall data points to obtain the current rainfall value 162 Expected cattle prices over time 164 Base digestibility values versus an age index 231 Seasonal digestibility factor versus season 231 Plant densities over the growing season 241 Rate of weight gain of male calf animals versus age 245 Forage densities over time as a result of basic climatic variables and grazing policies 251 Forage densities over time as a result of basic climatic variables and grazing policies 252 Forage densities over time as a result of basic climatic variables and grazing policies 253 Forage densities over time as a result of basic climatic variables and grazing policies 254 Forage densities over time as a result of grazing and harvesting policies 255 F r F r 1 Figure? ‘ i F i * Figurel b b Figure Figure Figure Figurr Figur F , r . i . .j Figu Figure Figure Figure Figure Figure Figure Figure Figure 7.5 7.6 Annual pattern of cyclic cattle prices used in demonstration runs Annual pattern of cyclic crop prices used in demonstration runs Annual pattern of solar radiation incident on the enterprise site Mature cow cohort populations through time for the three investment patterns Annual cash flow patterns through time for three investment cases Pattern of cash floWS through time over the three-year time horizon of the investigations for Cases 1, 2, and 3 Plot of the rankings of runs 6, 7, and 8 according to financial ranking and condition ranking Enterprise system diagram illustrating physical and information flows between components 262 263 270 274 287 292 gate a euterg beeu r roulc Tepar That in i and the CHAPTER I INTRODUCTION This thesis discusses the development of a simulation model of a beef cattle enterprise and the use of this model as a tool to investi- gate alternative management strategies available to the operator of that enterprise. Although a general model of an individual enterprise has been developed the emphasis of this thesis lies in the land extensive cow/calf operation. Coordination of this thesis with work in the Department of Animal Husbandry by Schuette has allowed incorporation of that work in the nutrient impact component allowing considerable detail in investigation of questions of nutrient feed intake on cattle growth and reproduction. An interactive management decision making component allows the user of this model to control the decisions made which affect the long range behavior of the enterprise. Chapter II discusses general concepts of systems, modeling, simula— tion and uses of simulation models. These ideas are the foundation upon which this thesis is based. Chapter III presents the general scope of the problem, reviews the relevant literature and previous approaches, and finally gives the detailed problem statement. Chapter IV gives a description of the model and its components without extensive use of mathematics. Chapter V presents the model and its components, both minor and major, in detail and lists all variable definitions in an appendix. The mathematical models used, the major assumptions made, and the exact equations upon which the computer programs are based are presented in this chapter. Chapter VI reviews the validation l grueedur properly sure mm of tires Chapter the ca] raking the at demon: siou grour spec bael procedures used to develop evidence that the model and its components properly represent their real world processes. Sensitivity testing of some model parameters is also presented to assist in identification of those capable of significantly altering the system's behavior. Chapter VII reviews the results of several demonstration examples of the capability of the simulation model to assist in various decision making evaluations. Chapter VIII presents the conclusions drawn from the construction and use of the model, reviews the results of the demonstration examples, and discusses areas of improvement and exten— sion of the model. It is anticipated that this thesis will be read by three distinct groups of persons-—systems science professionals, animal husbandry specialists, and management personnel. Because of the widely varying backgrounds of these three groups in terms of mathematical expertise and familiarity with computer simulation studies, not all of the mater— ial presented here is readily understandable by all readers. Specific chapter groupings will be recommended for each potential audience in order to facilitate understanding. Systems science professionals should read chapters 3, 4, 5, 6, 7, and 8. Animal husbandry special— ists should read chapters 2, 3, 4, 6, 7, and 8, while management per- sonnel should read chapters 2, 3, 4, 6, 7, and 8. An Appendix is provided which lists background references of particular interest to readers not familiar with systems science who wish to become acquaint— ed with it in more detail. M. m upueut lWEl'E these Serum ieuel EODVF CHAPTER II GENERAL CONCEPTS This chapter is written as an aid in understanding for those readers who do not have experience in systems studies or in the devel— opment and use of simulation models. Four important topics will be covered. First, some essential systems concepts will be presented; these form the basis for the entire approach used in this thesis. Second, modeling will be discussed in terms of the steps in model development, alternative modeling approaches, and useful diagrammatic conventions. Third, the idea of simulation will be briefly explained. Finally, simulation models will be examined from the point of view of the limitations and capabilities of different uses. This chapter will not be necessary for those having a systems background, or those with previous experience in simulation modeling. What is the rationale for using the "systems approach" in studying a beef cattle enterprise? The systems approach to problem solving is a general technique of analyzing needs to determine goals and evaluating alternative ways of achieving the desired goals. It offers the manager of an enterprise an opportunity to test management decisions to increase profitability, as well as the initial planning on which an investment is based. Although systems as an organized science is relatively new, the techniques and methods are drawn from the basic engineering disciplines. Systems studies and methods were first used in the aerospace industry because of the very complex problems typically encountered: the Apollo project is a notable example. During the 19603 business applications started as manufacturers realized that 3 umbilical solved u systems ratio-er beeoue scale 3 become are in the be thesis II.1 complicated production and inventory problems could be analyzed and solved using a systems approach. In more recent years the scope of systems studies has broadened rapidly. Extensions to economic and socio—economic systems, especially in large aggregate models, have become common. Systems studies are also being applied to smaller scale systems such as pest management and river ecology; these can become extremely complex when the myriad of details formerly neglected are included. The remainder of this chapter is devoted to building the basic foundation for the systems analysis application of this thesis. II.1 Systems Concepts This section will present some basic ideas and concepts of the systems approach to problem solving. These are the philosophy on which the methodology is based, the terminology most frequently used, and common graphical methods of representing the behavior of the system being analyzed. The Appendix to this thesis consists of a specialized bibliography that should be consulted for a thorough treatment of the material presented in this chapter. The methodology of the systems approach is based on a philosophy that views a problem globally, i.e., from the broadest possible perspective. The problem can be either one of design of a new system or of control of an existing system. To solve a problem one must first define it, but defining a problem requires knowledge of what one is attempting to accomplish. The desired goals of the problem must be r trauslat ‘ hurried: soluble iuuuivi seuera rue uu object the us parti grub] The translated into specific objectives that are to be accomplished. Knowledge of these objectives is vitally important to successful solution of the problem. These are found via a needs analysis involving all participants in the problem. If there appear to be several feasible ways to accomplish some of these objectives, then one must select from among the alternatives. Tradeoffs among various objectives must be resolved when they are in conflict. The designer, the manager, and the ultimate user should all have an opportunity to participate in assigning weights to the different objectives of the problem. This definition and selection process is highly iterative. The intent is to discover the solution which best meets the objectives, not merely to find a solution. The systems approach to problem solving is an organized methodology for developing specific problem objectives from the desired goals, and using these objectives to evaluate and select possible alternative solutions Clarification of the terminology used in any discussion greatly assists in the communication process. A number of words and phrases have meanings specialized to systems science; these are defined below along with an example in the context of a beef cattle enterprise. 1. System——a collection of components with some form of regular interaction. example——the beef cattle enterprise itself 2. Causality—-the concept of identifying the cause and its resulting effect, a cause precedes its effect in the time sequence of events 10. 6 example-—a decrease in rainfall (the cause) resulting in a slowdown in forage growth rates (the effect) Input variables—-variable consciously brought into the system to affect its behavior example—~feed crop purchases for use as cattle feed Output variables——those variables which represent the result of the system's operation and of the input variables. example--animal sales to the market Component—~an identifiable grouping of relationships between input and output variables. example——forage growth and removal through grazing Controllable inputs—~inputs to the system that are con- trollable by some policy action. example—-purchases of feed grains Uncontrollable inputs—~those desired inputs to a system which cannot be controlled. example-~amount of water naturally available on a pasture Exogenous inputs-—those inputs to the system as a result of the system operating within an environment with influences on the system. example——cattle prices System parameters-—particular values affecting the operation of the system through its structure. example—-basic photosynthetic efficiency of forage species Transient—~the time behavior of the outputs which result from zero level control inputs, commonly transient behavior decays toward zero with time. example—~changing age distribution of the breeding herd when culling is stopped or thau pert reg 1F ll. Steady state—~the output variable behavior after all transient effects have died away example-—the final age distribution of the breeding herd after changing effects are eliminated. 12. Asymptote-—the limiting value toward which a sequence of numbers approaches. example-—the total births of a calving season toward which the accumulated number of births approaches over time 13. Stochastic variable--a family of random variables, the elements of which are functions of some other variable, commonly time. example—-month1y rainfall as a function of time of year This list of terms is by no means exhaustive, but it does cover the major terminology that will be encountered in this thesis. Graphical representations are frequently more easily understood than verbal descriptions. Systems diagrams take advantage of this perceptual characteristic of humans to present very concise systems representations. The following diagrams are general ones outlining in diagrammatic form the terms defined above. 11.2 Modeling Modeling is the process of developing a mathematical description 0f the interrelationships between the input and output variables of the system. The model is the resulting description. To model an Object or a process, whatever its nature, requires one to specify the most important of that which is known about it. Not uncommonly, several variables are known to be important, but the exact relationships i are Fuel oneself huh gr attempt all in rent, and SI squat used sele uTF are unclear. The modeling process can be very helpful simply by forcing oneself to be specific about ideas not usually articulated clearly. Much greater understanding of a process can be gained by those who attempt to model it because they are required to integrate together all known features. This section will present steps of model develop- ment, validation of the model, various approaches to modeling itself, and some diagrammatic conventions. Only models using mathematical equations will be included in this discussion because that is the type used in this thesis. An all important first step in model development is concept selection. What is the concept we wish to model? Different concepts of the same object or process are possible depending upon the view point of the modeler. For example, consider the process of forage Plant growth. All concepts of this process must interrelate the exogenous inputs of weather to the outputs of plant growth. One concept of this process could take a mass and energy balance approach using the requirements of growth: solar radiation, carbon dioxide, water, and soil nutrients. This would require development of math— ematical relationships for root uptake of fluids, photosynthesis, resPiration, and transpiration. Another concept could take a less fundamental view by using experimental observations to allocate the energy intake from solar radiation to growth depending upon the relative "productive quality" of water, temperature, and soil conditions. A model based on the first of these concepts would be able to examine Cuntr Input Uueu lupr Fig Exogenous Inputs Controllable Inputs .____——————i— SYSTEM Output Variables _____—.-..g "——" ’ Uncontrollable I Inputs | System Parameters Figure 2.1 General system diagram Exogenous Inputs S stem Boundar __ _ _ __ I‘ ' Z,.. ' “'“ ‘y ----- '1 | i I l 1 r COMPONENT | TWO | Output g g Variables l I F r Input I I Variables : i | l | T F F i “A.” COMPONENT g ,7 I | ONE ' Output g T Variables L _________________ .1 Figure 2.2 System diagram illustrating multiple components certain : it would against oi plani requiri details and van detail better the CF “560 rela rel: rea ‘Fii 10 certain aspects better than the second, but by being more comprehensive it would be harder to develop. It would be more likely to run up against the unknown in the biochemistry of growth. These two concepts of plant growth would be used for different types of investigations requiring different levels of detail of the process. If a highly detailed model is desired, but the data to provide parameter values and variable relationships is unavailable, then the structure of the detailed model can be developed using crude working guesses until better information can be incorporated into the model. Selection of the concept, then, is important for it should be compatible with the use of the overall model one is developing. Once the concept has been selected, then the model itself must be developed. Here one must be careful to include variables and relationships which are relevant; one should exclude variables and relationships which are irrelevant. Both are equally important. The reason for this is simply that time and energy can be saved by excluding variables and relationships which have no direct bearing on the process being modeled. For example, in a model of the forage growth process, solar radiation, temperature, rainfall, and soil nutrients are important. Relative humidity has some effect on transpiration, but in the overall process of growth it is relatively unimportant and may be excluded. Relative humidity is irrelevant in this case. Variables and relation— ships which are pertinent should be included; in general, those not Pertinent should be excluded from a model. F or is that the lure of intro of ham large a approaa people are so i proee: deuel uieng once and sepe uitF den ll A common discovery in models composed of a number of components is that the individual components may be known rather precisely, but the interactions between them are indistinct. This is partly a result of increasing specialization in research, and partly from an inability of human beings to grasp the whole picture of what is happening in a large and complicated process. Fortunately, the use of the systems approach in multidisciplinary applications has alerted a number of people to this tendency in research, and one can hope that these gaps are soon eliminated. A basic decision to be made early in the model development process is whether the model will be broken up into components or developed in its totality. From a mathematical and computational viewpoint it would be better for the entire model to be developed at once. This is not the usual method. Most frequently a model can be, and is, divided into separate components with each component developed separately. The component model interactions can be remembered easily with the aid of diagrams such as Figure 2.2. An advantage to component development is that concept selection is easier for a less comprehensive process. Other advantages are the ability to use component models developed elsewhere, and the ability to test the model much sooner than if the entire model were developed at once. The last reason is particularly important in validation of the model. The model developed in this thesis follows the component organization pattern. Validation of the completed model is one of the most important steps in its development. Validation is the basic process of verifying that the describe correct, Federal logical able tc data in be gal iaiorn oi the as th regal when the ‘ iii 12 that the model correctly represents the real world it attempts to describe. One can never be fully satisfied that the model is totally correct, but use and experience with the model can bring confidence. Generally validation can be assured by checking that the model is logically consistent, follows intuitively obvious patterns, and is able to track historical data. In many instances there is no historical data to track, however, and confidence in a model's validity can only be gained with use. When the model is of an existing system of which information can be gathered, then a model can be validated by prediction of the future. This approach works well in business management models as the system--the business-~is monitored frequently, and data is regularly collected. Confidence in the validity of the model is gained when the model and the real world agree. When they do not agree then the model can be improved by using this newly recorded historical information. Only when a model has passed the validation stage should it be considered an adequate description of the system it represents. Two final steps in the model development process are sensitivity testing and stability analysis. In some senses these are both means of checking the validity of a model, but they go beyond merely that. / Stability analysis and sensitivity testing are procedures whereby l individual and groups of model parameters are adjusted and the model outputs observed. Sensitivity of a parameter is said to exist when small changes in the parameter are followed by large changes in the output variables. In the sense of validity testing the model behavior should mad world is ‘ taooot he very used heusitiv ol the s parameto of a mo lluotuz output: in des thistd helav would taint that the 13 should match the real world. In many cases, the parameter in the real world is unknown, so the knowledge that it is sensitive in the model cannot be used to validate the model directly. This information is very useful, however, in directing the research going on in this area. Sensitivity of a parameter indicates its importance in the operation of the system; if past and present research have overlooked this parameter, then redirection of efforts should be beneficial. Stability of a model or a particular parameter set is said to exist if transient fluctuations induced by sudden input changes diminish, so the model outputs adjust to a steady state. This aspect is particularly important in designing systems or in determining policies for controlling existing systems; undamped oscillations are usually very undesirable behavior. As an example, excessive seasonal fluctuations in herd size would be undesirable. As with formal validity testing, the information gained in sensitivity testing and stability analysis is used to verify that the model is correct or to improve the model. There are several approaches to modeling that can be selected by the modeler depending upon the system and the intended use of the model. A major characterization of a model is whether it is static or dynamic. A static model is one in which time plays no role; the model only describes variable interrelationships at one instant in time. For example, a static model is the optimization of crop sales and purchases for feeding uses at crop harvest time. A dynamic model includes time; variables may change their values from one instant to another. Even the relations between variables can change with passing time. An example i [ of a W . . and new i. o l a a for mode essentie of the ‘ variahl avallal esaopl doom to model the e the a rep: seal 14 of a dynamic model is the relation between quantity of feed consumed and amount of weight gained in growing cattle. Another classification for models is "black box" versus structural. "Black box" models essentially deal with input and output variables; the inner workings of the real world process are useful only as a guide in selection of variables to use. This approach can only work if historical data are available from which input/output relationships can be determined, for example by regression analysis. Structural modeling derives its name from models which incorporate the structure of the real world. Structural modeling is more complex than "black box" modeling, because frequently the exact structure is unknown. Nevertheless, structural modeling is the approach used when feasible because greater confidence in the model's representation of the real world exists. No model will ever be perfectly representative of the process it seeks to describe. Man's knowledge of the world is too limited for that to occur. Even the totality of man's understanding may not, and probably will not, be included in a model. There is simply too much information to use, and the relation of the parts to the whole are indistinct. One must always work with models that are imperfect in some way. As human understanding grows and is incorporated into models, they will perform better and have fewer shortcomings. Even imperfect models are still valuable because they are so much better than no model at all. This last point bears repeating; generally models are so much better than no model that the imperfections are 15 gladly tolerated. Decisions based on use of a model will be very useful and correct provided the limitations and imperfections of the model are recognized. Decision making is a process which involves uncertainty. This uncertainty arises from lack of understanding of the process in question, sampling errors in the data collected to represent the real world, and the presence of stochastic variables. Imperfect models are a result of this lack of understanding of the process, or perhaps conscious exclusion of factors thought to be irrelevant. Proper qualification of the predictions of models, and experience in using a model should reduce the uncertainty due to model imperfection. This still leaves the basic sources of uncertainty in decision making. These can be partially overcome by certain statistical methods (Monte Carlo usage of modelsl), but never completely eliminated. In short, decision making will remain risky. Modeling is greatly assisted by graphical aids that present concise visual representations of the interrelationships among variables. The human eye can readily discern the multiple interactions between variables in graphical form; whereas it is much more difficult to observe these relationships in a series of mathematical equations. M l. Hahn, G.J., "Sample Sizes for Monte Carlo Simulation", IEEE Transactions on Systems, Man, and Cybernetics, No. 1972, pp. 678—680. I|\ ——§D—* (l) (1) [dt .2. dt (1) + (3) + (2) (l) (3) Th (2) (1) (3) d (2) (2) (2) l6 add quantity (1) to quantity (2) to get quantity (3) multiply quantity (1) by quantity (2) to get (3) divide the numerator (l) by the denominator (2) to get quantity (3) integrate quantity (1) to get quantity (2) differentiate quantity (1) to get quantity (2) take the value of quantity (1) to use at location A Figure 2.3 Block diagram symbols for mathematical operations ll.l Simulat Develops: is an extrenn first step. step. is tl in finding : are theoret letted is n ldodels tha snail; for hedone than is a part connotes. is: beyond a he made alllltonin lniinit tineso Shula linuoo takes and t abild Shin allot 17 11.3 Simulation Development of a model as a mathematical description of a system is an extremely useful and beneficial undertaking, but it is only the first step. Solving the equations that make up the model is the next step. As the size and complexity of models increase, the difficulties in finding a solution become greater. While several possible methods are theoretically available for determining a solution, the one se— lected is nearly always closely related to the size of the problem. Models that are amenable to analytic solution are generally very small; for this reason modeling was not a very well-developed study before the introduction of the modern digital computer. Simulation is a particular method of solution for a model which employs the computer. A simulation model is a further abstraction from the real world, beyond a mathematical model, because certain approximations have to be made to make the mathematical model solvable. The most important approximation is one dealing with the time dynamics of a model. The infinite number of points lying between the starting and stopping times would present an impossible task if each were to be solved. A simulation model uses discrete time instances to approximate the con— tinuous time which actually exists. Using discrete instances in time makes the number of points where the model is to be solved finite, and thus open to reasonable computation. Simulation, because of its ability to easily handle time dynamics, is the method of model solution frequently chosen when a model includes time. The use of discrete time in place of continuous time is an approximation that inevitably introduces error into the solution that la ohtainu of deenea onooher oi as is ea: tnadeoif oodel. thunh c doothen iron I‘ asynpt that d glvln lllus Duh Vii 18 is obtained. The size of this error is generally a decreasing function of decreasing step size (the interval between time points used). The mmmer of steps required is inversely proportional to the step size, as is easily shown. The modeler is then confronted with a classic tradeoff; a tradeoff between simulation error and cost of running the model. There is no best step size to use, as a widespread rule of thumb calls for the smallest step size the project budget can afford. Another criterion frequently used employs the information obtained from running the model with successively smaller step sizes until an asymptote is reached for an important output variable. By assuming that the asymptote value is correct, one can choose the step size giving a value within the desired range of error. The following figure illustrates this idea. ouigut range of permitted variable error value __________-____-_______ J metote ~----°——o—-—--—-—--—--——-'--- o o ‘ | O | O I o I O I I l i l l l I I l maximum Simulation Step Size step size Figure 2.4 Relation between simulation step size and simulation error Siuul because it programmi the tonnpn ondary s handled oooponen nodels. dump a but oh lngiul than n is an Q0llipl sols l9 Simulation offers an advantage as a method of model solution because it is able to utilize the organization pattern of computer programming. A program is the main section of instructions directing the computer to perform specific operations. A subroutine is a sec— ondary section called from the main program. Simulation can be handled as a program with subroutines as components. Changes in component models then mean changes in the subroutine simulation models. A model can be made operational rather quickly by inserting dummy subroutines that have the proper interaction and connections, but which may not actually perform any calculations or have a mean— ingful effect. Debugging of each individual subroutine is much easier than would be debugging for one single program model. This advantage is another practical reason that the model development process uses component models more often than large integrated models. Simulation is then, first of all, a method of determining a solution to the mathematical model. In practice, a simulation model is a further abstraction from reality because it makes approximations that need not be made in the mathematical model. There are tradeoffs to be evaluated and made, mainly between simulation—induced error and cost of running the model. The advantages of component models in reducing the debugging time required are an asset to simulation. Finally, simulation is an extremely valuable technique for developing solutions to models possessing time dynamics. II.4 Uses of Simulation Models Simulation models can be used in a wide variety of ways. Important among them are testing specific cases, running "best versus toast" cash Emotions. attenistio dnon anon decision hllities oi the r parapet use al the no tria ahle dee pan 20 worst” cases, Monte Carlo applications, and optimization of objective functions. The goals of the user, the available budget, and the char— acteristics of a particular use are all important elements in selection from among these. This section will discuss alternative uses, the decision criteria for choosing among them and the limitations, capa- bilities, and costs of each. Costs will be viewed primarily in terms of the number of simulation runs required for that use. A fundamental use of simulation models is testing alternative parameter and input variable values to determine the outcome. This use allows a very creative interaction between decision makers and the model as they are able to test various ideas and intuitions. A major benefit that occurs is the learning experience on the part of the decision maker as he gains a better feeling for how the system works. The ability to interact with the model, learn from previous trials, and try new values not previously felt helpful is very valu— able. In this use the simulation model functions as a tool of the decision maker, extending his ability to evaluate the outcome of particular policies. An example of this use of simulation models can be found in a model of cattle grazing on natural pasture. A management policy over stocking rates is necessary to maximize the effective use of forage growth. A ranch manager can experiment with different stocking rates, and evaluate their effectiveness by using the weight gain of the cattle at the end of the season as the desired output variable. The manager can gain a better understanding of the grazing process by using the simulation model to test his intuition and the traditional methods of grazing control. This he cost 1' case. A 1 says very nandononen rated in not soil on cont llllCEIta to inpt l in a s that one o in t’ the 151 21 This first use of simulation models is extremely easy to organize. The cost is quite low, since only one run is required for each test case. A limitation is that a single run of a deterministic model says very little about the likelihood of that outcome if there is any randomness in the real world system. If randomness has been incorpo— rated into the model to reflect the real world, then a single run is not sufficient to allow choice between alternative values of parameters or control variables. An additional limitation is that one may be uncertain how to go about modifying the choice of control variables to improve upon the outcomes of a previous run. Perhaps the easiest way of dealing with aspects of randomness in a simulation model is to devise the worst and best cases of inputs that could be encountered. By running the model with these two cases, one can determine bounds on the outcomes that should not be exceeded in the real world system. All other outcomes will be contained within the "envelope" determined by the best and worst cases. This method is only slightly more difficult to use than the first. A limitation is that ”best and worst" cases may be difficult to determine, espe- cially when there are multiple sources of randomness in the real world system. Another limitation is that one still does not know the most likely outcome, only the bounds on the outcomes. Costs are quite low for this use of simulation models, since only two runs are required to evaluate each parameter or input variable desired. A much better method of dealing with randomness in the model is using the simulation model in a Monte Carlo mode. This requires that the probability distributions of each source of randomness be known; Perhaps a difficult requirement to meet. A run of the simulation model ls nade oi each stool corded. again did liter a tisties naniaoc he used on part tion on sequin cost iteqn SW VET 22 is made with particular values drawn from properly random sources for each stochastic variable. All output variables of interest are re- corded. Another run is made with a new set of random variables, again drawn from the proper distributions, and the outputs recorded. After a sufficiently large number of runs have been made, the sta— tistics of the outcomes can be computed with confidence. Mean and variance would be the most important. These computed statistics can be used to develOp the confidence intervals of the ponicy variable or parameter under study. The major difficulty with using a simula— tion model in a Monte Carlo mode is the large number of simulations required to properly determine the statistics of the outcome. The cost of Monte Carlo analysis for very large simulation models is frequently so high as to be prohibitive. The final category of simulation rodel use is optimization of specified parameters or control variables. This method can provide very beneficial results when an objective function can be specified for minimization or maximization. Optimization is very important in design of new systems. In problems of management policy or control optimization of those policies, the optimized values should result in the best system operation possible. A major difficulty with optimization of simulation models is the large number of simulation runs required: the same difficulty with Monte Carlo analysis. The cost of the required runs may preclude use of optimization. Another limitation of optimization in simulation models involves models that have unspecified or arbitrary exogenous inputs. Specification of that input allows optimization to be performed, but the values obtained as optimal mother v1 optimal a The in decid let us I ohjecti‘ pith a nainial the co and ti study about cons and the lion 23 as optimal are conditional on the specific exogenous input used. If another value is encountered, then the "optimal policies" may not be optimal at all. The objectives of the system study are the most important aspect in deciding which way a simulation model is to be used. For example, let us reconsider the cattle ranch possessing natural pasture. One objective could be to determine the most likely weight gain of cattle, with a specified stocking rate subject to the stochastic variables of rainfall and temperature which influence plant growth. A third objec- tive could be to optimize total revenue by controlling stocking rate and timing of animal sales. Each of these three objectives of a study w0uld use the same simulation model, but in different ways. Cost is an important factor to consider when making the decision about the way a simulation model is to be used. It should also be considered from the start of the entire modeling process, as the form and content of the model should be guided by the ultimate goals of the system study. If an objective can only be achieved by use of Monte Carlo analysis, for example, and the budget is tight, then the content of the model, or sometimes its depth, can be altered. An idea to keep in mind throughout is the relationship between the cost of the entire system study and the benefits obtained from it. 11.5 Summary This chapter has discussed the systems approach to problem solving, some topics of systems science, and simulation models. A review of the systems approach is, once again, an organized methodology of problem solving that seeks to achieve the ultimate problem goal by selection from the objects n ships am soteessl sion mal problem approan develo result tions tech and read det 50lll 24 selection of a solution which best meets specific objectives derived from the goal. A systems approach works best where the process and objects under study are well enough known that mathematical relation- ships among variables can be specified. Another requirement for successful application of the systems approach is that adequate deci— sion making power exist to carry out the solution, whether it is a problem of planning (design) or of management (control). The systems approach utilizing simulation models is not perfect. The models developed for the system under study can never be made perfect. The results they predict must be viewed with an eye to their imperfec— tions. On balance, however, the systems approach is a valuable technique for analyzing problems and developing solutions to them. By necessity the coverage of this chapter has been very rapid and it has only touched upon the high points of these topics. If the reader is interested in pursuing these subject areas in greater detail, then the references listed in the Appendix are an excellent source of introductory information. e lots stun l ranch M T ol the p of the 5 Ill 1 CHAPTER III THE PROBLEM STATEMENT This chapter will present four topics: the general description of the problem, a review of the literature, a rationale for selection of the simulation method, and the detailed problem statement. III.1 General Description of the Problem This thesis is a study of a beef cattle enterprise: a farm or ranch operation involving the breeding, growing, and selling of cattle for slaughter. There has been a steady historical progression of specialization in the industry to either breeding or fattening. Moreover, the availability of low—cost grains, chiefly corn, has led the fattening operation to lose its requirement of a land base. Feed lots are the result. Management of such feed lots has been extensively studied in recent years. The cow/calf operation, which is important in the production of feeder calves, has received little attention. The exception to this rule has been better weight gaining charac— teristics. The current depression in the beef cattle industry, due to both low cattle prices and the world food supply problem, is caus— ing rethinking about the organization and methods of operation of the entire cattle industry. This thesis concentrates its attention on a specific type of beef cattle enterprise: one which is characterized by a land— extensive cow/calf organization. Further, this operation is 25 to he small en its purchases he scale of methods of 51 individual e lot this ope This t? nanagementd received es individual ilcieucy h thesis vii tion of 1 0r produc this the: bleeding Pat is so a dition; Pledie alumina 0f the hunth 26 to be small enough to have no market effect from either its sales or its purchases. That is, prices remain unchanged by its activities. The scale of operation is sufficiently large, however, that the methods of systems and simulation can be profitably applied to this individual enterprise. Rough bounds on the number of breeding cows for this operation might be 50 to 1,000 animals. This thesis also concentrates on particular questions of management of therenterprise. Management is an area that has not received extensive study at the scale of the overall system operation. Individual aspects of production and certain forms of production ef- ficiency have received the bulk of the analysis here to date. This thesis will not make any analysis in the areas of genetics, alloca— tion of land to alternative uses, by—product or destination marketing, or production of crops (except grazing forages). The emphasis of this thesis is on Study of management policy in the areas of grazing, breeding, timing of sales, and general herd management. Part of the reason why study of the management of this enterprise is so attractive is the general low level of mathematics used in trar ditional practices. A major problem in the past has been a lack of Predictive capability. Extrapolation of past trends has been a dominant method employed. This is not sufficient when the dynamics of the beef production process have delays approaching two years. Another problem has been uncoordinated analysis of "separate" aspects 0f Production, rather than an integrated analysis of the whole. This tendency has lead to dominance of certain types of emphasis, such as genetic ma of husbanr‘ tore typir are pastt 'rrg leve‘ rrrket c of these perform 111.2 27 genetic manipulation. A general characteristic has been the dominance of husbandry practices and recommendations over economic considerations. Some typical problems unanswered by traditional management practices are pasture management to maximize efficient forage use, proper feed- ing levels to improve net profit, and the proper response to changing market conditions. This thesis seeks to contribute to the solution of these past problems and to increase the level of management performance. III.2 Review of the Literature An extensive literature discussing the beef cattle enterprise exists. Much of it is characterized as non-quantitative explanations of various subprocesses and reports of experimental observations. To a large degree, this literature is spread among variOus specialized disciplines such as animal nutrition, genetics, crop sciences, agri— cultural economics, and physiology. Relatively few studies have viewed the beef cattle enterprise as a system and rigorously ana— lyzed it from the systems science point of view. The quantitative studies which do exist have taken three broadly distinct approaches; namely, analytics, optimizing, and simulation. Of these, only a minority use the simulation approach; the one seeming to offer the most potential for significant gain in this complicated, involved Process of beef production by individual enterprises. The remainder of this section will review the previous work in the area of mathematical representation of, and management control over, beef cattle enterprises. The three approaches of analytics, optimizati unit, alt} M iii? studies limited some at trtoriz ships 1 new tr respo procr rati fror exa rat 28 optimization, and simulation will each be discussed as a separate unit, although there is some overlap between them. The Analytic Approach While a strict definition for analytic would include all of the studies referenced here, the term is used to describe those which are limited to functional representations between variables. There are some areas in which the analytic approach has been most prevalent: tutorial papers, analysis of data to determine values for relation— ships between variables, and attempts to offer advice to management. Representative of tutorial papers designed to present material new to the readership are development of general crop and livestock response functions[12], traditional descriptions of the cattle growth process[5], and the use of marginal analysis to optimize feedlot rations[4]. Uses of analytics to develop mathematical relationships from experimental data are too numerous to mention in detail. Several examples will serve as illustrations: algebraic growth models from cattle age-weight data[6], comparison of calving performance between drylot and pasture[37], and comparison of traditional versus modern range practices in Argentina[l3]. Carpenter's paper[9], developing an algebraic growth model, is typical of analytic means of offering advice to management over questions in the production process. The simple analytic approach to studies of the production process is inadequate for such complicated systems as this enterprise. While the analytic approach is not used appropriately in studies of the overall enterprise, it does continue to have value in determining the relationships between variables from experimental data. Mali} is 3 1 [or the la red uoulir he usual detemini return ir used a s ruu size Suhuablt uf resu Th have er method ire d be 1i tiue. prov dung tiu the 29 The Optimization Approach As a general technique, optimization has been extensively used for the last three decades. Linear programming, dynamic programming, and nonlinear programming are some important methods of optimization. The usual agricultural application of these techniques has been in determining the best combination of feed inputs to maximize economic return in beef cattle feedlots[29,53]. Long, et al.,[33,34,35] have used a simulation model in an optimizing mode to study the effect of cow size on productive efficiency in drylot and pasture operations. Schwab[49] has used linear programming to optimize the entire range of resources used in a cow/calf operation. The majority of applications of optimization in beef production have employed linear programming as the optimization technique. This method has certain restrictions which limit its usefulness in study- ing dynamic processes. Linear programming requires that the variables be linearly related, and that the relationships not be functions of time. A model that can be optimized using linear programming cannot provide an adequate description of the population and reproduction dynamics of a breeding cattle herd. All optimization algorithms possess some type of objective func— tion that is either maximized or minimized. While there is no doubt that the optimized variables are optimal for that objective function, there is doubt that the objective function correctly specifies what an enterprise manager wants optimized. For example, an important characteristic to some managers might be a flexibility of operation Which allows some room to maneuver if expected conditions do not mater- ialize. An objective function cannot measure an operations flexibility. hdirionall hat are at huot a m problems i he Simul Sim rudeis. is ireqr he mjr have be by usir psis i he he the p tart] 30 Additionally, the usual methods have trouble with exogenous variables that are arbitrary or stochastic. For these reasons, optimization is not a method of analysis which provides a way of solving all of the problems facing an enterprise manager. The Simulation Approach Simulation is a method for developing solutions to mathematical models. Because of its ability to handle time dynamics, simulation is frequently used where time is an explicit part of the problem. The majority of uses of the simulation approach in beef cattle studies have been macroeconomic in scale. Initiation of economic development by using the cattle industry[26,3l,42,43] and government policy anal— ysis in the agricultural sector are well represented in the literature. The World Bank has used this approach for several years in evaluating the potential value of economic development projects concerning the cattle industry[10]. Simulation of beef cattle at the enterprise level is beginning to appear more frequently in the literature. A pioneering work by Halter and Dean[23,24] studying management in an integrated range/ feedlot operation appeared in 1965. Cartwright and Long of Texas A&M University have participated in a number of works which concentrate 0n questions of cow phenotype optimization using a simulation model of cattle energy requirements[28,33,34,361. Witz[54] has applied Simulation to the feedlot ration optimization problem, while Afzal[l] has studied inventory modeling to respond to drought conditions. Many more references could be cited, but these serve as illustrations of the aspects of a beef cattle enterprise that been studied using the simulation purrogrart 111.3 Rat This abeef ca the great process ; acterist the sion in the tion tr hot on a turd intake Feed ture a up of e‘ the the 31 simulation approach. Joandetl has recently published an extensive bibliography of simulation applications involving beef cattle production. III.3 Rationale for Selecting the System Simulation Approach This thesis will use the system simulation approach in analyzing a beef cattle enterprise. Simulation was selected because it offers the greatest potential for understanding and solving the model of the process as involved as production of beef cattle. What are the char- acteristics of the enterprise that are so troublesome as to require the simulation approach? Most important are the time delays inherent in the growth process. In the U. S. the time required from concep— tion to marketing for slaughter animals may range from 11-30 months. Not only are time delays present, but the length of these delays is a variable which is partially under the control of management. Feed intake rates are an important control variable affecting growth rates. Feed intake also governs the reproductive characteristics of both ma- ture cows and grOWing heifer replacements. Reproduction, then, is also a dynamic process. In short, the entire herd population is in a state of dynamic flux. Herd population dynamics is a major complexity of the beef cattle enterprise requiring the power of the simulation approach. A similar source of complexity is the decision making that controls the enterprise's sales and purchases. Decisions involving breeding, because of the delay length in the growth process, will not affect sales revenue for nearly a year. The proper price to use in decision ________________________'___ lJoandet, G. E., and T. C. Cartwright, "Modeling Beef Production Systems," Journal of Animal Science, Vol. 41, No. 4, 1975, pp.1238—46. “king is that hmagememt der hat are to 0 tipated saleE decision makl his implies the beef prr is another Other tions of it into more ‘ gregated t applicable ies of th times are thermore “limit section “'11 st 111.4 an em breed POrtj emti 32 making is that which will be in force when the product is sold. Management decisions are conditional on the stream of expected prices that are to occur over the interval between the present and the anti— cipated sales date. To further add to the trouble of the enterprise decision maker is the well known volatility of agricultural prices. This implies that price expectations change rapidly. Control over the beef production process is very complicated and uncertain; this is another powerful impetus toward using the simulation approach. Other studies of beef cattle have developed partial representa- tions of the production process as a way of decomposing the system into more manageable pieces. These studies cannot, however, be ag- gregated together to form proper recommendations that are generally applicable. While simulation of the production and management dynam— ics of the enterprise is possible, it is not easy. Very long lead times are required to model and then simulate these processes. Fur- thermore, simulation models can never be perfect because of the approximations necessary to determine the solution. The following section will describe in detail the exact problem that this thesis will study. III.4 Detailed Problem Statement The system to be analyzed here is a beef cattle enterprise. Such an enterprise can be characterized as being land extensive, involving breeding of cows for calf production, and producing a significant pro— portion of the nutrient requirements of the herd through forage growth. It is of such a scale that it can be treated as an atomistic economic entity. The prr distinct gr his thesis itself. T1 spsten: ca rndnutrie managemen mamagemer prior in manager of exogn tapabil and in are de thhtrn grege the act tie tin 33 The problem addressed by this thesis is the construction of two distinct groups of components of this beef cattle enterprise. First, this thesis develops a dynamic simulation model of the enterprise itself. This includes models for the f0ur major components of the system: cattle demographics, forage growth, feed stock accounting, and nutrient impacts. Second, this thesis develops an interactive management algorithm which allows the user to investigate very detailed management control policies for ongoing operational decision making or for investment planning. This algorithm is structured so that the manager can evaluate alternative strategies to cope with the effects of exogenous variables on the enterprise. Three examples of the capability of the model to assist decision making for ongoing decisions and investment planning are presented. The simulation models of the four components of the enterprise are designed to be useful in analyzing various management policies to control the cattle herd and related processes. This requires disag— gregation to a level such that the herd can be controlled as discrete— ly by the user as could the manager himself. This also requires that the control variables of the model be devised so that a manager's actions can be simulated. Since a major weakness of previous simula- tion models of beef production enterprises has been a lack of popula— tion dynamicsz, and since these are required to correctly evaluate some herd control policies, the demographic model fully develops herd Population dynamics. Introduction of herd population dynamics requires that cattle nutrition and reproduction dynamics3 be included to ————__.____~____i____.___ 2Ibid., p. 1243. Developed by Margaret Schuette in partial fulfillment of M. S. requirements, Department of Animal Husbandry, Michigan State University. attentively a“ promth and fee magement con sent decision represent an enterprise tn time. Sever nun-physical component t and other e The no stem from cesses oi ponemt mo lanp cont thereby This Tar variable because them, 3‘ This in renal been an in l Waj “hit the 34 effectively and correctly describe births and deaths. The forage growth and feed stock components are constructed with the appropriate management control variables to allow the entire spectrum of manage- ment decision making to be exercised. These four component models represent an adequate description of the physical processes of the enterprise to allow realistic simulation of the system behavior through time. Several secondary components are developed to accomodate the non-physical aspects of the enterprise; an example is the financial component that determines the financial effect of purchases, sales, and other activities. The management algorithm has certain unusual requirements which stem from the nature of the simulation models of the physical pro— cesses of the enterprise. As an example, the cattle demography cone ponent model maintains populations on a very disaggregated basis. Many control variables must have the same level of aggregation, thereby requiring many values to actually control herd populations. This large number of effective control elements among a number of variables has the effect of eliminating all optimization techniques because of the extremely high cost of their use. Management control, then, is exercised exogenously when certain decisions are required. This necessitates an interactive control algorithm, but since the usual source for computer interaction--the teletype--is ruled out because of the large number of control elements that must be inputed, an interactive batch mode of operation of the model has been devised. A way to control the model's simulation thr0ugh time is developed which allows stops and restarts of the simulation whenever required. A beneficial feature of this interactive management algorithm is was—ea...“ we .. . that multil)1e used as the b algorithm reE runs, inputir decision to The to heel cattle common to 1 model is t puirement trol varir iluence t minimum aggregat in theh Specifi. this sy Tl brieil lust c The 35 that multiple management policies can be evaluated, and the "best" used as the base from which the simulation is restarted. Use of this algorithm results in a time simulation which is a series of short runs, inputing control variable values as required by the particular decision to be made. Specific Component Requirements The four components representing the physical processes of the beef cattle enterprise must meet certain requirements. A requirement common to all is that each be dynamic, because a major element of the model is the inclusion of time in an explicit manner. A second re— quirement common to all is that the component models contain the con- trol variables that allow the manager‘s decisions to direct and in— fluence the physical processes of the components. A third common requirement is that each model be constructed in a sufficiently dis— aggregated faShion that a real manager's decisions can be simulated in the behavior of the model. These general requirements and the specific requirements outlined below must be met if the objectives of this system simulation are to be realized. The individual requirements of the component models will be briefly summarized by listing the processes that each compenent model must contain. The cattle demography component must contain: (1) age, sex, and function disaggregation of the cattle herd, (2) births, deaths, and transfers of herd members, (3) birth rates a function of breeding activities, (4) weights for each herd subpopulation grouping. The forage growth component must include: (1) plant growth as a function of exogenous weather variables, (2) quart whicl (3) graz (h) mach (5) star he nutrient (1) del The fun deriven that 1 realis heats Dome cesr 36 (2) quantity and quality of forage disaggregated by land areas which are homogeneous, (3) grazing as a means of forage removal, (A) mechanical harvesting as a means of forage removal, (5) grazing consumption and wastage a function of stocking rates. The nutrient impact component must include: (1) determination of weight gains for herd subpopulations based on the difference between energy consumption and maintenance energy requirements, (2) determination of the onset of estrOus cycling in young heifers as influenced by nutrition from birth, (3) dynamic energy requirements for mature cows as related to reproductive condition, i.e., lactating, breeding, and stage of gestation, (4) determination of energy intake of cattle from multiple feed sources with individual TDN characteristics. The feed stock accounting component must include: (1) accounting for current feed stock levels as a function of purchases, sales, production, cattle feeding, and waste, (2) feed stock losses from such sources as spoilage, handling, waste, and pest contamination, (3) determiantion of the feed stock TDN value for individual feed stocks with particular rates of TDN decline over time. The four physical components' requirements, as outlined above, are derived from the basic intent of this thesis to develop a system model that includes population dynamics, and that is capable of offering realistic simulations that are useful to enterprise managers. The management decision making component has two general require— ments in addition to numerous specific ones. First, the decision making algorithm must itself be dynamic to control the physical come Ponent models which have included the dynamic elements of their pro- cesses. Second, the component must include the correct control variables to . actual olerat of the nuanagE (1) timr (l) sep (3) tin fen These reqn intent of to invest Changes : The infumat Pauses be email (1 l have ramp 37 variables to allow evaluation of realistic strategies to cope with actual operating conditions. The specific decision elements required of the management algorithm are: (l) timing and quantities of cattle sales and purchases, (2) separate feeding rates for each herd subgrOup, (3) timing and level of feed stock sales, purchases, cattle feeding, and production, (4) control over stocking rates in the individual land parcels, (5) control over stocking rates for the individual herd subgroups, (6) timing and level of mechanical harvesting of forage, (7) control over the timing and degree of breeding, culling, and weaning. These requirements, both general and specific, stem from the stated intent of this thesis to develop a simulation model that can be used to investigate alternative strategies of management response to changes in the environment in which the enterprise operates. The decision maker must make specific decisions based on the information that is known to him at the time. When the simulation pauses for exogenous control inputs, the following criteria should be available along with useful state variable values: (1 V accumulated cash flow to date, (2 V accumulated net profit to date, (3 v current levels of short and long term debt, (4 V current level of working capital on hand, (5) production efficiency measures, such as the ratio of the weight of cattle sold over the weight of feed used. In addition to the requirements of each of the components that have been discussed, the system model must accept certain exogenous variables that have strong influence on the system. Three weather variables- nhst he 8‘ the produ enterpris to plant vested n tattle, an mg of unce nanaget aimed ants Chant nat'n the 38 variables~—solar radiation, average daily temperature, and rainfall-— must be accepted because of their importance to forage growth. Since the production and use of feed stocks is important in beef cattle enterprises—-yet this modeling effort dOes not include the decision to plant and harvest crops——the amount and quality of feed stocks har— vested must be accepted as exogenous variables. Expected prices of cattle, production resources needed, and feed stocks must be from an exogenous source. These exogenous variables are the major source of uncertainty in the operating environment of the enterprise; the management strategies that the model will be able to evaluate are aimed at perfecting responses to fluctuations in these variables. The following chapter will describe in general terms the compon- ents developed to meet the stated requirements of this chapter. Chapter V will discuss these same components, but in the full mathe— matical detail needed to completely understand how they operate and the basis for their structure. .1 an 0 .V l n. nu D. m . am pr. 3 TL t e l. at m I“ «l. W D: 1 a 0 A» 2 mu. m m 9. Wm. B» d 1 s t m m h CHAPTER IV GENERAL DESCRIPTION OF THE MODEL This chapter will provide a general description of the model that was constructed to meet the requirements of the problem statement. The processes that have been modeled and the variables used in the model will be discussed for each of the components. The general orga— nization of the model will also be reviewed. Chapter V will cover the same material but in more explicit detail, by developing the mathe— matical relationships that are implemented in the computer programs. The present chapter will be limited to discussion of the modeling of components in verbal terms. IV.1 General Organization Chapter III enumerated the four components of the system model that represent the physical processes involved in a beef cattle enter— prise; these are the cattle demography component, the forage growth component, the feed stock component, and the nutrient impact component. Further, the management decision making that controls the enterprise constitutes another component. These are the five major components of the enterprise system that will be covered in this thesis. A land component, which would be required to fully investigate the entire spectrum of possible enterprise operations, is not included here. Figure 4.1 is a general system diagram illustrating the major physical 39 mkronu nNMhVONE .QZMHUH‘H 9333.53, sesame? omfipauouao wauumo moon. ago How Smuwmfin Emummm Hmwocoo Hie ouowwm moflom mono mommsohsm QOHU crustacean. Eoumtwm A susposmv 833m owmuoh mcflmo>nwm pooh h (UI-' I U I I -‘U mooim mcwuth «cocoafioo .I IIIIIIIIIIIII III..|-..III..I|.I macho acoEowdcmE IIIIIII-II|IIII|....|I-IIIIII wcmwooh mooCnH " wooooaxm r | ' ‘ gonad: ringwogoa 22.352 3mm fin mm 33MB mofimm «soda: @5on con—manofi: eel mBOHH Hmfihwudfi Ill mvaSOHSOH 0:qu .QZ HUHA ’ my. ’ 73....» W, and inform of inform The a three maj first is policy 0 and the cess is prenatu that th hatura‘ cannon dynan tesnc avert tail Syst not DE] 41 and information flows among components. If also shows the transfers of information and material across the system boundary. The component with the most central impact is cattle demographics. Three major processes are represented within this component. The first is the birth of calves. This is a function of the breeding policy of the manager, the age distribution of the mature cow herd, and the nutrient intake of breedable females. The second major pro- cess is the death of animals. Deaths are attributed to old age or to premature events, such as disease or accident. The final process that this component includes is maturation of animals within the herd. Maturation will be described using the variables of age and weight. Another important component is that of forage growth. This Component must represent the growth and harvest of forages on a dynamic basis; i.e., over the growing season. Plant growth is a response to the exogenous variables of weather-—solar radiation, average temperature, and rainfall. More important than actual rain- fall is the level of soil moisture that is available for the root SyStem to transport up to the leaf structure of the plant. Soil moisture must be determined through the influence of evaporation and Percolation of water down beyond an effective depth reachable by roots. Digestibility of forage is another important factor in determining the nutritional impact of grazing. Grazing and mechanical harvesting are the methods of forage removal that are included in this forage growth COmponent model. Another component of the enterprise model includes feed stocks. This aspect of the model is less complicated than any of the others, hetanse re various ft determine Feed sto< ailocati anong th this met this pa ievei ( detern stock. syste- this on t aniv gro nut 42 because relatively little is involved other than accounting for various feed stock activities. Feed stock purchases and sales are determined exogenously through the interactive management component. Feed stocks used as cattle feed are a result of the level of feed allocation to the herd and of the distribution of that allocation among the possible crops. Feed stock losses are also determined in this model. Crop production, although exogenously determined in this particular system study, is also a factor affecting the overall level of feed stocks. The last element of this component model is determination of the total digestible nutrient (TDN) value of the feed stock. TDN is used as the index of nutrient value because it is a system of measurement which has been widely accepted for many years. This long tenure means information in TDN is abundant. The nutrient impact component develops the effects of feed inputs on the cattle herd. The major processes included here are growth of animals through weight gain and reproduction resulting in births. The growth process is highly influenced by the quantity of digestible nutrients consumed by cattle. The requirements of body maintenance and of growth are met differently by the same feed input; this requires that the energy values for Weight maintenance and weight gain be avail— able for each feed stock consumed. When the requirements of body maintenance are fulfilled, any excess energy intake can be applied toward growth or toward other production processes. Reproduction is quite responsive to energy levels available to the breeding females, eSpecially in young heifers being brought into the breeding herd for replacements. The physical processes of estrous cycling in females, huth nature 4 productive u the control selected to Manage this enterr riding a n rent is vc of the he the quest this hel control] delays i reprodu and (3: the re exogev ever make cont Prov val 43 th mature and immature, are determined in this component.1 The oductive use of the reproductive condition of the herd is under e control of the herd manager through the breeding policy that is lected for use. Management decision making constitutes the last component of is enterprise model. Because this thesis is oriented toward pro- ding a manager with a tool to improve his decisions, this compo— nt is very important. The component is constructed as an expression f the belief that optimization is not an appropriate approach for he questions that this study attempts to address. The reasons for his belief can be briefly summarized as follows: (1) the decisions ontrolling the breeding action are highly dependent on the long time klays before productive revenue is possible (2) herd management of :eproductive animals is highly dependent on the future prices expected, and (3) animal prices are unstable and highly volatile (especially in the recent past). The substitute for optimization of decisions is exogenous supply of control variable values. This means that when- ever straightforward economic criteria cannot be explicity used to make a decision, that the management component requires an input of control values. The values used are up to the user. This technique prevides an opportunity for the user to explore alternative control values based on his intuition and knowledge of the behavior of the system. 1Developed by Margaret Schuette in partial fulfillment of the M. 8. requirements, Department of Animal Husbandry, Michigan State University. This the enter} live coup physical ration, T cperatio to prov: discuss variahi of the ill svstr the devc m This section has briefly discussed the general organization of e enterprise simulation model. The major processes modeled by the ve components of the system model have been reviewed. The four ysical components have been shown to involve cattle growth and matu— tion, plant growth, feed value energetics, and feed crop usage. The eration of the management component involves interaction of the user provide control variable values. The following section will again iscuss each component, but in a more formal sense, setting forth the ariables used in the mathematical model of the component. The use if these variables in the models will be explained. IV.2 Specific components This section will provide a description of the content of each system component as it has been modeled. The major variables used, the physical basis for the model, and the assumptions made in developing the model will be discussed for each component model. Cattle Demography Component Development of the cattle demography component has been heavily influenced by the requirement that population be maintained in a way that disaggregates on the basis of sex, age, and function. This has led the model to be constructed using nine cohorts to describe the members of the herd. These are: (l) mature females, (2) replacement heifers, (3) bred heifers, (4) mature bulls, (5) young bulls, (6) steers, (7) male calves, (8) female calves, and (9) slaughter heifers. Table 4.1 provides a more detailed explanation of the exact age/sex/ function disaggregation of these nine herd cohorts. The variable not nea‘ with re gregat: number at sut cisio descr heifv larg in v N W i ii). i(t) is the total population of cohort i. While this level of ording herd population may seem quite detailed, in reality it is nearly fine enough, as these groups do not behave homogeneously h respect to growth and reproduction. Accordingly, further disag— gation is provided by the variable SUBPOPi (t), which records the j er of animals in the jth subpopulation of cohort i. The number subpopulations for each cohort varies with the homogeneity pre— ion required. For example, the replacement heifer cohort must be cribed rather exactly, since the onset of first estrous in these ifers is an indicator of reproductive maturity. This requires a rger number of subpopulations than does the slaughter heifer cohort .which estrous cycling is of no interest. The variable KKi is the fiber of subpopulations used for cohort i. KK i IPi(t) = 3'21 SUBPOPij(t) (4.2.1) Each subpopulation of a cohort assumes that the level of maturity 5 homogeneous for that group. Age and maturity are rather highly arrelated in the typical operation in the U. S., but some variation oes exist. This variation leads to characterization of maturation s a distributed parameter process. The ages of individual members of group having the same level of maturity are distributed around a can. Some individuals are older than the mean, and some are younger. "young" e total length of time required for individuals to pass from ature cows to "old" mature cows no longer breedable is also a dis— ributed value. Thus, for mature cows (and for all the herd cohorts), 46 Table 4.1 Age and Sex Description of the Cattle Herd Cohorts Cohort male, 0—6 male 6—24 male, 24 month (a) male animals emale 0—6 6—2 24month x (b) female animals the n vith CESS pron equ be delay time is a distributed parameter process. DELAYi(t) describes he average length of time for individuals to pass through the range f maturity that is represented by cohort i. Flow rates of animals per year is a desirable way to describe he movement of animals through the various cohorts and subpopulations ithin cohorts. Inputs to delay processes and outputs from delay pro‘ cesses are easily visualized in this description. A distributed delay process may be represented mathematically by a kCh order differential equation: k k-l a 9_Xl£l + a 9———Z££l + ... + a y(t) = x(t) (4-2-2) k dtk k-l dtk-l 0 where: x(t) = the input rate y(t) = the output rate a1, a2, 33, ..., ak = constants. The order of the equation, k, determines the nature of the response for a particular delay time. The higher the value of k, the more tightly clustered will be the distribution of individual delay times. The value of k is a modeling parameter that is specified to most closely simulate the actual parameter distribution as recorded by experimental observation. Figure 4.2 illustrates the change in re— ‘sponse of different values of k in a distributed delay process. Figure 4.3 illustrates, for purposes of comparison, the behavior of a discrete delay process where all individuals have a common delay time. The variable Wij(t) describes the average weight of members of SUBPOPij(t). Weight gain as a result of energy intake above the 7A... lg. minimum r rich the that Still class of common these cohort breed equat enter vari beca encr dev PU 48 nimum metabolic requirement is represented by the variable DGAINij(t), th the subscripts again referring to the subpopulation of the herd at SUBPOPij(t) counts. The average slaughter grade or slaughter .ass of cattle described by SUBPOPij(t) is the value of the variable rADEij(t) . Births and deaths are the final processes of the demography >mponent that remain. BRi(t) and DRi(t) are variables representing rese processes, respectively, on the basis of an annual rate. Only )horts l, 2, and 3 have BRi(t) values because these are the only reedable female cohorts. The simple distributed delay model of quation 4.2.2 is flow conserving; that is, all of the flow which nters the delay subsequently leaves the delay through the output ariable. This does not appear to fit the case of cattle maturation, ecause deaths and sales before final maturity are commonplace occur— nces in cattle enterprises. This problem will be solved in the evelopment of the exact mathematical equations for the maturation mocess in Chapter V. Forage Growth Component Plant growth as a response to energy and nutrient inputs is the *UbjeCt of this component model. The biological growth process is exceedingly complex. Solar radiation——SOLAR(t), average temperature—- 5‘V(3Tl'fl’(t), and rainfall——RAIN(t), are the exogenous variables driving this Process. The model developed here is to some extent a modifica- tion of a grasslands model developed by Patton and Marshall [40]- While much more detailed models of this growth process could be 49 V Time ——-—.—_._._———~——-———_—~_———_—~—.—_——_ _.. —... __—_—.__—_———___——————q-_-_-.- t t Time gure 4.2 Graphical illustration of the response, Y(t), to an input, x(t), for a distributed delay process 50 K(t) Time Y(t) ...—-..—..-__...___._..__l.__-.__.__-_.__._- “l n tI = the delay length 0 n .—......____ H 0 Time Figure 4.3 Graphical illustration of the response, Y(t), to an input, x(t), for a discrete delay process develop of this A1 this 111 tion. impor is in into ing, gree pro‘ nut 51 :loped, the present effort is felt to be adequate for the needs :his thesis. Animal consumption is the primary end use for the forage growth 3 model describes, whether grazed or harvested for later consump— n. Since the nutrient value of animal feed inputs is of great ortance, the digestibility, DIGEST(t), and energy value of forages included in this model. The model divides the biotic component ,0 two materials, green plant material~-GRN(t)--suitable for graz- ;, and root storage——ROOTS(t). Grazing is strictly limited to sen material. The abiotic component consists of soil with two )perties of interest; these are soil moisture, SM(t), and soil trients, SNUT(t). SNUT(t) is a composite of all of the nutrients portant to plant growth: nitrogen, phosphorus, potassium, and trace tals. SM(t) is determined as a result of three variables interact- g over time; these are rainfall—~RAIN(t), percolation of soil mois— re—~PERC(t), beyond the depth that roots can reach, and evaporation—— AP(t). Because grazing management is an area that will be of im— rtance to evaluation of enterprise decision—making, the model is nstructed so that different land parcels are available for varying =es. The size of these parcels, LAND“, is arbitrary, but it is ‘sumed that they are each homogeneous with respect to all variables :ed in this model. The subscripts n on the variables of this com— lnent model refer to values for the nth land parcel; NLANDS is the ital number of these land parcels. The growth process within an individual land parcel begins with Le fundamental exogenous input—-SOLAR(t). Solar radiation does not affect 2 between in shad amount When p is tn radia energ pera prec' of em ct all green plant material equally. A major distinction is made een those areas of the plant that are in direct sunlight and those shade. This model approximates that distinction by limiting the unt of plant material available for photosynthesis per unit area. n plant density is higher than a specified threshold, the excess treated as being photosynthetically inactive. The input of solar iation on the active plant material determines a net photosynthetic rgy conversion——PHOTO(t). The relative quality of the ambient tem— ature, soil moisture, and soil nutrients are used to reduce this dicted quantity whenever these conditions are less than ideal. This net photosynthetic energy is available for growth in either two places—~forage greenery or root storage. The partition of this ergy is a function of the relative proportion between greenery and ot storage which already exists. A transfer of energy from root orage to greenery takes place according to the proportion of green— 'y and root storage quantities. Gross greenery growth rates, mDOTn(t), and net root storage growth rates, ROOTDTn(t), for the nth ind parcel are the result of the interplay between the partition of iotosynthetic energy and the transfer of energy from root storage to rage greenery. The net greenery growth rate, dGRNn(t)/dt, for land parcel n is etermined from PRODOTn(t) by subtracting the amounts representing nimal and mechanical harvesting. AHRn(t) and MHRn(t) are the model ariables for these latter two activities, respectively. The actual arvest, either by cattle or machine, has a greater effect on the quantity 0 occurs in to determ' from the waste fac rate wit waste fa waste f: green p ment oi QUITE“ age 1 this acct bri 1e: 53 ntity of plant material than the amounts harvested indicate. Waste urs in each of these activities; therefore, a waste factor is used determine a higher (and more accurate) amount of greenery removed m the land by both mechanical harvesting and grazing means. The to factor for grazing is modeled as a function of the stocking e within a particular land parcel; high stocking rates have higher te factors than do medium stocking rates, which in turn have higher te factors than do low stocking rates. The net growth rate of een plant material, dGRNn(t)/dt, is integrated over the time incre— t of the simulation to determine the quantity of GRNn(t) for the rrent period of time. The final mechanism of this component model is the determination 5 current digestibility, DIGESTn(t), of the green plant material in ind parcel n. This is computed through use of an index of the aver— ;e growing time to generate the forage existing at present. The higher iis value, the longer it would take at the current growth rate to :cumulate the present quantity of greenery. The digestibility of stage is aSSumed to decline with age. This assumption attempts to ing to the model the age dependent shifts in the proportion of plant eaf area, stem, etc. without having to actually model forage using ese different plant parts. An additional factor affecting digesti— ilities is the density of green plant material per animal grazing in particular land parcel. This density is determined by dividing urrent greenery by current stocking rate times days per simulation ime increment for the entire land parcel. When high forage densi— ies exist then the digestibility of the forage is assumed to be unchangEd are low ti to accoun that is i Forage d: the cur grazing Feed St- Ti curren- feed 5 cannot quant peric chan and and of 0W 54 nged from the value determined previously; however, when densities ow then the previously determined forage digestibility is reduced ,count for the fact that animals will be forced to eat material is included in GRNn(t), but which is of very poor quality. ge digestibility is then modeled as a function of current quantity, :urrent growth rate, season, and the current density of forage per ing animal. Stock Com onent The main function of the feed stock component is to provide a ent value for feed stocks on hand and the TDN value of each of these stocks. This model is quite simple in comparison to the other onent models. FSTOCKn(t) is the variable used to describe the atity of feed stock n on hand at the beginning of the current time .od. FQUALn(t) gives the TDN value of feed stock n. FSTOCKn(t) lges value through sales-—CSALESn(t), and purchases-—STKPURn(t), crop production-—CROPnl(t), quantities fed to cattle—-STKFEDn(t), through loss from pests and waste—~STKLOSn(t). Except for losses stocks, each of these variables are calculated elsewhere in the :all system model. For example, STKFEDn(t) is determined in the :ient impact component model as a result of the allocations of feed each cohort, and the distribution of the allocations among the sible feed stocks on hand. STKLOSn(t) is computed in the feed stock :1 by multiplying an annual fractional loss rate for the nth feed :k, FRCLOSn(t), by the current quantity of feed stock n on hand. The change in quality of feed stocks is highly influenced by rate of turnover of the stock itself. The variable SPOILn(t) is used to dete stock It. Ti user in the the new ave each of the time petio from the s the cure This the mock tepresen growth 5 tuttiett he went t: detail (inalit descr tine vain. iewe trot wal anf 0n 55 to determine the annual rate of decline in TDN values for feed L n. The particular values used are read in as data by the model in the initialization phase of the program. Each time period new average TDN value is determined by accounting for the TDN from of the following sources: carryover from the previous simulation period, new purchases, and crop production. The average amount ‘the sum of these three sources is the value used in FQUALn(t) for current time period. This component handles one other minor accounting task by totaling production of crops to get annual production to date. CROPGn(t) :esents the total amount of crop n harvested to this point in the nth season. Eight Impact Component Development of this component model has been guided by the require- t to describe reproductive links with nutrient intake in considerable ail. The model uses four variables to describe the quantity and lity of the nutrient allocation for each herd cohort. CNCALi(t) cribes the quantity of concentrates allocated to cohort i for one e period. TDNCi(t) describes the average total digestible (TDN) ue of the allocation to the ith cohort. FEEDALi(t) describes the el of roughages allocated to the ith cohort for one time period m voluntary feeding by management. TDNRi(t) represents the TDN ue of the roughage allocation. The nutrient requirements for mals of a particular cohort for weight maintenance are predicted the basis of the animals‘ metabolic weights. The feed consumption of anion the dif the ens energy Resear nine 1 cohor to ti heif int 56 ls is predicted to be a specified fraction of their weights. fference between the energy value of the nutrients consumed and ergy required for weight maintenance, and other requirements, is available for weight gain. The model uses the standard National ch Council2 relationships between energy and weight gain to deter— he rate of weight gain, DGAINij(t), for the jth subpopulation of i. eproductive considerations in the component model are restricted ee herd cohorts: mature cows, replacement heifers, and bred 3. Reproductive requirements for bulls are assumed to be zero ‘5 model. The energy requirements of mature cows vary depending age of the calf they are nursing, their current state in the tion period, and their breeding status. A beef cow has certain ological priorities which cause differential impacts if the over— energy intake is smaller than required, and the ability to be d is not the highest priority activity. Replacement and bred rs come into puberty as a function of age, weight, and rate of t gain; following puberty they have reproductive cycling just as 'e cows cycle. All three of these herd cohorts have the charac- tic that the success of breeding is a function of both current >rior feed energy intake rates. Breeding is the major event that the model uses to bring repro— ;ve dynamics into herd population dynamics. TBRDi and DURBi 2 National Research Council, Nutrient Requirements Of Domestic, als: Number 4——Nutrient Requirements of Beef Cattle, 4th Revised Lon, 1970. describe t represents the time : Ci’hTijk r ith serwi SUBPOP ii I In proporti from the Similar: by time the owe season the Cu each f the we Subpot latio down other rate tiat con 57 be the onset and duration of breeding for the ith cohort. lNBij tents the number of servicings for SUBPOPij(t). CTlMijk gives Lme of calving that corresponds to the kth servicing of SUBPOPij(t). ik represents the predicted accumulated calving rate after the :rvicing. Figure 4.4 depicts the expected calving pattern for ?ij(t) as a result of its breeding activity. in Figure 4.4 the first point (subscripted ij,l) describes the rtion of cows giving birth to calves during the interval of time the beginning of the calving period to time point CTIMij,l' arily, the second point describes the cow proportion giving birth me p01nt CTIMij,2' The final point (subscripted ij,5) describes verall proportion of cows giving birth over the entire calving n. BEGCAV(t) and ENDCAV(t) describe the beginning and ending of urrent calving season, respectively. A pattern of calving for female cohort capable of reproducing is determined by computing eighted average of the entire cohort using the number within each pulation of that cohort as the weight for the individual subpopu— n pattern. An adjustment factor, PADJST, can be used to scale the entire calving curve to reflect the incidence of disease and factors affecting breeding, if desired. The instantaneous birth at any point in time is determined for the cohort by differen— ng the cohort calving pattern. For details concerning the model for this component one should It Schuette[48]. 'h iccun Calw: Fra 58 umulated ving .tion I | I I I ___________ | I P I C ATij ’2 I l : I I | I I I ' I I | I I I | ———————— o-———o | . CPAT.. I I I I 13,1 I I | I I | | I I , I I I | I l I | I I I I : I I I . I I I . | I I I I L I . I T I I I I | CTIMij ,1 : CTIMij ,3 ' CTIMij,5 : Time I I I I BEGCAV CTIM . 13 ’ 2 CTIMi j ,4 ENDCAV igure 4.4 The accumulated calving fraction versus time me at the to the produr both high] tact: catt per: aho do 59 ement Decision Makin Com onent Management decision making is a very complex process, especially he detailed level used in this model. A factor that contributes he complexity of the decisions is the lengthy time delays of the uction process. Management must make decisions on the basis of h current and expected prices. This makes decisions of the manager hly conditional on the expected prices he has used. An additional tor is the well known volatility of agricultural prices, including tle prices. Since the expectations of the manager are apt to change iodically, the decisions be made under the previous assumptions ut future price behavior are likely to diverge from what he would under his current expectations. An additional factor is the mul— le desired goals of the manager which cannot be readily reduced to single objective function. Thus, the decision making of a beef :tle enterprise manager is not well suited to optimization proce- :es for these reasons. What information is necessary for the decision maker to use as a basis for his decisions? Certainly expected prices of cattle, Ips, and production resources are important. The time horizon over .ch these are needed is different for each type of decision. For ample, breeding is an activity based on decisions which use prices far forward in time as the likely date of sales of the resulting ’spring, on the order of two years. The decision of what stocking :es to use in various land parcels is likely to require only a few Iths of expected prices. The states of this system are a vitally tortant category of information. Current cohort populations, current parcel, system catego: crop 9 over I the c endog requ the Thi gen lat 60 rent feed stock levels, current quantities of forage in each land cel, and current financial conditions are examples of states of the tem which are important to decision making. The last important egory of information is expectations about future events. What p production is expected? What is the average forage production r the growing season? All of this information is needed to describe conditions under which decisions are made, whether they are made ogenously or exogenously. This component has been developed with regard to the general uirement that the control variables of the model be able to simulate decisions made by an actual enterprise manager operating a ranch. 's has meant that very detailed control variables are needed; in eral, much more detailed than have been used in agricultural simu— tion models in the past. The major decisions that the manager must ke to control his enterprise include: (1) timing and level of sales and purchases of cattle, (2) timing and level of sales and purchases of feed stocks, (3) feeding rates and types of feed for each cattle cohort, (4) stocking rates during the forage growth season, (5) timing and level of mechanical harvest of forage, (6) timing and extent of breeding, culling, and weaning. e following control variables are included in the simulation model carry out the above list of operating decisions. (1) ADDRT.(t) = the net annual rate of additions to the iFh 1 herd cohort (2) AGEMIN = the minimum age of calves to be weaned (3) CSALESn(t) = quantity of feed stock n sold this period 61 (4) CATTIN. (t) = the quantity of TDN of feed type m fed to ink a member of cohort i per day during the period of the 2th feeding plan (5) CNCFRC. (t) = the fraction of the concentrate TDN allocation in to be derived from feed stock n (6) CULFRC.(t) = the fraction of the population of the jth J subpopulation of cohort l to be culled (7) DURBi(t) = the duration of the breeding period for cohort i (8) FPLANS (t) = the time at which the 2th feeding plan ends (9) MHRn(t) = the rate of forage harvest from land parcel n (10) PADJST calving rate adjustment factor (11) PDSTRBin(t) = fraction of the population of cohort i that is to graze in land parcel n (12) REMOVL(t) = fraction of the forage to be harvested (13) RHGFRCi (t) = the fraction of the roughage TDN allocation n to be derived from feed stock n (14) TBRDi(t) = time of breeding onset for cohort i (15) TCULL(t) = time of mature cow culling (l6) TFRAC. (t) = the fraction of the weaned female calves from 31 the jth subpopulation transferred to cohort i (17) TMHRk(t) = time of the kth forage harvest (18) TWEAN(t) = time of weaning for calves (19) STKPURn(t) = quantity of feed stock n purchased this period. Decision making in this component model, as has been previously lained, is performed both endogenously and exogenously. When a ision can be made without recourse to exogenous input, it is so e. Such endogenous decisions include: (1) timings and quantities sales of the two slaughter cohorts—~steers and heifers, (2) the ision to harvest forage as guided by preseason planning, (3) sales purchases of feed stocks to obtain the winter feed base, and (I) the actu the herd as quantity pr< I.8 indicat of the four The re ting contr< (l) determ its which cohorts, of animal its tome types of events; and wean four eve iahles i calls f events the cut fallin the g! ments Which com the r 62 the actual feed allocations of concentrates and roughages fed to herd as guided by the feeding plans and adjusted to meet forage ntity problems as well as feed stock shortages. Figures 4.5 through indicate the decision mechanisms and the variables used in each the four endogenous decisions. The remaining herd decisions are determined by the user by input— ig control variable values exogenously. These decisions include: > determining stocking rates for herd cohorts, (2) breeding activ— y which includes timing, duration, and type, (3) weaning of the calf horts, (4) sales from or additions to the herd cohorts, (5) movement animals through the cohort structure following maturation beyond :5 former range, and (6) culling of the mature cows. Four of these rpes of decisions are clearly associated with particular physical Jents; these are the onset of spring forage growth, breeding, culling, 1d weaning. The component model has been developed so that these ur events, termed "decision points", are recognized and control var— bles are sought upon recognition. A fifth "decision point" also 11$ for exogenous input of control variable values when any of three ents occurs; these are (l) the quantity of feed stocks is less than e current rate of feeding times two periods, (2) working capital lling below $2000, and (3) the quantity of feed stocks at the end of e growth season being significantly different from the feed require— ents for the wintering season. Figure 4.9 is a blow up of Figure 4.1 hich shows the mechanism whereby the management decision making omponent recognizes that a decision point exists and acts to achieve he needed input of control variable values by the model user. 63 I Weights or Hard Legend: Subpopulation ® -- from Subroutine Weight. @ -- exogenous to the model ® -- decision variable Gain ® -- to financial accounting L 7 ‘ ‘- ism eding vols Weight Gains I Costs 3 and Revenue r‘op Prices - Expected Net Fri 6 Revenue c s Management Sale Decision-making Cohort Animal Sales igure 4.5 Decision mechanism for slaughter cohort sales Stock tat/as 64 Specified Harvest Tires ® -- exogenous to the model ® -- Forage Component 6)) -- decision variable -- from the main model poking ;es Forage Harvest Decision Making < > 3:35 Herve st Rates 'gure 4.6 Decision mechanism for forage harvesting 65 D Anticipated Legend: Net Sales (9 -- decision variable ® -- Feed Stock Component (9 -— exogenous to the model @ —- from the main model Feed Stock E Sales and Purchases >'—"_ ' ' Decision Making ement Time Time of Year Crop Purchases ‘ F Crop Sales Figure 4.7 Decision mechanism for feed stock sales and purchases to obtain wintering nutrient requirement \'\ Basin Rough Ratio Dee 1131 F1 66 > resired Population legend: Roughage Rations (::> -- Population Component (::) -- decision variable 61) -- Nutrient Impact Component (::) -- from the main model ® -- Forage Component ) Desired Concentrate Rations Requirements Rates Grazing Forage Shortage 7 Current Roughage and Ti Concentrate Allocation Decision-Making C > F°rag° Quality ’ Forage Quantity oughage Concentrate llocation Allocation Lgure 4.8 Decision mechanism for herd feeding allocations 01.6 uhnumOh manned :07:qu quzu‘un MR UZH :2 ZOHW~U Mann W Dozuuoxu no no coaamohu conumcmOUUu E I I 67 ucmcoafiou wcaxmé : can we mHHmumw HNGOHuosnm 303m Ou Emummflv Ewume Hmumwwwowm ucwEwwmcmE m.e enema ..m nuiqz; 55:3 Eugen—So .caEaunauz llllllllllllllll u - Eastman-1n nulnucsa uE-U 2.25:: HZWEMAH 072M; ZOHmHDMQ mDOZHUOQZM I I I I I I l J confine, 033.55 83:00 mucosa n on: oufluwo maowmauomo mome hon: meomamuou now: use >uhmu cu seduce L‘—-——-_.______‘ o: 02w unnumou 350A commwoov HZMENAH Q we we 07:? ZOHmHDH :ofimouu no.» a m WDOZHUONH . 03m: oooh .----_..-__..---.._-..-...._.‘ Figu making c0 decision to see i: values f passes t conditi< shown 1‘ model 5 is prir at his result which the c: contr event insu'. IQCO Sea \ the REG 0f 68 Figure 4.9 illustrates the operation of the management decision king component with the two elements-—exogenous and endogenous :cision making——logically related. The component constantly checks 3 see if any condition requires the use of exogenous control variable alues from the model user; if none is required, then logical control asses to the endogenous decision making element. When an event or ondition does trigger recognition of a decision point, then the steps :hown in the exogenous decision making element are followed. The mdel simulation is stopped after current state variable information is printed; this pause allows the user to study the current situation at his leisure and input control variable values to implement his resulting decisions. The model is restarted from the exact point at which it stopped and continues the simulation as before, except that the controlling actions of the user are substituted for any previous control decisions or decision making criteria. This sequence of events is followed in every time increment of the time simulation to insure that decision points requiring human decision making are recognized when they exist. Secondary System Components In order to simulate the beef cattle enterprise in the way that :he problem statement requires, several secondary components are also needed. A component performs the task if determining the proper value vf the exogenous variables—~crop production, prices, and weather—~for he current time. The main program reads in as data an entire year f exogenous variable values in its initialization phase; a subroutine iscalled the appro The inportan requires the none nation : process current Second, reache and in are or “(ICES Price In t' stre Done to. fuI pa 69 3 called at the beginning of each time iteration which determines he appropriate exogenous values to use. The financial effects of alternative policies are the most important outputs to use in choosing among the policies. This requires that a financial component be included that accounts for the monetary effects of sales, purchases, and payments. This infor— mation is used in two different ways. First, the ongoing management process requires information about present, immediate past, and current values of such variables as yearly net profit to date. Second, when the final time horizon of the simulation has been reached, the user wants summary information to help him to analyze and interpret what has OCCurred during the simulation. These tasks are carried out by a package of subroutines which perform all necessary financial accounting and analysis. The last secondary component is one which manages the expected prices that the manager uses as the basis for his decision making. In the initial program start and at each decision point, the future stream of expected prices for cattle and crops changes. This com- ponent reads in this new information as needed. A second task is to adjust the future expectations to account for the passage of time. As simulated time is incremented forward, this component shifts the future expectations toward the present and records the most recent past value in the pst expectations variable. It gives the decision maker the opportunity to know what price he expected to encounter at times that have come and gone. 1V.3 Sin Opel interact initial prise tr enterpr At this hrpm sions tion t procer run h the r compr been Pris sub cat cl: hi 70 .3 Simulation Model Operation Operation of this simulation model is heavily influenced by the teractive management component. Initially, the user reads in the Litial conditions and system parameters for the particular enter- :ise to be studied. The model will automatically simulate the nterprise behavior through time until a decision point is reached. t this point the model requires the user to input specific values or particular control variables. These values implement the deci— Lions that the user has made. The model then continues its simula- :ion through time until the next decision point is reached. The )rocedure is repeated until the final time horizon of the simulation sun has been reached. Figure 4.10 illustrates the organization of the management component and its relation to the other system components in the simulation loop. When the management component detects that a decision point has aeen reached, a detailed printout of the current status of the enter— prise is made. This printout typically includes the current herd subpopulation breakdown, current feed stock quantities, the expected cattle and crop prices over a future time horizon, and current finan- :ial status. The user must carefully study this information to make 118 decision about what is to be done at this point in time. The user Imy be investigating the effects of a particualr decision rule; he muld input control variable values to follow that decision rule. he user also inputs new expectations about future prices and events. At the true that the management component completes printing of he decision assistance information, it prints all variable values Figure 4.10 71 Enterprise Model ...... ._____________________1 I I I . Model Variables I Recognition UIdated I of a I Standard : Decision Point I Decision Maker I In-uts Controls : NO I I I I ‘ I I Model Variables l Recognition Updated I of a l Special | Decision Point I Decision Maker : NO In.uts Controls I | I I ‘ I I I I Endogenous I De C is ion S THE I MANAGEMENT ' COMPONENT : .J Flow diagram of the management decision making component in relation to the simulation model into a p to be re tions 3: making to read with tI ables E_nte_r; to cm chara to re new I po'm can PIOI per- 4.1 can is As 72 ' to a permanent file in the computer. This file enables the model 0 be restarted at exactly the same point and under the same condi— ions as when it stopped. When the user restarts the model after king his choice of decision variable values, the first action is to read from the permanent file to reestablish all model variables ith the proper value, the second step is to read the control vari- ables (using the formats given in the User's Guide to the Beef Cattle Enterprise Simulation Model), and finally to implement these values to control the enterprise simulation model. An extremely valuable characteristic of this interactive management component is the ability to return to a decision point that has already been passed by, input Inew control variable values, and continue the simulation from that point. At each decision point any number of control variable sets can be used by restarting the model with the desired control values, proceeding to the next decision point, printing all values into a permanent file, and returning to the original decision point. Figure 4.11 illustrates the decision tree of alternative control sets that can be evaluated. The key to this ability is the permanent file that is created to "freeze” the simulation as it was at a decision point. As long as the user does not dispose of these permanent files, he retains the option of returning to any one and proceeding forward in time with any control values that are desired. This evaluation of alternative strategies method of use will be employed later in this thesis to explore some management strategies for (1) age of weaning, (2) rate of herd development projects, and (3) general profit maximization. Use of the I The si evaluating Certain acI plex decis the user a evaluate I model wit‘ feed stoc etc. In to the 5‘ time. I mod\el, I reilfl deviate t0 invv Revisi the re to re; latio‘ etc. entev the I thrv se of the Simulation Model The simulation model is a tool for the user to employ in valuating specific alternatives confronting his own operation. ertain activities, such as breeding, weaning, etc., are highly com— lex decision and quite dependent on future prices and events. When he user approaches Such events in real time, he wants to begin to valuate his possible courses of action. He would then start up the odel with the current states of his cattle herd, forage in the field, eed stocks on hand, expected crop production, financial condition, tc. In effect, he initializes the simulation model to correspond to the states of his own enterprise. Then the model is used to explore the future consequences of his decision options through simulated time. After having explored the alternatives using the simulation model, the decision maker is in a much better position to make his real world decisions. At any time that conditions in the real world deviate from the user's previous expectations, he can use the model to investigate the effect of these differences on his enterprise. Revision of previously made decisions is possible on the basis of the results of these investigations. The user of the model is alerted to real thme events that require investigation through use of the simu— lation model by his own awareness of weather, prices, political events, etc. The simulation model is an assist to the manager of a beef cattle enterprise, not a substitution for the experience and intuition of the manager. The simulation model has simulated the dynamics of the enterprise through time but in ways that are not completely accurate. Two LEGEND 0 resta 0 rests 74 GEND restart file selected restart file not selected decision decision point point one two I I I I I I I I I I I I I I I I I I I current real time decision point three Figure 4.11 ”Decision Tree" of alternative strategies investigated using the simulation model's restart capability sources between in the errors tion t of ina do n0‘ reaso the s poin‘ mode vari beeI age his 1V 75 sources of inaccuracy provide the bulk of the reason for the divergence between real and simulated values. The first of these is inaccuracy in the simulation model formulation itself; this is due to modeling errors and approximations that simulation makes to develop a solu- tion to the mathematical model of the enterprise. The second source of inaccuracy is management decisions in the real world system that do not conform to those made in the simulation. For these and other reasons, the simulated enterprise behavior will never follow exactly the same path as the behavior Of the real world enterprise. At each point in real time requiring alternative decision explorations, the model should be reinitialized with the Current states of the model variables used to describe the enterprise. The simulation model has been constructed so as to be available for use by an enterprise man- ager at any time and to be as flexible as possible to accommodate his needs. IV.4 Summary This chapter has provided a general description of the simulation model that this thesis has developed. Five major and three secondary components of the model have been reviewed. The operation of the model by the user has been discussed. In the aggregate these com- ponents constitute the model that will be used to investigate stra- tegies of management decision making by this author and other users. Figure 4.12 provides an overview of the major components and the key control points that provide the means of implementing the decisions made endogenously by the management algorithm and exogenously by the user. Figure sequence of t horizon. Th! equations th computer prc 76 user. Figure 4.13 provides a pictorial description of the calling sequence of the model as it simulates the system over a desired time horizon. The following chapter will present the detailed mathematical equations that form the mathematical models that are simulated in the computer programs. _ ._ . ...—F-.._ ..:_.-.....4.—-.____,__.._.__——— "we mcc F msocowoxo 78 Exogenous Inputs Component I Expected Prices Component Livestock System Components I Management Decision Making Component Financial Component Figure 4.13 Calling sequence of the simulation model components to simulate the beef cattle enterprise through time This presented in the prI are the b Each of t separate section, A c be found volume from ei volume I.l C ulativ is th basis aver; ages func BIRI pre rat CHAPTER V DETAILED MODEL DESCRIPTION This chapter will build upon the general description of the model ;ented in Chapter IV. This chapter will take the variables defined :he previous chapter and develop the mathematical equations which the basis for the computer subroutines of the simulation model. 1 of the five major system components will be discussed as a arate section; the secondary components will constitute another tion, and the final section will summarize the enterprise model. A complete listing of the computer program and its subroutines can found in User's Guide to the Beef Cattle Enterprise Model, a separate ume from this thesis. Complete instructions for operating the model m either the initialization or restart mode may be found in this ume as well. Cattle Demography Component This component model uses four subroutines to provide the sim— tion of the birth, death, and maturation processes. Subroutine HDMOG4 the major element; it maintains herd populations on a disaggregated is by age, sex, and function. Subroutine WEIGHT determines the rage weight of particular subpopulations as the herd matures and s. Subroutine BIRTHZ computes an instantaneous birth rate as a ction of prior nutritional status and breeding activity. Subroutine AT is used by BIRTHZ to determine birth rates by combining the dicted birth information about each female subpopulation into birth as that apply to an entire cohort population. 79 The In maturation i Equation 4.2 it does not not conserVI and additio covered by the Laplace Y(s', whe Upon solv II This form as k 1st decompos >< A w v FiguI 80 The key to the model of maturation is recognition that ration is a process in which age is a distributed parameter. .tion 4.2.2 provides the basis for the model of this process, but oes not allow for cases in which flow is not conserved. Flow is conserved in the maturation process because of deaths, sales, additions to the herd. These occur throughout the range of age red by the cohort delay. Following Manetsch [38] one applies Laplace Transform to 4.2.2 to obtain k Y(s) n (Dis + 1) = X(s) (5.1.1) i=1 where: Y(s) = transform of the output time function X(s) = transform of the input time function .solving for Y(s), one obtains k 1 Y(s) = H ———————— X(s) (5.1.2) i=1 D s + l 1 :form suggests that the kth order distributed delay can be modeled L lst order delays in a cascaded form. Figure 5.1 illustrates this mposition. 1 l 1 l 7 w...- —’q D + l——-—’ ________ —» “"“' s ) DlS ‘I' 1 Rl(S‘ D25 + 1 R2(S) D38 + l R3(5) k Y(S) gure 5.1 Decomposition of a kth order distributed delay into a series of k first order delays Each of the as stages wi Sino is a loss f of the stag loss from I losses fr0I stages has conventior >< A m v Iigm 81 of the individual first order delays of Figure 5.1 are referred tages within the overall delay. Since flow is not conserved in the case of maturation, there .loss from the delay. In general there can be a loss from each he stages of Figure 5.1, which in the aggregate sum to the total from the entire delay. Figure 5.2 illustrates this idea of es from individual stages. Notice that the numbering of the ;es has been reversed in this figure to conform with the usual 'ention [32, 38]. _____’. . ..__). .———> ' K ' K—1 1 R =Y Rk Rk—l l (S) L Lk Lk—l 1 + + + L = total loss rate gure 5.2 Kth order distributed delay with losses At any I storage the lam Losse sinpl anal: The dif in be de 82 I any moment the total number of individuals in the ith stage (the ;orage in stage i), Qi(t)’ is a function of the flow rate, Ri(t)’ and 1e length of the ith delay, Di(t). This gives = * ._(t) Di(t) Ri(t) (5.1.3) 3 the total delay for the entire process, D(t), is spread uniformly Ier all k stages, then 5.1.3 becomes L(t) = Q(£)*Ri(t). (5.1.3a) K asses from this quantity can be individually specified, but as a implyfying assumption, let a common proportional loss rate, PLR(t), pply to all stages. Then the loss, Li(t)’ is i(t) = PLR(t)*Qi(t). (5.1.4) he net change in storage for the ith stage follows the following ifferential equation, (5.1.5) Q (t) _ 3t — Ri+1(t) — Ri(t) - Li(t) ecause the rate of change in the quantity is the rate of input to the elay minus the rate of output. Referring to Figure 5.2 gives the nput rate to state i as Ri+l(t)’ while the output rate is the sum of he output going to the next stage, Ri(t)’ and the loss, Li(t)- dfferentiating 5.1.3 gives Qi(t) = D(t)*dRi(t) + Ri(t)*dD(t) (5.1.5a) IQUating the right hand sides of 5.1.5 and 5.1.5a and solving for IRi(t)/dt gives the following, This is tribute length. SI equativ ith he by equ number SUBPOI The t P0P“: POP prov pub [IOI 83 {.(t) = K R (t) — R (t) l + D(t)*PLR(t) + dD(t) Elt— D(t) 1+1 1 K— ‘KdT- (5.1.6) 1is is the general expression for the ith stage of a kth order dis— :ibuted delay with proportional loss rates and a time varying delay angth. Subroutine DLVDPL is used to simulate a delay process which follows Juation 5.1.6. RINij(t) represents the jth stage flow rate for the :h herd cohort. The storage for that stage, SUBPOPij(t), is determined 7 equation 5.1.3a using the cohort delay length, DELAYi(t), and the Imber of stages in cohort i, namely KKi' Then .13 = * POPij(t) DELAYi(t) RINi.(t) (5.1.7) KK. 1 1e total storage ofIthe ith delay process (cohort i) is simply the ummation of the storage in each stage. This gives the total cohort qulation as KK. 1 )Pi(t) = 3'21 SUBPOPij(t) (5.1.8) A distributed delay is an excellent way of modeling the physical tocess of maturation. For some purposes, the actual age and weight ly serve better. this is the case for prediction of the oneset of merty in heifer calves. What is needed is a delay model that is It distributed; the output rate is a uniform delay of a prior input te. When proportional losses and the potential for a time—varying lay length are required, the delay process becomes more complicated an the simple illustration of Figure 4.3. 84 A simple discrete delay with no other characteristics can be esented by = X(t — d) (5.1.9) re: Y(t) = output rate at time t X(t) = input rate at time t d the time delay the time delay, d, is an integer number of DT time increments, then ) = X(t - N*DT) (5.1.9a) ice that this process can also be represented in terms of a cascade single stage delays, with each stage corresponding to a delay of : DT. Then, :) = R10; — DT) It) = R2(t - DT) 1(t) = [l - PLR(t)]*RN(t - DT) = X(t) (5.1.11) n the delay length, d, changes, then the number of stages separating input and output must change correspondingly, because the delay igth for each stage is fixed as DT. If the delay length increases, en there will be an interval with no output rate until the extra mber of required stages of DT delay are completed. If the delay ngth is decreased, then a large output rate will suddenly occur as e storage of the excess number of stages leaves the delay model. Subroutine DVDPLR has been developed to simulate this form of screte delay with proportional losses and variable delay time. Storage 5.1.7 I corres SUBPOI using (1150 PU age in this discrete delay is computed in the same way as equation 7 computes storage for a distributed delay. Since each stage esponds to a fixed delay of length DT, OPij(t) = DT*RINij(t) (5.1.12) For this simulation model cohorts l, 4, 5, 6, and 9 are modeled ng distributed delays and cohorts 2, 3, 7, and 8 are modeled using crete delays. The proportional loss rate, PLRi(t), that applies to the ith ort is made up of two loss factors. These are the death rate, ’(t), and the net rate of annual additions to the cohort, ADDRTi(t). Ice this variable is not on a proportional basis as is DRi(t), it divided by the current population of the cohort to obtain ‘ ADDRTi(t) “ {1(t) = DRi(t) - W (5.1.13) is loss rate is on an annual basis; therefore, within the delay )routines, the loss rate that applies for a single incremental e period is PLRi(t)*DT. If POPi(t) is zero and ADDRTi(t) is itive, a special mechanism adjusts the populations to reflect an ition of animals through the ADDRTi(t) control variable. DRi(t) r each of the nine herd cohorts is read into the model as data in e initialization phase of the main program. It is assumed constant er the entire length of the simulation. An improvement that could made in the model would be to internalize some part of the death te as a function of the weather variables,to simulate exposure, 6 asaafunction of the nutrient intake over time, to simulate weakenin constant variablv points ' PLRiIt) Tr repres DVLDPI the ar happe' as th the I to w wean ways intl coh de< 86 ening from poor nutrition. Since the death rates are actually tant, a more proper designation is DRi. ADDRTi(t) is a control able that the user can specify as desired at any of the decision ts which call for that as a variable for exogenous control. (t) is used in the cohort delay models as explained above. To reiterate, each of the nine cohorts of the cattle herd is :esented by a delay process, either subroutine DVDPLR or soubroutine )PL. The output flow rate from the ith delay, ROUTi(t), represents annual flow rate of animals from that cohort at time t. What pens to these animals? For example, what happens to female calves they age beyond the delay length given by DELAY8(t)? Presumably delay length is set to represent a natural weaning age which acts wean all calves which get to that age and still have not been ned. These weaned calves can be handled in a number of alternative 8: (l) entered into the replacement heifer cohort, (2) entered o the bred heifer cohort, (3) entered into the slaughter heifer ort, or (4) sold on the market as weaned calves. Management must ide what fraction of these calves are used for each of the above It uses. Three control variables act to carry out the manager's :ision; these are C3, CS, and C9. In a similar fashion the manager ;t decide how the output from each of the nine herd cohorts will be ad. Figure 5.3 is a block diagram of subroutine HDMOG4; in it are Lustrated the possible flows among herd cohorts, the control vari- Ies which direct those flows, and the general structure of the nographic model. 87 H Female Calves Marketed c_5 Heifers mketed Bir Births 2 skim . female heifers split I CALVES Cl der IS thv and t 5 are I I' i BIRTI RPOP a di matl tiov hei cob 88 Birth is the remaining process that this component model includes. hs are generated in HDMOG4 by multiplying an annual birth rate, t), by the reproductive population, RP0P1(t), for each of the three le reproductive cohorts——mature cows, replacement heifers, and bred ers. Total births, CALVES, is an annual input rate which must be t between male and female calves. 3 ES = 2 BR.(t)*RPOP (c) (5.1.14) i=1 1 i enotes the proportion of births which are female, and 1.0 - Cl he proportion which are male. The birth rate on an annual basis the reproductive population in the three female reproducing cohorts determined in subroutine BIRTHZ. The major function of subroutine IHZ is to determine the value of BRi(t) and RPOPi(t) at time t. Pi(t) is similar in definition to POPi(t) but is used ot maintain istinction between heifers giving birth to their first calf and ure cows who have calved at least once previously. This distinc- n is not properly developed by the POPi(t) variable because some fers in the reporduceable cohorts (numbers 2 and 3) leave those orts and enter the mature cow cohort before the calving interval been completed. Since the birth rate for the mature cow cohort based on a population of mature cows, a distinction must be drawn ween the two types of members residing in POP1(t). RPOP1(t) is actual number of mature cows in POPl(t) which have calved pre- usly. RPOP2(t) includes all members of cohort 2 which are preg- t and all former members now in cohort 1. RPOP3(t) includes all bers of cohort 3 which are pregnant and all former members now cohort l. LL .m e oft IIher RPOI bot coI 89 The variables BEGCAV and ENDCAV mark the beginning and the Iding of the calving interval. These are determined as a result 5 the previous breeding activity (roughly nine months in the past). Ien t is within the interval (BEGCAV, ENDCAV) values of BRi(t) and ’OPi(t) are usually non-zero, Outside of this interval they are >th zero. Subroutine BIRTHZ computes the birth rate for each breedable >hort by determining the derivative of the curve of accumulated irths over the time interval of calving. The points which make up 113 curve, BFRACi l = l, ..., INTCAV, are determined in subroutine 1, trat using the previous breeding activity. Figure 5.4 illustrates )w the birth rate is numerically approximated from the accumulated irth curve for cohort l. Subroutine IBRAT is called at the beginning of the calving season 3 establish the values for BFRAC Three variables provide a record 11' E the breeding history for each of the subpopulations of the repro- mcing cohorts. INBi is the number of servicings for SUBPOPij over 3' 1e breeding interval. CTIMijm is the expected calving date for UBPOPij(t) which results from the mth servicing; this time is the ummation of the actual time of breeding and the average gestation eriod. CPAT m is the accumulated fraction pregnant for SUBPOP (t) ij ij h fter the mt servicing. The BFRAC11 curve for a cohort is determined y these variables. The number of data points composing the BFRACil urve is = ENDCAV — BEGCAV NT CAV 0.025 (5.1.15) 90 cumulated Births slope = BR1(t) O O O BFRACl,ll I I I I I I I I l I I I l I I I I I ' I . I I 5 ° . . . I , I l I I I l I I I I I | I I I I I . I | : I . I I I I I l L ABFRACl,1 I i 1 BEGCAV t-dt t ENDCAV igure 5.4 Determination of the annualized birth rate by numerical approximation of the deriviative of the accumulated births curve ...“ ”an. —_.. . Q Figure 5.5 for each c: the accumu tration‘, o simulatior been bred tion, 3331 The cohor weighted Catt By assmn] ulation '. average determ: change j these a: occurs energy Pass be If the the 3:11 then t' do not the Inc Subjec grade: 91 re 5.5 illustrates the way in which the predicted calving variables each cohort subpopulation are used by subroutine BIRAT to determine accumulated calving curve for the entire cohort. For ease of illus- ion, only three subpopulations will be used whereas in a typical lation model run there are apt to be 5—10 subpopulations that have bred in each of the female cohorts. Again for ease of illustra- , assume that the number of animals in each subpopulation is equal. cohort average in Figure 5.5 is then simply the average of the ijk values for the three subpopulations without needing to be hted by the numbers of animals in each of the subpopulations. Cattle weights are determined through use of subroutine WEIGHT. ssumption, all animals represented by a particular cohort subpop- ion have the same weight. This value is, loosely speaking, an age weight of a fairly homogeneous group, but the value is not rmined by actually averaging individual animal weights. Weights ge in the cattle herd through two distinct but related processes; e are growth and aging. Within an individual subpopulation growth rs as a result of feed intake energy being in excess of maintenance gy requirements. Due to the passage of time, however, individuals between adjacent herd subpopulations, and even between herd cohorts. he animal that has just arrived in a subpopulation does not have same weight as the average of the subpopulation prior to its arrival the average must be recomputed to reflect its arrival. Animals ot continue to gain weight indefinitely; to account for this fact model has a constraint on weight for each herd cohort. The final ect of this submodel is price gradation of beef cattle. Fixed es have been assigned to each herd cohort. These remain constant Accumulat of Cohort 1.0 ' 0.5 ' 0L Figu 92 LEGEND umulated Fraction Cohort Calved —— subpopulation l subpopulation 2 —- subpopulation 3 .CIDO l l —— cohort average O 4 . : 4+— BEGCAV CTIMi32 ENDCAV gure 5.5 Determination of BFRACi values from the breeding history of each cohort subpopulation r unless the daily gain Then the e diverted t ulation. proportio' reflected The in the jl is SUBPOI mtpopul cussed e time acc who) If aniu the ave transfc 0f the that s ALPHAj Where Noti 93 ;s the maximum weight constraint is reached and the projected r gains would force weights higher yet but for the constraint. the excess of energy that has been directed to weight gains is rted to changing the price gradation of the cattle in that subpop— ion. This mechanism is assumed to reflect the changing muscle/fat artions that exist in cattle as weight increases and which are acted in changing prices for different levels of fat in meat. The variable Wij(t) represents the average weight of all animals 1e jth subpopulation of cohort i,i.e., those animals whose number JBPOPij(t). The daily rate of weight gain for animals within this opulation is DGAINij(t). If no aging (the maturation process dis— ad earlier) were to take place, then weights would change over according to t t) = Wij(t-dt) + J DGAINij(u)du (5.1.16) t-dt nimals aged, but did not change weight from changes in size, then average weight of a subpopulation would be solely a function of the sfer of animals between herd subpopulations. ALPHAi(t) is a measure 1e fraction of the animals in a subpopulation of cohort i that leave subpopulation in a single time increment. = * Ai(t) DT KKi DELAYi(t) (5.1.17) DELAYi(t) = the delay time of the ith herd cohort KKi = the total number of subpopulations in the ith cohort DT = the length of time of a simulation time increment. ce that ALPHAi(t) is bounded between zero and one. Because the ,; f KK. and DI 1 particula: of that c j of coho increment given by BETA..(t 1] where: C This ap} applied namely the eql is a WI these Mt) Equac Y0ung ferre each memb intc COhI Cal 94 and DELAYi(t) variables are constant over all subpopulations of a icular herd cohort, ALPHAi(t) is also uniform over all subpopulations hat cohort. BETAij(t) is the fraction of animals in subpopulation ‘ cohort i that were in the same subpopulation in the previous time 'ement; therefore an approximation to the true BETAij(t) value is :n by .ij(t) = SUBPOPi.(t)*(1 — ALPHAi(t)) (5.1.18) C 'e: = * _ * C SUBPOPij(t) (1 ALPHAi(t)) + SUBPOPi,j_l(t) ALPHAi(t). . approximation allows the simple form of equation 5.1.16 to be .ied to each of the two sources of animals in subpopulation j; :ly animals from subpopulation j and subpopulation j-l. Therefore equation for the weight of members of subpopulation j of cohort l weighted sum of the weights and rates of weight gain from both of ;e sources. t 't = BETA.. t * W.. t-dt + DGAIN.. d 5.1.19 I) 1J<)[ lJ( ) [Ht 13m T] < > t + (1 — BETAij(t))*[ Wi’j_1(t-dt) + I DGAINi,j_l(T)dT t-dt ition 5.1.19 applies to all subpopulations of cohort i except the Igest; the youngest is different because it involves animals trans— red between cohorts. The weight of the youngest subpopulation of l cohort,e.g. Wil(t), is determined from the weight of the previous ters of that subpopulation and the weight of the animals transferred I it from another cohort. An example of this transfer between Irts is the movement of heifers into the mature cow cohort. The f cohorts use a fixed birth weight for male calves, and another ‘h fixed' to net COW C( ble W impac herd of d DGA 95 weight for female calves. An improvement to the model would be ke these birth weights a function of the nutrient intake of the ohort through time, especially during gestation. Weight constraints for each herd cohort are imposed by the varia— STMAXi(t). The rate of weight gain determined in the nutrient t component is checked to see how it affects the weight of the subpopulations. This check results in the following redefinition e rate of daily gains based on the values obtained from NUTRN. f DGAINij(t), {t if Wij(t—dt) + J DGAINij(r)dr< WGTMAXi t—dt 0, ij(t) = 1 if Wij(t-dt) s WGTMAXi (5.1.20) * FRAC DGAINij(t), if Wij(t—dt) < WGTMAXi, t and Wij(t—dt) + I DGAINij(r)dr > WGTMAXi t—dt K WGTMAX. — W..(t—dt) FRAC = __t.___1__._1_3_.__ (5.1.21) DGAIN..(r)dT Jt—dt: 13 Grading of beef cattle is a complicated procedure that heavily ves subjective gradations of differences by human beings. Many of bjective measures used, moreover, are such cattle descriptors as size, height, etc. Since the gradation process and the objective 3 on which it is based are missing from this model the determina— of grades will quite arbitrary. An attempt will be made to relate yes in grade from those initial values entered as data by the user \\ to 0( the 1 for five sep nee 96 isions when the projected rate of weight gain conflicts with Kimum weight constraint. Individual price grades are specified 2h subpopulation of each cohort,i.e., GRADEij(t). There are iscrete steps of grades and associated with each of them is a te cattle price; this price structure is the reason for the or cattle grading. Over the entire five grades there might be :e difference of 100%. Grades are determined as follows, if wij(t—dt) = WGTMAXi (5.1.22) GRADEij(t—dt), { t GRADE .(t-dt) + [ DGAIN..(T)dr*FATFAC, 13 t-dt 13 Q.(t) = t if Wij(t—dt) < WGTMAXi FATFAC = a parameter relating predicted rates of weight gains at maximum allowed weight to changes in price grade through the mechanism of muscle/fat proportions GRADEi.(t) = an integer value (from 1 to 5) indicating the J proper price grade characterizing any animals in the jth subpopulation of cohort i; highest prices are for grade 1, and lowest prices for grade 5. v.2 dif CE] as 97 orage Growth Component ‘he grazing process is a complicating factor in the already :ult problem of modeling plant growth. The interaction of cattle -ants is not yet well understood. Cattle have preferences for in species of forages over others, but these preferences are dynamic a composition and digestibility of each plant species change over rowing season. To fully explore the dynamics of the grazing process e1 must include the following factors: (1) growth and maturation of individual species as determined by exogenous weather variables and field conditions of canopy height, competition, slope, soil type, etc. (2) digestibilities and consumption preferences for the major plant components——stem, leaf, new shoots, and seed (3) impact of grazing on growth, regrowth, reproduction, and regeneration of plant species. ly the above task is beyond the scope of this thesis. The model oped to meet the requirements of the forage growth component, as d in the problem statement section of Chapter III, is only an al step on the road toward a complete grazing model. The growth model developed for this thesis is an adaptation of sslands model of Parton and Marshall[40]. Net photosynthetic y conversion to plant material is a product of a transfer coef— nt per unit of material times the amount of effective material the level of incoming solar radiation times the minimum of three h quality condition indices. These indices measure the relative ty of temperature, soil moisture, and soil nutrients in terms of 98 ,imal growth conditions. The net growth predicted by the basic energy [version equation is partitioned into growth of greenery and increases root storage by a function of the prior quantities of these values. The amount of plant material that can effectively convert solar iiation to plant growth is limited due to shading. This constraint approximated by imposing a plant density limit on the quantity of ant greenery per unit area which can convert solar radiation. Figure 6 illustrates the determination of the photosynthetically active ount of plant greenery, ACTIVEn(t), per unit area. Three quality indices are used to describe the relative quality factors of plant growth other than solar radiation. The factors are :mperature, soil moisture, and soil nutrients. The index variables :scribing these are TPF(t), SMF(t), and SNFn(t), respectively. Use of Iese indices implies the assumption that certain levels of these vari— Iles are better for plant growth than others, regardless of the level 5 solar radiation. Figures 5.7 through 5.9 illustrate the relationship :tween each of these variables and its corresponding index value. The 1ta points for each of the four figures, 5.6-5.9, are derived from 1rton and Marshall[40] and Sauer[47]. Better values could be determined trough experimental study of weather conditions and plant growth at a Irticular enterprise location. The values of the three quality indices 7e determined by use of the table interpolation function TABLIE[32]. 1e value of the variable is the argument and the index value is the :sult of the interpolation between data points. The net growth rate of total plant material in any land parcel, [0T0n(t), is determined through the following equation, ..— ACTIV 1800 Fig 99 IVEn(t) .—._._—— —-——_ _——————.—_———_ 0 1800 GRNn(t)/LANDn (kg/heC) ure 5.6 Determination of the photosynthetically active quantity of plant material per unit area (t) ~ 0 O . o O . O O o 10 20 3o 40 50 60 7o AVGTMP (t) (° C) Ire 5.7 Plot of points determining the temperature quality index, TPF(t), from the average temperature .75 .50 100 [th) O O O O 5 O O 0 P O O O 5 . 0 O ’ . . - . . . 1 . 1, . 0 2 4 6 8 10 12 14 16 18 20 22 SM(t) (cm) .gure 5.8 Plot of points determining the soil moisture quality index, SMF(t), from the soil moisture [F (t) r O O O O 0 0 ,ST 0 0 ' O 5 _ O O 0 20 40 60 80 100 SNUTn(t) gure 5.9 Plot of points determining the soil nutrient quality index, SNFn(t), from the soil nutrient value 101 HOT0n(t) = ACTIVEn(t)*PHOMAX*SOLAR(t)*MIN (5.2.1) here: PHOT0n(t) = total plant growth rate in land parcel n-—kg/hec/day ACTIVEn(t) = active plant material in land parcel n—-kg/hec PHOMAX = conversion factor, 0.0004 kg/kg/langley/day SOLAR(t) = solar radiation in langley/day MIN = minimum value of TPF(t), SMF(t), and SNFn(t). his rate must be partitioned into the proportion going to greenery rowth and the proportion going to root storage. Since the quantity f greenery and indirectly the amount in root storage in each land parcel s a function of the previous grazing and harvesting activity, the pro— ortioning factor ZX3n(t) is determined separately for each land parcel. X3n(t) = PAR3 + PAR4*(1.0 - e‘PAR5*GRNn(t)/LANDn) (5.2.2) here: ZX3n(t) = proportion of PHOT0n(t) going to root storage GRNn(t) = quantity of greenery in land parcel n——kg PAR3, PAR4, PARS are constant parameters. transfer of growth from root storage to greenery occurs at a rate hat is largely controlled by the proportions of the total plant material hat are in root storage and greenery. This accounts for the spurt of rowth observed after a field has been heavily cropped as new shoots ppear from the crowns of the roots. - PAR2*LAND n(t) = PAR1*ROOTn(t)*e GRNn(t)/ n (5.2.3) here: F (t) = rate of material transfer from root storage to greenery n ——kg/hectare/day LANDn = area of land parcel n in hectares 102 ROOTn(t) = quantity of root storage per hectare—-kg/hectare PARl, PARZ are constant parameters. he net growth rates of greenery and root storage in a pure growth ituation are determined by equations 5.2.1 ~ 5.2.3; these result in GRN (t) —-‘i—— = (PHOTOn(t)*(1.0 « ZX3n(t)) + Fn(t))*LANDn (5.2.4) dt ROOT (t) ——n—— = PHOTOn(t)*ZX3n(t) - Fn(t). (5.2.5) dt This component model is not directly concerned with these pure rowth equations. It is, however, the base on which consumption by nimals and harvest by machine can be built. AHRn(t) describes the rate f consumption of greenery in land parcel n by grazing animals, while HRn(t) describes the rate of greenery removal through harvesting. Both f these variables have units of kilograms per day. Neither grazing nor arvesting is a pure consumption process; there is waste from trampling, eces, dropping of cuttings, etc. A waste factor is used to model this ccurence. MWF is a mechanical waste factor, with value 1.10, that is onsidered constant for all levels of forage harvesting. AWFn(t) is a aste factor for the grazing process; its value is a function of the tocking rate in a particular land parcel n. Stocking rates represent he number of animals per unit area of land. This model uses the control ariable, PDSTRBni(t), to describe the proportion of the population of ohort i grazing in land parcel n. Equation 5.3.6 determines the tocking rate for land parcel n. 9 rocmna) = “13m * 2 PDSTRBni(t)*POPi(t) (5.2.6) n i=1 5 103 :e: STOCKLn(t) = the stocking rate of grazing animals in land parcel n —— #/hectare PDSTRBni(t) = the proportion of the population of cohort i which is grazing in land parcel n POPi(t) = the population of cohort i LANDn = size of land parcel n —- hectares. ire 5.10 illustrates the relationship between stocking rates and grazing waste factor, AWFn(t). As the stocking rate increases, waste factor is also increased reflecting greater impacts of cattle :he land. AWFn(t) .5- O O O O O .0” . 5 D 0 l. 2. 3. 4. 5. STOCKLn(t) (animals/bee) 'igure 5.10 Relationship between grazing waste factor, AWFn(t), and the stocking rates, STOCKLn(t) 104 .roduction of these sources of greenery consumption and their ;pective waste factors concludes the factors which affect the tnge of greenery in pasture. This gives the net growth rate :d in the model, [lg—(L) = (memo (t)*(1.0 — 2x3 m) + F (Mum) - AWF (t>*AHR (c) . K n n n J n n n — MWF*MHRn(t) (5.2.7) The quantity and the quality of forage growth consumed by animals 1 equally important. DIGESTn(t) describes the quality of forage in 1d parcel n in terms of total digestible nutrients (TDN) on a dry sis. Digestibility is a dynamic variable that reflects time changes average plant composition and in actual changes in the energy value the individual parts. Since this model does not maintain forage intities on the basis of constituent parts, a less rigorous approach digestibility must be used. Digestibility is modeled as a function three factors: relative age of green material, time of season, 1 density of greenery per animal grazing. The base digestibility .ue (in terms of Z TDN on a dry matter basis) is determined by a relative age of green material in a particular land parcel. This lel formulation assumes that the major element in the change of rage digestibility over time is a process of decline associated :h the length of time since growth first started. Time in the iwing season and greenery density per grazing animal are Viewed as :ondary factors modifying the base digestibility value. Mathemati- .ly, forage digestibility is described by iESTn(t) = BASEDGn(t)*FRQUAL(t)*FDENSEn(t) (5.2-8) 105 re: DIGESTn(t) = forage digestibility of green plant material in land parcel n as Z TDN BASEDGn(t) = base forage digestibility determined from the relative age of forage greenery in land parcel n FRQUAL(t) = seasonal digestibility factor modifying the base digestibility value BASEDGn(t) FDENSEn(t) = forage density digestibility factor modifying the base digestibility value BASEDGn(t). The base digestibility of forage in any land parcel is deter— sd by linear interpolation between data values representing digesti— Lty versus relative age. The argument used to obtain the resultant ige digestibility value is the variable PINDEXn(t). This variable In index representing the amount of time that would be required the current forage quantity in a land parcel to be grown at the :ent rate of growth existing within that land parcel. Mathemati— Ly, _ GRN (t) )EXn(t) — n (5.2.9) dGRNn(t) dt : index is able to represent the major situations in plant growth Ising the ratio of current quantity over the rate of change of : quantity. Figure 5.11 graphically plots the relationship between and digestibility which is the key to this model of plant quality. :an be seen from the graph, the plot is a monotonic decline in Lge quality with increasing relative age. This formulation in unction with equation 5.2.9 is able to correctly represent the ,ity of forage over time. For example, if during a period of high 'th mechanical harvest were to severely reduce the quantity on hand, \\ 106 BASEDGn(t) 1.0 .75» O O O I O O I .50' . Q .25' 0 A I A A l A A 1 I A o 10 20 3o 40 50 PINDEXn(t) Figure 5.11 The intermediate digestibility, BASEDG (t), as a function of pseudo—age, PINDEXn(t) n an the relative age determined by equation 5.2.9 would be quite v. Using the plot of Figure 5.11 gives the correct fact of relative- high forage quality. Others cases encountered in forage growth :uations are also handled by this model of forage digestibility well. Although the relative age of forage is important in determin— ; digestibility, it is not sufficient. There are qualitative differ- :es in the new growth over the length of the growing season and in the rticular growing conditions encountered. This factor is incorporated :o the model through use of a multiplicative forage quality factor EUAL(t). FRQUAL(t) is determined by integrating the value of TPF(t) tr the length of the growing season and using that integral value determine an index of the degree of plant maturity attained. ,1 5‘ 107 FRQUALn(t) 1.01L A A A A A A A ‘ A A .75 ' .50 . .25 h 0 a A A A # L l A A A 0 50 100 150 200 250 SUMTPF(t) igure 5.12 The growth quality factor, FRQUAL(t), as a function of the integral of the temperature index over the growth season TPF(t) = TPF(T)dr (5.2 10) 4TSPRNG SUMTPF(t) = the integral of the temperature quality factor over the growth season to current time t TPF(t) = the temperature quality factor as determined from use of Figure 5.7 and the average temperature TSPRNG = time at which spring growth begins. TPF(t) is used as the argument for another use of the table func— 1 TABLIE[23] to determine the value of FRQUAL(t). Figure 5.12 indi— as the relationship between SUMTPF(t) and FRQUAL(t) graphically. 108 FDENSEn(t) 1.0t O O D O C .75- o .50' .25’ 0 5 10 15 20 25 30 GRNDENn(t) Figure 5.13 Plot of the forage density digestibility factor against per animal density The final factor modifying the base digestibility value is NSEn(t)-—the influence of the current amount of forage per animal digestibility. The value of this density factor——GRNDENn(t)—-is d as the arguement to obtain FDENSEn(t) by linear interpolation ween data values as represented in Figure 5.13. GRN (t) 2 DENn(t) = _.__n_____ (5. .11) STOCKLn(t)*LANDn*DAYS re STOCKLn(t) = animals grazing in land parcel n—-#/hectare LANDn = number of hectares in land parcel n GRNn(t) = greenery in land parcel n——kg GRNDENn(t) = forage density per grazing animal—-kg/anima1. \\ 109 This development concludes the description of the model for etermination of forage digestibilities; three factors have been hown to be important to this modeling effort: relative age of orage in land parcels, season of the year, and the density of orage available for grazing in each land parcel. Two state variables remain to be defined before this component vodel's description is complete. SM(t) is a variable describing the mount of soil moisture. SNUTn(t) is a variable describing the amount f soil nutrients. The rate of change of soil moisture is dependent n three factors: SM(t) dt = RAIN(t) — EVAP(t) — PERC(t) (5.2.12) here: SM(t) = soil moisture at its current level—-cm in effective root depth RAIN(t) = rainfall in cm/day EVAP(t) = evaporation in cm/day = a*AVGTMP(t)2 PERC(t) = percolation beyond root depth in cm/day —SM(t)/c II b* 1.0 —e a, b, c are constant parameters. 'he rate of change of soil nutrients is a function of the rate of hotosynthetic energy conversion and of fertilizer application to the oil. This implies that the greater the growth rate, the larger is .he withdrawal of nutrients from the soil. iSNUTn(t) = — *PHOTO t (5.2 13) dt SPFERT B n( ) 110 SNUTn(t) = current level of soil nutrients in land parcel n SPFERT = rate of applicatin of fertilizer per unit of land PHOTOn(t) = rate of photosynthetic energy conversion in land parcel n B = a constant parameter. This forage growth component model is relatively crude. It Jides the basic elements needed to evaluate realistic grazing agement strategies but could be improved considerably. This model also very empirical, as is evidenced by the extensive use of TABLIE Look up interpolations between data values. The data values used uld, and could, be tuned to a particular location. Figure 5.14 a block diagram of the entire component model. This figure is a venient way to trace the interconnection of the model variables they have been described in this section. lll ..TElE some , L f(GRN,TEMP,SOIAR,SM£IQr PHOTO I ‘ H 1.0 + zr3 - Z M / Home) I H STOCKL + f(GRN,ROOTS) + 2 _ 11 £213. F _ GRNDOT H Elli 5““ MWF GRN r %,GRNDOT,TEMP ,STOCKL) 'igure 5.14 Block diagram of the forage growth component 112 Feed Stock Component The feed stock component consists of basic accounting for the .ous activities involving feed stocks. Chapter IV, section 2, :ribed these activities. Subroutine FEEDS performs this accounting :he enterprise simulation model. Quantities of feed stocks are represented by the variable FSTOCKn(t). equation determining current feed stock levels uses the previous :1, the amounts bought and sold, the amounts fed to cattle, and the Int lost through waste and pests. Thus )CKn(t) = FSTOCKh(t—dt) + STKPURn(t-dt) + CROPn(t—dt) CSALESn(t—dt) - STKFEDn(t—dt) FRCLOSn*FSTOCKn(t~dt)*DT (5.3.1) FSTOCKn(t) = current quantity of feed stock n —- kg STKPURn(t) = quantity of feed stock n purchased —— kg CROPn(t) = quantity of feed stock n produced —— kg CSALESn(t) = quantity of feed stock n sold —- kg STKFEDn(t) = quantity of feed stock n fed to cattle —— kg FRCLOSn = annual rate of loss of feed stock n DT = length of simulation time increment —— years. of the sources of increase or decrease for FSTOCKn(t) except losses determined in other model components. FRCLOSn is a constant frac— al loss rate assumed to be uniform over the entire year. The digestibility of feed stocks in terms of a TDN value are lied by FQUALn(t). Changes in quality are a result of purchases, production, and spoilage. Spoilage is assumed to mean that the 113 :y of the crop declines over time. SPOILn is an annual frac— l decline in quality of crop n; like FRCLOSn it is assumed to nstant over time. FQUALn(t) is determined by .n(t) = FQUAN*FQUALn(t—dt)*SPOILn*DT Fsmocxnu) + PRQUAL7n (t) * STKPURn (t) + CROPn (t ) *CROPQLn (t) FSTOCKn(t) (5.3.2) FQUAN = FSTOCKn(t-dt) - CSALESn(t—dt) — STKFEDn(t—dt) PRQUALn(t) = the TDN value of feed stock n purchases CROPQLn(t) = the TDN value of feed stock production. The final variable determined in this component is the total Ial crop production of feed stock n, namely CROPGn(t). This re gives a running total of the production to date and could be >ared with XCPRODn(t) to obtain an estimate of the quantity of 1 stock n that might yet be harvested in this growing season. ?Gn(t) = CROPn(T)dT (5.3.3) I t TSPRNG ‘ Figure 5.15 is a block diagram illustrating the total inter— ationship between variables that are included in this component 31 of feed stocks. 114 we CROPQL CROP FRCLOS Z + FSTOCK + STKPUR PR UAL SPOIL ’d n 1 ~ m + RT. FQUAL Figure 5.15 Block diagram of the feed stock component. 115 I Nutrient Impact Component The model for this component develops the impacts of nutrient take in two areas: reproduction and growth. The nutrient impacts .at are modeled are those related to energy; protein requirements I well as vitamins and trace minerals are assumed to be available in ifficient quantities that energy is the predominant variable of in— erest. The majority of this component is contained in subroutine UTRN, and is described in Schuette[48]. This thesis will not present hose results except as they interact with the remainder of the system Iodel. Feed quantities and qualities are a prime concern of this compon- ent. Feed is described in terms of four variables for each herd cohort: (l) CNCALi(t)——the quantity of concentrates allocated to cohort i, (2) RHGALi(t)——the quantity of roughages allocated to cohort i, (3) IDNCi(t)——the average TDN value of the concentrates allocated, and (4) TDNRi(t)—-the average TDN value of the roughages allocated to "ohort 1. These variables are inputs to subroutine NUTRN which then determines the impact of these feed inputs. NUTRN returns several variable values: TDMICi(t)—-quantity of concentrates actually consumed by cohort i, TDMIRi(t)——quantity of roughages actually consumed by cohort i, and DGAINij(t)—-the rate of daily gain by the jth subpopula- tion of cohort i as a result of the interaction of its feed intakes of energy and requirements for energy. Total roughage allocation to any cohort is composed of feeds from grazing as well as those from feed stock allocations. Usually these sources are disjoint, but there are some conditions under which grazing vould be supplimented by feed stocks. The procedure used here is to 116 nine the roughage allocation as a result of the grazing distri— n of the herd, and add to it the voluntary feed allocations. The . grazing roughage allocation for the ith cohort is then * * NLANDS PDSTRBni(t) POPi(t) GRNn(t) iL (t) = i n21 (5.4.1) 9 Z PDSTRB n it 1:1 POP (t) l FORGALi(t) = current forage available to the ith cohort as a result of its distribution over the land parcels .. kg GRNn(t) = total quantity of greenery in land parcel n —- kg PDSTRBni(t) = fraction of the population of cohort i which is grazing in land parcel n POPi(t) = current population of cohort i. digestibility of this allocation in terms of TDN is a function of forage digestibilities in each land parcel. NLANDS IPDSTRBni(t)*POPi(t)*GRNn(t)*DIGESTn(t)I TDNi(t) = 2 9 ‘ n=1 2 PDSTRBn2(t)*POPl(t)*GRNn(t) r=1 re: (5.4.2) FORTDNi(t) = digestibility of the FORGALi(t) allocation in TDN DIGESTn(t) = the TDN digestibility value of the forage in land parcel n. Addition of the roughage from grazing and the voluntary allocation es the total roughage allocation to cohort i, ALi(t) = FORTDNi(t) + FEEDALi(t) (5.4.3) ch has a digestibility value determined by the weighted average of different sources of the roughage. 117 = * (t) FEDTDNi(t) FEEDALi(t) + FORTDNi(t)*FORGALi(t) (5.4.4) FEEDALi(t) + FORGALi(t) FEDTDNi(t) = the average TDN value of the roughage voluntarily allocated to cohort i FEEDALi(t) = the roughage allocation voluntarily given to herd cohort i —— kg/DT. Determination of the value of FEDTDNi(t) and TDNCi(t) requires ; the allocation of TDN from roughage and of TDN from concentrates :elated to the actual physical quantities of feed stocks that are This results in * * . (t) = CATTINli£(t) POPi(t) DAYS (5.4.5) 1 CNCALi(t) CATTIN . (t)*POP (t)*DAYS TDNi(t) = 212 1 (5.4.6) FEEDALi(t) re: TDNCi(t) = average TDN value for the concentrate allocation to cohort i FEDTDNi(t) = average TDN value of the roughage allocation to cohort i CATTINl.£(t) = quantity of TDN allocated to each member of l cohort 1 per day from concentrates under plan 2 = quantity of TDN allocated to each member of CATTIN cohort i per day from roughages under plan 0 212(5) CNCAL (t) = physical quantity of concentrates allocated to the i . ith cohort —— kg/DT FEEDALi(t) = physical quantity of roughages allocated to the ith cohort -— kg/DT DAYS = number of days in a simulation time increment. units of the variables FEDTDNi(t) and TDNCi(t) are kg TDN/kg feed, 118 lowing directly from their definition as the average TDN value of uantity of feed stocks fed to cattle. Determination of the actual quantity of feed stocks fed to cattle each time increment of the simulation requires knowledge of the ired distribution of the TDN allocations over the possible feed stocks sting. The control variables CNCFRCin(t) and RHGFRCin(t) are used management to distribute the cohort allocations of concentrates and ghages over the range of available feed stocks. STKFEDn(t) is the iable used to represent the actual quantity of feed stock n used for tle feeding in the current time increment of the simulation. This iable is not a rate but a quantity, as are all variables affecting current quantity of feed stock n on hand——FSTOCKn(t). Since the d stock variables include both roughages and concentrates in a single ting, the following equation determines the amount of feed stock n d for cattle consumption in this time increment. 9 ...- _. * .FEDn(t) — DAYS*iEl POPi(t)* CATTIN1i (t) CNCFRCin(t) - " 1 *-——————- 5.4.7 + CATTIN2i (t)*RHGFRCin(t) FQUALn(t) ( ) re STKFED (t) = quantity of feed stock n fed to cattle in this DT n CNCFRC. (t) = fraction of the TDN allocation of concentrates to in cohort 1 derived from feed stock n RHGFRC. (t) = fraction of the TDN allocation of roughages to In cohort i derived from feed stock n FQUALn(t) = current TDN value of feed stock n POPi(t) = current population of cohort i. Equation 5.4.7 brings together the actual quantities of feed Icks that have been used to feed cattle. It combines the consumption roughages and concentrates in one single equation, but Since this isa proi str' men mat 41“ {I555 119 a mutually exclusive description there is no overlap to cause )blems. Therefore, the matrices CNCFRC(t) and RHGFRC(t) are con— ructed so that when the inth element of one is nonzero, the inthele— it of the other is zero. Additionally, the rows of both of these trices must sum to one if any allocation is to be made. This require- nt exists because either a row of CNCFRC(t) or RHGFRC(t) represents e fraction of the cohort allocation that is drawn from each of the ed stocks. A final note concerning the use of the feed stocks for nsumption——this model assumes that the entire amount of the feed ock allocated for consumption is used. Any quantities allocated to e herd or a specific cohort which are not consumed are assumed to be mpletely wasted. The consumption characteristics of the herd cohorts determine ether the allocations to the herd are completely consumed or not. en the herd is not grazing, any excess allocation over consumption is tally wasted. When the herd is grazing, however, a differentiation st be made between nutrient intakes from grazing and from voluntary eding. The following priority of cattle consumption preferences is sumed to hold: (1) cattle prefer concentrates to roughages, and (2) Lttle prefer feed stock roughages to grazed roughages. Differential feferences among feed stocks are assumed to be negligible, although Iey do in fact exist. The consumption of roughages by the herd which 7e not covered by voluntary roughage allocations comes from grazing. Iis grazing removal of forage must also be distributed across the Lfferent land parcels being grazed. The variable PDSTRBni(t), which is used to obtain the stocking rates in the land parcels, is used here 3 distribute the forage removal by grazing over the land parcels. 2‘ fl 120 5 HR (t) = n i=1 [PDSTRBin(t)*(TDMIRi(t) — FRAC*FEEDALi(t))] (5.4.8) here: TDMIRi(t) = actual roughage consumption of cohort i AHRn(t) = grazing removal of forage from land parcel n FRAC = proportion of roughages consumed that are from voluntary feed sources SPLIT, if RHGALi(t) > FEEDALi(t) 1.0, otherwise SPLIT = a parameter based on cattle preferences. Variables internal to subroutine NUTRN maintain the reproductive ondition of each subpopulation of the breedable female cohorts. See chuette[48] for a complete discussion of this model. Reproductive ffects on male breeding cohorts have been ignored up to this point in he model's development. When breeding activity occurs, the variables PATijk and CTIMi. describe the resulting pattern of pregnancy rate jk nd timing for the jth subpopulation of cohort i from its kth servicing. ne variables are used in Subroutine BIRAT of the population demography omponent to develop overall cohort birth curves. The simulation model of this component is the subroutine CONSUM. 1e subroutine develops values for STKFEDn(t), AHRn(t), DGAINij(t), and eproductive impacts reflected in the values of CTIMijk and CPATijk ien breeding occurs. Figure 5.16 represents a block diagram of the 1broutine CONSUM. CATTINl H H n n llll I'll lqg F UAL n d TDNC .FORTDN d EEEAL FORGAL + + f(FEEDAL,RHGAL) FRAC I I 7 SUBROUTINE NUTRN - o " TDMIR + DGAIN reproductive AHR TDMIC effects CATTINZ lllll —————4 POP 11 .- FEDTDN FEEDAL Figure 5.16 Block diagram of the nutrient impact component. PDSTRB .5 Management Decision Making Component This section will provide the full details of the model of the ‘ nagement component. As outlined previously, management decisions [:ve been separated into two discrete groups; those decisions which fan be determined endogenously using standard economic criteria and :hose whose nature is not so obviously solvable. This latter group If decisions is acted on by the user of the model through exogenous supply of control variable values. The purpose of this section is :o explain the details about how decision points-—times when exogenous :ontrol is exercised-—are recognized, to give the form of the equa— :ions for carrying out specific decisions, and to review the criteria for endogenous decision making. Four of the five decision points are organized around physical :vents; these are the onset of spring growth, breeding, weaning, Ind culling. These will usually occur in the order listed above, but not necessarily so. The fifth decision point is the occurrence of an :vent requiring the attention of the model user. These are fall feed :tock supplies unsynchronized with feeding plans, feed stocks unable :o carry the herd for two DT time increments, and working capital falling below acceptable levels. One characteristic which makes decision making difficult at :hese five points is the fact that the prices the manager expects Ilay a key role in the action he takes. An additional complication ,s the uncertainty in knowing how to act with regard to husbandry rctivities such as weaning, culling, and breeding. The very long 123 time delays involved in the production of marketable cattle (at whatever stage of maturity) and in the introduction of replacement breeding animals imply that the manager must guess the future market and hope he does so correctly. The response of many managers is to go with the current price as a guide for future desired production. This results in wide industry supply fluctuations (and thereby price fluctuations) over a multi-year cycle. This cyclical behavior exists to some extent in all livestock products and is partly a result of the uncertainty (due to the delay lengths) that long production lead times bring to decision making. This component model is there— fore organized to recognize decision points, to allow the user to make his own judgmental decision, and to determine all other decisions internally using standard economic criteria. Organization of this component model can be easily understood through review of Figure 4.9. In every time increment of the simu- lation model subroutine INQUIR attempts to determine whether any of the five forms of decision point currently exist. Figure 5.17 illustrates the construction of INQUIR. In each DT simulation time increment the current time value is checked against each of the five possible time points that constitute decision points-—TSPRNG, TBRD(1), TBRD(Z), TWEAN, and TCULL. Additionally, the three criteria that call for a special decision point are checked. If any of them are true, then the special decision point is called by setting IFLAG equal to 2, while IFLAG is set to 1 at any of the regular decision points. If the IFLAG variable value is not equal to zero at the conclusion of the decision point checks, then INQUIR creates a permanent file of 124 Ill values that are needed to preserve the simulation in its current ;tate. This permanent file is available for access at any time to :estart the time simulation exactly where it left off. If the IFLAG value is still zero (as it was set as the initial action in the sub— :outine) at the end of all decision point checks, then control skips from subroutine INQUIR to subroutine NORMAL. Subroutine INQUIR is constructed (again see Figure 5.17) so that each decision point is checked every DT increment of simulated time so that multiple decision points at any one time can be detected. in example of this is the event of weaning and culling occurring at the same time. INQUIR will have gotten all the way thrOugh its structure to the point of checking whether time is equal to TWEAN; since this is true, the if statement branches to YES, IFLAG is set equal to one, and the detailed states of the model and any other information that the decision maker might need to make a decision are printed onto the output file for the user to observe. Following the Logic of INQUIR, the checking of time for equality to TCULL is the next operation. The if statement check is true, then the logic path of YES is followed. This logic branch results in additional printing of state variable and other decision information, but the printing for each decision point is tailored for that decision point resulting in some new information being made accessible to the decision-maker. In practice the values of TBRD(1) and TBRD(2) are also likely to be equal; that is why the logic of INQUIR bypasses part of the YES branch of the TBRD(2) if statement check when time has already been 125 IFLAG = 0 START YES PRINT = TSPRNG IFLAG=1 . 1. price expectations 4. financial status PRINT IF ————‘ N0 T = TBRD(1) IFLAG=1 price expectations herd populations subpopulation weights feed stockl ml financial status WL‘WNH A T PRINT IF IF “-—— T = TBRD(1) price expectations herd populations subpopulation weights feed stock levels . financial status T = TBRD(2) IFLAG=1 mpmmb—a m ... . price expectations . calf cohort numbers and weights for each subpopulation . financial status N La J PRINT H price expectations . mature cow cohort numbers and weight for cat ch of the subpopulations . financial status n T = TCULL YES IFLAG=1 N 1...) N0 Figure 5.17 Flow chart of subroutine INQUIR of the management component 126 gRINT . price expectations NH . financial status 4‘ T v COMPUTE FLEVEL (t): the e n in this increment IF PRINT 1. rice ex ectations LEVEL>FSTOCK IFLAG =2 2. herd numgers and wghts 2 . for all subpopulations 3. feed stock levels 4 feed requirements N0 5 crop expectations 6 financial status YES COMPUTE FWLONG : the win- N0 1.price expectations 2.herd numbers and ugh: for all subpopulations 3.feed stock levels 4.winteting feed needs 5.crop expectations 6.financial status = irLAG=2 WRITE NO CREATE THE RESTARTFILE—>. BY SUFFERING OUT ALL VARIABLE VALUES WHICH ARE NEEDED TO RESTART THE SIMULATION AT THIS YES EXACT POINT Figure 5. l7 (cont 'd.) 127 recognized as equal to TBRD(1), thus setting the value of IFLAG to one. The logic of INQUIR has been developed so that multiple decision points existing at a particular point in simulated time can be detected and acted upon. Subroutine RESPNS, which reads in control variable values, has a parallel structure to subroutine INQUIR. Exogenous Decision Making When a decision point has been detected, the decision information printed, the permanent file constructed, and the simulation stopped, the user makes his decisions based on any decision rule or intuitive feeling that he desires. Once he has made his decisions for a partic- ular decision point and the data cards for the control values are pre— pared, the user restarts the simulation by reading from the permanent file to fix all model variable values just as they were when the deci— sion point(s) was detected. The next step is to read the control vari- able values; subroutine RESPNS accomplishes this action. Figure 5.18 is a flow chart illustrating the construction of subroutine RESPNS. Subroutine RESPNS is constructed parallel to the structure of subroutine INQUIR; the control variables that are read in to direct the enterprise's operation at each decision point follow the input formats in User's Guide to the Beef Cattle Enterprise Simulation Model. When current time is equal to a decision point time, or when the value of IFLAG is two, then the logic path follows the YES branch from the if statement. The first action is to read the control variable values which apply for that decision point from the user supplied data cards; the second action is to perform any execution statements that are peculiar to that decision point. An example of such specialized exe— cution statements is the action to wean the calf cohorts; here the READ expected prices grazing controls expected growth forage harvest time harvest controls UIbHNH READ expected prices breeding duration feeding plans TWEAN, TCULL b~u N H READ . expected prices . breeding duration . feeding plans . TWEAN, TCULL #1 READ COMPUTE YES 1. expected prices 1. weaned calves ‘2. weaning controls 1. transfers to 3. minimum age heifer coh. 4. CULL 3 record time and ages and weights t No ean n A. weaning costs READ COMPUTE YES 1. expected prices . cull nature 2. culling pattern ____.. cow cohort 2' :::§ agiitions 2. determine ' s s culling costs READ . expected prices . herd transfers . feed stock actions . feeding plans @ Figure 5.18 Flow chart of subroutine RESPNS of the management component [‘ 129 control variable TFRAC£i(twean ) is used to transfer a fraction of the 2th female calf subpopulation to the ith heifer cohort. Similar exe- cution statements are required to cull the mature cow cohort using the control variable CULFRCj(tcull). When any necessary execution state— ments have been completed for a particular time related decision point, the logic path proceeds with the determination of the existence of other decision points in this simulation time increment. If a special decision point has been detected by subroutine INQUIR, then the value of IFLAG is two. If IFLAG has a value of two, then subroutine RESPNS branches to the YES side of the final if statement check, reads in any control variable values required, and performs any needed execution statements. At the conclusion of this step, the YES branch logic path rejoins the main flow within the subroutine; the last action is then to exit from the subroutine and proceed with the normal logic flow of the overall simulation model. The final element in the management component is subroutine NORMAL; this element will be discussed in considerable detail in the last part of this section under the subtitle "Endogenous Decision Making." Control is transferred to NORMAL from two points; first, from subroutine INQUIR if no decision points have been encountered in the current time increment. The second referral point for subroutine NORMAL is after subroutine RESPNS has completed its reading of control variable values and executed any statements carrying out the user's decisions. Spring growth onset and breeding decision points require the user to specify the value of various expectation and control values that will have long term effect on the enterprise's behavior. At the time spring .. I'llA .[Illllllfl 130 growth begins the user must specify several control variables that will influence the use of forage and its growth over time. PDSTRBni(t) con— trols the stocking rates on each of the land parcels through specifi- cation of the fraction of the population of cohort i that grazes on land parcel n. Equation 5.2.6 illustrates this effect. TMHRk specifies the times at which forage is to be mechanically harvested, while REMOVL controls the amount harvested. XCPRODn and XCQUALn provide information to the model about the expected amount and quality (TDN) of crop n that will be harvested in this growing season. This includes forage growth as well as feed crops. Breeding is the activity of the enterprise that is most affected by the uncertainties of cattle and feed stock prices. At this time the manager must decide how his reproductive females will be bred, the duration of the breeding period, and which animals are to be bred. These decisions are very complex since they are ultimately conditional on the prices that will occur in the market for cattle and for feed stocks in the interval between the beginning of this breeding period and the soonest date for calf sales. Moreover, the relationships between feed stock intake rates over time and the success of breeding, and between feed stock intake rates and weight gains of growing cattle are so complicated that only use of a simulation model of the herd is likely to correctly develop the impacts of breeding and feeding policies on calf numbers and weights. Knowledge of the physical outcomes of different feeding and breeding policies would then be combined with expected prices to obtain predictions of financial outcome conditional on the prices used. This simulation model can accomplish this invest— igation through use of the decision point mechanism developed to permit 131 just this sort of evaluation of control strategy alternatives. When the breeding decision point is reached, the management com— ponent recognizes this fact and creates the restart file of simulation variable values. It has also printed current state variable values to allow the user to have a complete picture of the current condition of the enterprise. The user makes the decisions about breeding and feed- ing policies he wishes to follow and restarts the model to determine the results of his decisions over time. When the next decision point of the simulation is reached this procedure is repeated. The user proceeds in this manner until the final time horizon has been reached for evaluation of the breeding and feeding policies he is investigating. He can then return to the restart file from which he started and repeated this entire procedure using a different set of control variable values representing a different breeding policy. By utilizing this sort of procedure to investigate the alternative breeding and feeding policies the user should be able to select one that is best for the price expec— tations and other assumptions he is using. Figure 4.11 illustrates the sort of alternative routes that can be developed using the decision point mechanism. The control variables that the user enters at the breeding deci- sion point are (l) DURBi(t)——the duration of the breeding interval for the ith female cohort, (2) FPLANSz(t)——the times at which the feeding plans for the herd end, (3) CATTIN (t)--the feeding rates for the ikl cohorts under the different feeding plans, and (4) DISTFDin£(t)-—the proportions of the desired TDN feeding levels to come from specific feed stocks under the alternative feeding plans. 132 Weaning is an event which requires extensive calculation to carry out the user's decisions. At the time of weaning, the population of calves (both male and female) will be spread among numerous sub— populations of cohorts 7 and 8. The user specifies how he wants these calves used through the variables AGEMIN and TFRACzi(t). AGEMIN is the minimum age that is to be weaned. This variable is included to allow control over weaning in extended calving seasons, where calves may be present in the entire range of subpopulations of the calf cohorts. TFRAC£i(t) is the fraction of female calves in subpopulation 2 that are to be transferred to cohort i. In all cases the index Q refers to the 2th subpopulation older than the minimum weaning age, and 1 refers to cohorts 2 or 3——the replacement heifer and bred heifer cohorts. An important consideration in the mechanics of weaning is to preserve the age distribution that has been maintained in the calf cohorts following transfer to the heifer cohorts. This age distri- bution is important to future determination of puberty in heifers, which is largely a function of age and weight. Discrete delays are used in the calf and reproductive heifer cohorts to allow the population to be recorded strictly on the basis of age, not on the basis of relative maturity as is done in the other herd cohorts. The individual stage delay length for the delay models of the calf and replacement heifer cohorts are equal so as to keep the same level of fineness of the age distribution. Figures 5.19 and 5.20 illustrate the way in which the variables AGEMIN and TFRACi(t) can select from the female calf cohort to create the new entries in the replacement igBPFP8j(twean) 20 ' o 9 l . ‘ : AGEMIN 9 : I o I l | I l I | . I I ' I I ‘ : 1o ' : . l i . , I I ' I I I : : l : : i 9 1 i : I o ' ' I I I : ' I t ‘ I l ' ' ' I I ' 3 I 0 l n n 4 l L n L L 2 4 10 12 14 16 18 Figure 5.19 Subpopulation values from the female calf to weaning action by management j index cohort prior gngpP2j(twean) TFRAC12(twean) 0 TFRAC22(twean) 0 TFRAC32(twean) = 0.5 20 ' TFRAC42(twean) = 0.5 TFRAC52(twean) = 1.0 Q TFRAC62(twean) = 1.0 2 U 10 ' O O : O I I I I I : z : : ' : l I ' : : i : I I I I O A. - 4 6 8 10 12 14 16 18 Figure 5.20 j index Subpopulation values from the replacement heifer cohort following weaning action 134 heifer and bred heifer cohorts. Figure 5.19 depicts a female calf cohort with six subpopulations older than the minimum age for weaning, AGEMIN. The table of values listed for TFRAC£2(t), £= l,...,6 indicate the fraction of the subpopulation indexed in the 1th position greater than the index corresponding to AGEMIN that are transferred to the replacement cohort. The result of the initial female calf cohort subpopulation distribution, and the indicated values of AGEMIN and TFRAC22(tL is the replacement heifer cohort subpopulation distribution given by Figure 5.20. Mathematically, SUBPOP (t) = SUBPOP (t)*TFRAC (t) (5.5.1) 22 £2 8,m+£ where: SUBPOP ij(t) = the population of the jth subgrouping of cohort i TFRAC£ 2(t) = the fraction of the m+£ subpopulation of cohort eight transferred through weaning to the replacement heifer cohort m = the subpopulation index corresponding to the oldest sub— population younger than age AGEMIN. Similarily for the females calves weaned and sent to the bred heifer cohort, = * 5.5.2 SUBPOP3£(t) SUBPOPS’m+£(t) TFRAC£3(t) ( ) where: TFRAC£3(t) = the fraction of the m+£ subpopulation of cohort eight transferred through weaning to the bred heifer cohort. The remaining calves in the female calf cohort older than AGEMIN are sold on the market, as are all male calves not retained for bull replace- ment S o 135 Sales of cattle is partially accomplished through use of the cohort delay output variable ROUT1(t). Subroutine RESPNS makes use of this variable when carrying out sales of the weaned animals not saved for the reproductive or fattening cohorts. The current value of ROUTi(t) has been computed in subroutine HOMOG4 in the demography component. Subroutine RESPNS redefines this value to include sales animals from the entire cohort rather than only those leaving the cohort due to satisfying the required discrete delay time. Then 1 m7 R0UT7(t) = [-—J E SUBPOP7 (t) (5.5.3) DT j=m j ROUT ( ) — i “:8 SUBPOP (t)*[l - TFRAC (c) — (5 5 4) 8 t ' DT , 8j j+l—m,2 ' ' J=m TFRACj+l_m’3(t)] where: ROUTi(t) = annual rate of output from the ith cohort delay process TFRAC£i(t) = fraction of the female calf subpopulation indexed m+£ transferred to the ith heifer cohort m = the subpopulation index corresponding to the oldest sub- population which is younger than AGEMIN DT = time increment of the simulation——fraction of a year. ROUTi(t), as redefined above, is used to transfer cattle between cohorts using the control variables C2, C3, C4, C5, C6, C9, C10, and C11. These equations have aggregated together many animals of different ages, and hence weights; the weight of the final subpopulation of the calf cohorts must be redefined to reflect this aggregation. Therefore 136 KK7 * jgm SUBPOP7j (t) w7j (t) W7,KK7(t) = (5-5-5) 27 SUBPOP7j(t) j=m 8 w _ _ * E SUBPOP8j(t) [1 TFRACj+1-m,2(t) TFRACj+1vm,3(t)] W8j(t) W8 K (c) = 3"“ 3 8 m8 * _ _ E SUBPOP8j(t) [1 TFRACj+1_m,2(t) TFRACj+1_m,3(t)] (5.5.6) j=m where: wij(t) = average weight of the jth subpopulation of cohort i. SUBPOPi (t) = the number of animals in the jth subpopulation ij of cohort i KKi = the number of subpopulations in the ith cohort TFRAC£ i(t) = fraction of the female calf cohort subpopulation indexed nH£transferred to the ith cohort. Equation 5.5.5 is simply the weighted average of weights of male calves weaned. Equation 5.5.6 is the weighted average of female calves weaned, but which are not selected for transfer to either of the reproductive heifer cohorts. Using the variable TFRAC£1(t), the simulation model user can be selective about what calves are retained for use as replace— ment animals. A common strategy of enterprise operators is to choose the oldest calves for such replacements to maximize the likelihood of successful initial breeding. The final action of these special execution statements at calf weaning is to set all SUBPOP7j(t) and SUBP0P8j(t) values to zero (for j subpopulations older than AGEMIN) to reflect the fact that these calves have been weaned and are gone. 137 Culling is a relatively simple process mechanically, but it has wide ranging implications on future herd reproductive characteristics. CULFRCj(t) is the fraction of the jth subpopulation of the mature cow cohort that is to be culled. This control variable allows very detailed control over culling-—control that can be used to remove cows which are not reproductively useful from the herd. The following equation imple— ments culling for each subpopulation individually, SUBPOP1j(t) = SUBPOPlj(t)*(l-CULFRCj(t)) (5.5.7) where: SUBPOPlj(t) = the number of animals in the jth subpopulation of the mature cow cohort CULFRCj(t) = the fraction of the jth subpopulation culled. In a fashion similar to 5.5.3, the output rate of the mature cow delay is used as a vehicle for funneling the culled cows through the financial mechanism. 1 “I ROUT (t) = ROUT (t) + -—- Z SUBPOP (t)*CULFRC (t) (5.5.8) 1 1 _ lj j DT j—l where: ROUTl(t) = the annual rate of output of animals from the mature cow delay process CULFRCj(t) = fraction of the jth subpopulation culled DT = the time increment of the simulation~—fraction of a year. Equation 5.5.8 differs slightly in form from 5.5.3 and 5.5.4 in that the previous value of ROUT1(t) is retained, with the culled cows being added to the former value. The reason for this difference is that in the case of culling, the cows culled are in addition to those which are leaving the delay process through old age. Weaning is an action 138 which is all inclusive, so that the animals exiting the calf cohort (if any) when weaning takes place are counted in the equations 5.5.3 and 5.5.4, where they would be counted twice if the form of 5.5.8 was followed. Since the mature cow cohort is assumed to have relatively small weight differences among the various subpopulations there is no need to compute a psuedo average weight for the output rate as is done in equations 5.5.5 and 5.5.6. The weight W l,KK1(t) is assumed to be a sufficiently accurate value. Figure 5.21 illustrates an example of this culling process guided by the input of control values by the model user. Solid square points represent the overall calving rate achieved by the various subpopulations of the mature cow cohort during the previous calving season. Circular points represent a particular culling fraction to be applied against that subpopulation. This example culling pattern is guided by the desire to eliminate cows from the herd which have not calved in the previous year. The likelihood of cows failing to calve increases with previous calving failures, so elimination of such animals from the herd should raise the overall herd calving fraction. The reason for desiring a higher calving fraction is the simple goal of reducing feeding costs to achieve the same physical output. Other strategies can, of course, be followed by model users. The culling fraction for the youngest subpopulation is lower than the value indicated by the above strategy to reflect the fact that these animals were not yet mature at their first calving experience. 139 Accumulated Birth Rate or Culling Rate 1.0 - I I I I I 0.8- I ' I 0.6 ' I I 0.4 - . O O 0.2 - 0 o 0 o ‘ o o 0 L n 1 . J 2 4 6 8 10 j index . —— accumulated birth rate for jth subpopulation . -— culling rate for jth subpopulation Figure 5.21 Graph of culling rates for each subpopulation of the mature cow cohort using accumulated birth rates as a guide 140 The manager could also be interested in adding to his herd, as well as culling some or all of it. The control variables ASELLi(t) and C2, C3, C4, C5, C6, C9, C10, and Cll are available to the user at the culling decision point to carry out any herd changes desired, whether positive or negative. Figure 5.3 illustrates how the herd maturation flow variables CZ, ..., C11 affect the herd. ASELLi(t) is the pro— portion of the current herd population of cohort i that is to be sold in this time increment. Then -ASELLi(t)*POPi(t)*(l-DRi(t)*DT) ADDRTi(t) = (5.5.9) DT where: ADDRTi = the annual rate of additions to cohort i P0P1(t) = the current cohort 1 population DRi(t) = the annual cohort 1 death rate DT = the simulation time increment ASELLi(t) = proportion of the population of cohort i that is to be sold in this time increment. "Special decision point" is the term used to designate three events that require the interactive attention of the model user. The first of these is post-harvest feed stock levels imcompatible with the planned feeding rate through the end of the wintering season. The second of these is feed stocks insufficient to meet the planned feeding over a short interval into the future, eg. one to two simulation time periods. Finally, the third decision point testing criteria is that working capital falls below $2000.00. Each of these criteria is checked every time increment of the simulation just as the Current 141 time is checked for equality with one of the time related regular decision points. In the first simulation time increment following the completion of fall harvesting, subroutine INQUIR determines the requirements for each feed stock by using the current cohort populations and the current values of the time based feeding plans. This results in a value of ACTEDln(t)—-the total quantity of feed stock n required by feeding plans from the present through the end of the wintering season. ACTED1n(t) is determined in much the same way as STKFEDn(t) is determined except that ACTEDln(t) is integrated over time in order to obtain the entire wintering requirements. This gives FPLAst—l l 9 ACTED (t) = f —-—-—————-* 2 POP (t )*[CATTIN, _ (t)* In tfall FQUALn(t) i=1 i fall 11% 1 * CNCFRCin(t) + CATTIN12£_l(t) RHGFRCin(t)] dt tspring l 9 +-f -————————- Z POP (t )*[CATTIN (t)* FPLANS£~1 FQUALn(t) i=1 1 fall 112 *RHGFRC t dt CNCFRCin(t) + CATTIN122(t) in( )1 where: (5.5.10) ACTEDl (t) = the season long requirements for feed stock n based n on the present feeding rates and feed distribution planned CATTIN. £(t) = quantity of TDN fed to individual members of cohort 11 1 per day from concentrates under feed plan 2 CATTIN12£(t) = quantity of TDN fed to individual members of cohort i per day from roughages under feed plan 1 CNCFRC. (t) = fraction of the concentrate allocation of TDN to in cohort 1 derived from feed stock n 142 = DISTFDin£(t), if LABEL1?n = 1 0, if LABELFn = 2 RHGFRC, (t) = fraction of the roughage allocation of TDN to 1n cohort 1 derived from feed stock n = DISTEDin2(t), 1f LABELFn = 2 = 0, if LABELFn = 1 FQUALn(t) = fraction of TDN per kg feed for feed stock n POP(t ) = the population of the ith herd cohort at t fall fall FPLANS£ = time when the 1th feed plan ends and the 2+1 plan begins 1, if t < FPLANS 1 2, if FPLANSl : t < FPLANSZ z ' 3, if FPLANSZ : t < FPLANS3 4, if FPLANS3 : t < FPLANS4. When the values of ACTED1n(t) have been computed, the following equation is the criteria used for deciding that a special decision point is needed to allow the user to adjust feed stock levels and planned feeding patterns to be compatible. A special decision point is required if FSTOCK (t ) - ACTED (t ) FTOLl < n fall In fall (5.5.11) ACTEDln(tfall) where: FTOL1 = fraction feed stocks can differ from the computed feed stock requirements implied by the feeding plans FSTOCKn(t) = the current quantity for feed stock n on hand ACTED1n(t) = feed stock n requirements of the present feed plans. The second special decision point criteria is checked each time increment that feeding is taking place. A special decision point is required if 143 FINFLT*ACTED2n(t) > FSTOCKn(t) — STKLOSn(t) (5.5.12) where: ACTED2n(t) = feed stock requirement for feeding cattle during the present time increment for feed stock n FSTOCKn(t) = current quantity of feed stock n on hand STKLOSn(t) = quantity of feed stock n lost thrOugh spoilage in the current increment of time FINFLT = 2.0, factor by which the feed stock requirement of the current time increment is inflated as a safety margin. ACTEDzn(t) is determined in the same manner as is STKFEDn(t) in sub— routine CONSUM (described in section 4). Then, 9 DAYS — _—_ ____.___ * * * * ACTED2n(t) FQU n(t) iglPOPj-(t) LCATTINiu(t) DISTFDin£(t) 1N1 * * + CATTINi 2(t) DISTFDin £(t) 1112] (55.13) where: ACTED2 n(t) = the quantity of feed stock n required by the current feeding plan for this time increment POPi(t) = current population of cohort i FQUALn(t) = fraction of TDN per kg feed for feed stock n DISTFDin(t) = fraction of the feed allocation to cohort i derived from feed stock n CATTINil(t) = quantity of TDN allocated to cohort i from concen— trates per day under feed plan £ CATTIN. (t) = quantity of TDN allocated to cohort i from 12 roughages per day under feed plan 2 INl = indicator of type of feed [I 1, if LABELFn = 1 II = 0, if LABELFn 2 144 IN = an indicator of roughage feed 1, if LABELFn = 2 0, if LABELFn = 1 DAYS = number of days in a simulation time increment. If the level of working capital falls below the arbitrary value of $2000, the model also signals a special decision point. This event is indicative that large cash outflows have drained the enterprise of working capital. A possible solution is short—term borrowing. When any of these events signal a special decision point, the model follows the normal decision point procedure and awaits a restart with control variable values. Since the problem is not fully specified by events leading to a Special decision point, the entire range of con- trol variables is available to the user for inputting new values. Even if relatively little adjustment is made, the user is required to input all control variables; the values are left to user discretion. The User's Guide to the Beef Cattle Enterprise Simulation Model specifies the exact format of the data cards containing control varia— ble values required for each decision point. Figure 5.22 is a typical message printed by the computer at a decision point indicating the control variables which are required to restart the simulation. Endogenous Decision Making The management component model is constructed so that if none of the five decision points is encountered, then logical control shifts to subroutine NORMAL. This subroutine carries out all endogenous deci- Four categories of decisions sion making within the component model. comprise the contents of this subroutine; these are: All.” 145 EXOGENOUS INPUT REQUIREMENTS IN ORDER APFUTR(5,20) BPFUTR(8,10) CPFUTR(NSTOCK,10) PDSTRB(NLANDS,9) XCPROD(NSTOCK) XCQUAL(NSTOCK) CFINAL REMOVL TBRD(1),TBRD(2),TWEAN,TCULL TMHR(5) SEE THE USER GUIDE FOR INPUT FORMATS Figure 5.22 Printed message at the spring growth decision point requesting values for specified control variables 146 l V timing of slaughter cohort cattle sales, 2 V fall crop sales and purchases to obtain wintering feeding requirements, 3) mechanical harvesting of forage growth, and 4 v concentrate and roughage cohort allocations. These decisions are obviously more routine than those which must be made at each of the decision points. The criteria upon which some of these areas of endogenous decision making are based are standard economic ones; e.g., selling if the marginal cost of holding exceeds the marginal revenue. Some of the parameters used here are supplied at decision points for use over an extended time period. Mechanical harvest of forage growth is an example of using exogenously supplied parameters over an extended period of simulated time. Three key variables are inputted into the model at the spring growth onset decision point; these are TMHRk, REMOVL, and CFINAL. TMHRk specifies certain simulated times at which the forage growth then existing in various land parcels should be harvested. REMOVL specifies a fraction of the "excess" over and above current herd needs that is to be harvested at each of these harvest times. CFINAL specifies the fraction of the standing quantity of forage that is to be harvested at TFALL-—the end of the growth season. MHRn(t) is the daily rate of forage removal from land parcel n. It is determined by: 0, if t {[TMHRk, k=l, ..., 5] MHRn(t) = 0, if GRNnt) < FLEVELn(t) (5.5.14) REMOVL __._.* .— DEYS [GRNn(t) FLEVELn(t)] 147 where: GRNn(t) = the quantity of forage material in land parcel n DAYS = the number of days in the time increment UT 9 FLEVELn(t) = LOW*DAYS* X PDSTRBni(t)*POPi(t) (5.5.15) i=1 = a minimum quantity of forage needed to supply the animals grazing in land parcel n for one time increment PDSTRBni(t) = fraction of the population of cohort i grazing in land parcel n LOW = 5.0, a minimum daily quantity of forage intake POPi(t) = current population of cohort 1. When the end of the growth season arrives (t = TFALL), then the manager controls the final forage harvest through CFINAL. GRNn(t)*CFINAL DAYS (5.5.16) MHR (t) = daily rate of mechanical harvest of forage from land n parcel n GRNn(t) = quantity of forage available in land parcel n CFINAL = fraction of the forage in land parcel n to be harvested at the end of the growth season DAYS = number of days in a simulation time increment. Another end—of—growth-season activity is sales of the crop production that is produced by the enterprise. The predicted sales amounts have been determined at a breeding decision point; and if no extraordinary crop price changes have occurred that were not foreseen, then the predicted quantities are either sold or purchased. The resulting feed stock levels should be adequate to meet the desired 148 feeding rates that the manager has also specified at a breeding decision point. CSALESn(t) = - REACTDn, if REACTDn < 0 (5.5.17) STKPURn(t) = REACTDn, if REACTDn > 0 (5.5.18) where: CSALESn(t) = quantity of feed stock n sold this period STKPURn(t) = quantity of feed stock n purchased this period REACTDn = the predicted net purchases of feed stock n. REACTDn is controlled by the user at the various decision points which are appropriate for such feed stock predictions. An extremely important decision in a cattle enterprise is timing of animal sales. This decision is complicated by the reproductive inr plications of most sales; the uncertainties of the future market make decisions in this area quite complex. Two herd cohorts--steers and slaughter heifers--do not share this general difficulty. These cohorts have already been designated as sales animals only; the near term market condition and the weight gaining characteristics of these animals are the major considerations of the decision maker. The approach taken in this model is to determine whether or not the expected marginal revenue of retaining these cattle for one time increment exceeds the expected marginal costs. When net marginal revenue is positive no sales are made; the entire cohort population is sold when net marginal revenue is negative. A further check over a future interval of cattle prices is made before a decision to sell on the first griterion 15 actually carried out. A key element in determination of expected marginal revenue is the weight gain expected from the input of feed to the animal. The 149 method used here parallels the development of Schuette[48] in the nutrient impact subroutine NUTRN. The quantity and quality of feed fed to each slaughter cohort animal is the primary determinant of the weight gain of that animal. A constraint on cattle intake of feeds allocated is the capacity of the animal to consume the feed; this is modeled as a maximum fraction of the animal's body weight. Then _ 0.75 DMIij(t) — min[ a*wij(t) . ALLOCi(t) 1 (5.5.19) where: DMIi.(t) = the daily consumption of feed of members of the J jth subpopulation of cohort 1 (dry matter basis) Wij(t) = weight of members of the jth subpopulation of cohort i ALLOCi(t) = CNCALi(t) + RHGALi(t) (5.5.193) POPi(t)*DAYS = the total allocation of feeds to members of herd cohort i a = parameter based on the body weight of the cattle CNCALi(t) = concentrate allocation to cohort i —— kg/DT RHGALi(t) = roughage allocation to cohort i —- kg/DT POPi(t) = population of cohort 1 DAYS = number of days in a DT time increment. Equation 5.5.19 has fixed the quantity of feed that is available for digestion by a member of a particular weight animal, while equation 5.5.20 determines the average TDN value of this quantity. DMITDNij(t) = CNCALi(t)*TDNCi(t) + [DMIij(t) - CNCALi(t)]*TDNRi(t) DMIij(t) (5.5.20) where: DMITDN (t) = the average TDN value of the feed consumed by a ij member of the jth subpopulation of cohort i 150 TDNCi(t) = average TDN value of the concentrate allocation to cohort i TDNRi(t) = average TDN value of the roughage allocation to cohort i. This equation assumes that the concentrates are consumed in their entirety before any roughages are consumed. Once DMITDNij(t) is determined it may be used as the basis for computing the energy values of the DMIij(t) feed consumption for two purposes, i.e., energy used for maintenance of body processes and energy used for weight gains. EGAIN represents the energy value of the feed consumed in terms of mcal energy for gain/kg feed, while EMAIN represents the energy value of the feed consumed in terms of meal energy for maintenance/kg feed. These values are determined from DMITDNi (t) using Subroutine CNVRT, which uses a method of conversion 5! between these different systems of units authorized by the National Research Council. The amount of energy used for weight gains is that which remains after the animal's maintenance requirements have been subtracted from its energy intake. Since there are two energy values per feed intake (one for body maintenance and one for weight gains) the determination of the energy available for weight gains must be determined from the quantity of feed used for maintenance purposes. This results in EFORGij(t) = EGAIN*[ DMI where: 0.75 ij(t) — 0‘077*wij(t) /EMAIN ] (5.5.21) EFORG. (t) = energy available for weight gains for an animal lj in the jth subpopulation of cohort i —— mcal EGAIN = gain energy value of the feed intake -- meal/kg EMAIN = maintenance energy value of the feed intake -— mcal/kg. 151 Having the value for EFORGij(t) is the next to the last step for determining the actual rate of weight gain for animals in the jth subpopulation of cohort i. DGAINij(t) is computed by the following equation a. : b.*EFORG..(t) DGAIN1'(t) = i _l_ 1 _1___0_1_;5_ - ci (5.5.22) 3 d. w..(:) ‘ 1 1:] where: DGAINi.(t) = the rate of weight gain for members of the jth J subpopulation of cohort i —— kg/day ai, bi’ Ci’ di = constants for the ith herd cohort. The upper signs are used when EFORGij(t) is positive, and the lower Signs are used when it is negative. This equation, then, gives posi— tive weight gains for positive energy for gain values, and negative weight gains for negative energy for gain values. Determination of the rate of weight gain that will result from the planned quantity and quality of feed allocation allows the margin- al revenue of that weight gain to be calculated. This value is an expected revenue gained from feeding the animal to obtain the weight gain computed by equation 5.5.22. KK FREVi= Xi DGAINij(t)*DAYS*SUBPOPij(t)*XPECTAk(t+dt) (5.5.23) i=1 where: FREVi = expected marginal revenue from the ith cohort —— $ DGAIN..(t) = the rate of weight gain for the animals in the 13 jth subpopulation of cohort 1 ~- kg/day SUBPOPi (t) = population of the jth subpopulalation of cohort i j XPECTAk(t+dt) = expected price of cattle of grade k at time t+dt -- $/kg k = GRADEij(t) = price grade of animals within SUBPOPij(t) 1. .1. 1.11.7”. EJAfiJY‘WWE .4‘... ‘1. 152 KKi = the number of subpopulation within cohort i DAYS = the number of days in a DT time increment. Once the expected marginal revenue of the weight gain has been determined the remaining factor is the marginal cost of the weight gain. Here, however, the marginal cost of maintaining the animal another DT time increment is more than just the cost of the feed used for weight gain. To be precise, the marginal cost is the cost of all of the feed fed to the animal, the costs of delivery of that feed, and any other costs associated with retaining the animal. The model for marginal costs used here breaks costs down into two factors, cost of the feed allocated, and cost of delivery of that feed. Other factors are assumed to be negligible in comparison to these two factors. NSTOCK CPRICEn(t) FCOST. = POP (t)*DAYS* Z -—————*[ CATTIN. (t)* 1,feed 1 n=1 FQUALn(t) ill * CNCFRCin(t) + CATTIN12£(t) RHGFRCin(t)] (5.5.24a) where: FCOST. f d = the cost of the feed stocks fed to members of 1’ ee cohort i in the current time increment —- $ CPRICEn(t) = current price of feed stock n —- $/kg CATTINil£(t) = allocation of TDN from concentrates per animal per day to members of cohort 1 under feeding plan 2 -— kg TDN/day CATTIN12£(t) = allocation of TDN from roughages per animal per day to members of cohort 1 under feeding plan 1 —— kg TDN/day CNCFRC1 (t) = DISTFD.n£(t), proportion of the feed from con— n centrafes delivered to cohort 1 derived from feed stock n RHGFRCin(t) = DISTFDi (t), proportion of the feed from rough— ages deTivered to cohort 1 derived from feed stock n IL 153 FQUALn(t) = the average TDN value of the nth feed stock. The cost of feeding these quantities of feeds is given by FCOSTi,feeding = [ CNCALi(t) + FEEDALi(t) ]*HRFED1*BPRICEl(t) (5.5.24b) where: = cost of feeding cohort i for one DT time increment from labor use -- $ FCOSTi,feeding CNCALi(t) = quantity of concentrates fed to cohort i —— kg/DT RHGALi(t) = quantity of roughages fed to cohort i -— kg/DT HRFEDl = hours of labor for feeding cattle -— hours/kg BPRICEl(t) = cost of labor -- $/hour. The total marginal cost of retaining the ith cohort of the herd for another DT time increment is then the summation of the two factors explained above——the feed itself and the labor to feed it. Then FCOSTi = FCOSTi,feed + FCOSTi,feeding (5.5.24c) where: FCOSTi = the total marginal costs of retaining cohort i for another DT time increment -— $. When the expected marginal revenue is discounted apprOpriately, and then still exceeds the expected marginal costs, then no members of the cohort will be sold in the current time increment. Discounting the expected marginal revenue reflects the uncertainty that exists in the future cattle price values. If marginal revenue happens to be smaller than marginal costs, then expected cattle prices are searched over some interval into the future in an attempt to discover the highest price within this time interval. The period of time into the future searched by subroutine NORMAL is a control variable set by the model user. If the highest value of cattle prices within the interval 154 [t, t+RANGE) falls within the interval [t, t+dt) then the entire cohort population should be sold. The revenue gained from sales will be highest if the animals are sold immediately, rather than retaining them for the future. The control variable ADDRTi(t) is used to carry out this sales action; therefore _ 7‘: _ * POPi(t) (1 DRi DT) ADDRTi(t) = -—-——-—Efir———-——-- (5.5.25) where: POPi(t) = population of cohort 1 DR = death rate of cohort i in fraction per year i ADDRTi(t) = annual rate of additions of animals to cohort i f -— animals/year DT = the basic simulation time increment -— years. If the highest price occurs at a time exceeding t + dt then no sales are made in this period, because more revenue can be gained by waiting for this higher price before selling. The endogenous decision mechaniSm for selling the slaughter cohorts has utilized the idea of marginal net revenue to guide the decision to sell or to retain animals. When marginal revenue (net) is positive then the proper decision should be retention of the cattle since the costs of holding are exceeded by the anticipated revenue of holding. Alternatively, when marginal net revenue is negative, the proper deci— sion is to sell immediately. This mechanism is a straightforward application of standard economic theory. The last area that is included in this model of endogenous deci- sion making is the allocation of feed stocks to the herd cohorts. This also includes the distribution of the allocation among specific feed 155 stocks using the control variables CNCFRCin(t) and RHGFRCin(t). Feed stocks are allocated to the herd under two separate circum- stances: during the wintering period and during the growing season if the quantity or quality of forage falls too low. The alternative feeding plans used by the manager of this enterprise are modeled using the variables FPLANS, CATTINik(t), and DISTFDin(t); these are the times when feeding plans change, the quantity of TDN allocated to members of herd cohorts per day, and the distribution of the TDN allocation among feed stocks, respectively. The allocations of con- centrate and roughage to each cohort is determined by the allocation to each animal, the distribution of the allocation among the feed stocks, and the TDN value of each feed stock. This results in NSTOCK = * it CNCALi(t) “Z CNCFRCin(t)/FQUALn(t) CATTINill(t) POPi(t)*DAYS (5.5.26) [NSTOCK = * * * FEEDALi(t) 1 E1 RHGFRCin(t)/FQUALn(t) CATTINi2£(t) POPi(t) DAYS (5.5.27) where: CNCALi(t) = the quantity of concentrates allocated to cohort i for one DT time increment——kg conc./DT FEEDALi(t) = the quantity of roughages allocated to cohort i for one DT time increment—~kg/DT = allocation of TDN from concentrates per day to CATTIN11£(t) members of cohort i under feed plan 1 CATTIN.2£(t) = allocation of TDN from roughages per day to 1 members of cohort i under feed plan 2 POPi(t) = current population of cohort i DAYS = the number of days in a DT time increment 156 FQUALn(t) = the current average TDN value of feed stock n DISTFDinl(t)’ if LABELFn = l CNCFRCin(t) = 0, if LABELFn = 2 (DISTFD. (t), if LABELF = 2 1n£ n RHGFRCin(t) = 0, if LABELFn = 1 (1, if 0 j_t < FPLANSl(t) 2, if FPLANSl(t) j_t < FPLANSZ(t) 3, if FPLANSz(t) §_t < FPLANS3(t) 4, if FPLANS3(t) j,t < FPLANSA(t) The nutrient impact component subroutine CONSUM uses the variables CNCALi(t), CATTINil£(t), FEEDALi(t), CATTIN12£(t), and RHGFRCin(t) to determine the average TDN value of concentrates, TDNC1(t), and of roughages, TDNRi(t). This is more fully explained in section V.4. During the grazing season there may be instances where supplementary feeding is necessary to provide nutrients that grazing alone cannot deliver. Drought, with extremely low growth rates or none at all, is an example. Subroutine NORMAL checks for such don- ditions by determining the quantity and quality of forage available. Normally the feed allocations to all cohorts are zero; i.e., CNCALi(t) = FEEDALi(t) = 0.0, i=1, ..., 9. A minimum herd feeding rate, RGRAZE(t), is computed to compare against forage stocks, with NLANDS RGRAZE(t) = MIN*DAYS* Z STOCKLn(t)*LANDn (5.5.28) n=l where: RGRAZE(t) = a minimum quantity of forage necessary to Sustain the grazing herd over one DT time increment—-kg/DT 157 STOCKLn(t) = the stocking rate in land parcel n—-#/hec LANDn = area of land parcel n--hectares MIN = a minimum daily ration of forage per animal--kg/day DAYS = the number of days in a DT time increment. The quantity of forage available, TGREEN(t), and the associated average TDN value of that forage, VGREEN(t), are given by NLANDS TGREEN(t)= X sauna) (5.5.29) n=1 1 NLANDS VGREEN(t) = m * Z GRNn(t)*DIGESTn(t) (5.5.30) n=l where: GRNn(t) = quantity of forage in land parcel n DIGESTn(t) = quality of forage in land parcel n. If TGREEN : RGRAZE(t) and VGREEN(t) 3 0.40 then no supplemental allocation is made; otherwise, the following roughage allocation is made FEEDALi(t) = RAL*DAYS*POPi(t), i=1, ..., 9 (5.5.31) where: FEEDALi(t) = quantity of roughage allocated to herd cohort i—— kg/DT POPi(t) = current population of cohort i RAL = 5.0, a minimal supplementary feeding level to assist an individual animal--kg/day. The entire allocation of supplement must come from stored forage retained from the previous wintering season, or perhaps from forage already harvested and stored this growth season. RHGFRCil(t) = 1.0 for all herd cohorts. 158 Summary The management decision making component model has been constructed to meet the requirements specified in the problem statement (Chapter 111). Because the complexity of many decisions makes them extremely difficult to reduce to computationally solvable forms, the model has been developed to seek some control values exogenously. Such deci- sions have been termed decision points and usually are associated with specific herd actions; e.g., breeding, weaning, culling, and starting grazing. Some less complex decisions can be determined endogenously within the model using standard economic criteria. Examples are the decision to sell slaughter cohort animals, forage harvesting, and feed allocations to the herd. The mechanical process of seeking exogenous values for some control variables involves using the simulation model interactively in a batch mode. The user studies the simulation rec- ord over time and the specific information printed at each decision point to provide a basis for action. Evaluation of multiple decision strategies at a particular decision point is possible through the use of permanent files which "freeze" the simulation as it is at a decision point. The user can review the results at his leisure and then con- tinue the simulation by simply attaching the appropriate permanent file once he has made his selection. This feature makes exploration of intuitive feelings and alternative exogenous variable assumptions quite feasible. The decision making component has been developed to be realistic in its treatment of actions taken by operating managers of ranchs. The fine level of control variables makes many different types of strategy evaluation possible. Annual decision patterns, such as culling and weaning, can be investigated, as can investment planning projects. In short, the decision making component has been con- structed to be as flexible as possible in meeting the potential de— sires of managers interested in evaluating the consequences of alternative decisions. 160 v.6 Secondary Component Models Chapter IV, section 2, provided a brief introduction to the three secondary components of this system; these are exogenous variable determination, management of expected prices, and enterprise financial accounting. This section will review the purpose and detailed con- struction of each of these components' model. Exogenous Variable Component The simulation model has three types of exogenous variables which influence the physical variables and management decision making. These are crop and animal prices, climatic factors-~temperature, solar radia- tion, and rainfall, and feed crop production. Although a totally comprehensive model of a beef cattle enterprise would endogenously determine the crops to be planted as a function of prices and available resources, this modeling effort is restricted to exogenous specification of feed crop production. A year long stream of crop production is specified with uniform spacing of values. This process is illustrated graphically in Figure 5.23. The subroutine EXOG determines crop product— ion for each possible feed crop in the interval ( t — dt/2, t + dt/2 ) by comparing the current simulated time T to the specified harvest time of each crop. The effect of this structure is an impulse whose height is the amount harvested in the interval. CROPn1(t) is the variable name used for this quantity harvested in the time interval. CROPQLn1(t) gives the TDN value of this quantity, CR0Pn2(t) specifies the quantity of crop n residue that is harvested, while CROPQLn2(t) gives the TDN value of that residue. 161 CROPnl(t) 40,000 35,000 30,000 25,000 20,000 15,000 5,000 0 ‘ 10,000 ‘ I .] llTll 0.0 .65 .7 .75 1.0 time Figure 5.23 Yearly pattern of crop harvest as exogenously specified Current prices and the current weather variables are also determined by subroutine EXOG. These are treated somewhat differently than crop production because both prices and weather are continuous variables, while crop production is not. Instead of searching the input values for a match between current time and some specified time point to get an associated data value, prices and the weather variables are determined through linear interpolation to compute the proper value. function TABLIEl is used to accomplish this interpolation. The utility Figure 5.24 shows how this procedure works for the weather variable RAIN(t). The following exogenous variables are determined in Similarily: SOLAR(t), AVGTMP(t), APRICEk(t), k=1,...,5 and CPRICEn(t), n=l,...,NSTOCK. l. Llewellyn, R.,FORDYN: A Industrial Dynamics Simulator, privately published, Raleigh, N. C., 1965, p. 4—20. 162 Rainfall (cm/day) . O O .\ RAIN(t) ‘3 \. I . O O O p C O o o t Time Figure 5.24 Linear interpolation between exogenously specified annual rainfall data points to obtain the current rainfall value Currently the model assumes that the weather and crop variables are spaced uniformly with an interval of 0.05 years between values. Further, the model assumes that the weather variables are repeated year after year with no change. This assumption requires that one view the weather variables as long run averages rather than specific values from a stochastic process. A change in the component model to actually use stochastic processes to represent the three weather variables would be quite simple, but since at this time the entire model uses deterministic variables rather than stochastic ones, this change will not be made. Other spacings between the Specified annual pattern would also be quite easy to implement, even non—uniform ones. 163 Expggted Prices Component Expected prices are an important base for the decision making of the enterprise manager. These prices are also characterized by their volatility as the agricultural sector of the economy; and the beef cattle industry, in particular, reacts to economic and political events throughout the world. The subroutine RESPNS has provisions for reading in up to two years' expected prices for cattle and up to one year's expected prices for crops at each decision point. During routine operation of the overall model as it simulates the passage of time, subroutine GENERT acts to make available the proper stream of future expected prices from the current point in time. Re— call that at each decision point the model user is asked to respecify the stream of prices for the feed stocks and for all five cattle slaughter grades through the future. The array CPFUTRn(t) holds the stream of future prices for feed stock n spaced in intervals of 2DT's from the present, to, through simulated time t + 20DT's. O APFUTRk(t) holds the stream of future prices for cattle grade k spaced The array in intervals of 2DT's from the present,to, through simulated time t + 0 In an effort to make this specified set of prices available to ODT's. the model as simulated time progresses forward from the time at which the values were specified, GENERT advances the entries in the arrays CPFUTRn(t) and APFUTRk(t) on alternate DT increments of simulated time. Figure 5.25 provides an example of this procedure using the variable APFUTR3(t). In Figure 5.25a the points specify the expecta— tions of prices for cattle of grade 3 for a time span of 24DTs. The labelled points p , p , p specify grade 3 cattle prices at times 1 2 3 164 APFUTR3(t) 0.80 0.70 . . o o 0601; . .pl .pz .133 . o I . . g - t1 1 94k . A n . . . L to t0+2DT' to+6DT t0+20DT TIME Figure 5.25a Expected cattle prices over time starting from t0 APFUTR3(t) 0.80 , 0.70 _ . o o 0 o 0.60 1:5 p O c 1 2 p3 . . O I / t1 tl+2DT tl+6DT tl+20DT TIME Figure 5.25b Expected cattle prices over time starting from t1 165 t + 4DT, to + 6DT, and t0 + 8DT, respectively. The point to 0 represents the current value of simulated time while pl, p2, and p3 are specified distances of time into the future (from the point of view of simulated time). When simulated time advances as the simu— lation proceeds, the stream of points needs to be shifted to the left to bring the proper value of expected price into correlation with the present value of simulated time. Lets suppose that simulated time has advanced 4DTs from to; this is the value t1. Figure 5.25b de— picts the proper positioning of value of APFUTR3(t) with tl being the present simulated time. The curve given by data points in Figure 5.24s has been shifted 4DTs to the left as it should to preserve the relationship between pl, p2, and p3. Notice that the price values p2 and p3 are still the proper 2DTs and 4DTs into the future of simulated time from tl in Figure 5.24s as they were in Figure 5.24a. Subroutine GENERT maintains this price positioning for all five cattle price grades, and for all NSTOCK feed stock prices. Subroutine CBOX[23] is used in subroutine GENERT to perform this shifting of price array entries 0n alternate DT time increments. CBOX also cycles the price value being removed to the furthest most array location meaning that a price cycle of duration 4ODTs is assumed for cattle prices, and a price cycle of 20DTs is assumed for feed stock prices. Of course the model user has the opportunity to respecify the entire stream of prices at each decision point, so the price cycle assumption implied by using CBOX is quite weak. The major reason for this complex movement of price values as simulated time advances is the ease with which current prices and future expected prices from the current time may be determined. The three 166 functions XPECTA(k,T), XPECTB(1,T), and XPECTC(n,T) are used to determine the expected price at time T for grade k cattle, resource 1, and feed stock n, respectively. Each of these function subprograms are designed to linearly interpolate between array entries to deter- mine the correct price at any desired time point within the range of times for which price expectations are valid. Current prices are determined from these function subprograms by using the current time as the argument of the function; this gives APRICEk(t) = XPECTA(k,t), BPRICEl(t) = XPECTB(1,t), and CPRICEn(t) = XPECTC(n,t). The function subprogram TABLIE(from Llewelyn) is used within each of the expectation functions to perform the linear interpolation required. Financial Accounting Component This component performs two tasks: first, accounting for all purchases, sales, repayments, and taxes which occur during simulated time; and second, computing present values of certain variables at the end of a multiyear simulation run. Subroutine FINANC is the main element of this component, others are subroutines REVENU, PRCOST, CAPTAL and TAXSUB. The name of each of these subroutines is il- lustrative of the function it performs. This component will be explained in terms of four types of variables: revenues earned, costs incurred, interest on debt and debt repayment, and taxes. Subroutine REVENU determines the revenue earned from sales of cattle and crops. Physical amounts sold times the market price equals the revenue incoming to the enterprise. Because crops are assumed to be homogeneous, crop revenues are much easier to determine than cattle revenue . 167 Then CREVn(t) = CPRICEn(t)*CSALESn(t) (5.6.1) where: CREVn(t) = revenue from sales of crop n —- $/DT CPRICEn(t) = market price of crop n —— S/kg CSALESn(t) = quantity of crop n sold —- kg/DT. Total crop revenue i8 the summation of each crop's revenue giving V (t) = NCROPS NCROPS + 1 Z CREVn(t) (5.6.2) n=1 CRE where: CREVNCROPS + 1(t) = total crop revenue earned in this time increment--$/DT CREVn(t) - crop revenue earned from sales of crop n in this time increment—~S/DT NCROPS = the total number of possible crops. Determination of cattle revenue is complicated by the fact that not all sales animals have the same weight or the same price grade. Quantities of cattle sold from each of the nine herd cohorts are given by two variables, ROUTi(t) and ADDRTi(t) represents the annual output rate of cohort i, while ADDRTi(t) represents the annual rate of additions to the herd (which can be either positive or negative). When ADDRTi(t) < 0, then there is a negative addition to the cohort; i.e., sales. These additions are assumed to affect each cohort subpopulation as a uniform percentage of its current subpopulation. Section 1 explains this process more completely. Certainly not all of the animals leaving cohort i are sold-—some are only being trans— ferred from one cohort to another. An example is the output from the replacement heifer cohort; here virtually all of the output is being transferred to the mature cow cohort and not being sold. 168 Figure 5.3 illustrated the structure of the herd demographic model and the use of the control variables CZ, C3, C4, C5, C6, C9, 010, and C to direct the flows of cattle leaving each cohort. The variable CONi(t) represents the proportion of the output flow rate, ROUTi(t), that is sold. Study of Figure 5.3 will show that C0N1(t) = 1.0 (5.6.3) CON2(t) = 1.0 P C10 CON3(t) = C11 CON4(t) = 1.0 CON5(t) = C6 CON6(t) = 1.0 CON7(t) = C4 CON8(t) = C9 CON9(t) = 1.0 Combining the sales resulting from negative "cohort additions," and exits from the herd due to satisfaction of individual cohort delay times, gives the total animal sales in the current time increment as ASALESi(t) = DT*[ROUTi(t)*CONi(t) + Ui(t)] (5.6.4) where: ASALESi(t) = number of animals sold from cohort i in this time period~—#/DT DT = length of the simulation time increment in years ROUTi(t) = cohort i delay output rate——#/year CONi(t) = proportion of ith cohort output rate sold Ui(t) = 0, if ADDRTi(t) 3 0 = —ADDRTi(t), if ADDRTi(t) < 0 169 ADDRTi(t) = annual rate of additions to cohort i-—#/year. Since cattle prices are in terms of $/kg, the above expression is not yet sufficient for determination of revenue. The weight of each of the cohort sale quantities must yet be determined. Additionally, there are five cattle price grades; this requires that the average grade for each cohort's sale animals also be determined. The variables AVGWi (t)and AVGGRDi(t) specify the average cohort sales weight and average price grade, respectively. DT*ROUTi(t)*CONi(t)*wi,KKi(t) AVGW, = + 1“) ASALESi(t) [ASALESi(t) — DT*ROUTi(t)*CONi(t)]*ALWTi(t) (5.6.5) ASALESi(t) where: .th . Wi.(t) = the average weight of cattle in the j subpopulation of J cohort i ALWTi(t) = the average weight of the ith herd cohort —- kg/animal KK. * £1 Wij(t) SUBPOPij(t) (5.6.5a) j=1 POPi(t) where: KK. = the number of stages in the delay model of cohort i .th . SUBPOP..(t) = number of animals in the j subpopulation of 13 cohort i POPi(t) = total population of cohort i. The average grade of cattle sold from cohort i is * DT*ROUTi(t)*CONi(t) GRADE1,KK,(t) (5 6 6) AVGGRD.(t) = 1 + . . 1 ASALESi(t) [ASALESi(t) — DT*ROUTi(t)*CONi(t)]*ALGDi(t) ASALESi(t) if j 170 where: GRADEi.(t) = the price grade of animals in the jth subpopulation J of cohort i ALCDi(t) = the average grade of animals in cohort i KKi GRADE..(t)*SUBPOP..(t) 13 11 z (5.6.6a) J=l POPij(t) ROUTi(t) = the output rate of the ith cohort delay model——#/year ASALESi(t) = the number of cattle from cohort i sold in this time period——#/DT CONi(t) = the fraction of the output rate of cohort i sold SUBPOPij(t) = the population of the jth subpopulation of cohort i POPi(t) = population of cohort i DT = the time increment of the simulation—~fraction of a year. AVGGRDi(t) is further constrained to be an integer, since there are specific price grades, not a continuum of prices based on slight physical differences. The revenue earned from cattle sales, AREVi(t), is then AREVi(t) = ASALEsi(t)*AVGwi(t)*APRICEAVGGRDi(t)(t) (5.6.7) where: AREV.(t) = revenue earned from sales of cattle from cohort i in 1 the current time increment——$/DT ASALESi(t) = the number of animals sold from cohort i in the current time period——#/DT AVGWi(t) = the average weight of cattle sold from cohort i——kg APRICEk(t) = the current price for cattle of grade k——$/kg AVGGRD.(t) = the average price grade of cattle sold from cohort 1 i in this time increment. The total revenue earned from cattle sales in the current time period : 171 is the summation of the revenue earned from sales in each cohort, giving 9 AREVlO(t) = 121 AREVi(t) (5.6.8) where: AREV (t) = total revenue earned from sales of cattle in this 10 time increment -— $/DT AREVi(t) = revenue earned from sales of cattle from cohort i in this time increment -- $/DT. Subroutine PRCOST determines the costs of production for cattle and for crops. Cattle production costs are determined by valuation of the quantities of eight production resources used for various activi— ties associated with the cattle herd. Examples of such activities are breeding, weaning, culling, feeding the herd, etc. Nine categories of production cost are used to allow realism in this model. A list of the nine categories of cattle production cost along with the variable name and the units used is given in Table 5.1. Crop production costs are determined in a much less rigorous manner since crop production is not modeled as such in this simulation model. Crop production costs are basically determined by assuming that the cost of production to the enterprise is a fixed proportion of the market price at harvest time. This costing mechanism is recognized as being unrealistic, but it is tolerable here because of the place crop production assumes in the overall priority of this project. ACOSTk(t) and CCOSTk(t) are the variables which are used to represent specific categories of cost for cattle and crops, respectively. Table 5.1 defines these variables. 172 Table 5.1 Definition of Crop and Cattle Expense Variables Variable Definition Units ACOSTl(t) Labor expenses $/DT ACOST2(t) Repair expenses $/DT ACOST3(t) Utility expenses $/DT ACOST4(t) Veterinary and breeding expenses $/DT ACOST5(t) Fertilizer and seed expenses $/DT ACOST6(t) Leased land expenses $/DT ACOST7(t) Animal feeding expenses $/DT ACOST8(t) Cattle purchase costs $/DT ACOST9(t) Miscellaneous, overhead, etc. $/DT .CCOST;(t)I—Feed.Ergp—pfoduction.expenses ———————— $7DT-_'—-_ CCOST2(t) Mechanical harvesting of forage costs $/DT Feed stock purchase expenses $/DT CCOST3(t) 173 Crop costs are determined directly from crop Production and feed stock purchases. Since all crops except forage are exogenously speci— fied by the user, determination of costs is quite difficult. The method used here is to simply deflate the current market price to get a hypothetical production cost. Then NSTOCK CCOST (t) = 2 CROP (t)*CPRICE (t)*DEFLAT (5.6.9) 1 “:2 nl n NLANDS ] CCOST2(t) = DAYS* X MHRn(t) *[HRHARV*BPRICE1(t) n=1 J + SPHARV*BPRICE3 (t) ] (5.6.10) where: CCOST1(t) = production of the current feed crop costs CCOST2(t) = current costs of harvesting forage CROPnl(t) = current production of crop n -— kg/DT BPRICEl(t) = current price per unit of production resource 1 HRHARV = hours of labor per kg forage harvested SPHARV = units of utilities required to harvest forage -— units/kg MHRn(t) = daily rate of forage harvest from land parcel n DAYS = number of days in a DT time increment CPRICEn(t) = current market price of feed stock n —— $/kg DEFLAT = scaling factor to determine crop production costs in terms of current price. Costs of crop purchases are simply the summation of the quantities purchased times their current price, giving NSTOCK CCOST3(t)= Z STKPURn(t)*CPRICEn(t) (5.6.11) n=1 174 where: STKPURn(t) = quantity of crop n purchased this time period CCOST3(t) = cost of feed stock purchases in the current period CPRICEn(t) = current price of crop n-—$/kg. Total crop costs in this period (feed stock costs) is the summation of each of these individual cost factors; CCOST2(t) is excluded from this summation because the forage production costs will be added into the animal production costing. This is done simply to keep all animal—related factors together. CCOST4(t) = CCOSTl(t) + CCOST3(t) (5.6.11a) where: CCOST4(t) = the total crop-related production costs in this time period. Table 5.2 Definition of Resource Variables VARIABLE I DEFINITION RESORCl(t) Hours of labor RESORC2(t) I Units of veterinary supplies RESORC3(t) Units of utilities RESORC4(t) Kgs of fertilizer RESORC5(t) Kgs of seed RESORC6(t) Hectares of land leased RESORC7(t) Units of breeding supplies RESORC8(t) Units of repair materials The quantities of each resource variable, RESORCl(t), used in each time increment of the simulation are a function of the differ- ent events and activities which have taken place within that DT time 175 period. To write conventional equations listing the total amount of each resource used would require as many equation sets as there are combinations of events and activities. To avoid this difficult task, the quantity of resources required for each event or activity will be defined separately for each event or activity which exists. The total quantity of each of the eight resources used in any par- ticular DT time increment is determined by the particular set of events which has occurred within that time increment. This means that the sense of each of the equations used here is RESORCl(t) = RESORC1(t) + a specific event—related amount. In the simulation model this organization is quite natural because of the ease with which IF statements can be used to shunt the logi- cal path of flow through series of equations which apply only for specific events or specific times of the year. When an equation of form 5.6.12 through 5.6.25 is encountered, the quantity calcu- lated is simply added to whatever sum already existed for that variable. The only requirement needed to perform this addition is initialization of resource quantity used to zero at the beginning of each DT increment of the simulation. Feeding is a major activity requiring the use of resources, primarily labor. The amount of labor used for feeding is NSTOCK 1215501201“) =HRFEDl* Z STKFEDn(t) (5.6.12) n=1 where: RESORCl(t) = hours of labor required to feed cattle this time increment——hr/DT STKFEDn(t) = quantity of feed stock n fed to cattle—~kg/DT HRFEDI = hours of labor per kg of feed stocks fed. 176 Another major source of production costs is breeding, especially if artificial insemination is used. This activity requires labor as well as breeding supplied, Furthermore, breeding is not a single event but an activity covering an extended period of time; i.e., DURBi(t) for the ith cohort. The number of hours of labor required for breeding is modelled as a function of the number of servicings experienced by each herd cohort subpopulation. The variables DT and DURBi(t) act to spread the quantity of labor hours required uniformly over the interval DURBi(t), i=1, ..., 3. They act simi— larly in spreading breeding supply resources used uniformly as well. KK 3 1 INBij(t)*SUBPOPi,(t)*DT RESORC1(t) = HRBRED* Z DURB (t) J (5.6.l3a) i=1 j=l 1 KK. . 3 1 INBij(t)*SUBPOPi.(t)*DT RESORC7(t) = SPBRED* Z Z J (5.6.13b) i=1 j=l DURBi(t) where: RESORCl(t) = hours of labor used for breeding in this time increment——hours/DT RESORC7(t) = units of breeding supplies used for breeding in this time increment—-units/DT INBi.(t) = number of breedings for the jth subpopulation of J cohort i DURBi(t) = duration of cohort i breeding HRBRED = hours of labor required per servicing SPBRED = number of units of breeding supplied per servicing SUBPOPij(t) = number of cattle in the jth subpopulation of cohort i. Calving costs are modeled as using labor exclusively; equation 5.6.14 determines the number of hours of labor used in a DT time 177 increment by summing the number of newly born male and female calves and multiplying by.the labor resource parameter HRCALV. Then RESORCl(t) = HRCALV*[SUBPOP7l(t) + SUBPOP81(t)] (5.6.14) where: RESORCl(t) = hours of labor required for calving assistance in the current time period SUBPOP7l(t) = number of newly born male calves SUBPOP81(t) = number of newly born female calves HRCALV = hours of labor required per calf born. When calves reach a specified age, all calves are vaccinated and most male calves are castrated. These events are modeled as using labor and vaccination supplies, giving RESORCl(t) = HRCAST*(1 - C2)*P0P7(t) + HRVACC*[POP7(t) + P0P8(t)] (5.6.15) RESORC2(t) = SPVACC*[POP7(t) + POP8(t)] (5.6.16) where: RESORC1(t) = hours of labor required in the current DT time increment to perform castration of male calves and vaccination of both male and female calves-— hours/DT RESORC2(t) = units of veterinary supplies used for vaccination in this DT time increment-—#/DT SPVACC = quantity of veterinary supplies used per calf vaccinated POPi(t) = total population of the ith cattle cohort C2 = fraction of male calves cohort output being retained for bull replacement HRCAST = hours of labor per calf castrated HRVACC = hours of labor per calf vaccinated. 178 Weaning and culling are herd management activities which only require labor. Weaning labor requirements are: KK7 KK8 ‘ RESORCl(t) = HRWEAN* X SUPBOP7j(t) + X SUBPOP8.(t) j=IAMIN j=IAMIN 3 (5.6.17) where: RESORCl(t) = the number of hours of labor required to wean calves and transfer them to desired uses—-hours/DT SUBPOP7j(t) = population of the jth subgroup of male calves SUBPOP8j(t) = population of the jth subgroup of female calves KKi = the number of subpopulations of the ith herd cohort HRWEAN = number of hours of labor required per calf weaned—— hr/calf IAMIN = the subpopulation index of the calf cohorts which is the smallest index greater than the age represented by AGEMIN. Culling is another herd management procedure which can be modeled as utilizing labor resources exclusively. The quantity of labor required for culling is directly proportional to the number of cows culled; this gives KK RESORCl(t) = HRCULL* Z SUBPOPl,(t)*CULFRC.(t) (5.6.17a) j=1 J J where: RESORCl(t) = the number of hours of labor required for culling of the mature cow herd th . SUBPOPl.(t) = the number of cows in the j subpopulation of J the mature cow cohort CULFRC.(t) = fraction of subpopulation j of the mature cow J cohort culled HRCULL = hours of labor per cow culled. 179 Reseeding and fertilizing pasture lands is one way that a manager can increase forage production. The costs of this activity are in the use of labor to perform the activity and in the physical supplies used. NLANDS RESORCl(t) = (HRSEED + HRFERT)* X LANDn (5.6.18) n=1 NLANDS RESORCS(t) = SPSEED* Z LANDn (5.6.19) n=1 NLANDS RESORC4(t) = SPFERT* Z LANDn (5.6.20) n=1 where: 1 RESORCl(t) = hours of labor used in reseeding and fertilizing of herd grazing lands—~hours/DT RESORC5(t) = quantity of seed used in reseeding grazing lands—— kg/DT RESORC4(t) = quantity of fertilizer used in fertilizing grazing lands of the herd—~kg/DT LANDn = area of land parcel n--hectares NLANDS = number of distinct land parcels HRSEED = hours of labor to seed per hectare HRFERT = hours of labor to fertilize per hectare SPSEED = units of seed applied per hectare SPFERT = units of fertilizer applied per hectare. Another event highly land oriented is harvesting of forage growth through mechanical means. or for cash sale. This would be either for storage Harvesting requires labor and utility usage, since fuel is an element of the composite resource ”utilities." 180 NLANDS RESORCl(t) = HRHARV*DAYS* Z MI-IR (t) (5.6.21) n=1 n NLANDS RESORC3(t) = SPHARV*DAYS* Z MHR (t) (5.6.22) n=1 n where: MHRn(t) = mechanical harvest rate from land parcel n--kg/day DAYS = the number of days in a DT time increment HRHARV = hours of labor per kg forage harvested--hour/kg SPHARV = units of utilities per kg forage harvested-~units/kg RESORCl(t) = hours of labor required for forage harvesting in this time increment--hour/DT RESORC3(t) = units of utilities used in forage harvesting in this time increment--units/DT. RESORC6(t) is the quantity of land which is leased, either for grazing or crop production. HLEASE is this quantity; it is assumed to be constant over a single year. Two activities are assumed to take place on an ongoing basis throughout the year. Repair of buildings, equipment, and land improvement, ie. fencing, is one. A second is the use of utilities to warm buildings and provide power for lights, ventilation, etc. The amount of utilities required is certainly a function of the time of year and of the activities that are seasonal because of climatic variations. 9 NSTOCK RESORC3(t) = UNIT*[SPUTL1* Z POPi(t) + SPUTL2* ): FSTOCKn(t) i=1 n=1 ) (5.6.23a) if t < TSPRNG, or t > TFALL 181 9 NSTOCK ‘ RESORCB(t) = UNIT* SPUTL3* Z POPi(t) + SPUTL4* ' X FSTOCK (t) i=1 n=1 n (5.6.23b) if TSPRNG : t _<_ TFALL where: RESORC3(t) = quantity of utilities used in this time increment to provide heat, power, lights, etc.--units/DT POPi(t) = current cohort i population FSTOCKn(t) = current feed stock n quantity--kgs TSPRNG, TFALL = the onset and stoppage of plant growth, respectively SPUTLl = units of utilities per animal per month—-units/ animal/month SPUTL2 = units of utilities per kg of feed stock per month—- units/animal/month SEUTL3 = units of utilities per animal per month—~units/animal/ month SPUTL4 ? units of utilities per kg of feed stock per month-- units/animal/month UNIT = DT/.O83333. Equations 5.6.23a, b represent the quantity of utilities needed to provide power for a DT time increment to buildings during wintering and grazing periods, respectively. SPUTLl, SPUTL2, SPUTL3, and SPUTL4 are monthly time-based parameters; the variable UNIT is a scaling factor to adjust the use of utilities to the time interval DT that has been used. Repairs are also an ongoing activity that must be adjusted to get a quantity of resources used per DT time increment. Labor and repair materials are the only resources that are used in the equation modeling these repairs. 182 [ 9 NSTOCK RESORCl(t) = UNIT* HREQIP* Z POP.(t) + HRFCAP* Z FSTOCK (t) l i=1 1 n=1 n NLANDS + HRMAIN" 2 LAND (5.6.24) n=1 n [ 9 NSTOCK RESORC8(t) = UNIT* SPREP1* X POP.(t) + SPREP2* Z FSTOCK (t) l ._ 1 n 1-1 n=1 NLANDS 1 +SPREP3* X LANDSn (5.6.25) n=1 where: RESORCl(t) = quantity of labor used for repairs this time increment--hours/DT RESORC8(t) = quantity of repair material used in this time increment--units/DT LANDn = area of land parcel n--hectares HREQIP = hours of labor per animal per month--hour/anima1/month HRFCAP = hours of labor per kg of feed stock per month--hour/ animal/month HRMAIN = hours of labor per hectare per m0nth—-hour/hectare/ month SPREPl = units of repair material per animal per month/units/ animal/month SPREPZ = units of repair material per feed stock per month-— units/kg/month SPREP3 = units of repair material per hectare per month--units/ hec/month POPi(t) = current pepulation of herd cohort 1 These two equations model repair resource consumption as functions of the number of cattle, the quantity of feed stocks on hand, and the amount of land area in use. 183 Determination of animal production costs is completed by costing out the quantity of resources used in each DT time increment of the simulation. RESORC£(t), 2 = l, ..., 8 gives the quantity of each of the resources used. BPRICE£(t), R = l, ..., 8 gives the price per unit of resource as the resource variables RESORC(t) were defined in Table 5.2. Multiplication of the quantity of physical resource used by its current price is the cost of supplying that resource quantity in the current time increment. The production cost vari— ables ACOST£(t) give the costs of animal production in the current time increment by appropriately grouping the production resources. This gives the production costs as defined in Table 5.1 as ACOSTl(t) = RESORCl(t)*BPRICE1(t) (5.6.26) ACOST2(t) = RESORC8(t)*BPRICE8(t) (5.6.27) ACOST3(t) = RESORC3(t)*BPRICE3(t) (5.6.28) ACOST4(t) = RESORC2(t)*BPRICE2(t) + RESORC7(t)*BPRICE7(t) (5.6.29) ACOST5(t) = RESORC4(t)*BPRICE4(t) + RESORC5(t) (5.6.30) ACOST6(t) = RESORC6(t)*BPRICE6(t)*DT (5.6.31) NSTOCK ACOST7(t) = 2: STKFEDn(t)*CPRICEn(t) (5.6.32) n=1 9 6 _ i: . . ACOST8(t) — DT*iE1ADDi(t)*APRICEPGRADEi(t)(t) PAWATEi(t) (5 33) ACOST9(t) = OVHEAD*DT (5.6.34) where: ACOSTk(t) = production cost in category k in the current time increment--$/DT PAWATE.(t) = average weight of animals purchased and added 1 to cohort i in this period-—kg 184 BPRICE£(t) the current price per unit of RESORC (t)--$/unit the quantity of resource used in the current time increment of the simulation as determined from all events and activities which have occurred in the DT increment--units/DT. RESORC£(t) DT = the simulation time increment in fraction of a year NSTOCK = the number of feed stocks in use by the enterprise STKFEDn(t) = quantity of feed stock n fed to cattle in the current time interval--kg./DT CPRICEn(t) = current price of feed stock n——$/kg. APRICEk(t) = current price of grade k cattle——$/kg. PGRADEi(t) = grade of cattle purchased as a addition for cohort i OVHEAD = annual miscellaneous production costs——$/yr ADDi(t) = annual rate of herd additions 0, if ADDRTi(t) 5 0 ADDRTi(t), if ADDRTi(t) > 0. Total animal production costs for this time increment are ACOST10(t) = .E ACOSTi(t) iii ACOST7(t) values are excluded from this total because they are counted in CCOST3(t) where the purchases of feed crops are itemized. Double counting of expenses would occur if ACOST7(t) were included as a part of the total running operating costs of the enterprise. ACOST7(t) is useful, however, because the current value of feed stocks fed to the herd is a valuable item of status information for management. Table 5.3 itemizes the entire parameter list used in deter- mining production costs and presents a sample value. Other values 185 Table 5.3 Parameter Values and Definitions for the Production Cost Element of the Financial Component Parameter Value Definition OVHEAD 5000. annual overhead and miscellaneous costs--$ DEFLAT 0.91 ratio of crOp production costs to prices SPBRED 0.5 units of breeding supplies per cow servicing SPVACC 1.0 unlts of vaccination supplies per calf HRCALV 0.05 hours of labor per calf born HRFEDl .00047 hours of labor per kg. feed fed HRBRED .25 hours of labor per cow servicing HRVACC 0.05 hours of labor per calf vaccinated HRWEAN 1.75 hours of labor per calf weaned HRCULL 0.2 hours of labor per cow culled NRCAST 0.2 hours of labor per male calf castrated HRSEED 0.04 hours of labor per hectare of land seeded HRFERT 0.04 hours of labor per hectare of land fertilizer HRHARV .0021 hours of labor per kg. forage harvested HREQIP 0.03 hours of labor for equipment repair per animal per month HRFCAP .00002 hours of labor for repair per kg. feed stocks per month HRMAIN 00475 hours of labor for repair per hectare of land per month SPUTLl 2.5 units of utilities per animal per month during wintering SPUTL2 .0001 units of utilities per kg. feed stock per month during winter SPUTL3 0. unlts of utilities per animal per month during grazing SPUTL4 .00001 units of utilities per kg. teed stOCK per month during grazing SPUTL5 0.02 units of utilities per kg. forage harvested SPSEED 0.0 units of seeds used per hectare sowr SPFERT 0.0 units of fertillzer per hectare fertslized SPREPl 0.10 units of repair material per animal per month SPREPZ .000013 units of repair material per kg. feed stock per month SPREP3 0.01 unlts of repair material per hectare of land per month HLEASE hectares of leased land per year 186 can, of course, be used by reading in any desired value during program initialization and start up. Detailed cost examination of individual enterprise records and operation should serve to obtain better values than these given. A final note--OVHEAD repre— sents miscellaneous costs and the salary that a manager would obtain that is separate from hours of physical labor that he might perform. Debt of the enterprise has been divided into the arbitrary classifications of short-term and long-term. This has been done to draw the needed distinction between borrowing for land purchases, e.g., mortgages, and borrowing for operating capital. Long-term debt is assumed to only decrease or remain constant during individual runs of the model. Short—term debt can be increased by borrowing at special decision points, the interest rate and repayment schedule can also be renegotiated, whereas these are fixed for long-term debt. Long-term debt is therefore handled by the following equations. CAPTL1(t) = CAPTL1(t-dt) — PMONTH(t) (5.6.36) PINTER(t) = CAPTL1(t)*CAPTL3*DT (5.6.37) where: CAPTL1(t) = current long—term debt——$ CAPTL contracted monthly repayment of long-term debt—-$/month 2 CAPTL3 = annual interest rate for long-term debt PINTER(t) payment of interest on long-term debt in this time increment--$/DT repayment of principal on long—term debt in time increment—-$/DT PMDNTH(t) 0, if CAPTL1(t) = 0 CAPTL1(t), if CAPTLl(t) f CAPTL2(t) CAPTL2(t), if CAPTL1(t) > CAPTL2(t). 187 Short-term debt is handled analogously except that additional loans can be obtained, and the interest rate and repayment variables are functions of time. SDEBT(t) = SDEBT(t-dt) — SMONTH(t) + SLOAN(t) (5.6.38) SINTER(t) = SDEBT(t)*SDEBTR(t)*DT (5.6.39) where: SDEBT(t) = current short-term debt-—$ SDEBTR(t) = current annual interest rate on short—term debt SLOAN(t) = short-term debt incurred this time period--$/DT SINTER(t) = payment of interest on short-term debt--$/DT SREPAY(t) = monthly repayment of short-term debt--$/month SMDNTH(t) = repayment of principal on short-term debt--$/DT 0, if SDEBT(t) = 0 SDEBT(t), if SDEBT(t) f SREPAY(t) SREPAY(t), if SDEBT(t) > SREPAY(t). The total payments of principal and interest, both long— and short- term, are aggregated into CAPTL5(t). CAPTL5(t) = PMONTH(t) + PINTER(t) +~SMONTH(t) + SINTER(t) (5.6.40) Inclusion of debt in its various forms and the required repayment of debt in an important aspect of this model, because of its significant contribution to negative cash flows. Debt repayment on common terms is a constant cash outflow, whereas cash inflow is highly concentrated and seasonal. This leads to the common occurence of negative cash flows during the bulk of a year interrupted by singular periods of heavy cash inflow. Short-term debt is a key feature allowing oper— ating capital to remain positive, thus permitting continued enterprise operation. 188 The final financial element included in this model of enterprise finances is taxation. Taxes take two forms—~income taxes on the gross profits of the enterprise, and property taxes on its real property. Since taxation is a highly localized process, the model developed here is rather basic and should be made more specific for actual operation in a fixed environment. Income taxes are assumed due in one payment on April 1 for gross income earned in the preceding calendar year. Property taxes are paid twice yearly, on January 1, and on July 1. Depreciation is taken on depreciable assets (assumed to have a cemmon tax life) at the time of tax payment and is computed using sum of the digits. Thus taxable income is accumulated from January 1 using TAXINC(t) = TAXINC(t-dt) + TPGRP(t) - CAPTL4(t) (5.6.41) where: TAXINC(t) = accumulated yearly taxable income—-$ CAPTL4(t) = annual depreciation--$ TPGRP(t) gross profit in this time period--$/DT. Equation 5.6.41 is used to determine the current contribution to taxable income from the gross profits of the current simulation time increment. Taxable income is an accumulated variable which begins with an initial value of zero on January 1 of each year; in simula- tion runs which last longer than a single year, the variable TAXINC(t) is reset to zero when the simulated time equals January 1. At this time two actions occur: TAXLIB(t) is set equal to the current value of TAXINC(t), and TAXINC(t) is set equal to zero. TAXLIB(t) repre~ sents the taxable income from the previous year which is subject to 189 represents the average depreciation lifetime of these assets. The sum-of-the digits method of depreciation is used to determine the actual depreciation charge that can be made in each tax year. This computation is made by the model on April 1 of each simulated year and is subtracted from taxable income at that point. CAPTL4(t) is the variable name used to represent this depreciation variable; it is determined by equation 5.6.42 to be TLIFE — INT(t-t0) .-.- * CAPTL4(t) LIFE LVALUE (5.6.42) 2 INT(t—t ) 0 t=tO where: CAPTL4(t) = current amount depreciated in this time increment--$ LIFE = the depreciable lifetime allowed by tax law--years INT(S) a function which integerizes the value 3 LVALUE the original purchase price being depreciated//$. CAPTL4(t) is equal to zero at all times other than April 1 of each year. Property taxes are paid on the actual real estate actually owned. The variables LANDn, giving the area of the nth land parcel, are not necessarily owned; seme could be leased. The true area subject to tax is the difference between total land and leased land. HLANDS ) TAx(t) = 0,50*pR0TAx* t Z LANDn — HLEASE (5.6.43) n=1 if t = January 1, or July 1 190 income tax in the current year. As long as income taxes have not yet been paid, the variable TAXLIB(t) remains at the value that it was set to on January 1. On April 1 tax is paid at the current rate payable by corporate farms; TAX(t) is the income tax paid. Then 0.0, if t t April 1, January 1, July 1 TAXCt) = TRATE*TAXLIB(t), if t = April 1 where: TAX(t) = the dollars in taxes paid this time increment of the simulation--$/DT TRATE = 0.5, the tax rate on gross taxable income TAXLIB(t) = the gross taxable income from the previous tax year--$. A significant factor involved in determination of taxable income is depreciation-any allowable depreciation is subtracted off of gross profits when computing taxable income. Depreciation is itself a complex subject which is highly controlled by rules and regulations of various tax authorities. Additionally, it is completely specific to the individual enterprise being studied since the heart of depre— ciation is the quantities of depreciable assets owned by the enter- prise and the allowable depreciable lifetime. This financial model makes a very modest effort to include depreciation primarily because of its potential significance in causing differences in the dynamic computation of cash flow and net profit. The model of depreciation adopted here employs the general idea of an average purchase value of assets owned and a corresponding average depreciable lifetime. LVALUE represents the total purchase prices of all depreciable assets owned by the enterprise, while LIFE where: PROTAX = annual tax on land--$/hectare—year LANDn = area of land parcel n —— hectares HLEASE = area of land leased -— hectares. The combination of taxes on property and on income is handled by the variable TAX(t). TAX(t) is zero at all times of the year except for the dates January 1, April 1, and July 1. Subroutine FINANC provides the overall organization of the financial component by calling the above simulation models of the various financial elements. Gross profits in the current time increment is the difference between revenues and payments. TPGRP(t) = AREV10(t) + CREVNCROPS+ft) — ACOST10(t) - CCOST4(t) - CAPTL5(t) (5.6.44) where: TPGRP(t) = current period's gross profit-~$/DT AREV (t) = total animal revenue earned in this 10 . per10d——$/DT CREVNCROP 3+1 ('3) = total crop revenue earned in this period--$/DT ACOST10(t) = total animal production costs in this period--$/DT CCOST (t) = total crop production costs in this 4 . per10d—-$/DT CAPTL5(t) = total debt payments in this period—-$/DT. Cash flow in this period is the difference between gross profits and tax payments, if any. CASH(t) = TPGRP(t) — TAX(t) (5.6.45) where: CASH(t) = current time increment cash flow——$/DT 192 TAX(t) = taxes paid in this time increment, either income or property——S/DT TPGRP(t) gross profits of the enterprise in this time increment--$/DT. Cash flow is an important management decision-making variable because it reflects the actual transfer of cash into and out of the enter- prise. As explained earlier, beef cattle enterprises are character- ized by steady net cash outlfows during most of the year, interrupted by heavy cash inflows when products are sold. These seasonal sales products are, of course, weaned calves and excess forage and crop production. Finally, net profit in the period is the difference between cash flow and depreciation, giving TPNP(t) = CASH(t) — CAPTL4(t) (5.6.46) where: TPNP(t) = net profit in this time increment--$/DT CAPTL = depreciation taken in this time increment-~$/DT. 4 Operating capital of the enterprise is, of course, affected by cash flow and borrowing, giving WCAPT(t) = WCAPT(t-dt) + SLOAN(t) + CASH(t) (5.6.47) where: WCAPT(t) = current level of operating capital-—$ short-term loans incurred in this time increment-—$. SLOAN(t) Yearly accumulated profits, both gross and net, and cash flows are important variables in analyzing the overall operation of this enterprise. The subroutine FINANC determines such annual variables by summing the individual incremental values for each calendar year. 193 This gives TAGRPm(t) = TAGRPm(t-dt) + TPGRP(t) (5.6.48a) TANPm(t) = TANPm(t-dt) + TPNP(t) (5.6.48b) TACASHm(t) = TACASHm(t-dt) + CASH(t) (5.6.48c) where: TAGRPm(t) = accumulated gross profits to date in the mth year of operation—-$ TANPm(t) = accumulated net profits to date in the mth year of operation—-$ TACASHm(t) = accumulated cash flow to date in the mth year of operation-—$ m = index of the year of operation since the beginning of the simulation 1, if 0 ft~<1 2,if1_<_t.<2 DUR, if DUer f t < DUR DUR = time specified as the final time horizon of model time simulation. Computation of present discounted values of these three annual variables is extremely helpful in evaluating the impact of specific management policies over time. Undiscounted values can be quite misleading since the time value of money and the impact of inflation are net accounted for. When the simulation has reached the final time horizon (t = DUR) the following values are computed for a variety of discount rates. M TAGRPm(t) (5.6.49a) m=1 (1 + DFLATRi)m 194 6 M TANP (t) PRV21 = 2 1“ (5.6.49b) m=1 (1 + DFLATR1)m M TACASH (t) PRV3i = X m (5.6.49c) m=l (1 + DFLATRi)m where: PRV11 = present discounted value of annual gross profits using discount rate 1 PRV21 = present discounted value of annual net profits using discount rate 1 PRV3i present discounted value of annual cash flows using discount rate i DFLATRi = discount rate 1 M = the integer number of years of the simulation run. The financial component is an important part of the simulation model of the enterprise; it is central to evaluation of alternative management strategies because it reduces physical events and activi— ties to dollars and cents. Most applications using the model will be interested in the financial aspects of management strategies rather than the specifics of the physical events that have occurred as a result of the strategy. The financial model developed in this section is certainly adequate for the purposes of this thesis, but in use by specific enterprises there will need to be certain changes to tune the financial model to the Specific enterprise operating environment. Tax rates and the parameters listed in Table 5.3 are the main means of tuning the model; these are entered during the initialization of the computer program. The details of this initialization can be found in the User's Guide to the Beef Cattle Enterprise Simulation Model, Chapter Three. 195 V.7 Simulation Model Calling Structure The previous sections of this chapter have defined the mathemat- ical models that describe the various processes of the system. The simulation models that implement these equations are written in FORTRAN, and can be found in the User's Guide ..., Chapter Four. Figure 4.13 and others have referred the reader to the general calling structure of the simulation model. This section will indicate the exact sub- routine calling structure used, noting calls to subroutines from within other subroutines by indentations to the right. MAIN . . . . . . . . . . the main program GENERT EXOG HDMOG4 . . . . . . . . population demography * BIRTHZ . . . . . . birth rates BIRAT DVDPLR DLVDPL DDPLR WEIGHT . . . . . . . . cattle weights FEEDS . . . . . . . . feed stock accounting FORAGE . . . . . . . . forage growth INQUIR . . . . recognition of decision points RESPNS# . . . . . . . . reading control inputs NORMAL . . . . . . . . endogenous decision making * developed by Margaret Schuette[48] # called only when a decision point is encountered by INQUIR 196 MAIN, continued CONSUM . . . . . . . . nutrient impacts NUTRN . . . . . . shell for Schuette * subroutines COWCYC * ALAC * GROFEM * GROMAL FINANC . . . . . . . . financial accounting REVENU PRCOST CAPTAL TAXSUB PRINTR . . . . . . . . intermediate simulation increment printing. This listing of subroutine calls is only a brief reference to the organization of the computer program and its subroutines which actually simulate the system behavior. The full and complete details of the FORTRAN programming can be found in Chapter 4, User's Guide to the Reef Cattle Enterprise Simulation Model. v.8 Summary This chapter has developed the mathematical models upon which the simulation model is based. These models meet the requirements of the problem statement, but considerable room for improvement and exten— sion remain. The four physical system components--cattle demography, forage growth, feed stock accounting, and nutrient impacts--provide the model user with the necessary detail to be useful in investiga- ting management decision making strategies. The management decision making component itself has been shown to involve both endogenous decisions made by the model and exogenous decisions made by the model user. Several secondary components have been explained in terms of handling necessary details such as correct selection of exogenous input variables and financial accounting. The following chapter will outline various means used to validate the model, that is, to verify that it properly describes the beef cattle enterprise. 198 v.9 Glossary of Variables The following alphabetical listing of variables includes all variable names used in the preceding sections of this chapter. These same names are used whenever feasible in the computer program and subroutines, as listed in Appendix 1. Parameter names are not included here, but may be found in the text as they are used. 1. 10. ll. 12. l3. l4. ACOSTm(t) = ACTEDkn(t) ACTIVEn(t) ADDRTi(t) = AGEMIN = AHRn(t) = ALGDi(t) = ALLOCi(t) = ALWTi(t) = APFUTRmk(t) = APRICEk(t) AREVi(t) = ASALESi(t) ASELLi(t) = the produEfiion cost in the current DT time increment for the m cost subdivision--$/DT quantity of feed stock n required for feeding in the current DT(k=2), or the winter season (k=1)--kg forage density photosynthetically active in the nth land parce1--kg/hectare the annual rate of addition of animals to herd cohort i—-#/yr the minimum age of calves weaned--years the rate of forage harvest by grazing in land parcel n—- kg/day the average price grade of cattle in cohort i currently total quantity of feed allocated to individuals in cohort 1 per day--kg/day the current average weight of cattle in cohort i--kg current expectation of the value of the kth grade cattle price m DT's into the future--$/kg the current market price of cattle of grade k—-$/kg revenue earned in the current time increment from sales of cattle from cohort i--$/DT the number of cattle sold in the current time increment from cohort i--#/DT the proportion of the current pOpulation of cohort i that is to be sold in this time increment 199 15. AVGGRDi(t) = the average price grade of cattle sold in this time increment from cohort i 16. AVGTMP(t) = the average temperature in the current DT time increment--°C l7. AVGWi(t) = the average weight of cattle sold in this time increment from cohort i--kg l8. AWFn(t) = the animal harvest wastage factor in the current time increment for land parcel n l9. BASEDGn(t) = the basic digestibility factor in the current time increment for land parcel n 20. BEGCAV = the time in the year at which cows begin to calve 21. BFRACil = the 1th point in the curve describing the accumulated calving pattern for cohort i 22. BPFUTle(t) = the current expectation of the value of the 1th production resource price m DT's into the future—— $/unit 23. BPRICE1(t) = the current market price for the 1th resource of production--$/unit 24. BR (t) = th current birth rate on an annual basis for the 1 E i herd cohort--#/yr 25. CAPTLm(t) = the current value of the mph financial variable 26. CASHl(t) = current cash flow in this time increment—-$/DT 27. CATTINik1(t) = the quantity of TDN allocated to members of herd cohort i per.day from concentrates (kel), or roughages (k=2), under feed plan 1 28. CCOST1(t) the crOp pfioduction cost in the current time increment for the l crop cost subdivision-—$/DT 29. CFINAL = the proportion of the existing forage in each land parcel that is to be harvested at the end of the growth season (TFALL) 30. CNCALi(t) the quantity of concentrates allocated to cohort i for the current time increment-~kg/DT 31. CNCFRCin(t) = the fraction of the concentrate TDN allocation to cohort i to be obtained from feed stock n 32. CPATijk = the fraction of the jth subpopulation of cohort i to have calved by CTIM.. 13k 33. 34. 35. 36. 37. 38. 39. 40. 41. 42. 43. 44. 45. 46. 47. 48. 49. 50. CPFUTR (t) = mn CPRICEn(t) CREVn(t) = CROPnl(t) CROPGn(t) CRQUALn1(t) CSALESn(t) CTIMijk(t) CULFRCj(t) C2 = C3 = C4 C5 C6 C9 C10 C11 = DAYS = 11 200 the current expectation of the value of the nth feed stock price 2m DT's into the future--$/kg the current market price of feed stock n--$/kg the revenue earned from sales of feed stock n in the current time increment-—$/DT crop production in the current time increment for feed stock n--kg/DT annual crop production to date of feed stock n——kg the TDN value of the current quantity of feed stock n being harvested this time increment the quantity of feed stock n sold in the current time increment--kg/DT the time of year that calves will be born of female animals from subpopulation j of the ith cohort as a result of the kth servicing--years the fraction of the jth subpopulation of the mature cow cohort that is to be culled at TCULL fraction of the output of the male calf cohort trans- ferred to the young bull cohort after weaning fraction of the output of the female calf cohort transferred to the slaughter heifer cohort fraction of the output of the male calf cohort sold on the market as weaned calves fraction of the output of the female calf cohort transferred to the replacement heifer cohort fraction of the output of the male calf cohort sold on the market after weaning fraction of the output of the female calf cohort sold on the market as weaned calves fraction of the output of the replacement heifer cohort transferred to the mature cow cohort fraction of the output of the bred heifer cohort sold on the market the number of days in the DT time increment 51. 52" 53. 54. 55. 56. 57. 58. 59. 60. 61. 62. 63. 64. 65. 66. DELAYi(t) = DFLATRi(t) DGAINij(t) DIGESTn(t) DISTFDin1(t) DMIij(t) = DMITDNij(t) = DR. = 1 DT = DUR = DURBi(t) = EFORGij(t) = EGAINi(t) EMAINi(t) ENDCAV = EVAP = 201 the current length of time required for the average member of cohort i to pass through the maturity interval modeled by that cohort——yr the ith discount rate used to compute the discounted present values of financial variables at the end of the simulation run the rate of weight gain for the jth subpopulation of cohort i—-kg/day the TDN value of the forage in land parcel n at the current time the fraction of the cohort i allocation of concentrates (if n is a feed concentrate), or of roughages (if n is a feed roughage) to come from feed stock n under feed plan 1 the dry matter intake per day of a member of subpopulation j of cohort i--kg/day the average TDN y lue of the dry matter intake of members of the j subpopulation of cohort i the annual death rate for herd cohort i members-- fraction/year the time increment used in the simulationn-year the time horizon over which the simulation is to be run--years the duration of the breeding season for members of cohort i——years quantity of energy available for weight gain after maintenance needs satisfied-~mcal energy value of the feed intake in terms of growth for cohort i__mca1 gain energy kg feed energy value of the feed intake in terms of maintenance for cohort i__mcal maint. energy kg feed the time at which the calving season is complete-—year the equivalent height of water evaporated per day—~cm/day 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. Fn(t) = FATFAC = FCOSTi(t) = FDENSEn(t) = FEDTDNi(t) = FEEDALi(t) = FLEVELn(t) = FORGALi(t) = FORTDNi(t) = FPLANSl(t) = FQUALn(t) FQUANn(t) FRCLOSn(t) = FREVi(t) = FRQUAL(t) = FSTOCKn(t) = FTOLl = GRADE1 (t) = J GRNn(t) = 202 the rate of transfer of growth from roots to greenery in land parcel n—-kg/hectare/day parameter relating predicted dailygains for cattle at their cohort weight maximum into price grade increases the cost of retaining the p0pulation ofcohort i for an additional time increment at the planned feeding schedule--$ forage digestibility factor based on forage density per animal in land parcel n the average TDN value of the roughage fed to cohort i from feed stock sources the quantity of roughage allocated to cohort i for the current increment of time--kg/DT the minimuquuantity of forage needed to sustain animals grazing in land parcel n per DT time increment--kg the quantity of roughage allocated to cohort i as a result of grazing policy-~kg/DT the TDN value of the forage from roughage allocation to cohort i the time at which the 1th feed plan for the herd is completed——year the current TDN value of the nth feed stock the quantity of feed stock n in the current time increment which is carried over from the previous period's stocks-~kg the annual loss rate of feed stock n—-fraction/year expected marginal revenue gained from retention of cohort 1 animals an additional DT time increment--$ forage digestibility factor relating current forage digestibility to time in the growth season the current level of feed stock n supplies--kg the absolute fractional deviation between current feed stock levels and seasonal requirementsallowed . . .th the current price grade of cattle in the 3 subp0pulation of cohort i the current quantity of forage existing within land parcel n—-kg 86. 87. 88. 89. 90. 91. 92. 93. 94. 95. 96. 97. 98. 99. 100. 101. 102. 103. GRNDENn(t) = IFLAG = INB.. = 13 INTCAV = min) = LAND = n LIFE = LVALUE = MHRn(t) = MWF = NCROPS = NLANDS = NSTOCK PDSTRBni(t) = PERC(t) = PGRADEi(t) = PHOT0n(t) = P = INDEXn(t) 203 the quantity of forage greenery available per animal per day in land parcel n——kg/animal/day an indicator flag describing the existence of regular and special decision points the number of servicifigs during the breeding season for animals in the j subpopulation of cohort i the number of intervals of length one-half month within the calving interval the number of subpopulations within cohort i the area of land parcel n--hectares the average depreciable life of the mix of depreciable assets owned-—years the purchase price in total of the depreciable assets owned—-$ the rate of mechanical harvest of forage growth in land parcel n—-kg/day the wastage factor in mechanical harvest of forage the number of feed stocks potentially used by the enterprise the number of land parcels the number of feed stocks potentially used by the enterprise the fraction of the population of cohort i grazing in land parcel n the equivalent height of water percolating down below effective root depth--cm/day the average grade of animals purchased and added to cohort i the net growth rate of forage in land parcel from photosynthetic energy conversion——kg/day relative age of forage existing in land parcel n-—days 104. 105. 106. 107. 108. 109. 110. 111. 112. 113. 114. 115. 116. 117. 118. 119. 120. PINTER(t) = PMONTH(t) = POPi(t) = PROTAX = PRQUALn(t) = PRmG = RAIN(t) = REACTDn(t) = REMOVL(t) = RESORCl(t) = RGRAZE(t) RHGALi(t) = RHGFRCin(t) RINij(t) = ROOTn(t) ROUTi(t) = RPOPi(t) = 204 payments of interest in the current time increment on long—term debt-—$/DT payments on principal in this time increment for long— term debt--$/DT the current population of cohort i the current property tax rate on an annual basis-- $/hectare/year the TDN value of feed stock n purchases the discounted present value of the annual values of gross profit (m = 1), net profit (m=2), or cash flow (m=3), using discount rate k the current rate of rainfall-~cm/day the current intentions of feed stock n net purchases at the end of the growth season (TFALL)—-kg the current fraction of the existing forage in each land parcel to be harvested at each harvest time the quantity of production resource 1 used in the current time increment-~units the minimum quantity of forage required to maintain the cattle herd grazing in the current time increment—— kg/DT the overall roughage allocation to herd cohort i in this time increment--kg/DT the fraction of the roughage TDN allocation (from feed stocks) to cohort i to be obtained from feed stock n the intermediate output rate in the delay model of cohort 1 corresponding to subpopulation j the current density of roots in land parcel n-— kg/hectare the current output rate of animals from herd cohort i--#/year the population of cohort i which is reproductively characteristic of cohort i behavior 121. 122. 123. 124. 125. 126. 127. 128. 129. 130. 131. 132. 133. 134. SDEBT(t) = ll SDEBTR(t) SINTER(t) = SLOAN(t) = SM(t) = SMF(t) = SMONTH(t) = SNF(t) = SNUTn(t) = SOLAR(t) SPLIT = SPOILn(t) = SREPAY(t) STKFEDn(t) STKPURn(t) STOCKLn(t) SUBPOPij(t) = the current level of short—term debt-—$ the annual interest rate required for short—term debt payments of interest made in the current time increment for short—term debt--$/DT current quantity of short—term debt acquired this time increment-—$/DT the current level of soil miosture within an effective root depth-—cm the current value of the soil moisture quality index the current repayments of principal of short~term debt——$/DT the current value of the soil nutrients quality index the current level of soil nutrients in land parcel n—-units the current average daytime rate of incoming solar radiation-~langleys/day the fraction of roughage consumption derived from feed stock roughages when an excess roughage allocation has been made the annual rate of TDN decline in storage for feed stock n the current monthly payment to principal required for the outstanding short—term debt-—$/month the quantity of feed stock n allocated to herd consumption in the current time increment--kg/DT the current quantity of feed stock n purchased in this time increment-—kg/DT the stocking level of grazing cattle in land parcel n—~#/hectare . . .th _ the current number of animals in the j subpopulation of cohort i 206 138. SUMTPF(t) = the current value of the integral of the temperature quality index (TPF)-—units 139. TACASHm(t) = the annual cash flow in year m of the enterprise simulation--$/year 140. TAGRPm(t) = the annual gross profit in year m of the enterprise simulation—-$/Year 141. TANPm(t) = the annual net profit in year m of the enterprise simulation——$/year 142. TAX(t) = the number of dollars of taxes paid in this time increment to all authorities—-$/DT 143. TAXINC(t) = taxable income in the current calendar year to date-—$ 144. TAXLIB(t) = income from the previous calendar year liable for income taxes currently--$ 145. TBRDi(t) = the current value of the date on which breeding of cohort i is to commence-—year . 146. TCULL(t) = the current time at which culling of the mature cow cohort is to occur—~year 147. TDMICi(t) = the quantity of concentrates consumed by herd cohort i in the current time increment—-kg/DT 148. TDMIR_(t) = the quantity of roughage consumed by herd cohort i 1 in the current time increment—-kg/DT 149. TDNCi(t) the average TDN value of the concentrate allocation to herd cohort i 150. TDNRi(t) = the average TDN value of the roughage allocation to herd cohort i 151. TFALL = the time at which the growth season staps——year 152. TFRACmi = the fraction of the weaned female calves in the mth cohort subpopulation older than the youngest subpopu— lation weaned which is transferred to cohort i 153. TGREEN(t) = the total quantity of greenery available for grazing by cattle at time t--kg 154. TMHRl(t) = the time at which the 1th mechanical harvest of forages is to occur——year 155. 156. 157. 158. 159. 160. 161. 162. 163. 164. 165. 166. 167. 168. 169. TPF(t) = TPGRP(t) TRATE = TSPRNG = TWEAN(t) = VGREEN(t) = Wij(t) = WCAPT(t) = WGTMAXi = XCPRODn(t) XCQUALn(t) XPECTAk(t) XPECTBl(t) XPECTCn(t) zx3n(t) = 207 the current value of the temperature quality index total dollars of gross profit earned in the current time period——$/DT the rate of taxation for_businesses on taxable income the time at which the growth season is begun-—year the time at which the calf cohorts are to be weaned—— year the average TDN value of the total forage currently available for grazing the average weight of members of the jth subpopulation of cohort i--kg the current level of working capital on hand-—$ the maximum weight that members of cohort i can achieve—-kg the expected production of feed stock n during the growth season--kg the expected TDN value of the production of feed stock n expected the expected price of cattle of grade k at time t-— $/kg h the expected price of the It resource of production at time t--$/unit the expected price for the nth feed stock at time t~— $/kg the fraction of the photosynthetically converted energy growth rate used for root growth CHAPTER VI MODEL TESTING AND VALIDATION Validation is, in its essence, a process of verifying that a model correctly represents the real world process that it is supposed to represent. Model validation is an essential step in model devel— opment because use of an invalid model c0uld easily be worse than no model at all. Validation of simulation models is more difficult to achieve than validation of other model forms, because of the great complexity that simulation models are commonly used to portray dy- namically. This chapter will seek to review some approaches to model validation in general, present evidence to demonstrate the simulation model's validity, and summarize more sophisticated procedures for developing user confidence in this model. V1.1 Approaches to Validation In some respects the process of validating a model is a search for truth. What is desired is that the model truly represent the real system. Unfortunately, demonstration of truth has been subordi— nated to questions about truth and truthfulness themselves in the literature. Philosophers through the ages have failed to resolve these questions in any generally acceptable manner. In the realm of economic models there are three approaches to establishment of model 208 209 truthfulness about the real world.1 Robbins [44] espouses the thinking that models must be ultimately based on unverifiable basic assumptions that have to be accepted or rejected on their own merits. The model should correctly evolve from them, but the basic assump— tions themselves are not testable. Hutchinson [27] totally rejects this approach and maintains that nothing is proved true until it can be empirically tested. This includes the basic model assumptions; they are suspect where assumptions cannot be supported by data. Freidman [17] represents the Positive Economics school of thinking, which would find truth if a model correctly predicts behavior. Some question this line of thinking by stating that it leads to use of models which may predict behavior correctly, but which are based on obviously false assumptions. Some also draw a distinction between positive and normative models,2 but this beef cattle enterprise simulation model contains both positive and normative characteristics. We are led to the conclusion that this model must produce the proper behavior but must also be developed according to realistic and valid assumptions. In a practical sense validation is a two—phase process for simulation models. First, there is the problem of validating, or 1T. H. Naylor, J. L. Balintfy, D. S. Burdick, and K. Chu, Computer Simulation Techniques (New York: Wiley, c. 1966), Chapter 8. 2Definitions frOm Gilmour [18], Chapter 1: Positive~—a model which must show reasonable correspondence to the real system Normative——a model which indicates a desirable level of operation for the real system which may or may not be currently achieved. 210 verifying, that the simulation model correctly represents the mathematical model. Here the major difficulty is debugging com- puter programs and checking that approximations made to achieve a solution are giving acceptably low errors. Second, there is the more difficult problem of verifying that the mathematical model really does represent the real world system. The less well understood the real world system is the greater the difficulty in validating the model. A common occurrence in development of simulation models is discovery of areas that have been ignored by conventional researchers, even though strong understanding is needed to develop a model of an entire system. This uneven level of understanding of parts of a system leads to difficulties in validation, because it is the entire system behavior that is of interest. The "weak links" which exist, because the necessary exploratory research has not been done, inhibit generation of confidence in a system model. A beneficial result of development of simulation models is the discovery of these poorly understood areas, if then resources for research can be reallocated to these problem areas. Several specific proposals for validating simulation models have been developed. These borrow from one another rather heavily, and perhaps this is due to the type of problem that the developer was familiar with at the time he proposed his procedure. Gilmour [18] proposes the following steps: 211 1. determination of face validity 2. determination of output validity through a. analysis of stability b. historical comparison c. comparative analysis of output after making assumption changes. Popper [41] suggests that the goal of such exercises is increasing confirmation in the truthfulness of the model. Therefore the proper procedure is to make tests of the model; the more tests the model passes, the greater the degree of confirmation in its truthfulness. A procedure which seems generally applicable to many situations proposes a hierarChy of testing stages. The procedure is to work from easy to difficult as confidence in the model's validity is developed. These validation steps, in order of increasing difficulty, are: l. logical consistency 2. tracking historical data 3. satisfaction of expert eyeballing 4. prediction of the future. One accomplishes these steps through study of model output over time by varying parameters to observe the direction of output changes, by determining that the parameter sensitivities reflect real system sensitivities, and in general by immersing oneself in the model for lengthy periods to understand how the simulation behaves. Satisfac- tion of experts in apprOpriate areas is an excellent indication of validation. This satisfaction can be achieved, in all likelihood, only after an iterative series of presentation reviews, criticism, 212 and model changes. Prediction is the final test of model validity, to be confirmed or rejected by real events. Complete certainty in model validity is never possible; this means that models are used with less than perfect confidence in their predictions. Model use and continuing refinement should be linked together to insure that the model increases in validity as time passes. Chapter 8 of this dissertation will suggest several areas wherein greater sophistication in the component models would prove beneficial to both generation of confidence and usefulness to model users 0 V1.2 Validation Tests This section will define and present examples of validation tests for the four steps reviewed in section one: logical consistency, tracking historical data, satisfaction of expert eyeballing, and pre— diction of the future. It should be understood that validation testing has not been limited to examples presented here. Logical Consistency Logical consistency is the requirement that the model satisfy elementary system characteristics. Among these are satisfaction of system identities, variables uniquely defined, consistency with known laws, etc. Also classified here are basic tests to insure that behave ioral modes of the model correspond with real world behavior. Identi— ties that exist in the real system must be modeled and simulated as identities. If the change of a parameter in a real system would change 213 an output variable, then the same direction of change should occur in the simulation model. If the real world system is stable under certain conditions, then the simulation model must be stable as well. Varying model parameters to generate alternative output dynamics, sensitivity testing of parameters, multiple input condition sets, and hand calculation are all means of determining logical consistency in a simulation model. Examples illustrating logical consistency in six areas will be described here. These will concern: 1. stability of the simulation approximations by tests of DT size 2. verification of population identities 3. verification of feed allocation and feed stock consumption identities 4. verification of financial identities 5. exploration of parameter values controlling forage digestibility 6. exploration of parameter values controlling forage growth. These examples will constitute partial evidence of the validity of the simulation model developed to describe a beef cattle enterprise. An important characteristic of the simulation model is stability with respect to the DT step size used in the time simulation. The following section of this chapter will present the results of two tests of the simulation model's stability with respect to DT size. The ap- proximations that are made in the differential equations of the popu- lation model to make them solvable in the computer via difference equations are directly related to the size of the DT step increment selected. Theory indicates that the errors of approximation should 214 tend toward zero as DT tends toward zero. First, during the early deve10pment of the demographic model five different step sizes were investigated to use in selecting the most appropriate DT step. Sec- ond, an exhaustive comparison of two runs of the most recent version of the simulation models was made using different DT sizes. These two examples will demonstrate that the DT step size selected for use is sufficiently accurate and need not be reduced to attempt greater accuracy. An early version of the demography component HDMOG4, referred to as HDMOG3, was tested with five DT step sizes; the sizes tested were 0.01, 0.02, 0-03, 0.04, and 0.05 years. Three points in the simulated time output printing were selected for analysis; these times were 0.15, 0.50, and 1.0 years. Time equal to 0.15 corre- sponds to the most common value printed before the run using a DT value of 0.01 was prematurely aborted. Time equal to 0.50 corre- sponds to the point in the simulation horizon where maximum population was achieved. Time equal to 1.0 corresponds to the final time value. Table 6.1 presents the results of these simulation runs in terms of the total herd population at these three specified times for the five DT sizes investigated. Examination of these results shows that there is very little difference in the pepulation values obtained for any of the DT sizes investigated. Therefore, to economize in the use of computer execution time, the size of 0.05 years was selected for use in this simulation model. 215 Table 6.1 Total Herd Population at Selected Points in Time Versus DT Step Size DT size T = 0.15 T = 0.50 T = 1.00 0.05 215.17 349.43 243.31 0.04 216.15 347.94 242.65 0.03 215.17 349.35 243.40 0.02 216.30 352.01 244.98 0.01 215.17 -- -- An attempt at validation of the population demography component is important because of the central position held by this part of the model. An identity exists in the demographic equations which will be exploited in this section to verify that the demographic component does perform correctly. By definition, net population change = births - deaths — sales + purchases (6.2.1) over any time interval, where each of these are quantities not rates. This equation may be solved for the variable, deaths, and this new equation used to compute deaths in each time interval. These death values computed from computer run output printing will be compared with the death values computed using the theoretical death rates assumed to apply over time. Equation 6.2.1 can be solved for the variable, deaths, and transferred into the proper variable names as given in section v.9, with the following result: 216 t+dt 9 deaths = SUBPOP71(t+DT) + SUBPOP81(t+dt) + X APURi(t+dt) t i=1 9 9 — Z ASALESi(t+dt) + Z [POP.(t) — POP.(t+dt)] (6.2.2) . 1 1 1=1 i=1 where: t+dt deaths = total herd deaths over the time interval (t,t+dt) t SUBPOP7l(t+dt) = male calf births over the interval (t,t+dt) SUBPOP81(t+dt) = female calf births over the interval (t,t+dt) 0, if ADDRTi(t) : o APURi(t+dt) = ADDRTi(t)*DT, otherwise = animals purchased for cohort i during the time interval (t,t+dt) ASALESi(t) = animals sold from cohort 1 during the time interval POPi(t) = population of the ith cohort at time t. The theoretical deaths that should occur over the time interval (t,t+dt) are determined by the individual cohort death rates and the cohort populations at time t. The death rates are constant over time and are specified as data entries in the simulation model initiali— zation phase. Therefore, deaths should follow the following theoreti— cal relationship: t+dt 9 deaths = DT* E POP (t)*DR. (6.2.3) 1 1 t i=1 where: t+d t deaths = total herd deaths over the time interval (t,t+dt) t DRi = annual death rate for cohort i. 217 These two equations will be used to compute the simulated value and the theoretical value of deaths compared for a typical run of simula— tion model over the interval (0,0.35) in DT increments of 0.05 in Table 6.2. Table 6.2 Comparison of Computed(Corrected) and Theoretical Herd Deaths Over Time t+dt t+dt t sim. deaths theo . deaths error t t t l 0.00 0.424 (0.325) 0.325 0.099 (0.000) 0.05 0.419 (0.320) 0.308 0.111 (0.012) 0.10 0.422 (0.323) 0.297 0.125 (0.026) 0.15 0.844 (0.745) 0.692 0.152 (0.053) 0.20 1.020 (0.921) 0.886 0.134 (0.035) 0.25 1.089 (0.990) 0.921 0.168 (0.069) 0.30 1.063 (0.964) 0.956 0.107 (0.003) 0.35 1.083 (0.984) 0.953 0.130 (0.031) 1 interval totals 6.364 (5.562) 5.336 1.028 (0.226)\ interval error ——— -~- 19% (4%) The results presented in Table 6.2 indicate a consistently high estimation of death rates by the death rate equation 6.2.2, as com— pared to the theoretical values computed using equation 6.2.3. Since equation 6.2.2 is subject to errors from many sources, it is not sur— prising that there is some error present. A factor contributing to 218 these consistently high estimates is the way in which the variable ASALESi(t) is computed. This variable is the number of cattle sold in the time increment (t-dt, t) from cohort i. As explained in section V.6, this variable is integerized so as to give even integer numbers of animals sold from each herd cohort. This integerization process loses the true value, but the rationale is that the errors of inte— gerizing will sum to zero over the entire herd. In the results ana— lyzed in Table 6.2, the ASALES4(t) value is always zero, but it is zero because the integerizing process makes it zero. Without such integerizing there w0uld be a small positive value for each ASALESA(t) through time. When this factor is included in the error deternfination, the resulting percentage of error in accumulated deaths over the time span (0,.35) drops from 19% to 4%. In table 6.2 the values in paren- theses indicate the value of simulated deaths and the error of this value, taking into account the integerization SOurce of error. The sample output analyzed here verifies that the demographic component is performing as it should. The population identity relating births, deaths, sales, and purchases has been shown to be simulated quite closely to the theoretical values. When the errors are viewed in terms of their magnitude compared to the total herd population at any time, then such errors are extremely small (on the order of 0.2%). The only significant effect of the error is a very slightly smaller herd population than w0u1d be predicted if the theoretical death rates held true. A slight error is made in the financial component by using this integerization process, but it is not significant. 219 A second area where identities may be used to determine proper functioning of the model is in feed allocations and feed stock con— sumption. Management allocates feed to the cohorts on the basis of specified desired levels of TDN to be delivered per animal per day. Depending on the current TDN value of each feed stock and the desired proportion of the TDN allocation to come from each feed stock, the physical quantities allocated are determined. CNCALi(t) and FEEDALi(t) are the quantities of concentrate and roughage allocated to cohort i, respectively. These allocations are in units of kg/DT. Equation 6.2.4 represents the total feed (in kg/DT) allocated to the herd. 9 ALLOCATIONl = i§l[CNCALi(t) + FEEDALi(t)] (6.2.4) The mix of the TDN allocation to each cohort from feed stock n and the population of each cohort reSult in the quantity of feed stock n fed in the current time increment—-STKFEDn(t). Equation 6.2.5 repre— sents the total quantity of feed stocks allocated to the herd. NSTOCK ALLOCATION2 = z STKFEDn(t) (6.2.5) n=1 Logical consistency requires that these allocations be equal to one another; that is, the allocation in terms of cohort allocations must be the same as the allocation in terms of feed stocks used for feeding. A typical model run from initialization (t=0.0) to the spring decision point (t=.25) was selected for analysis to veryify this feed stock identity. Table 6.3 presents the results of output printing over a time interval of 4 DT's. The allocations according to 220 equations 6.2.4 and 6.2.5 are given along with the difference between them. Even cursory examination will show excellent agreement between these two forms of feed stock allocations. Table 6.3 Feed Stock Allocations t ALLOCATIONl ALLOCATION2 DIFFERENCE1_2 0.05 54,037.3 54,039 ~1.7 I 0.10 54,552.4 54,555 —2.6 \ 0.15 54,742.0 54,742 —-- \ 0.20 61,192.3 61,192 0.3 J Source: Computer Printout Number MV57864 A second feed stock identity used to verify that the accounting process of subroutine FEEDS works properly is the basic individual feed stock balance equation. FSTOCKn(t) = FSTOCKn(t-dt) — CSALESn(t) - STKFEDn(t) - STKLOSn(t) + CROPn1(t) + STKPURn(t) (6.2.6) Briefly stated, present feed stock level equals former feed stock level minus sales, quantity fed to cattle, losses and plus crop pro- duction and stock purchases. Detailed explanation can be found in section V.3. This equation must be followed for the feed stock component of the simulation model to be working correctly. By drawing on the same printout used above, the validity of the feed stock component can be demonstrated. A single feed stock was 221 selected for study and followed through the same time interval used above; i.e., (0.0,0.25). Table 6.4 presents the results of the come puter printout and the calculated value of FSTOCKn(t) using equation 6.2.6. FSTOCKp represents the simulated value printed, while FSTOCKC represents the calculated value. Excellent agreement was found between these two values at each time point. The simulation model component FEEDS does work properly. A third testing point for logical consistency in the feed allocation and feed stock consumption area involves comparison of the desired TDN to be fed to the herd and the TDN actually fed. Allo- cation of TDN is controlled by the user through the values of CATTINil£(t), CATTIN12£(t), and FPLANSR. These are the kg TDN to be delivered to each member of cohort i from concentrates per day, the kg TDN to be delivered to each member of cohort i from roughage per day, and the times at which feed plan f is to be completed, respectively. Desired TDN delivery is then 9 DESIRED TDN = DAYS*iZlPOPi(t)*[CATTIN11£(t) + CATTIN12£(t)] (6.2.7) Actual TDN delivered to the herd is simply the product of each feed stock delivered times its TDN value, summed over all feed stocks. This gives NSTOCK ACTUAL TDN = Z FQUALn(t)*STKFEDn(t) (6.2.8) n=1 An example run for a single DT time increment is presented in Table 6.5; this is the same run analyzed previously. There the popu- lation and desired feed levels are reported for each herd cohort as 222 womnm>z can Hounaaoollmouoom o H o H o -- a-uuosuo maesmfi Nausea assess «chaos omkmmm .. ueuoeme -- em as mos Has omNN monasm -- o o o o 6 some .. momma amass amass enema o amaaem -- o o o o o mameem .. o o o o o mmuamo wowemfi assess assess fleeces omsNNN OOOmNN axooeme mN.o oa.o mH.o on.o mo.o oo.o 6 oEHH Ho>o poummfiou mosam> xuoum boom «.0 oHan 223 Table 6.5 Sample Run of Feed Stock Use and Desired Allocation i/n POPi(t) CATTINill(t) CATTINi21(t) FQUALn(t) STKFEDn(t) 1 228.800 5.0 1.0 0.49 15,937 2 39.939 4.0 1.0 0.76 7,425 3 ——— 4.0 1.0 0.83 ——— 4 19.837 6.0 2.0 0.46 16,977 5 ——— 5.0 1.0 0.76 13,700 6 ——— 5.0 1-0 ——— ——- 7 --- ___ ___ 3 --- ___ --- 9 --- 4.0 1.0 Source: Computer Run MV57864 DESIRED TDN = 31,594.3 kg/DT ACTUAL TDN = 31,673.5 kg/DT error = +0.25% 224 well as the quantity fed and quality of each feed stock. Only a very slight difference exists between actual and desired TDN; this further indicates logical consistency. A final area where logical consistency can be verified through identities is within the financial component. Variable values are computed within four subroutines specialized in a particular area, such as production costs, and are brought together to form higher level variable values, such as gross profit, in subroutine FINANC. Three equations will be used to verify that the financial component is properly bringing costs, revenues, depreciation, debt repayment, and taxes into computation of gross profit per period, annual gross profit, and discounted gross profit at final project termination correctly. The fundamental equation of interest is that one determining gross profit per DT time period. This is (t) - CCOST TPGRP(t) = AREVlO(t) + CREV4(t) — ACOST (t) NACOST+1 NCCOST+1 — CAPTL5(t) (6.2.9) In this equation gross profits in the current time increment are the animal and crop revenues minus the animal and crop production costs minus debt servicing. This gross profit figure contributes to tax~ able income for the current year, as do depreciation charges. This rationale gives a second equation: TAXINC(t) = TAXINC(t—dt) + TPGRP(t) — CAPTL4(t) (6.2.10) where TAXINC(t) is the taxable income earned in the calendar year to date. Finally, at the conclusion of the simulation time horizon, the discounted annual gross profit figures are used to compute the 225 discounted present value of this profit stream over time. Thus TAGRPm(t) = TAGRPm(t) + TPGRP(t) ‘ (6.2.11) MYEAR TAGRP PRVUL: Z ———m—m (6.2.12) m=l (1+DISC2) For a detailed explanation of these equations and individual variable definitions, see sections 6 and 9 of Chapter V. A test case used to analyze whether the financial component is working properly is represented in tabular form in Table 6.6. A non—interactive management algorithm was used in this run to speed up the simulation process; this allowed a run over a one—year time horizon in a single model run. A DT time increment of 0.03846 (two weeks) gives 26 time increments over the time interval (0.0,l.0). Table 6.6 lists all financial variables needed to compute equations 6.2.9—12 by hand. Computing TPGRP(t), as given by equation 6.2.9, and comparing it with the simulated values given in Table 6.6, reveals no instances where the computed value and the simulated value differ by more than one cent. This excellent agreement indicates that TPGRP(t) is simu- lated correctly from the constituent elements of equation 6.2.9. Table 6.7 summarizes the comparison of the simulated and calculated values of TAGRPm(t), TAXINC(t), and PRV through PRV 1i 16' Examination of Table 6.7 will reveal excellent agreement between the printed values obtained from the simulation and the calculated values determined from the elements of the equations. Again, the conclusion is reached that the financial component is properly 226 00.00000- 00.0000- 00.0 00.0000 00.0 00.0000 00.0 00.000 000.0 00.00000- 00.0000- 00.0 00.0000 00.0 00.0000 00.0 00 000 000.0 00.00000- 00 0000- 00.0 00.0000 00.0 00.000 00.0 00.000 000.0 00.00000- 00 0000- 00.0 00.0000 00.0 00.000 00.0 00.000 000.0 00.00000- 00.0000- 00.0 00.0000 00.0 00.000 00.0 00.000 000.0 00.00000- 00.0000- 00.00000 00.0000 00.0 00.000 00.0 00.000 000.0 00.00000- 00.0000- 00.0 00.0000 00.0 00.0000 00.0 00.000 000.0 00.0000- 00.0000- 00.0 00.0000 00.0 00 000 00.0 00.000 000 0 00.0000- 00.0000- 00.0 00.0000 00.0 00.000 00.0 00.000 000.0 00.0000- 00.0000- 00.0 00.0000 00.0 00 000 00.0 00.000 000.0 00.0000- 00.0000- 00.0 00.0000 00.0 00.000 00.0 00 000 000 0 00.0000- 00.0000- 00.0 00.0000 00.0 00 000 00.0 00.000 000.0 00.0000- 00 0000- 00.0 00.0000 00.0 00.000 00.0 00.000 000 0 3 V 0 0 w o 0 00 000000 00000 00000 00000 s w >000 >000 0200 NL NI 3 V D 3 0 O S S l L + + T. .L umww one we cou0uom mafia COHuQHDEHm m Hm>o moanmflum> Hmfionmnflm vmuooaom 0.0 @0300 227 00.00000 00.0000- 00.0 00.0000 00.0 00.000 00.0 00.000 000 0 00.00000 00.0000- 00.0 00.0000 00.0 00.000 00.0 00.000 000.0 00.00000 00.0000- 00.0 00.0000 00.0 00.000 00.0 00.000 000.0 00.00000 00.0000- 00.0 00.0000 00.0 00.000 00.0 00.000 000.0 00.00000 00.0000- 00.0 00.0000 00.0 00.000 00.0 00 000 000 0 00.00000 00.00000 00.0 00.0000 00.000 00.0000 00.00000 00.000 000.0 00.00000- 00.0000 00.0 00.0000 00.000 00 000 00.0 00.0000 000.0 00.00000- 00.0000- 00.0 00.0000 00.0 00.000 00.0 00.000 000.0 00.00000. 00.00000 00.0 00.0000 00.0000 00.000 00.0 00.00000 000.0 00.00000- 00.0000- 00.0 00.0000 00.0 00.0000 00.0 00.000 000 0 00.00000- 00.0000- 00.0 00.0000 00.0 00.0000 00.0 00.000 000.0 00.00000- 00.0000- 00.0 00.0000 00.0 00.0000 00.0 00.000 000.0 00.00000- 00.0000- 00.0 00.0000 00.0 00.0000 00.0 00.000 000.0 3 V 0 3 0 0 w w 0 00 oszmmo >mm< 0509 m m D 3 0 0 m m + + I I U.ucoo 0.0 mafimH 228 Table 6.7 Comparison of Calculated and Simulated Values of Financial Variables Variable Cgmguted I Simulated Difference TAXINC(tf) $48729.08 348729.07 $—0.0l TAGRPl 59479.08 59479.07 —0.0l PRVll* 59479.08 59479.08 -0.01 PRV12 58312.82 58312.82 0 PRVl3 57191.42 57191.42 0 PRV14 56112.34 56112.33 —0.01 PRV15 55073.22 55073.22 0 1 % the discount rates used for PRV 11 through PRV 15 are 0.0, 0.02, 0.04, 0.06, and 0.08, respectively 229 simulating the revenues and costs involved in the enterprise's operation. A final note concerning validation of the financial component directs the reader's attention to Table 6.6. The typical pattern of highly irregular cash flows from cattle and crop sales are well illustrated in this run. The effect on gross profit in each time increment is just as mentioned previously in this thesis——a cattle enterprise is characterized by negative cash inflows, punctuated occasionally by very heavy positive inflows. Even though the overall profit earned in this model run is positive, there are only three DT time increments where positive cash inflows have occurred. These times are 0.692, 0.769, and 0.808. These times correspond to sales of weaned calves, sales of culled cows, and sales of forage harvested beyond projected herd needs for the winter. The animal production costs reported in the ACOST (t) column are indicative of the NACOST+1 many different events occurring through the course of the year requiring different levels of resource consumption and, hence, overall production cost. Proper determination of forage digestibilities is an important aspect of the validation of the forage growth component, because of the vital influence digestibilities have in the nutrient impact com- ponent. The initial modeling attempt for this component, following Anway [2], failed to properly predict forage digestibilities. Some reasons for this failure are the lack of a factor relating forage density in terms of quantity of forage per animal grazing to 230 digestibility and that the curves represented by the arrays BDGEST and QUALTY were not well specified. Digestibility of forage in each land parcel is determined bv equation 6.2.13. It is the product of the basic digestibility, the seasonal adjustment factor, and the forage density adjustment factor. DIGESTn(t) = BASEDGn(t)*FRQUAL(t)*FDENSEn(t) (6.2.13) Each of these three factors is determined by linear interpolation between data values to obtain the proper factor value corresponding to the argument value. A complete discussion of this formulation and the determination of the function value corresponding to the input argument can be found in section V.2. Figures 6.1 and 6.2 depict several trial sets of values for the elements of the arrays BDGEST and QUALTY used in the determination of BASEDGn(t) and FRQUAL(t), respectively. Table 6.8 lists the re— sults of nine runs investigating the effect of alternative trial shapes of the curves defined by the elements of the arrays BDGEST and QUALTY. Run 1 is the result of the original curves as taken from Parton and Marshall [40], Sauer [47], and Anway [2]. Runs 2 through 6 investigate alternative shapes of these curves as pic— tured in Figures 6.1 and 6.2. Runs 7 through 9 illustrate the effect of alternative weather patterns on the original model (run 1) and the best model (run 6). All runs have zero—level mechanical harvest- ing and grazing to remove these influences from the range of digesti- bilities produced by the natural response of the forage growth component to climatic input factors. 231 1.0 BASEDGn(t) o 3 10 15 20 25 3o . . . . . PINDEXn(t) 35 40 45 50 55 (day) Figure 6.1 Base digestibility values versus an age index. 1.0 FRQUAL(t) .9_ .8_ n 7 ‘.~"..-‘- ' ‘\\ """ ‘ ““““ i «-... Run 2 \ .6_ \ \ \ ‘\ .5, . \ \ .4, \ \ \ .3 , 'L\ \ ‘v0 \ .2 . ‘.‘ ‘\ .1 ~‘.,Run 1 . l . . . 0 . 00. . L . SUMTPF(t) O 25 50 75 100 125 150 175 200 225 250 275 Figure 6.2 Seasonal digestibility factor versus season. 232 Table 6.8 Forage Digestibility Factors Investigated to Obtain an Improved Digestibility Range Run Definition Digestibility Range 1 original 29 - 46 Z 2 first QUALTY set 32 — 46 3 first BDGEST set 35 - 51 4 both 2 and 3 38 - 51 5 second QUALTY set 38 - 51 6 both 3 and 5 45 — 54 \ 7 1 and first of \ climate changes 28 — 53 8 6 and first of \ climate changes 45 — 56 9 6 and second of climate changes J 45 — 51 X Runs 2, 3, 4, and 5 represent intermediate improvements over the base run. A better run than all of these is run 6, which incor— porates the changes made in run 3 and run 5. Forage digestibilities ranged from 45 to 54% TDN in this run. This range closely corresponds to digestibilities of forage grasses obtained through field trials and experimental observation in temperate climates [11, 14, 15]. If the forage material is alfalfa or other high—energy plant species, then this range in digestibilities is excessively small. A scaling factor to adjust forage digestibility base levels might be adapted to provide a more flexible tool for evaluating different grazing environments. This improvement will be left for future development. Runs 7, 8, and 9 represent an attempt to determine the responsiveness of the forage growth model to alternative climatic patterns. Logical consistency requires that forage growth and digest— ibility have some degree of response to levels of solar radiation and rainfall. Run 7 evaluates the effect of a first alternative weather pattern on the "best" model, as determined in run 6. Run 9 evaluates yet a second alternative weather pattern on the "best" model. Table 6.9 indicates the weather conditions, range of forage digestibility, and forage production for runs 1, 6, 7, 8, and 9. Runs 1 and 7 should be compared to each other, as should the group of runs 6, 8, and 9. Comparison of runs 1 and 7 in Table 6.9 reveals that under conditions of reduced rainfall and reduced solar radiation that forage production has dropped significantly from run 1 to run 7. 234 Table 6.9 Weather Effects on Forage Growth and Forage Digestibility run variable 1 7 6 8 9 average solar radiation(ly) 221 186 221 186 186 average rain— fall(cm/day) 0.84 0.69 0.84 0.69 0.70 average tem- _‘_.___—___ perature(°C) 8.5 8.5 8.5 8.5 8.5 digestibility range(%) 29—51 28—53 45—54 45—56 45-51 total forage production 4449 3813 4449 3813 4418 (kg) 235 The range of digestibilities has also broadened marginally. These effects are what would be expected, since forage growth is primarily related to solar radiation levels. In fact, the percentage of reduc— tion in forage growth is roughly midway between the percentage of reduction in solar radiation and the percentage of reduction in rainfall. Comparison of runs 6 and 8 in Table 6.9 reveals that the shift from the standard climatic variable pattern shared by runs 1 through 6 to the climatic variable pattern shared by runs 7 and 8 has a very similar effect to that reviewed above for runs 1 and 7. Again, the range of forage digestibilities has broadened slightly. The same drop in forage production has occurred, becaase the growth prediction aspects of all nine of these runs are common. Run 9 shares the same factors determining forage digestibility with runs 6 and 8 but has yet a different weather pattern. The aver- age values of the climatic patterns listed for runs 8 and 9 are nearly the same, but the pattern across the year (and especially the growing season) is different. This difference is all important, as the results of run 9 indicate. Run 9 has only a slightly decreased level of forage growth from run 6, whereas run 8 had quite a large drop. Forage digestibilities have also decreased in the range ex- perienced in run 9, compared to an increase in range in run 9, using run 6 as the basis for comparison. This illustrates the obvious fact that the distribution of rainfall is as important to plant growth as is the overall level of annual rainfall. 236 Forage growth is sensitive to a number of parameters in the growth model, as explained in section V.2. The original parameter values result in a split between greenery and root storage, which appears to favor storage excessively. The following paragraphs will discuss some sensitivity studies of the forage.growth component to obtain an improved proportioning of energy transformation between greenery, growth, and root storage. Five parameters, which are used in a total of two equations, were tested singly and together in an effort to achieve a more real- istic forage growth characteristic. Equation 6.2.14 indicates the return flow rate of greenery growth from energy storage in the roots to greenery. -GRN (t) /PAR2*LAND Fn(t) = PARl*ROOTn(t)*e n n (6.2.14) Equation 6.2.15 determines the proportion of the overall plant growth rate being apportioned to root storage, as opposed to greenery growth. -PAR5*GRNn(t)/LANDn] J (6.2.15) ZX3n(t) = PAR3 + PAR4*[1.0 - e Table 6.10 lists the results obtained from 14 runs testing various parameter values in an attempt to achieve an increased proportion of total growth directed to greenery instead of to root storage. The run numbering used here continues that started in the runs listed in Table 6.8. Runs 10 through 22 represent changes in the noted parameter(s) from.the base values listed for run 6. Run 6 is the same as the run 6 referred to in Table 6.8. It includes the original parameter values 237 Table 6.10 Sensitivity Testing of the Parameters of the Forage Growth Mbdel Run Description Forage(kg/hec) Roots(kg/hec) 6 base values 4449 3221 10 PAR5=0.0008 4595 3143 11 PAR5=0.0005 4926 2919 12 PAR5=0.0003 5280 2614 13 run 12, PAR4=.20 5463 2446 14 run 12, PAR4=.20 5651 2272 15 run 12, PAR4=.15 5844 2093 16 run 15, PAR1=.15 5915 2108 17 run 15, PAR1=.20 5956 2116 18 run 15, PAR2=400. 6033 2090 19* run 18, PAR1=.20 6402 2212 20 run 18, PAR3=.15 6452 1764 21 run 18, PAR3=.10 6886 1426 22* run 21, PAR1=.15 7035 1457 * an unstable parameter set which caused the quantity of roots per hectare to approach zero 238 obtained from the first developers of this model form (Parton and Marshall, Sauer, and Anway) and the improved digestibility factor relationships described earlier in this chapter. The goal of these investigations was to obtain improved parameter values, which would decrease the quantity of root storage, while increasing the quantity of greenery growth. The quantity of forage and root storage reported in Table 6.10 are the average kilograms per hectare existing at the conclusion of the growing season. No animal or mechanical harvesting was allowed in these runs. Runs 10, 11, and 12 evaluated successively smaller values of PARS from its base value of 0.001. The effect of this parameter is to decrease the weight of the current greenery in the equation con- trolling the split of growth going to root storage and growth going to greenery. The observed effect was, in fact, successively higher forage levels and lower root values. Runs 13, 14, and 15 evaluated successively smaller levels of PAR4 from its base value of 0.30. The effect of this parameter is to control the maximum fraction of photosynthetically converted growth which is directed toward root storage as opposed to greenery. The result of these decreases in the value of PAR4 were increased forage growth and decreased root storage, as predicted by the equa- tions themselves. These three runs were performed while holding PARS at the value of 0.003, so the base level of comparison for them Erm2rmMrmmrm6. Runs 16 and 17 are unprofitable attempts at departures from run 15, with increased levels of parameter PARl. Rather than 239 decreased root storage, as might be expected, there were very slight increases in the season end figures. Run 19 evaluated the addition of PAR2 = 400, while retaining the other values of run 17. It results in unacceptable behavior, in which root storage quantities drop from their initial spring level to zero. All three of these runs are actually unreasonable, since the average root storage quantity drops nearly to zero in runs 16 and 17. Parameter PARl is quite sensitive in early forage growth stages, and any levels higher than the base level of 0.10 reSult in unrealistic model behavior. For this reason, PARl = 0.10 will be retained as an element of all parameter sets. Run 18 evaluates the effect of PAR2 increases on the best run obtained so far; i.e., run 15. Increasing PAR2 from its base level of 200 to 400, decreases the weight given to forage greenery levels in predicting energy transfer from roots to greenery. As expected, increased forage growth and decreased root storage result. Runs 20 and 21 evaluate the effect of decreased levels of parameter PAR3 from the basis of run 18. PAR3 controls the minimum level of the proportion of photosynthetically predicted storage going to roots. Decreases from 0.20 to 0.15 and 0.10 result in significantly increased forage growth and decreased root storage. Run 22 again attempts decreases in the value of parameter PARl, with the same unstable results as in runs 16, 17, and 19. This rein- forces the conclusion that paramter PARl should be maintained at a value of 0.10, regardless of other parameter values. To do other— wise allows excessively high transfer from roots to greenery, 240 resulting in the quantity of root storage being driven to zero. This cannot be permitted to happen. As a result of these investigations, the parameter set evaluated in run 21 was adopted as the new basis set to be used in any subse— quesnt use of the model. This set includes PARl = 1.10, PAR2 = 400, PAR3 = 0.10, PAR4 = 0.15, PARS = 0.003, and PAR6 = 15.0. End-of-the* growth—season forage quantities are 6,886 kg/hec, while root storage quantities are 1,426 kg/hec. Figure 6.3 [ 3,20,40] plots the dynamic growth path followed by forage greenery in run 21, along with several wide-ranging indications of relative forage growth. Goudriaan [20] reports two growth patterns very different from the growth path fol— lowed by run 21. Simulation models were developed for each of these curves, which duplicated the observed data quite closely. The range of forage growth values reported by Parton and Marshall [40] repre~ sent outcomes of alternative climatic conditions for their grassland model. Unfortunately, the weather variables they used were quite different from those resulting in run 21; in fact, the seasonal rain- fall was much less than that encountered in run 21, which is reflected in the much slower growth levels experienced. The range of growth values reported by Baker [3] are indicative of Venezuelan conditions. This range completely brackets the result of run 21. The various references cited here, while unable to verify that the forage growth pattern is accurate, are indicative that the proper range of response has been achieved with the parameter set used in run 21. 241 Forage Density (kg/bee) 15000 ' Goudriaan 2 10000 ' Goudriaan l 5000- Baker simulated J - Parton and Marshall 0 I, A A ‘_ n l J 4 90 120 150 180 210 240 270 Time (days) Figure 6.3 Plant densities over the growing season 242 Tracking Historical Data An important part of the validation process is the tracking of historical data by the simulation model. This ability (on the part of the model) is a significant achievement, which contributes substantially to confidence in the simulation model. Demonstration of the model's ability to track multiple historical time series is further evidence (and of a higher order than logical consistency tests) that the model really does represent the real system it is supposed to represent. The following paragraphs will clarify the process of tracking historical information, discuss some potential problems and difficulties, explain the relationship between tracking the past and predicting the future, and explain the historical tracking performed here. The purpose of having a model track historical time series is, except for academic studies into past behavior, determination of the model's valid representation of the structural relationships over the period in question, with the common sense feeling that this validates the model for the future as well. Obviously this linkage of the past and the future supposes that the relationships between variables is constant through time. In other words, the system has not changed. We can see immediately that this supposition is questionable as a general rule of behavior. We need, therefore, assurance that the past and the future systems are the same, if we are to trust a validated model of the past in the future. When we are satisfied that tracking historical information is a valuable thing to accomplish, how does one go about it? How many 243 time series should be used? What means of reconciling deviations between simulated values and historical values should be used? How closely must the simulation series track the historical series to be proclaimed satisfactory? What criteria can be used to distinguish between time series which are "important" and those which are "unim— portant?" Of these many questions, little of a definitive nature can be said. Current practice leans heavily to such measures as sum of squares and total sum of squares of the deviations between historical and simulated series as indicators of disagreement between data and model.3 Gilmour [18] reviews many statistical techniques which can be used in principle but offers little direction about criteria for choosing among them for specific examples. This is an area needing research and attention as simulation models become more widely used and more important to decision—making at both the micro and macro economic levels. Supposing that one had available relevant time series and an appropriate measure(s) of error, there remains the thorny problem of fine tuning the model to reduce this error measure. In the compli- cated simulation models now being constructed, there are extremely large numbers of parameters which are candidates for change to make the model perform better. No general advice is possible here, as every model is unique at this point. It takes a deep familiarity with, and intuition about, the model to select parameters for adjustment to 3G. L. Johnson, G. E. Rossmiller, and T. W. Carroll, "Problems of Verification and Validation of Large—Scale Simulation Models," Statistics Seminar, Michigan State University, January 20, 1976. improve the error measure. This is a time-consuming problem which should be viewed in a long-term perspective as an ongoing process. This brings one up to the point where one must realize that models are rarely ever clearly right or wrong. They have gradations of rightness and wrongness which can be changed over time through improved parameters, improved data to track against, and improved structural relationships. Models (in the positive sense) are built to predict the future, not to duplicate the past. A model must be installed and used and then trusted to greater degrees as its per- formance improves. The "new" historical information recorded as time passes becomes data which can be tracked against, even if there were no "old" historical data in the first place. If the model structure is adequate, the passage of time affords the user/developer the op- portunity to track against valid time series to fine tune the model parameters. The model will gradually improve, and confidence in its predictions will then increase. Eventually good models will be trusted and used because of their great ability; bad models will be either made good or shelved. The simulation model of a beef cattle enterprise developed in this thesis has no historical data to track against. This is a weak— ness which will inhibit verification and validation of the model, but not fatally so. The very extensive work on logical consistency reported earlier in this chapter is a partial substitute for this lack of data. The use of experts to "eyeball" model results is an additional step in the validation process which can be relied upon to complete the determination of validation for this model. 245 Expert "Eyeballing" As the third step in the hierarchy of validation procedures, the satisfaction of expert "eyeballing" is more sophisticated than logical consistency and tracking historical data. At the same time this step is also less rigorous than the first two. It is quite possible for a model to have cleared the first phases of validation and still be found wanting by experts of the system's behavior. Logi— cal consistency and tracking of historical data are levels of valida— tion which do not require extensive knowledge of the system under study. A model can still contain errors of significance which can only be detected by having an expert review the model output for cer- tain initial conditions and control modes. The intuitive knowledge of the expert is highly useful at this point because it is this source which can say, "This just doesn't look right!" More detailed study of such problem areas can reveal whether the model is correctly simu- lating the real system that it is supposed to represent. If found wanting, then the model can be improved and returned to the expert for appraisal; this iterative process of review may cycle several times before an acceptable model emerges validated. An example of expert "eyeballing" discovering modeling misbehavior is the case of the rates of weight gain for calves. Well into the validation process, the discovery was made (by an expert in animal husbandry) that the predictions of male and female calf weight gains were inconsistent with common experience. Figure 6.4 illustrates Rate of weight gain 246 (kg/day) 1.2 > 1.0 ' Beltsville 0.8 . simulated 0.6 . four breed average 0.4 _ 0.21 0 - * ‘ Calf Age 0 05 .10 15 .20 .25 .30 (years) Figure 6.4 Rate of weight gain of male calf animals versus age 247 the pattern of weight gains produced by the model and two typical weight gain patterns.4 Examination of Figure 6.4 shows that the modeling error is the pattern of weight gains starting at excessively high levels for very young calves and decreasing with age. The real behavior is to have low rates of weight gain initially, with subsequent increases as the calves grow older. Realization of this error resulted in a modifi- cation to one of the routines imbedded within subroutine NUTRN. Details of this final version can be found in Schuette[48]. A second example of the use of expert "eyeballing" of output reSults to determine model validity is given by a series of simula— tion runs exploring alternative grazing and harvesting policies for their effect on forage growth. This series of runs explores timing of forage harvest, intensity of grazing, and combined grazing/har— vesting schemes for their effects on dynamic growth and total forage available for harvest. Additionally, the effect of weather differ— ences on the effect of these policies is evaluated through use of second runs in many cases. Table 6.11 presents the results of 15 model runs reporting total forage harvestable with the described management action, and with no harvesting activity at all. The units reported are are kg forage harvested and harvestable per hectare, and kg TDN forage harvested and harvestable. The effects in terms of kg TDN/hec are more crucial to the impact on the cattle herd because TDN values change radically as new growth appears after harvesting actions. 4USDA, Beltsville Growth Standards for Holstein Cattle, Technical Bulletin No. 1099, Washington, D. C., 1954, p. 5. 248 Table 6.11 Effect of Grazing and Harvesting Policies on Forage Growth and Obtainability Forage Quantity (kg/hec) J- kg TDN/hec I Run Description Action No Action Difference Difference l 80% harvest at T = 0.4 5,790.02 6,886.34 —l,096.32 —4S4.22 2 80% harvest at T = 0.5 6,251.11 6,886.34 —635.23 -205.67 3 80% harvest at T = 0 6 6,175.32 6,886.34 ~7ll.02 —298.79 4 80% harvest at T = 0.7 6,410.52 6,886.34 —475.82 -214.11 S 80% harvest at T = 0.4 anda tT= 0.7 5,357.11 6,886.34 -l,529.23 —649.03 6 80% harvested at T = 0.4, = 0.5, T = 0.6, T = 0.7 2,863.26 6,886.34 -4,023.08 -l,672.55 7 80% harvested at T = 0.4 and at T = 0.7, w t different weather from run 5 3,827.04 5,831.44 -2,004.40 -862.87 8 80% harvested at T = 0. 4, T = 0.5, T = 0.6,T =0 7, with different weather from run 6 2,016.39 5,831.44 —3,815.05 —l,651.83 9 Animal grazing removal of 500 + 2,000T kg/day This action is infeasible with season—long grazing. 10 Animal grazing removal of I 500 + 2,000T kg/day, with different weather from run 9 This action is infeasible with season—long grazing. ll Grazing 5.0 kg/hec—day 4,613.6 6,886.34 —2,272.74 —915.05 12 Grazing 5.0 kg/hec—day, with different weather from run 11 4,674.20 5,831.44 —l,lS7.24 -430.43 13 Grazing 5.0 kg/hec—day + 80% harvested at T = .4, with different weather from the base This action is infeasible with season—long grazing. l4 Grazing 5.0 kg/hec—day + 80% I harvested at T = 0.5, with different weather from the base 2,381.81 5,831.44 —3,502.63 —l,446.27 15 Grazing 5. 0 kg/hec—day + 80% harvested at T = O. 7, th . different weather from the base 3,649.10 5,831.44 -2,182.34 -887.17 249 Runs 1, 2, 3, and 4 explore the effect of a single harvest removing 80% of the existing forage by weight at times 0.4, 0.5, 0.6, and 0.7, respectively. In terms of minimum effect on physical quantity potential harvestable (amount actuallylharvested and final forage value) run 4 is superior. In terms of TDN, run 2 is slightly less harmful than is run 4. In all cases the effect of harvesting is to reduce the quantity of potentially harvestable forage. Runs 5 and 6 investigate multiple harvesting schemes, with the result that two harvests is superior to four but that both are worse than any of the single harvests evaluated in runs 1 through 4. Runs 7 and 8 duplicate the harvesting action of runs 5 and 6, respectively, but under somewhat less favorable weather conditions. The overall harvest possible using only a single end of the growth season harvest (T=O.75) drops from 6,886 kg/hec to 5,831 kg/hec simply because of weather differences. Run 7 turns out to be worse than run 5 in both physical effect and TDN effects, while run 8 is superior to run 6 in both physical and TDN effects. The overall average rainfall, average solar radiation, and average temperatures are important to growth as well as the distribution lying behind such averages. Dis— tributional effects are surely responsible for the improvement of run 8 over run 6 in light of run 7 being inferior to run 5. Runs 9 through 12 shift from forage harvest to cattle grazing. Runs 9 and 10 explore the consequences of a heavy grazing pressure on a land parcel of 100 hectares. In both the standard weather con— ditions (run 9) and less favorable weather conditions (run 10), the 250 given rate of forage removal through grazing is untenable with season— long grazing. The pressure is so great that all forage is consumed. Figures 6.5 and 6.6 depict the dynamic forage levels which occur in runs 9 and 10, respectively, with grazing and without, for comparison purposes. Runs 10 and 12 illustrate the effects of'a successful level of grazing on forage production. Although this policy is successful in terms of being possible over the grazing season, the net effect of grazing is reduced forage production. This effect is quite sensitive to weather, as the difference between runs 11 and 12 indicates. Run 12 has an overall growth reduction of 1,054.9 kg/hec from basic weather conditions but a marginal increase in forage potential under the grazing policy tested. Figures 6.7 and 6.8 illustrate the dynamic pattern of forage growth in runs 11 and 12, respectively. Runs l3, l4, and 15 illustrate the effects of combined grazing and harvesting policies. The grazing intensity remains constant at 5.0 kg/hec/day, while the timing of 80% harvesting shifts from 0.4 to 0.7. Figure 6.9 illustrates the dynamic effect of those policies on forage density. Run 13 is an untenable policy, as the early har~ vest combined with grazing pressure exhausts all plant material in midsummer. Runs 14 and 15 are feasible, but neither would be par— ticularly profitable, since the quantity harvested during the growth season is not as large as the drop in the final density at T = 0.75. The fifteen runs reported here present a picture of the current operating characteristics of the forage growth component, with 251 Forage Density (kg/hec) 4000 - 3000' 2000 1000 ' no grazing grazing A L Jr 0.20 0.30 0.40 0.50 0.60 0.70 Time (years) Figure 6.5 Forage densities over time as a result of basic climatic variable inputs and grazing policies 252 Forage Density (kg/hec) 4000, 3000. 2000 1000. grazing - l n A 0.20 0.30 0.40 0.50 0.60 0.70 Time (years) Figure 6.6 Forage densities over time as a result of basic climatic variable inputs and grazing policies 253 Forage Density (kg/heC) 4000 . 3000 no grazing 2000 1000 grazing ‘ O ” | l 1 A IL I 0.20 0.30 0.40 0.50 0.60 0.70 Time (years) Figure 6.7 Forage densities over time as a result of basic climatic 5 variable inputs and grazing policies 254 Forage Density (kg/heC) 4000 f 3000 . no grazing 2000 [ 1000 _ grazing 0 4 I I l l I II 0.20 0.30 0.40 0.50 0.60 0.70 Time (years) Figure 6.8 Forage densities over time as a result of basic climatic variable inputs and grazing policies 255 Forage Density (kg/bee) 4000 3000 climate induced growth pattern 2000 run 15 1000 0 . 0.20 0.30 0.40 0.50 0.60 0.70 Time (year) Figure 6.9 Forage densities over time as a result of grazing and harvesting policies 256 respect to grazing and mechanical harvesting policies. The effects of the policies investigated here are conditional on the weather patterns used in each run. Other weather patterns are, of course, possible; and the results of the specific policies could change with such different weather patterns. Finally, more exhaustive policy tests would be certain to reveal superior policies yet. Prediction Section one of this chapter indicated that prediction is the final step in model validation. Prediction is not attempted until the model has passed through the preceding three steps in the vali- dation process. Once the model is thought to be ready for use, it is used, but on a tentative basis. The model will make time pre- dictions that can be verified simply by waiting for real events to catch up to simulated events. An extremely helpful action at this point would be careful data collection oriented to the variables used in this model and predicted by it. This allows tracking pro— cedures to be used to identify time series which are not as well modeled as others and, thereby, indicate areas where parameter changes or new model structure is needed. This final stage of validation is also a time when users can begin to develop confidence in the model as they see it work and improve over time. Gradually the model will be more and more reliable, and the validation effort can be concluded. V1.3 Summary This chapter has presented a hierarchy of validation steps which a model must pass through before it can be considered verified or 257 validated. These steps are logical consistency, tracking historical data, satisfaction of expert "eyeballing”, and prediction. A large number of examples of logical consistency have been presented here for the beef cattle enterprise model, as well as some discussion of satisfaction of expert "eyeballing.' Tracking of historical data has not been attempted simply for lack of suitable test cases. Predic— tion must wait for an actual model user willing to devote the neces— sary resources to this final step in the validation process. In short, the model in its current state is well dowu the road of vali- dation, but it has not yet completed the validation stage of model development. Chapter VII MANAGEMENT STRATEGIES This chapter will be the primary report of uses of the simulation model to investigate management strategies of decision—making. Three topics will be covered here: an orientation to management strategies as used in the model, several illustrative demonstrations of the model and its capability, and a summary of what has been learned from these sample demonstrations. VII.1 Strategies of Management Decision Making Although this model contains many quite detailed control variables to allow realistic simulation of an actual enterprise, its main useful— ness lies in investigations of broader questions of management decision— making. As used here the word strategy refers to an approach to manage- ment or operation which embodies a multiple of individual decisions. An example of a management strategy is the question of whether to sell calves as they are weaned, or to retain them for sale later as yearlings. A relatively fundamental difference exists between these two modes of operation. Quite different cash flow patterns would be expected, as well as rather different wintering feed requirements, etc. Detailed decisions to be made within each of these strategies include how much and what type of feed to allocate to the herd cohorts, how to effective— ly use forage growth, and what sales date should be followed. The simulation model accomodates these types of decisions through the mechanism of decision points where the user has the choice of alter— native actions to take to control the behavior of the enterprise. These 258 ———__1 259 decision points occur as often as complex decision requiring the input of a decision—maker are encountered. VII.2 Demonstration Examples of Strategy Investigations This section will discuss specific uses of the simulation model in investigating alternative strategies of management decision—making. Three different and distinct areas of management strategy will be evaluated here: (1) early versus late weaning of calves (2) the rate of development appropriate in increasing the steady state size of the breeding cow herd (3 V general profit maximization for a given breeding cow herd size. These three examples have been chosen as the means of illustrating the capability of the simulation model because of the wide range of decision-making that they encompass. The reader can think of numerous other examples which might be just as appropriate as these. In order to cover these three examples in the limited time and space available, the testing will be under typical conditions but will not be exhaus— tive or complete. The conclusions drawn here should be viewed as tentative and preliminary and subject to further testing before being confirmed. The major criterion for comparing the strategies tested here will be financial; this is in keeping with the subject being discussed in this thesis—~a profit maximizing agricultural business. Other cri— teria will be used when appropriate to reflect differences not handled or poorly handled by financial considerations. The three demonstration examples to be discussed in the remainder of this chapter require numerous exogenous variable specifications 260 through time. These variables fall into four groups: expected prices, climatic variables, crop production, and miscellaneous. Expected prices over time are required for the five cattle price grades, the eight production resources, and for each of the feed stocks. Climatic variables required are the solar radiation, average daily temperature, and rainfall patterns through time. Crop production specifications include the timing and quantity harvested for each feed stock. The miscellaneous category includes the quantity of grazing land and its divisions into homogeneous parcels, the property tax rates, as well as other initial conditions. All three demonstrations will use an expected price pattern which assumes an annual cycle of price fluctuations for cattle and feed stocks. Expected prices for the eight production resources are assumed to be constant through time. Figures 7.1 and 7.2 illustrate the pattern and degree of fluctuation for expected prices of cattle of grade 1 and feed stock 1 (forage), respectively. A pattern of constant prices will be used in the weaning timing demonstration for comparison with the cyclic prices usually assumed. Such prices will be roughly midway between the highs and lows of the cyclic price pattern. Figure 7.3 illustrates the annual pattern of solar radiation incident to the enterprise site. Forage harvesting will be performed at the end of the growth season as well as within the growth season at T = 0.55. Finally, the onset of spring growth and the end of the growing season are assumed to be 0.25 and 0.80, respectively, throughout all runs for these three examples. The many other initial conditions required to operate the model will not be specifically mentioned; unless other~ wise noted such initial conditions are uniformly applicable over all the model runs discussed. 261 Amummkv mzHH 0.0 1|| m.o w.o menu coaumuuwcoEmo G0 now: mooflum wauumo uaaoku mo suouumo Hmncc< 0.0 muzw0m om.o mo.0 OH.H 000000 0000000000 000mm>v mzHH 0.0 0.0 [14‘ c 262 mean cowumuumcoaop :0 pom: mooflua mono 000006 mo :uwuumo Hmscc< N w .0 m»: we omo. mqo. 000000 0000000000 263 C00 0200 0.0 ouflm wwwuanwusw map so uzwm0ou0 COHumfipmu umaom mo cumuuma as: 0.0 1 0.0 ‘1 m.o J 0.0 m.o 1; cc< m.m wusmflm 0.0 0.0 0 . . 0 CON 00m . 0mm A>mm\0mawcmav Auvmom moamm 30Hm mononsm nomwmm mo ummco pooh HmuOH Mano cosmos 5mmo Hmdcq< coaumfluumwn mam mo>amo mo mcwamoz oumg mamuo> haumm mo muoowwm onu waflumwwumm>SH msdm oasfimw mo moanmmm H.m magma 267 to the fact that calf nutrients could be partially supplied by grazing during the early fall period, but were mainly supplied by lactating cows which themselves were grazing and therefore not consuming feeds which are counted in the quantity listed in Table 7.1. Additionally, the ages of the heifers at first estrous, both for replacement heifers and bred heifers, were younger for late weaned calves than for the early weaned calves. In both late and early weaning the oldest group of female calves were selected for replacements, and the second oldest were selected for bred heifers. All of the other calves were sold on the market. The age difference between the oldest and second oldest female calf groups selected accounts for the difference in the ages of first estrous between replacement and bred heifers. The age difference , for both replacements and bred heifers, between those weaned early and those weaned late is only 0.0359 years, or approxi— mately two weeks. This difference would have some effect on the rate of pregnancy achieved for the two heifer groups during their first breed- ing, but only in a marginal way. The overall conclusion drawn from this set of four runs of the model is that late weaning is to be preferred to early weaning even under adverse price expectations. Only when the expected fall in prices for cattle during the fall period is so severe so as to more than offset the increase revenue from heavier animals should early weaning be con- sidered. Even in this instance the consequences of additional purchased feed consumption and delayed estrous onset should be carefully considered before early weaning is adopted. 268 Rate of Development This example of the capability of the enterprise model illustrates the ability of the model to assist in project investment planning. The three cases considered here are ones which would be considered among the range of possible approaches to achievement of the goal. The pro- ject to be investigated here is achievement of a steady-state breeding cow herd of 200 animals within three years of project initiation. The three cases tested are (1) purchase of 200 bred heifers and no other animals, (2) purchase of 250 mature cows and no other animals, and (3) purchase of an entire ongoing herd having a steady—state cow population of 100 animals. Certainly other possible initial states exist, but these three will be the ones tested in this model demonstration. Each of the three runs discussed here shares the same common exogenous variable environment. The standard weather pattern used in the majority of the simulation runs of this thesis is again present here. Figure 7.3 illustrates the solar radiation pattern of this weather set over the calendar year. Annual cycle of cattle and crop prices is expected with peak crop prices in the spring and minimum prices in the fall. Similar timings of the cattle price fluctuations are assumed. Figures 7.1 and 7.2 illustrate typical levels of price fluctuation and the timing of these fluctuations. Production resource prices are assumed constant throughout the three-year time interval evaluated. All runs begin at T = 0.0 (corresponding to January 1 of the calendar year), and spring calving was used for all three runs. The grazing land parcels were identical to those of the weaning demon- stration, as were the feed stock production quantities and timings. 269 Property taxes were again $50 per hectare per year and due in Semi—annual installments. Figure 7.4 indicates the convergence of the mature cow populations of each of the three cases toward the target value. Case 1 population increases rapidly from zero to 200 because the initial conditions were 200 bred heifers which quickly had their calves and became mature cows. Subsequently this population drops below 200 because there are no replacement animals to take the place of cows which die or are culled. Only as replacements are generated from the first calf crop does the population again approach the desired level of 200. Case 2 exhibits a steady drop from the initial level of 250 cows that one would expect because culls and deaths are not replaced with incoming heifers until the first calf crop matures to replacement heifer age. Case 3 follows a slow but steady path of population increase that is expected of a herd which began at a level of 100 and which is growing solely through retention of weaned female calves. Culling is at a very low rate in this case to maximize cow population. All female calves are retained for replacement heifers to minimize the time required to reach the target population. By T = 3.0 Cases 1 and 2 are quite near the desired population target, but Case 3 is still somewhat short of this goal. Had the plot of Figure 7.4 been extended to T = 3.25, then all three cases would have been slightly in excess of the target population. The considerable rise in the Case 3 population would have been due to the large number of replacement heifers (retained from previous calf crops) which would be changing cohort designations as they become two years old. All three cases can be considered to have reached the target. i.e— 270 J mcuwuumm unmaumw>cfi mousu How oEHu nwsounu mcowuwasaoa unocoo 300 wHSumz q n w . mu: am Amummzv oEHH N.m o.m w. 1 q 1 J a N rm .2 N.N 9N m: sq .3 «A o; m. a. a. om OOH omH oom {---}: omN coaumaamom Boo muses: - ‘ [lulIIIIIL Selection of a "best" rate of development from among a group of alternatives is most likely to be made on the basis of financial return. The selection of criteria to determine this return is a decision which should be made quite carefully. Operating characteristics as well as some measure of the change in value of property and assets should be made. For this demonstration example the cash flow of the runs and the change in net worth of assets other than land will be the criteria used for selecting the best alternative. Cash flow will be discussed in terms of overall accumulated flow, annual cash flow, and per period cash flow. Net worth will be discussed in terms of long and short term debt, change in value of feed stocks on hand, and change in value of the cattle herd. Operating capital levels will also be included. Figure 7.5 illustrates the cash flows on an annual basis for the three cases under consideration. All of them share the characteristic of being highly negative; this indicates the current poor condition of the cattle industry faithfully. High feed prices and low cattle prices are the cause of this situation. Case 1 is less negative than the other two alternatives in years 2 and 3, but case 3 actually made money in the first year. Case 3 then plunges down to the worst position of the three for years 2 and 3. This drop can be attributed to low revenues associated with retention of all female animals to maximize calf births in the following years. A reason for the highly negative cash flows of the three alternatives is the debt repayment made necessary by borrowing to meet the needs of the enterprise during the long intervals with negative cash flow. Typical loan conditions used here was repay— ment within one year at 10% interest on the unpaid balance. These terms 272 momma unwaumo>afl moon» pow mafia swsounu mauoouma 30am ;mmo Hound< m.m w w u: we . 00H: m ommo Ampmohv wane . ON Aomms\oooaw v Beam ammo Hmscn< 273 insure that the loans must be rolled over since the entire three year interval produces no windfall of revenues to enable the enterprise to eliminate them entirely. If more favorable terms could be arranged then the cash flows would not be quite so bad, but then the debt situation would not be as good. More detail on the timing of cash inflows and outflows contribut— ing to the annual patterns of Figure 7.5 are given by Figure 7.6. Here the individual DT time increments of cash inflow have been aggregated into values for intervals of 0.20 year. Cases 1 and 2 show quite similar patterns of cash flow over the entire three year project duration. Case 3 differs somewhat from the other two, but becomes more similar as time approaches the end of the project. The working capital requirements of the enterprise can be determined through integration of the consecutive periods of cash outflow. Such outflows must be covered by working capital on hand at the beginning of the outflow interval, or by borrowing during the outflow interval with repayment to be negotiated between the borrower and the lender. The change in net worth of the enterprise is a function of the change in herd composition and age structure, changes in the feed stock inventory, changes in the level of working capital, and debt load. Table 7.2 illustrates the net worth changes which have occurred over the three year time horizon of these investigations. The initial cattle herd valuations are strictly a function of the composition of the herd at T = 0.0, as initial feed stock valuations are solely determined from the quantities of each feed stock on hand. All three cases shared a 30000 15000 -15000 —30000 -45000 -60000 \ l l w I 30000 / 15000 -15000 -30000- -45000 —60000 30000 15000 0 -15000. ~30000 -45000 —60000 274 r ($/ .2yr) I I L7 3. Time V‘VzA W Case 1 .($/-2yr) ‘ L I ‘ l: . I A J . l . Li 3: Time W (years) \/ Case 2 ($/-2yr) r- ;\7 AJV'L UW3. Time (years) - Case 3 Figure 7.6 Pattern of cash flows through time over the 3 year time horizon of the investigations for Cases 1, 2, and 3 275 Table 7.2 Change In Net Worth Over Time Financial Variable Case 1 Case 2 Case 3 Initial--T = 0.0 Value of cattle 32000 92500 64900 Value of feed stocks 21060 21060 21060 Working capital 45000 45000 45000 Long term debt 62100 122600 95000 Short term debt 0 0 0 Net Worth of liquid assets 35960 35960 35960 Final-—T = 3.0 Value of cattle 80200 83000 83050 Value of feed stocks 20400 18220 25070 Working capital 43370 13330 —7l60 Long term debt 0 50600 23000 Short term debt 41200 46400 29000 Net Worth of liquid assets 102770 17550 48960 Change in net worth +66810 —l7510 +13000 L common level of net worth of assets and liabilities. The widely varying herd valuations were offset by different debt levels which would represent the borrowing necessary to acquire such herds. Overall net worth at T = 0.0 was set at $35,960, exclusive of land related indebtedness and valuation. After three years, Cases 1 and 3 have increased net worth while Case 2 has suffered a significant drop in net worth. Cattle and feed stocks were valued at the appropriate prices in effect at T = 3.0 to obtain these end of project figures. Due to the annual cattle and crop price patterns used throughout this dissertation, the prices in effect at T = 3.0 happen to be identical to those in effect at T = 0.0. Case 1 had such a large increase in net worth because the herd increased dramatically in value-~animals which were initially valued as replacement heifers were valued at the conclusion of the project as mature breeding cows. Overall debt shrank in all three cases, but the rapid repayment of long term debt was offset by large increase in short term debt. The short term debt levels are a result of the negative cash flows which have had to be funded through borrowing. An overall combination of cash flow and changes in net worth gives a rather gloomy picture for all three development cases. Case 1 has an overall undiscounted present value of —$34,790, while Case 2's value is —$l69,210 and Case 3’s is -$l30,200. Clearly Case 1 is the best of the three tested by a wide margin. Therefore, these tests would conclude that purchase of bred heifers is the best choice from among these three. However, the fact remains that even Case 1 lost nearly 277 $35,000! Only land speculation or desires for tax losses to offset other taxable income would induce an investor to make this investment under the conditions tested with these runs. Profit Maximization The final demonstration of the capability of the simulation model to evaluate and assist in decision making will be a search of selected control values and strategies to maximize profits for a given breeding herd size. Here the breeding cow population will fluctuate around 200 animals with appropriate replacements to maintain this herd size. Three areas of potentially significant financial effects will be investigated: use of artificial insemination versus natural breeding, sales of weaned calves or retention for later sales as yearlings, and herd feeding plans. Monetary considerations of alternatives will be supplemented by other criteria when appropriate in selecting among possible control value sets. Cattle and crop prices will be assumed to fluctuate on an annual cycle as illustrated by Figures 7.1 and 7.2. The weather variables will be assumed to follow the pattern given by Figure 7.3——a plot of solar radiation levels over the calendar year. Property taxes are assumed to be $20 per hectare per year, and due in equal installments June 30 and December 31. Total land area allocated to grazing is 660 hectares divided into five land parcels. No land is leased. Small quantities of shelled corn, rye grass, and oat silage are produced for winter consumption. The decision to select either artificial insemination (AI) or natural breeding over the other has elements of financial cost, breeding 278 cow conception rates, and genetic change involved. This enterprise simulation model excludes genetic factors completely. Currently, the overall conception rate from AI is the same as natural breeding as long as the onset and duration of breeding are equal. This leaves the direct financial costs as the only viable means of making comparisons between these breeding methods. Two model runs were made to investigate the financial effects of the method of breeding. One had a bull/cow ratio of 1:20 with no AI used, while the other used AI exclusively with zero bull population. Both runs covered the time interval from T = 0.0 through T = 1.0. Table 7.3 summarizes the results obtained. Use of natural breeding is superior in terms of financial cost by a margin of 4,900 dollars. AI supplies and labor are a large expense which more than exceeds the feed costs and replacement costs of maintain— ing a bull herd. AI supplies were assumed to be priced at $10 per ampule, which is a typical value. In actual practice there might be the additional factors of different cow and heifer conception rates and genetic changes to modify these findings. However, the financial costs of the two methods of breeding clearly favor natural breeding; enterprise profit maximization implies the use of natural breeding. Disposition of weaned calves is a question which cattle ranchers face yearly. Should they be sold after weaning, or retained for sale later as yearlings? Immediate sale had the advantage of lower winter feed requirements, less labor in general animal care and feeding, and a known current market price. Retention for later sale speculates that the increased revenue of selling heavier animals will offset the 279 Table 7.3 Breeding Method Cost Comparison Costs of Costs of Factor AI Natural Breeding Breeding ($) ($) Labor and AI supplies 10,980 ——— Bull feed costs -—- 1,270 Bull replacements ——— 4,800 Total 10,980 6,070 costs of keeping these slaughter cohort animals. Three separate runs were made to compare the overall financial effects of (1) sale immediately following weaning, (2) retention of female heifers not selected as breeding replacements, and (3) retention of all weaned calves. All animals retained for later sale were sold at T=l.15 when yearly cattle prices Were at their peak; animals sold at that time were from 10 to 12 months old. Table 7.4 indicates the results of these three runs. Total cash flow and net profit are reported for the period (0.0, 1.15). Additionally the values of the feed stock inventories at T=l.20 have been compared and the difference of the two retention runs from the base given. Finally, total cash flow and net profit are adjusted to include the feed stock inventory valuation differences. These adjusted values are a true basis for comparison among the three runs. As the 280 values of adjusted cash flow make clear, the more weaned animals that are retained the worse off the enterprise becomes. Under the feed plans used, and the basic price cycle specified, the profit maximizing operator would sell all weaned calves not retained for replacement of breeding cows and save none for later sale. Table 7.4 Financial Effects of Weaned Calf Retention Policies Factor Sell All Sell Males Sell None ($) ($) ($) Total Cash Flow +4770 +2790 +100 Total Net Profit -5980 -7960 —10650 Feed Inventory Value 0 -3620 -1850 Total Adjusted Cash Flow +4770 -830 —1750 Total Adjusted Net Profit -5980 —11580 ~12500 The final aspect of profit maximization to be investigated in this thesis is the area of cattle feeding plans. Feeding of cattle is a very significant part of the overall cost of a cattle enterprise and offers room for financial savings if feed reductions can be made. The simula— tion model is well suited to these investigations because of the great flexibility which exists in the quantity and quality of feed that can be allocated to the individual herd cohorts over time. For example, the model user can specify up to four different feeding plans through time and control the time at which allocations shift from one to another. Since these plans are changeable at most decision points the time interval 281 during which each feeding plan is in effect can be quite short. Within a particular feeding plan the user can specify the quantity of TDN he wishes allocated to each cohort's animals for both concen- trates and roughages per day. The units of the control variables of the feeding plan are kg TDN/animal/day. Further, the user can specify what fraction of the roughage and the concentrate allocation is to be obtained from what feed stock. This model organization gives the model user nearly as much flexibility as an actual enterprise manager in feeding cattle. A series of eight simulation runs of one year's duration was made, attempting to reduce feed allocations withOut hurting cattle performance. The base run had a series of six feeding plans cover- ing the entire one—year interval drawn from rough adherence to rec— ommended guidelines. Two different sources caused misallocations where improvement was obviously attainable. First, several instances occurred in which the allocation to animals of a particular cohort was more than they would consume. This excess is considered totally wasted in the model (see section V.4 for details). Second, several other instances occurred in which the allocation to animals of a cohort gave a projected daily weight gain rate which conflicted with the weight maximum constraining that cohort. In such instances any daily gains exceeding the weight maximum are disallowed (see section V.l for details). Thus, the immediate goal was to eliminate these sources of excess feed consumption from the feeding plans. Table 7.5 indicates the results of the eight model runs over the time interval (0.00,1.25). Run 1 is the base from which improvements 282 l11l‘l\l\1\\n\ 7‘4 n.mw+ N.o~+ o.mm+ w.m+ m.m+ s.aa+ 5.0+ . mama Hm>o unmam>0HaaH N .s . . .o Osman enema Osaka Osman seams QAHNH amass oases m + a ommN+ oste+ omms+ o~o~+ ossa+ oo~a+ oss+ I mama tm>o mucmnwcuae .m .m Osaka oaoo ommaa omma ommm eases amass cameo scam ammo Hmseea .s omasm osmmm ommem oNNsN osmma oomNN osoaa oosam usam> sucuam>ea .m oaaam oaaam oaaem owoqm omoqm oHamm comma ohomm mmmaam capo camew osmea cases oomoN oomom osmoa osaow omoma owmaam umm ossmm oammm owsmm oqmma oawsm owesw oeemm cases away» a owswm comma omoos cease omoos omoos oomaq ommse Hmaa cmmnsom oases cacao oooqm cease comes owoos cease cause coco amassew ceases owamwa osaama oasama omwmoa omosma ommama camama mmwmuom Aoénavhoucmtfifi comm .N ONNHM oNNHm emaam Canes oamss oases omams cease mwmaaw coco onmaa ommsa ommsa oNsou omqou oesoN oomou osmom wmmaam use canoe omaoq omaoq ooamm oommm swans ommma ommma mmmum may amass cocoa osoam amass owes“ oases oases oqams same emuasom oamws cameo omswm caswe comma oases oases ocean :uou sesamsm enemas osomaa ossmaa csaasa enemas cameos commas oewasa mmwwtoo soumuoaaa some .a Inn w cam n cam m cam m cam a and n max N cam Amwmnvfl cam cowumfluomwa ; unwam mdflvomm o>HumnuouH< mo muoommm Hmfioamcwm n.n manmh 283 were to be made. Run 2 increased the mature cow allocations during calving to attempt more rapid weight recovery, while also decreasing allocations to replacement heifers and mature bulls. Both these latter two cohorts were being allocated excessive feed quantities; heifers were allocated more than they would consume, while bulls were fed so much that the resulting daily gains took them beyond the weight constraint of 600 kg. A small improvement in the overall financial measure of Run 2 was shown——approximately 6.7% better than the base run in terms of the combination of annual cash flow and the excess feed inventory on hand. Run 3 attempted to the correct the problems still remaining after Run 2 by further decreasing replacement heifer feed allocations in the early part of the year, and by reducing the allocations of concentrate to bulls throughout the year. These changes resulted in another small increase on overall profitability as total feed allocated was reduced. An 11.7% improvement from the base run was achieved. Runs 4 and 5 departed significantly from the feeding plans of the first three runs. Early year cow and heifer pre—calving feed rates were increased to reflect the common practice of boosting energy intake prior to calving. Run 4 also added a schedule of feeding calves supple— mentary concentrates during the later summer months. These increased concentrate allocations are reflected in the higher concentrates consumed values in Table 7.5. The overall financial result was still slightly better than the base run, by 3.5%, but a drop from the previous runs. Run 5 eliminated the summer concentrate supplements to calves of Run 4, but was otherwise identical to it. Its overall financial result was 284 a gain from the results of Run 4, but still worse than Run 3. A 9.8% improvement over the base run was obtained. Run 6 made another significant change by extending the grazing season, where no supplementary feeding is made, from T = 0.70 through T = 0.80. This latter time value is the beginning of the wintering season as defined with the use of the variable TFALL. This increase in grazing period is significant because of the large number of cattle which are present in comparison to the rest of the year. Calves are able to eat quite a lot with full weaning being only a short period into the future (TWEAN = 0.85). Elimination of this period of feeding saved expensive feed stocks by substituting forage which is relatively low in cost. The overall financial effects were startlingly good—— a 58.6% improvement over the base run. Runs 7 and 8 use Run 6 as a base and differ from it by increasing or decreasing post—calving cow concentrate supplementation, respectively. Observation of the overall concentrate consumptions listed in Table 7.5 shows roughly 10,000 kg increases in Run 7, and 10,000 kg decreases in Run 8, from the base levels of Run 6. Overall financial effects reflect these levels of change in concentrate consumption, with Run 7 dropping to only 26.2% better than the base run, and Run 8 increasing to 83.5% better than the base run. The above discussion of the effects of alternative feeding plans has concentrated on the financial effects without mention of some more subtle feeding effects. One of these effects is the condition of the breeding cow herd as related to calving, lactation, and rebreeding. Table 7.6 gives an idea of the feeding plan effects on cow condition 285 by presenting ranges of weights of cohort l cows (at T = 1.0) for each of the eight runs. Three basic groupings may be made from the eight runs: (1) Runs 1 and 8, (2) Runs 2, 3, 4, 5, and 6, and (3) Run 7. Runs 1 and 8 presumably have comparable levels of breeding cow condition as the weight ranges are very similar. Run 8 is vastly better than Run 1 financially, so it would be preferred to Run 1. Runs 2 through 6 have very similar weight ranges implying a uniform cow condition among them, while Run 6 has much better financial outcomes than the others. Run 6 should then be preferred over the others. Finally, Run 7 has a somewhat better condition than either of the other two groups. Choice of the best overall run of these eight is difficult because of the inverse relationship between the two measures of outcomes. Figure 7.7 plots these conflicting results. Decisions about what is preferable among these groups of runs—-numbers 6, 7, and 8—— must be made on the basis of the weights assigned to the two measures of performance. If the weight differences are relatively unimportant, then Run 8 is best. If the weight differences are very important, then Run 7 is best. Perhaps Run 6 is best under moderate importance of both measures. The decision maker must decide for himself. Profit maximization (or loss minimization) is an open ended pro— position. This demonstration of the capability of the model has touched on three areas which seem likely to be important to achievement of the goal. Of course, many additional aspects have not been examined at all. The results of these model runs do tentatively support the proposition that profit maximization implies natural breeding not AI, sales of weaned 286 calves immediately not retention, and a tight level of feeding. ‘Run 6 is probably characteristic of such feeding levels. Other conclusions could possibly be drawn here if the assumptions about price, weather, and crop production change significantly. Additionally, a more sophis— ticated model which includes an improved model of the energy intake/ reproductive potential interactions might obtain somewhat different and more conclusive results with regard to the effects of the feeding plan alternatives discussed here. Table 7.6 Mature Breeding Cow Conditioning Effects From Alternative Feeding Plans Weights at T = 1.0 Run Average Minimum Maximum (kg) (kg) (kg) 1 386 358 419 2 405 372 440 3 406 371 440 4 406 370 440 5 406 370 440 6 396 367 440 7 417 380 443 8 385 351 [ 423 287 Financial Rank best . Run 7 middle . Run 6 worst . Run 8 worst middle best Condition Rank Figure 7.7 Plot of the rankings of runs 6,7, and 8 according to financial ranking and condition ranking 288 V11.3 Summary Three demonstration examples have been presented to illustrate the capability of the simulation model to assist decision makers in ongoing operational decisions and in investment planning. The question of early versus late weaning was decided in favor of late weaning. 0f the three paths tested for achieving a steady—state breeding cow herd of 200 animals, the best method was to begin with 200 bred heifers and generate herd replacements internally. Profit maximization turned Out to imply natural breeding as opposed to artificial insemination, sales of all weaned calves immediately following weaning rather than retention for later sales, and a tight level of herd feeding. All of the conclu- sions determined through these demonstrations are tentative, because the purpose of these investigations was merely to illustrate the capa— bility of the model, not to come to a definitive conclusion about a particular decision rule of management. Confirmation of these findings could be achieved through more extensive testing with the simulation model in conjunction with improved data. The deterministic nature of this model and its output variables could be modified to allow running in a Monte Carlo mode of operation. This would provide confidence intervals on the Output variables as related to randomness in the various variables of this system. For example, the weather variables used in the runs at present are deter- ministic not random as they are in nature. Forage growth should, therefore, be a stochastic variable responsive to the random variable values of solar radiation, average temperature, and rainfall. The financial variables of the model would then be stochastic in response to different levels of forage growth for any given management policy. 289 Confidence intervals could be determined around the financial outcomes indicating the degree of fluctuation of them from weather variation. Managers might prefer policies which while less attractive on the average have less variation around the mean. The preferences of the manager toward uncertainty could then enter into his decisions regard— ing investment and management policy. These three demonstration examples do confirm the usefulness of the simulation model in assisting management decision makers. Both ongoing operating decisions and project investments of a realistic nature can be evaluated with the model in its current form. A tool such as this model should allow decision makers to perform better since a much wider range of alternatives is open to analysis and objective comparison. Information is generated which the manager can use to significantly reduce the risk and uncertainty which are associated with particular management strategies. While the development of this model is costly, the operating expenses of using it are extremely small in comparison to the potential advantages that can be gained from better responses to economic forces and the environment. Many questions of significance could be answered in a day of programmer time and $50 worth of computer time, as measured on the MSU CDC 6500. The great flexibility of the model should make it of interest to many different users. CHAPTER VIII SUMMARY AND CONCLUSIONS This thesis will end with a summary of the preceding seven chapters, the conclusions to be drawn from the development and use of the model, further extensions and improvements to the model, and implementation of the model. VIII.1 Summary The preceding chapters have discussed in great detail the development and use of a system simulation model of a beef cattle enterprise for the purpose of investigating alternative management decision making strategies. Chapter IV developed the general de— scription of the modeling approach that has been taken, while Chapter V discussed the details of the mathematical model as implemented in the FORTRAN subroutines listed in the User's Guide to the Beef Cattle Enterprise Simulation Model, a separate volume from this thesis. The details of the growth prediction and reproductive impact component can be found in Schuette [48]. This simulation model represents a strong attempt to fully model the time dynamics and reproductive dynamics of a beef cattle herd; this is an area which has been ne— glected up to this time. Furthermore, the intent of this thesis has been development of a practical tool for decision makers which can be helpful in providing evaluations of alternative decisions. 291 A general model of a beef cattle enterprise which leans toward the land-extensive cow/calf form of operation has been viewed as con— taining five major system components. These are ..- cattle herd demographics—-a description of the herd at each point in time by age, sex, function, and, weight; forage growth component—-dynamic plant growth of forage as influenced by climatic variables and mechanical and animal harvesting of plant material; feed stock component——accounting for the quantity and quality of cattle feed stocks through time; nutrient impact component-~determination of the response of cattle to feed intake rates in terms of growth and of reproduc tive po tent ial; management decision-making component——the decisions and actions needed to manage the herd on both long- and short- term bases. Additionally there are several secondary components which determine the financial effect of actions, generate proper values of exogenous variables through time, etc. Figure 4.12 is repeated here to summarize the components of this system and the interconnecting variables. A key feature of the management decision making component is the distinction between decisions which the model itself can determine endogenously and those which must await exogenous control by the model user. Currently there are seven instances in which the model must stop and be restarted with appropriate control variable values determined by the user after study of the preceding simulation results and the current states of the system. These seven events are organized into five "deci— sion points" which correspond, in part, to natural cattle phenomena. These decision points occur when: mung—omega somzuwn, mBoam coaumfiuofifi was Hmoflmhsa wnfiumnuwSHHH Emuwmfio Sway? . meHQHmud—W Nch wHSmH .h m05§0>0h HZHZOEEOU UVGFZDOUUAaV mumoo Ill/II comuaESncoo msumuw dsUZ¢EQS XUOHm IHBOmO ”35.0 k 0. 928522 l .0. noun a. moumu W com an0>ud£ MU