THESfii 113mm mchigan Stat. University This is to certify that the dissertation entitled INFORMATION VALUE USING VARIABLE PRECISION DATA TO DELINEATE NHEAT EXPANSION AREAS IN SYRIA presented by Scott G. Witter has been accepted towards fulfillment of the requirements for Ph.D. Resource Development degree in qupmfs Date {49%;}; /¥; /7£L? MS U is an Affirmative Action/Equal Opportunity Institution 0- 12771 RETURNING MATERIALS: bViESI.) Place in book drop to LJBRARJES remove this checkout from .—_—. your record. FINES Will . be charged if book is returned after the date stamped below. Mada» I '1130 ‘QJ ‘iiztlilfiflfllt‘ 4179’ 00% (DijfllillLifii-Jlfiis f." INFORMATION VALUE USING VARIABLE PRECISION DATA TO DELINEATE WHEAT EXPANSION AREAS IN SYRIA By Scott George Witter A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Resource Development 1982 ABSTRACT INFORMATION VALUE USING VARIABLE PRECISION DATA TO DELINEATE WHEAT EXPANSION AREAS IN SYRIA By Scott George Witter For centuries wandering bands of nomads throughout Syria have created problems not only for the settled peoples, but also for the governing bodies trying to rule them. In the past these wandering bands of nomads have contributed little to Syria's overall economy. Current (1980) Syrian government goals for full employment, full agricultural land utilization, and the confinement of the nomadic move- ments in Syria required a complete inventory and evaluation of potential land resources available for agricultural de— velopment. However, two significant questions arise when inventory data are used to evaluate potentials for agricul- tural land development: "Howreliable is the information?" and "What will the consequences be of identifying crop expan- sion areas and potential yields using these data?". The purpose of this research was to investigate the value of information derived from using variable precision data to delineate potential wheat expansion areas. To es- timate the monetary effects of using data of variable preci- sion to forecast wheat yield (multiple regression model), a quadratic loss function was used. Loss function analysis was used to adjust the maximum potential gains obtained from the unadjusted yield model for the three Syrian study areas of Alleppo, Swedia, and Hasseke. Maximum potential dollar losses from a combined data precision error were calculated for series of actions and states. Minimum potential gross gains for each area by each action and state were estab— lished by subtracting maximum potential dollar losses from the maximum potential gains. The area with the highest minimum potential gross gain for all actions was Hasseke. To determine what the minimum potential net gain would be, a modified land rent analysis was used. Information value was established by measuring the dif- ference between the maximum potential gain (minus the total costs) and the adjusted minimum potential net gains (minus the total costs) according to the probability of occurrence based on a 20 year record. The information value equaled the potential loss from overestimating the net gains from a given expansion action. The estimated information value for the Hasseke area ranged from $2,721,776 for a 25% expansion to $7,687,099 for a 100% expansion into the available range- land during any given 1 year period. ACKNOWLEDGMENTS The author would like to extend a special thanks to Dr. Daniel Chappelle for his direct support and guidance throughout the author's program and dissertation work. Thanks is also extended to each member of the doctoral committee, Dr. Milton Steinmueller, Dr. Eckhart Dersch, Dr. Delbert Mokma, and Dr. Michael Chubb, for their guidance and comments during the writing of this dissertation. Special recognition is also extended to Dr. Ronald Shelton, Dr. James Johnson, John Putnam and Bill Enslin who, combined with the Syrian Government, gave the author the opportunity to travel and work in Syria. The deepest gratitude is extended to my wife, Brenda, whose hard work and continual support helped us to achieve all our original goals in ten years. Most importantly, how- ever, was her gift of our son, Gavin, who made meeting all goals truly meaningful. ii TABLE OF CONTENTS Page List of Tables . . . . . . . . . . . . . . . . . . . . V List of Figures. . . . . . . . . . . . . . . . . . . . vii I. Introduction . . . . . . . . . . . . . . . . . . 1 Problem Setting . . . . . . . . . . . . . . . 1 Land Assessment . . . . . . . . . . . . . . . 2 Purpose and Objectives. 3 Procedure 7 Research Hypotheses and Models. . . . . . . 9 Assumptions . . . . . . . . . . . . . . . . . ll II. The Study Area . . . . . . . . . . . . . . . . . 13 Agricultural Resources. . . . . . . . . . . . l3 Swedia. . . . . . . . . . . . . . . . . . . . l7 Hasseke . . . . . . . . . . . . . . . . . . . 21 Alleppo . . . . . . . . . . . . . . . . . . . 25 III. The Value of Information by the Establishment of Variable Precision Levels. . . . . . . . . . . . 29 The Loss Function . . . . . . . . . . . . . . 33 Variable Precision. . . . . . . . . . . . . . 38 Remote Sensing . . . . . . . . . . . . . . 38 Soils. . . . . . . . . . . . . . . . . . . AA Soil Moisture. . . . . . . . . . . . . . . A5 iii iv Page IV. Identification of Crop Expansion Areas . . . . . 57 Method. . . . . . . . . . . . . . . . . . . . 57 Soils. . . . . . . . . . . . . . . . . . . 57 Soil Moisture Storage. . . . . . . . . . . 61 Yield Equation. . . . . . . . . . . . . . . . 66 Final Yield Equation. . . . . . . . . . . . . 7A Swedia . . . . . . . . . . . . . . . . . . 7“ Alleppo. . . . . . . . . . . . . . . . . . 7“ Hasseke. . . . . . . . . . . . . . . . . . 75 Loss Function . . . . . . . . . . . . . . . . 78 Assumptions. . . . . . . . . . . . . . . . 79 Actions and States. . . . . . . . . . . . . . 80 Loss Function Comparisons Between Sites . . . 81 Probability of Moisture Occurrence. . . . . . 85 Monetary Risk . . . . . . . . . . . . . . . . 89 Expansion Site. . . . . . . . . . . . . . . . 90 Cost Versus MPGG. . . . . . . . . . . . . . . 91 Action Versus Budget. . . . . . . . . . . . . 9“ Information Value . . . . . . . . . . . . . . 95 V. Summary, Conclusions and Recommendations for Future Research. . . . . . . . . . . . . . . . . 98 Information Value . . . . . . . . . . . . . . 101 Recommendations for Future Research . . . . . 101 Appendix A . . . . . . . . . . . . . . . . . . . . . . 10“ Appendix B . . . . . . . . . . . . . . . . . . . . . . 105 Bibliography . . . . . . . . . . . . . . . . . . . . . 11A 10. 11. LIST OF TABLES Page Estimated Soil Category Purity and Associated Mapping Scale. . . . . . . . . . . . . . . . “5 Swedia Test Site-—A Comparison of Soil Areas Being Used for Rain Fed Agriculture and Rangeland (by Montika) . . . . . . . . . . . . . . . . 58 Alleppo Test Site--A Comparison of Soil Areas Being Used for Rain Fed Agriculture and Rangeland (by Montika) . . . . . . . . . . . . . . 59 Hasseke Test Site--A Comparison of Soil Areas Being Used for Rain Fed Agriculture and Rangeland (by Montika) . . . . . . . . . . . . . . . . 6O Yearly Wheat Acreage (hectares) Correction Pro- cedure for the Swedia Montika. . . . . . . . . . . 62 Mean Precipitation, Adjusted Potential Evapotrans- piration, and Soil Moisture Storage Values for 1955- 1969 for Wheat Growing Period by Study Area. . . . 6“ Percentage Differences From the Mean Region Response for the Selected First Order Stations by Month, Precipitation, Adjusted Potential Evapotranspiration and Potential Soil Moisture Storage (1955-1969). . 67 Mean Monthly Precipitation, Adjusted Potential Eva- potranspiration, and Soil Moisture Storage Values for 1970- 1977 for One Representative Station in Each Test Site . . . . . . . . . . . . . . . . . 68 Multiple Regression Wheat Yield Estimates for Swedia, Alleppo, and Hasseke for 1970-1977 Storage and Soils First Run. . . . . . . . . . . 71 Multiple Regression Wheat Yield Estimates for Swedia, Alleppo, and Hasseke for 1970—1977—- All Variables-—Second Run. . . . . . . . . . . . . 73 Multiple Regression Wheat Yield Estimates for Swedia, Alleppo, and Hasseke for 1970-1977-- Selected Variables—-Third Run. . . . . . . . . 76 l2. 13. 14. 15. 16. 17. 18. 19. vi Quadratic Loss Function Analysis of the Swedia Test Site Showing Maximum Potential Gain, Maximum Potential Dollar Loss, and Minimum Potential Gross Gain . . Quadratic Loss Function Analysis of the Alleppo Site Showing Maximum Potential Gain, Maximum Potential Dollar Loss, and Minimum Potential Gross Gain . . . . . . . . . . . . . . . . Quadratic Loss Function Analysis of the Hasseke Site Showing Maximum Potential Gain, Maximum Potential Dollar Loss, and Minimum Potential Gross Gain . . . . . . . . . . . . . Minimum Potential Gross Gains Adjusted by the Probability of Moisture Occurrence for the Swedia Site by Each State and Action. . . . . . . Minimum Potential Gross Gains Adjusted by the Probability of Moisture Occurrence for the Alleppo Site by Each State and Action. . . . . . . . . Minimum Potential Gross Gains Adjusted by the Probability of Moisture Occurrence for the Hasseka Site by Each State and Action. . . . . . . Minimum Potential Net Gains (MPNG) Minus the Variable and Analysis Costs for the Hasseke Site Information Value Using Variable Precision Data to Delineate Wheat Expansion Areas in the Hasseke Test Site. Page 82 83 8A 86 87 88 93 96 NOW-1:00 (I) LIST OF FIGURES Syrian Test Sites. Precipitation Patterns in Syria. Euphrates Reclamation and Drainage Projects. Soils in the Swedia Test Site. Soils in the Hasseke Test Site Soils in the Alleppo Test Site Interactions Between Decision, Information, and Data Systems Monthly Water Budget of Alleppo, Syria Average Daily Potential Evapotranspiration as Es- timated by Lysimeter Growing Deep—Rooted Grass- Legumes and as Computed by Thornthwaite, Penman, and Blaney-Criddle vii Page l5 l6 l9 23 27 30 A7 A9 CHAPTER I INTRODUCTION Problem Setting Nomadic peOples have wandered throughout Syria for thousands of years. These peoples have had no permanent residences and have lived almost exclusively in tents as they roamed the deserts looking for water and pasture for their animals. Most of these groups have been governed only by tribal laws, tribal chiefs, and tribal courts. Be- cause no centralized governing body has had complete control over these groups, Arab history has been filled with hosti- lities between the settled peoples and the nomads. "In the nineteenth century, the Ottoman [Empire] began a program of agricultural development and forced settlement of nomadic groups" (Bates, 1971, p. 121). Tribal leaders were offered large tracks of land to settle their people, however, little was known of the ability of these land tracks to provide sufficient subsistence to keep these people set— tled. As a result, it was continually necessary to use military force to keep these people on their designated land. Problems between the tribes and the Ottoman peaked during World War I when the Arab revolt took place. The revolt enabled several of the larger tribes to gain indepen- dence from the Ottoman. This new-gained freedom, however, was short—lived as the tribes were not able to contend with the better—equipped French army which controlled Syria until just after World War II. The current Syrian government has sought to establish a clearer policy for dealing with the nomads. Because an immediate solution, other than military, does not exist, a series of small-scale social experiments are being under- taken. These experiments include better animal husbandry, establishing permanent water wells, and the development of new strains of drought-resistant crops to be grown on cur- rently non-productive lands. To meet the data needs for these experiments and the goals of the fourth Five Year Economic and Social Development Plan, full employment and full agricultural land utilization, it is necessary to first inventory and evaluate land resources available for agricultural development. Land Assessment Agricultural scientists have for decades conducted in- vestigations aimed at classifying land areas and climatic conditions into information systems suitable for the pre- diction of potential crOp yields. Agricultural land classi- fication relies heavily on the technological meshing of a number of disciplines: a soil scientist provides soil boundaries and chemical property breakdowns; a meteorologist provides data concerning temperature and moisture regimes; a remote sensing specialist uses repeating tonal and textural patterns as keys to agricultural use; an agronomist provides information concerning the most adaptable species of crop; an agricultural economist determines supply and demand trends associated with current and future monetary returns. These approaches pose two significant questions: "How reliable is the information?" and "What will be the conse- quences of identifying crop expansion areas and potential yields using these data?". Purpose and Objectives Risk is involved in every decision. The way people react to this risk, however, can be extremely varied. Some individuals try to avert risk while others are either neutral or risk seeking. Information about a given situation tends to lessen or eliminate some of the inherent risk involved with decision making. In general, the greater the quantity and quality of information, the less risk involved. To most adequately deal with the risk involved in de— lineating crop expansion areas, the decision maker should specify the precision level of the data needed to make a par- ticular decision. As a result, he or she has defined the amount of risk he or she is willing to take. The precision level used, then, places a value on the information by speci- fying a sample size and by placing a percentage estimate that can be related to the potential monetary losses or gains associated with a given decision. One problem is that most decision makers do not have sufficient information to specify precision levels which will yield the desired results. Statistical decision theory, through the use of a loss function, provides a means for determining potential losses from using variable precision data in the decision making process. The loss function al- lows the decision maker to assess or alter a common utility function1 to more realistically reflect the outcome of a given decision. Three Syrian test sites will comprise the study area (Figure 1). To illustrate the role of the loss function in assessing potential wheat expansion areas for domestic use, three study objectives were identified. The objectives are: 1. To identify potential areas suitable for wheat produc- tion within the Syrian test sites using remote sensing techniques, water balance equations, and the most cur- rent soil data. 2. To establish the monetary risks of making wheat yield estimates using variable precision data with a quad- ratic loss function.2 3. To compute the monetary gains or losses associated with crop expansion into the test site with the highest potential return using 1977 dollar values reported in the 1980 Syrian Agricultural Sector Asses— ment. The data and spelling of place names used in this study come primarily from the ]978-l980 Syrian Agricultural Sector Assessment. The author was a member of the Comprehensive lUtility Function refers to a group of individuals' choices aggregated into one common choice or goal. 2Quadratic Loss Function is used to determine the Optimal estimates of a central value based on past mean values. C) Syrian Test Sites. Figure 1. 2‘— 54.4% /.,!32=o=x 9 .lIJ..I.IqII.. . 00 O _0050_ 2.0.255. mmtw hmmh Zm «Ear. «Sax. LON am :0: 6 Ch- O>~C 00. UL‘Q 3“ uoou< 33 v .nao:< >325... oxoooa: Resource Inventory and Evaluation System (CRIES) research team which was responsible for land cover/use, soils, and economic analysis during the assessment. CRIES was designed to explore basic questions about agricultural resource plan- ning. The agencies involved at the time of the assessment were the United States Agency for International Development (AID), the United States Department of Agriculture (USDA), the Economic Surveys and Systems of the USDA, and Michigan State University (MSU). CRIES was designed around two general objectives (CRIES, 1980): 1. To apply a consistent approach to land resource clas- sification that is adaptable to many countries and suitable for agrotechnology transfer, and 2. To provide the training and technical assistance necessary to classify and inventory resources, to evaluate crop adaptability and productivity, and to assist in developing food strategies in participating countries. CRIES personnel work closely with in-country representa— tives from the host country to inventory and classify land resources, determine present land use patterns, determine current potential agricultural production, establish computer- based information systems, and determine important socio- economic characteristics that might effect future agriculture production. Every effort is made to fully utilize all existing data sets and where necessary aid the collection of the primary data needed to meet each countries' project goals. Procedure A suitability classification procedure was used to select land areas within each test site that were not presently being used for crop production, but would be suitable for wheat pro— duction. The initial phase of the study combined the classi- fication of current (1978) land cover/use patterns Within the three test sites. The three areas selectedvmnwaSwedia, Alleppo, and Hasseke. Certain montikas (county level) were eliminated from the Alleppo and Hasseke Mohafazas (state level) to main- tain a closer approximation of site size and cropping pattern. These areas were selected because of their similarities and dissimilarities. Similarities among the sites were area size, cropping patterns, reliability of published statistics, and availability of areas for crOp expansion. Dissimilarities included rainfall, crop yield, and soil types. The second phase combined soils data at the suborder and great group level with data collected in phase one. Cross tabulations between soils and land use data were run to establish which soils were producing the rain fed crops. Pro- cedures for soil assessment are discussed in Chapter IV of this work. The Thornthwaite (1955) water balance procedure was used to establish past trends in soil moisture availability during each month of the growing season. Actual evapotranspiration (APE) figures derived from the Thornthwaite equation were re- placed with APE figures calculated using the Penman (Safely, 197A) equation as the Penman equation is more precise in dry climates (Chapter III). Each station was plotted by the soil unit and the montika it was located within to establish their regional representation. Each data set was overlayed, both cartographically and digitally, to determine potential wheat expansion areas. The mapped information provided the location data while the digi- tal data provided area estimates and category composition, by percentage, for each variable within expansion areas. An expansion area was defined as an uncrOpped area (rangeland) capable of supporting a wheat crop. The expansion areas were then ranked within each test site according to the amount (kilograms/hectare) of potential wheat each could produce. The highest ranking went to those areas capable of producing the largest crop the highest percentage of the time. A dol- lar value was assigned to each potential wheat crop by mul- tiplying the estimated kilograms per hectare of wheat by the 1977 government supported prices for wheat in Syria. Because each step of the procedure, as well as the yield forecast, has a potential error, it was not feasible to as- sume that all of the expansion areas' yield could be identi- fied. A loss function equation was used to estimate what the absolute minimal gain would be. Potential losses resulting from using variable precision data were calculated using a quadratic loss function. Weight- ing values for the loss function components were derived as composite values from published situations. Maximum potential losses from making a decision to expand wheat production into a region were derived by combining the effects of the compounding variable error (kg/hectare x $ value) plus the cost of conducting the study. The minimum potential net gains were derived by subtracting the total variable cost plus analysis cost from the minimum potential gross gain. The possible expansion areas were then ranked according to their potential return. Because each decision maker within the Syrian Agricul- tural Sector might have a different utility function, it was necessary to aggregate them and consider the entire sector as one client. The common utility function was as— sumed to provide an estimate of the utility associated with identifying potential wheat expansion areas to increase wheat revenue with the highest potential return and at the least risk. The potential Syrian farmers were also considered as one client to eliminate the problem of adding utility func— tions over individuals. The farmers' utility function was assumed to provide an estimate of the utility associated with identifying potential wheat expansion areas to increase wheat revenue with the highest potential return and at the least risk. Research Hypotheses and Models Two primary models were used during the study. The first was a linear regression model used as a traditional yield model The yield model is: 10 i = lel + b2X2+b3X3+Uu = wheat yield in kilograms/hectare y y b = regression coefficient X = millimeters of potential available soil moisture X = acreage of a given soil within a given wheat producing region by study site X = acreage of rain fed wheat within each study site U“ = error term The associated hypotheses are: Null hypothesis HO: wheat yield for Syria can— not be estimated with this model at a 90% confidence interval Alternative hypothesis H wheat yield for Syria can be predicted with this model at a 90% confidence interval 12 The rationale for using a linear regression yield model with a 90% confidence interval was to obtain a yield esti— mate similar to that considered acceptable by the Large Area Crop Inventory Experiment (LACIE). This allows a direct com- parison with a major ongoing international program which does not take into consideration the potential estimation error involved with variable precision data. To determine what the maximum dollar (Syrian) loss that might be experienced by utilizing this model to estimate yield in expansion areas, the following loss function equation was used: Maximum Potential Loss = [(YP) l-(X b °X b2°X3b3°Xubu)2-(V)] l l 2 yield model probability YP X b l 1 soil moisture (mm) availability times soil moisture estimate precision ll X2b2= soil category classification area (hectares) percentage times the soil boundary estimate of precision X2b2= Landsat delineated range area (hectares) per- J 4 centage times the classifying precision level of the Landsat scanning system Xubu= wheat yield estimate in kg/hectare times as- signed precision level of Syrian governmental statistics V = the 1978 Syrian value of a kg of wheat This procedure was also meant to be hypothesis genera- ting from the standpoint of estimating the reliability of yield information and in determining economic consequences of using variable precision data in yield modeling. Assumptions The major assumptions of this study were: 1. Acreage and yield statistics reported by the Syrian government were 95% precise (an estimate based on a re— view of variations in reported statistics). 2. Syrian cultural and/or farming practices would not change yield characteristics from one study site to the next. 3. Predetermined precision levels assigned to each variable were representative (based on Chapter III). A. Syrian recorded moisture and temperature values were 90% precise (refer to Ward, 1967, catchment error). 5. Adjusted evapotranspiration corrections calculated for each station were 92% precise (refer to McGuinnes and Bordre, 1972, p. 15, for comparisons of Penman and ly- simeter values). 12 The assumptions made in this study represent a composite of those conditions necessary to utilize a quadratic loss function to study the monetary effects of using variable precision data to forecast wheat yield. Precision levels of the data used and the representativeness of the information values established based on these assumptions may vary dra- matically if used under different physical settings and scales of investigation intensity. However, they are con- sidered to be necessary for this study and representative of the conditions found in Syria by the author. CHAPTER II THE STUDY AREA "Over the centuries, man has shown great ingenuity in using climate, soil, and other agricultural resources of the Middle East" (Clawson et al., 1971, p. 1). Syria is no exception. Vast portions of the country receive less than 200 mm of rain per year. Limited rainfall combined with only scattered reservoirs of ground water greatly limit the development of agriculture in Syria. This is extremely im- portant to a country where "65 percent to 75 percent" (Lieftinck et al., 1956, p. 7) of the population derive their livelihood directly from agriculture. Lieftinck et a1. (1956) found that crop production and livestock production accounted for between “5% to 50% of Syria's national income. This situation has not changed appreciably. Syria, which does not share its neighbors' oil wealth (Iran and Iraq),rmnfi:rely heavily on a limited agri- cultural base, therefore, the reliable assessment of natural resources is important. Agricultural Resources Two of Syria's principal agricultural resources are un- doubtedly land and water. 0f Syria's approximately A6 million acres of land surface, approximately one-half is mountains, rocky areas and desert. Approximately 15 million acres of land have sufficient rainfall to support crops without the 13 1A aid of irrigation. Refer to those areas receiving average annual rainfall of 200 mm or more in Figure 2. The coastal regions receive rainfall in excess of 600 mm and are considered a Mediterranean climate. Numerous orchards are found in the coastal region which produce oranges and olives. General soil usage patterns in the coastal region are: l) the best soil is used for citrus and cereal grains; 2) the next best for olives; 3) the third best for refores— tation of pine; and A) the remaining areas have scrub oak and assorted types of brush. These priorities are based on discussions with the Agricultural Director of the Homs region during December of 1978. In the areas experiencing precipitation ranging from A00 to 600 mm, wheat, barley, chickpeas, and assorted orchard crops are produced. The region extending from south of Homs to just north of Alleppo is one of the most productive soil regions in all of Syria. The areas on Figure 2 representing between 200 mm to 300 mm of precipitation are devoted primarily to the produc— tion of wheat, barley, and highbred olive trees. The domin- ate soils in these areas are Entisols and Aridsols. The arid conditions found throughout the remaining por— tions make the development of better moisture monitoring techniques and irrigation systems essential. The major watershed found in Syria is the Euphrates River (Figure 3). Historically, the Euphrates River has flooded annually, covering its banks with fresh layers of silt. When the Tobuqa l5 Precipitation Patterns in Syria. Figure 2. «OE-=2: :- 00.5 = oouuoopus com-occue oo . . 3.3?» 0 can a coopucconu Quiccnlm . 000 r ..o>OI _. <_m>m 2. mzmmhhZ 28 Elevation in the western portion of the Alleppo site ranges between 350 meters (south) to 650 meters along the west-central section. Temperatures are slightly cooler than the eastern section, ranging from 15° to 17°C, while preci- pitation is comparable. The major non—irrigated crops in- clude wheat, barley, olives, corn, figs, pistachios, and assorted types of fruit trees. Small fields of well irrigated lettuce, cabbage, tomatoes, beans, cotton, and poplars are also found throughout this section of the Alleppo site. The major soils in this section are Xerorthents (Entisols in the Orthents suborder). These soils are primarily shallow, less than 50 cm, and are characterized by high clay content. Because of the high clay content, surface cracking occurs during dry periods. These soils are not very permeable and tend to have low moisture holding capacities. CHAPTER III THE VALUE OF INFORMATION BY THE ESTABLISHMENT OF VARIABLE PRECISION LEVELS "Regardless of the investment criterion used in evalua- ting the worth of information (e.g. alternative criteria in- clude present net worth, internal rate of return, benefit- cost ratio), we must always be concerned with two types of factors: (a) the costs of acquiring information and (b) the benefits that accrue from having the information" (Chappelle, 1976, p. 1A5). "Information only acquires value in the con- text of a decision, i.e., the use of information in economic decisions determines its value (Arrow, 1962, p. 615). The more the risk and the greater the potential return, the more valuable the information becomes. Information problems and value of the data depend on the identification of the vari- ables to be included in the information system and on how much data should be collected concerning these variables. "One general rule based on economic reasoning is that we should collect data until the marginal cost of the informa- tion is equal to the marginal benefit generated by the infor- mation which is developed from the data" (Chappelle, 1976, p. 1A5). To determine information cost it is necessary to sum unit costs of each input required, plus the costs of the analysis. Chappelle (1976) graphically illustrates the flow of information, inputs and products (Figure 7). 29 30 Information, and Figure 7. Interactions Between Decision, Data Systems. f" —————————————— 1 : INFORMATION SYSTEM : I In formation I DECISION I INFORMATION I SYSTEM (“T Retrieval STORAGE I I I A ' : L __ __ _ ____' Processing : I Specification I Analysis I I 0/ Decision Variables I Evaluation l l (Feedback Loop) : Hedge/m” I I I. Interpretation 1 __.__I'_ __________ "Z. ZZZZZZZZ‘ r- :I/ ‘ ‘1 DATA C/OSSI/Ieallon DATA COLLECTION "'-'—'" Cad/n? —‘>‘ STORAGE Source: Keypunching DATA SYSTEM Chappelle, 1976, p. 1A3. I I I I I 31 From the information diagram depicted in Figure 7 one can see the logical flow of data inputs and manipulation. Values are easily attributed to these procedures. However, benefits are often much more difficult and complicated to derive. "The value of the information to the decision—maker or its purchaser is unknown until he has the information, but to make a decision on its value the purchaser in effect must obtain the information without cost" (Riemenschneider, 1977, p. 7). Often if an individual is forced to place a value on an information set before receiving it, the value must be established from prior experience with a similar data source (i.e., consultant). Normally, the more prior experience a consultant has, the more reliable or precise the advice. New data sources which may or may not be as reliable as pre- vious sources must be automatically devalued until proven otherwise. Arrow (1962) explains that these and other prob- lems such as the indivisibility of information and its non— appropriability (i.e., wrong scale or outdated) all tend to cause suboptimal allocation of resources. Information used totally for private use can be organized under three basic methods. "Each individual or firm could collect the information that it needs or purchase it from other firms, or firms could work collectively to gather in— formation and make it available to all the firms in the or- ganization, or finally government could collect the informa- tion and supply it to all of the firms" (Riemenschneider, 1977, p. 7). Problems, however, are encountered at each 32 stage. Primary data collection generally has high fixed costs, thus limiting many firms and individuals from collect- ing it. Competition between firms using similar data might induce mistrust, potential monOpolizing and misrepresentation. Government expenditure of public funds for specific sectors (a subsidy) normally is characterized by underproduction and underutilization of the information. Because of one or more of these problems, firms are forced to use whatever data (information from secondary sources needed to make a decision on) are available. While the quantity of these data may be vast, the quality or ap- propriateness (suitable and precise enough data to base a decision on) of the information may vary greatly. The firm may be faced with deciding just exactly what they need to know. Once the firm decides what they must know, they must address the apprOpriateness of the data needed. A common measure of data quality is the level of precision attained in the estimate. Eisgruber (1972) describes precision as the magnitude of the error of the estimate. Kendall and Buckland's definition of precision is: ...a quality associated with a class of measurements and refers to the way in which repeated observations conform themselves; and in a somewhat narrower sense refers to the dispersion of the observations, or some measure of it, whether or not the mean value around which the dispersion is measured approximates to the 'true' value. (Kendall and Buckland, 1960, p. 22A). Cochran (1977) explains the difficulty of ensuring no unsuspected bias enters into the estimate. Precision is nor- mally used as a measure instead of accuracy to limit unsus- pected bias being entered into the sampling procedure by 33 inferring a measure of accuracy which generally cannot be measured. "Accuracy refers to the size of the deviations from the true mean, u, whereas precision refers to the size of deviations from the mean m obtained by repeated applica- tion of Sampling procedure" (Cochran, 1977, p. 16). Deming (1960) points out that statistical theory is useful in avoiding errors either caused by attaining more precision or insufficient precision than the decision maker requires. Many decision makers and analysts alike have chosen to use secondary data or to infer data-decision relationships (past experience) established in other studies. This may or may not pose a serious problem depending on the data precision used. It may become necessary for the decision maker to alter the original utility function to accommodate data choices which allow for larger potential risks from data precision than originally envisioned. The risks and costs involved with variable precision decision models can be quantified using a loss function. "The loss function is an increasing function of 'errors' or discrepancies between values of the endogenous variables as determined by the model and the forecasts of them" (Fisher, 1969, p. 23). The loss function represents an aggregated utility function for all involved in the decision. The Loss Function Decision making under conditions of uncertainty is re- ferred to as statistical decision theory or Bayesian decision theory. "Statistical decision theory has developed into an 3A important model for the making of rational selections among alternative courses of action when information is incomplete and uncertain" (Hamburg, 1970, p. 61A). The theory provides the principles and methods needed to make the most appro- priate decision given a certain set of goals and conditions. However, the theory does not provide an actual description of how the decisions are made. "A useful concept in the analysis of decisions under uncertainty is that of'opportunityloss'" (Hamburg, 1970, p. 62A). The Opportunity loss analysis or loss function is used to identify the loss incurred because of failure to make the best possible decision. In statistical sampling problems, the optimal sample size for a given decision should be set at a magnitude where the loss plus the cost of data collection is minimized. Cochran (1963, p. 82) pre- sents the following formula for this type of sample size de- termination: C(n) + L(n) C(n) the cost of sample size n L(n) the expected loss for sample size n, in this instance this would be to set n to minimize the loss and is equivalent to: L(n) 1(z)f(z,n)dz 1(z) the loss incurred by a decision with an error in the amount of z in the estimate "Although the actual value of z is not predictable in ad- vance, sampling theory enables us to find the frequency dis- tribution f(z,n) of g, which for a specified sampling method 35 will depend on the sample sizetf'(Cochran, 1963, p. 82). The sample then, if properly taken should reduce the potential loss associated with a given decision. The problem of measuring precision and associated loss is greatly aggregated when dealing with multi-purpose and multi-user studies. No definite answer can be given to the question, how much precision. The amount of precision will depend on the purpose at hand. In statistical decision theory this is taken into account by the introduction of a loss function. A loss function approach has, however, limited usefulness in survey sampling. In multivariate surveys, where data are gathered with a specific purpose in mind, the loss function approach may be valid. In multi-purpose surveys (such as those conducted by public agencies) where the potential users of the data are not known, it is apparent that no general loss function can be constructed (Chatterjee, 1968, p. 532). As a result it is necessary to restrict the utility function to a desired precision deemed necessary by the decision maker. The decision maker, in turn, especially when dealing with agencies, must be disaggregated to one entity. The first problem then in developing a loss function is to determine who will be the decision making entity. In the case of a public agency, even though we deal with it as one entity, we must, as Chappelle (1975) indicates, consider the remote clients involved. When dealing with a nation's natural resources (forests, water, etc.) the remote clients are those who by being citizens own a share of the resources. "Just as the determination of the level of significance (a) used in statistical testing is not a statistical question, the 36 formulation of the appropriate loss function appropriate to the allowable cut decision [of timber] is neither a statis- tical nor a technical forestry question; rather it is a public policy question" (Chappelle, 1975, p. 25). To for- mulate a loss function which represents the public interest it must reflect an aggregated utility function for all in- volved, not an easy task by any means. Theil (196A) presents the following "Committee Loss Function" to deal with group decision making. G lc(x) = Z dg lg(x) e=l committee loss function where: lc(x) dg = load of the raw loss function g = committee member G = number of members of the committee lg(x) = loss function of individuals (Theil, 196A, p. 337). The committee loss function represents an aggregation of individual loss functions of each group member. Normally these would be limited to the committee or group responsible for the decision. This concept of a committee loss function could be used to help satisfy the needs of the remote client concept. If the loss function is considered a social welfare func- tion and each manager or agency representing the remote clients have been sampled to determine their choices or utility func— tion, the loss function could be formulated as before 37 (Cochran, 1963): L(n) = 1(z)f(z,n)dz "1(z) is the loss incurred by the individual remote client through a sampling error of amount z in the estimate" (Chappelle, 1975, p. 38). To determine the total expected loss, the results from each loss function would be summed. This could be done by groups of importances and an appro- priate weighting factor assigned to each set of losses. Chappelle (1975, p. 38) adjusts Theil's committee loss func— tion to reflect this grouping: G lc(x) = 2 dg lg(X) g=1 (Theil, 1964, p. 337). loss function for the committee of interest groups Where: lc(x) dg = loading of the interest group loss function (the degree to which the interest group 'counts' in the deci- sion) g = interested group G = number of interest groups involved in the decision lg(x) loss function of the interest group g It is apparent that if this procedure was adopted it would be time consuming, costly and the chance of bias enter— ing through manager interpretations of group needs could be considerable. Chappelle (1975, p. A0) offers two ways to limit these problems with reference to making an allowable timber cut decision. 38 1. Base the loss function only on the judgement of the appropriate forest officers, thereby making the officer the sole client and internalizing the expected loss to within the origanization. 2. Base the loss function only on technical forestry considerations, rather than socio-economic vari- ables. By using only one representative client and then basing the loss function on more quantifiable physical data, the measurement becomes more general. The physical loss of tim— ber, water, soil, etc. could be multiplied by their current market value and the total expected loss could be determined. Variable Precision Remote Sensing Over fourteen years of research have been conducted using satellites to collect agricultural crop area data and yield estimates. Early attempts to use oblique photography taken during the Gemini missions allowed only limited crop cate- gory definition. Acreage estimates from the Gemini photo- graphy ranged from 50% to 60% area and crop type accuracy. The first Landsat satellite was launched into orbit in June of 1972. Since 1972, two more Landsat satellites have been launched into space, making thousands of images avail- able to scientists from all over the world. The accuracy of area and yield measurements have ranged from 50% to 100% de- pending on the scientists' skills, procedures used and the area under investigation. Remote sensing-based investigations using area and yield measurements can be separated into two primary groups: whole 39 area inventories and area samples. These measurements are usually taken from two platforms: aircraft or satellites. For the purpose of this study, discussion of data produced from aircraft platforms will be omitted. Syria strictly for- bids any aerial inventory work because of the sensitivity of military installations. This situation is typical of most middle eastern nations where political tension is greatly heightened by present world politics. The National Academy of Science (1977) in its investiga- tion of remote sensing techniques for developing nations, con- cluded that total crop acreage estimates could be made from Landsat with a 95% accuracy level. NASA (1973) published acreage estimates ranging from 70—95% accuracy. Higher ac- curacy percentages were attained when several dates of coverage were used to map the area in question. Adnam (1975) reported 95% accuracy in measuring wheat, oats, and barley acreages from Landsat. Bauer et a1. (1973) and Baumgardner et a1. (197A) while working with the Labora- tory for Applications of Remote Sensing, reported 90% and 95% accuracies, respectively, in corn, wheat, pasture, and fallow field measurements. Others yielding similar levels of accur— acy include Myers and Moore, 1978; Worcester and Moore, 1978; Thomas and Hag, 1977; Hanuschak, 1979; Witter and Hill—Rowley, 1979; and Witter, Schultink, and Lusch, 1980. By drawing on these Sources, a justifiable case can be made for believing that the precision of using Landsat imagery to delineate major crops and their acreages is 95%. This figure is, of course, A0 tempered by the procedures outlined by these authors (i.e., repeat coverage, collateral data, image enhancement, and a suitable mapping medium). Several large scale Landsat based projects utilized other data sets in conjunction with crop acreage estimates from Landsat imagery. The most notorious studies are the Large Area Crop Inventory Experiment (LACIE), a joint program for Agriculture and Resources Inventory Surveys through Aero- space Remote Sensing (AgRISTARS), an Evaluation of Remote Sensing with Repsect to Crop Acreage Estimation in Canada and Programme Plan for Developing the Capability of Forecasting Crop Production. The Large Area Crop Inventory Experiment (LACIE) was es- tablished in 197A as a joint effort of NASA, the USDA, and NOAA to utilize remote sensing technology on an experimental basis for wheat production forecasts. "Three years of inten- sive evaluation of LACIE estimates for the U.S. crop and 2 years of experience in estimating the Soviet crop indicated that accuracy commensurate with USDA performance goals for foreign wheat production forecasting was achievable in re- gions where fields are sufficiently large to be resolved by Landsat" (MacDonald and Hall, 1980). The LACIE in-season 1977 wheat forecast was within 10% of yield statistics sub- sequently reported by the government of the Soviet Union. Current accuracy prediction for the Soviet wheat crop, 9 U.S. states and India, China, Australia, Argentina, and Brazil exploratory studies have results ranging from a 23/90 early Al season estimate (U.S.S.R.) to a 100/90 preharvest estimate (U.S.). Yield forecast procedures used by LACIE include the use of regression models which incorporate weather variables from World Meteorological Organization network. These models are based on multiple linear regression equations of histori- cal yields and monthly averages of temperatures and precipi- tation. Once crop type and acreage estimates are made using Landsat digital data, the LACIE procedure uses the following mathematical yield determination model (MacDonald and Hall, 1980, p. 673): Yield preceeding year's yield for average weather) : 8 (yearly adjustment for technology trend) + C (effects of current weather) LACIE's procedures have been criticized by Baumgardner (1980) for not considering soil variations and by Thomas and Hag (1977) for the cost of the sampling design. Both criti- cisms are justified. However, if the project receives future funding for continued research, these problems will become better defined and the development costs less. LACIE's direct application to most developing nations is very limited. "In the developing countries cropland frequently is interspersed with noncropland, fields are small and irregularly shaped, numerous crops have similar spectral responses" (National Aca- demy of Sciences, 1977). In a 1975, Program Plan for Developing the Capability of Forecasting Crop Production, the FAO outlines a pilot crop forecast framework for major food crops of the world. The approach begins with the delineation of ecozone maps of wheat A2 producing regions. A multiterrain approach is taken for the collection of key data which includes climate, botany, hydro- logy, and pedology to determine ecozone boundaries. Crop calendars are combined with historical synoptic weather sta- tion data to establish yield trends. Daily METSAT data are prepared into rainfall distribution maps. Periodic acquisi- tions of Landsat data are used to confirm vegetational re- sponses to reported environmental stimuli. Data bases are then divided into 25 nautical mile squares and assigned pro- duction potential values for the Food Information System or soil moisture values for the Global Early Warning System. The final step incorporates a computer run establishing daily yield estimates cell by cell. The major drawbacks with a sys— tem such as this are: data at several different scales are used, a large data base is needed, substantial time is needed to establish the system, considerable cost is required to operate the system, and reported yield estimates vary drama- tically from country to country. The Program for Agriculture and Resources Inventory Surveys Through Aerospace Remote Sensing (AgRISTARS) is the most recent and largest yield forecasting project. AgRISTARS began a six-year program of research develOpment, evaluation, and application of aerospace remote sensing for agricultural resources during the fiscal year 1980. The program represents a COOperative effort among the USDA, NASA, USDC, USDI, and AID. For a discussion of each agency's responsibilities, refer to the management/organization plan for AgRISTARS (Kibler et al., 1980). A3 AgRISTARS' main goals are to establish the usefulness, cost, and extent to which aerospace remote sensing data can be integrated into the USDA systems. "The overall approach comprised a balanced program of remote sensing research, development and testing which addresses domestic resource management, as well as commodity production information needs" (Kibler et al., 1980, p. 1). The technical program for this approach is broken down into 8 major phases (Kibler et al., 1980, p. l): 1. Early Warning/Crop Condition Assessment; 2. Foreign Commodity Production Forecasting; 3. Yield Model Development; Supporting Research; Soil Moisture; Domestic Crops and Land Cover; Renewable Resources Inventory; and (DNONUW Conservation and Pollution Neither the Management/Organization Plan (Kibler et al., 1980) or the Technical Program Plan discuss actual models used, therefore, a review of the technical aspects of AgRISTARS is impossible. However, it does appear that the base technical program considerations are sound and have great potential for future crop forecasting. Current difficulties with the AgRISTARS, Domestic Crops and Land Cover Project includes poor quality Landsat data, late 1980 wheat estimates, and the possibility of non-available Landsat data for 1981, due to technical problems aboard the satellites. AA Sells. Soil, as a term used in U.S. Soils Taxonomy (1975, p. 1), refers to the "collection of natural bodies on the earth's surface, in places modified or made by man of earthy materials, which contain living matter and supports or is capable of supportingpflennmsin natural environment." Soils data for this study came from information already collected by the Comprehensive Resource Inventory and Evaluation Sys- tem (CRIES). W.J. van Liere's (1968) 1:200,000 and 1:500,000 soil maps provided the principal source of soils information for Ken Ackerson, CRIES soil scientist. Ackerson worked on conjunction with several Syrian soil scientists who provided supplementary data as well as in-country expertise to help refine soils data to meet CRIES project needs. The map units and soil descriptions are based on specific criteria set forth in Soil Taxonomy, 1975. Ackerson (1980) lists five soil orders, six suborders, and fourteen great group categories. These categories are, in turn, used to make 81 soil classifications on the l:750,000 base soil map for Syria. Perhaps the best measure of a soil classification pre- cision would be its category purity. Purity refers to the degree of homogeneity of the soil mapping units. In dis— cussion with Dr. Delbert Mokma (1982) brief comparisons of the purity of soil classifications were made at a detailed county, regional, and state or country level using county and state soil maps of Michigan. The estimates, based on Michigan A5 data, are as shown in Table 1: Table l.-—Estimated Soil Category Purity and Associated Mapping Scale. Map Type Percentage Scale Detailed 85% l:20,000 County 70% 1:200,000 Regional 65% l:500,000 State or Country A0% l:1,000,000 The soil categories listed by Ackerson on the Syrian maps were at approximately the same level of detail as the regional and state or county maps reviewed. Because both the Michigan data and the Syrian data were classified using U.S. Soil Taxonomy, purity values were considered to be syn- onymous for this study. Consequently, the purity of soil categories on the available 1:750,000 Syrian soil maps was determined by adding the regional and state or country purity percentages and dividing by 2. The result was a purity level of 52.5% (65 + A0/2). Soil Moisture Numerous systems, procedures and models have been es- tablished to estimate soil moisture availability for crop growth. "The major problem which occurs when calculating moisture budgets for individual localities is the difficulty of assessing soil-moisture storage and actual evaporation" A6 (Barry, 1973). One of the most reliable means of establish- ing available soil moisture is accomplished with a lysimeter. Observations are made at regular intervals with weight changes, precipitation, evaportranspiration, and percolation measured (refer to King et al., 1956). Use of this method to measure the available soil moisture, although accurate, is too costly and too time consuming for a countrywide assessment. Consequently, it is necessary to utilize other methods for establishing evapotranspiration rates and estimates of available soil moisture. Methods de- termining available soil moisture will be further confined by the meteorological data available in developing nations. Data will be limited to monthly and yearly temperature plus preci- pitation amounts for all but 11 stations in Syria, therefore, the moisture balance equation used must be a simple one. One of the most widely used equations was developed by Thornthwaite (19A8) and revised by Thornthwaite and Mather (1955). The for- mula is: 100(S-D) PE Im = Where: Im = moisture index in millimeters S = annual moisture surplus in millimeters D = moisture deficit in millimeters PE = potential evapotranspiration in millimeters (Barry, 1973). A monthly moisture budget calculated by using this ap- proach and data collected by Thornthwaite and Hare (1965) for Alleppo, Syria is presented in Figure 8. A7 0 inches Em O\\l LL) Figure 8. Monthly Water Budget of Alleppo, Syria. The status of the moisture availability in the Thornthwaite model can be restated as: Potential Evaportranspiration = (Precipitation - Deficit) + Surplus : Storage Change Potential Evapotranspiration--the evaporation and transpiration loss under optimum moisture conditions, or soil continuously at field capacity. Precipitation-~water falling onto the earth's surface from the atmosphere as rain or snow. Water Surplus--the difference between precipita- tion and potential evaporation when soils are at field capacity. Water Deficit--the difference between potential evapotranspiration and actual evapotranspiration. Soil Moisture Recharge (change)—-the difference between precipitation that exceeds evapotranspiration when soils are not at field capacity (Thornthwaite and Mather, 1958, pp. 18-19). A8 The individual components of the equation are more easily com— pared over a long term record when presented in this manner. Ward (1967) explains the pOpularity of the Thornthwaite equation in part results because the formula expresses PE as a function of mean air temperature and day length, two quan- tities which are independent of the rate of evaporation, and which are applicable over a wide range of conditions. A criticism of the Thornthwaite approach to water budgets is that it is empirically based (Lee, 1978; Terjurg, 1976). "Although I strongly support arguments for more rigorous and/ or systematic research in climatology, it can be said that many of the critics of empirical water budgets have misinter- preted the purpose and utility of regression--broadly defined (Wilmott, 1977). Willmott (1977) indicates that researchers may correctly use such models when data for more rigorous analyses are missing or not obtainable (the case in most developing nations) and when the physical—biotic mechanisms that produce the desired answers are either well-known, un- known or unimportant. McGuinness and Bordne (1972) indicate water budgets have been very successful at satisfying the only criterion on which they should be judged--accuracy. An example of average daily potential evapotranspiration esti- mated by a lysimeter and as computed by the Thornthwaite (as described), Blaney-Criddle, and Penman methods is shown in Figure 9. The estimates are very close during the cooler months, but significant variations occur between April and August. “9 Figure 9. Average Daily Potential Evapotranspiration as Es- timated by Lysimeter Growing Deep-Rooted Grass- Legumes and as Computed by Thornthwaite, Penman, and Blaney-Criddle. Oct. Nov. , Dec. Aug. Sept. Feb. Mar. April May June July Jan. 0.30 ._ 0.20 .— 0.10 0!) Potentlal EvapotranSpiratlon In Inches 50 These variations can be minimized by taking the best esti- mate from the methods that reduces the variation between the lysimeter value and the calculated value (refer to Appendix A). Other empirical methods which could be used to adjust Thornthwaite's potential evapotranspiration values have been developed by Penman, Blaney-Criddle, Eagleman, and Hargreaves. "Penman has made the most popular compromise of the energy balance method by eliminating factors which are difficult to measure and by substituting empirical rela— tions where necessary to avoid complicated equipment and measurements" (Carter, 1958, p. A1). Penman considers three stages important in estimating evapotranspiration: 1. The determination of a hypothetical Open body of water E0; 2. The use of an empirical seasonal correction factor of E0 to convert potential evapotranspira- tion ET, over a surface covered with vegetation; 3. The value derived from part two can be further altered for the depth of a vegetation's root system. Thus, the availability of moisture during a deficiency period could possibly be monitored. Safely (197A) illustrated the Penman equation in an expanded form for better clarity. This equation is presented as: VP # 3%? Ra(l—r) (.18+.55n/N) (to determine vapor pressure) SVP = 7%? oTu(.56-0.092 VP (.10+.90n/N) (to determine saturation vapor pressure) 51 PE = Kg? .35 (1.+W/100) (SVP-VP) (both SVP and VP are needed to determine Penman's estimate of PE) where: A = slope of the saturation vapor pressure curve at mean air temperature in milli- bars per degrees C; 6 = constant of the wet and dry bulb psy- chrometer equation; R = Angot's value (the theoretical income shortwave radiation at the outer limits of the atmosphere); r = albedo; n = actual hours of sunshine; N = theoretical duration of sunshine; CT = black body radiation at mean temperature (T) in degrees Kelvin; VP = mean vapor pressure at mean air tempera- ture; SVP = saturation vapor pressure at mean air temperature; W = run of wind at standard height of two meters (miles per day); and PE = evapotranspiration (Safley, 197A, p. 7). Penman's model has several major disadvantages which the Thornthwaite model does not have. Penman's formula re- quires data for radiation, humidity, and wind which are usually only collected at first-order (primary) weather stations. Consequently, large portions of underdeveloped nations normally do not have sufficient data collected. The conversion of calculated PE to ET over vegetation is very difficult to precisely calculate. In addition, the equation describes areas where optimum supply of soil moisture is 52 maintained. Other criticisms of the Penman model are dis— cussed by Carter (1958). Blaney and Criddle (1950) created a simplified model estimating water requirements by various crops in arid por- tions of the United States. The model is presented as: U = ktp consumptive use (potential evapotranspira— tion in inches); where: U k = a crOp water use coefficient; t = the monthly daytime temperature in °F; and p = the monthly percent of daytime hours on a yearly basis. Blaney and Criddle developed their consumption coefficients from field and lysimeter studies. The Blaney-Criddle method's main advantages are that it is easy to use and the required data is normally available at any class of weather station. However, the crop coeffi- cients were derived from small test sites of the sort that could absorb inordinate amounts of energy, thus producing more evaporation per unit area than would be possible from large farming areas. Other problems with the model are found in assumptions made concerning the data. The assump- tions are: 1. There is non-limiting water supply to the plants; 2. Consumptive use varies directly with daytime hour percent and monthly mean temperature; 3. Fertility does not vary among areas; and A. Length of the growing season is an index to con- sumptive use (Safley, 197A, p. 8). 53 Another problem involved in using this method is that the crop coefficients have only been derived for selected- areas in the United States. Consequently, it would be necessary to either set up test sites or to interpolate the coefficient values taken in the United States to the country being studied. Eagleman (1971) investigated the linear potential eva— potranspiration rate equations by Thornthwaite and Mather, 19A8; Van Bavel, 1953; Viehmeyer and Hendrickson, 1955; Denmead and Shaw, 1962; Eagleman and Decker, 1965; and Van Bavel, 1967. The relationships among these models were determined under various climatic conditions. "It was found that they could be combined into a single regression model which may be useful for calculating actual evapotranspira- tion (AE) rates for specified amounts of soil moisture and atmospheric demands" (Eagleman, 1971, p. l). Eagleman (1971) found that each situation tested resulted in a cur- vilinear relationship between soil moisture content and a ratio of actual evapotranspiration divided by potential eva- potranspiration (PE). A uniform moisture ratio (MR) was es- tablished in order to compare model results. MR = (SM-WP)/(FC-WP) where: SM = measured soil moisture content WP = moisture content at wilting point FC = moisture content at field capacity Regression coefficients were calculated for four data sets using the eight models. The coefficients produced were 5A very similar. The regression equation used was: AE/PE = A+B(MR)+C(MR)2+D(MR)3 where A, B, C, D = regression coefficients. These data were plotted and correlation coefficientsi compared. The coefficients for each data set were then used as expressed functions of PE. Actual evapotranspiration rates were then calculated in terms of MR and PE. This pro— cedure was then tested against actual field measurements over a 25-day period. Results between measured soil moisture and estimated soil moisture varied by 5 to 6%. Hargreaves (1977) took a slightly different approach by first establishing precipitation probabilities for a given location using these equations: F m/n+l and P lOO-F Where: m the order number assigned to the date; n = total numbers of data points; F = frequency number; and P = percentage probability of occurrence. Precipitation data published by the World Meteorological Or- ganization for a 30-year period between 1931-1960 was used to obtain mean values and the 97, 79, 60, A0, 21, and 3% probabilities of occurrence. Hargreaves (1977) maintained that crOp water requirements could be obtained using pub- lished climatic data for ambient air temperature and solar radiation. "The best relationship between these elements and crop water requirements exist when mean temperature is 55 expressed in degrees Fahrenheit, TMF, and incident solar radiation, RSM, is expressed in equivalent mm of water eva- poration" (Hargreaves, 1977, p. 3). RSM data, however, is Often not directly measured. Hargreaves offered a conversion equation using tabular values of extraterrestrial radiation (RMM) in equivalent mm of water evaporation per month with the percentage of possible sun— shine (S) occurring at a given location. The equation reads as: RSM = 0.075 RMM x S 1/2 Percentages of S were determined from actual duration of S in hours (SH) from day length (DL) and from the number of days in the month (DM). The equation then reads: S = 100 SH/DL x DM If SH data were not available, S could be approximated from mean monthly relative humidity (HM) using: 3 = KS (lOO-HM)l/2 KS = adjustment coefficients (using a computer program developed by Hargreaves, 1977). Hargreaves (1977) estimated potential evapotranspiration (ETP) using mean temperature in degrees Fahrenheit (TMF) and RSM. Values for ETP are derived in mm per month using: ETP = 0.0075 RSM x TMF a 75% probability of precipitation occurrences was considered dependable precipitation (PD). "The moisture availability index, MAI, is a moisture adequacy index at the 75 percent probability level of precipitation occurrence" (Hargreaves, 1977, p. A). The MAI was defined as PD/EPT. If a value of 1.00 was attained, it means PD = ETP. 56 The Thornthwaite potential soil moisture values were derived using adjusted potential evapotranspiration figures for each test site using the Penman equation. The actual potential evapotranspiration (APE) recorded by the Penman equation were used to replace the PE values obtained using the Thornthwaite equation. The difference between the values derived for available soil moisture using APE and the Thornthwaite model derived PE values were compared. The percentage difference between the Penman equation and Thornthwaite approach was calculated and used as a correc- tion factor to alter the potential soil moisture values ob— tained for all other stations using the Thornthwaite model. CHAPTER IV IDENTIFICATION OF CROP EXPANSION AREAS A crop expansion area can be defined as an area having the same soils as a producing region with similar moisture availability and potential to produce wheat. For the pur- pose of this study it has been assumed that farming prac- tices for dry land wheat would not be significantly different for the three test sites and would not change for any future expansion. Method agile The initial step was to reduce the Syrian master data file, a 185,000 cells per data set (soils, land cover/use, political boundaries, etc.), down to those variables and test sites pertinent to this study. Cross tabulation of soils by land use were run for each montika to determine which soils were producing the rain fed wheat and where simi- lar soils were currently used for rangeland. Comparisons of soils currently producing wheat and the same soils currently in rangeland for each test site are listed in Tables 2—A. By reviewing Tables 2—A it was Obvious that not all the soils available for crOp expansion were also capable of producing rain fed crops. Those soils not currently (1977) producing crops were not considered as potential crop expan- sion areas because there would not be comparable yield 57 58 .m xfipcooo< CH mQOHOOHmomOO HHOm Opoaosooxxx pcmflowcmm N m** pom :Hmm N mmx pcmaowcmm 0mm: pom cfimm mmm: mmHH mmw Hmmfi mum mesa MQOH 0mm mmza mmm :HOH proe om omm me p m w an n» moa mooa Nmfi mom Hp o m 0 mm Ho 0 H mm m o no m: wad mm ozm mm o mmm mma ooa m: 0 mo om m 0 0mm mm :mH m: Ham 0 was 0 moma m mm wmm amp moa mwa mHH mwm mm mm 0 mm ma am :H H mm mod 0 ma ***mHHom A.Ex .Umw A.Ex .va A.Ex .Umw A.Ex .Umw A.Ex .vmw m mm m mm m mm m mm **m *mm on: used onzxfinm nossnm naooem nANH e.nnm .Amxfiucoz mov ocmaowcmm can OLSp IHSOpr< pom :Hmm now pow: wcflom mmmp< Hfiom mo comfimeEoo OH EE 00H pom coupoaomv meLOOm manpmfioe Hflom N um HCOHpmhflomcmhuoom>m Hmwpcouom Umpmznp< N mm< MCOHQNOHOHOOLQ N pam* 6A O.mO O.O O.OH H.em O.NH O.ON mm = H.OO O.OO N.NO 0.0m O.Hm m.NO Ono: nnoaa HH< O.ON N.e O.H O.ON O.e O.e am e O.Hm O.OO N.Om N.HO 0.0N 0.00 onoz oxonnnm 0.0m O.O N.NH O.ON m.m O.OH mm = m.ON O.me O.OO O.HO O.MN O.OO Ono: OOOOHHO O.NO H.OH O.MN m.em NH O.Om eeom e H.OO m.NO O.NO N.Nm N.mO N.Ne Ono: OHOoem O.O O.HO Om O.NH N.Om HO HHOm Hm m O.mm m.mm em O.Hm N.ON mO nosne Hoe 3 OO NO mm mm ON OO oxonnnm m OOH N.NO ON OOH m.ON OO NONH N.OO N.NO Om O.He O.ON OO OOOAHonx OOH N.NO Om OOH O.ON He enaHeHonnoz v OOH mm OO OOH O.Nm OOH OnonooeonO n H: mm ON OO O.ON Om page OO<.m O.HN mm Om O.OO O.ON HO ponndnex.d OOH O.Hm OO OOH O.OH Ne nnOHonnnO o OOH O.Hm OO m.OO O.OH OO noose: NO m: mm NO ON NO oaOOHH< OOH HO OO NO Om ONH OHE O.OO O.Om OO H.NO m.ON HO OOONHOO m.Om H.OO OH O.OO O.mm ON cocncnm m N.ON N.mm NO O.NN O.OO OO onsonm Hoe m m.Om H.OO NO O.mm O.OO He OANH I m.mO H.OO NO O.em O.OO mm NHOozm 2 SE EE EE SE SE SE coapmpm Om ma< Odd Om ONO Odae mpmzpnom mumscmh .mop< zUSpm an poasom mcfisopo upon: on» pom momalmmmH pom mosaw> owwhoum endpmfioz HHom cum .sofiummfiomcmhuoow>m HOHOCOpom UOOmSnO< .COHpmpHQHOOHm smozll.w OHONB 65 .COHOOO>OQ Ohwpcmum** .AHO>OH EE ooa LOO OOOOOOONV meOOOO magpmaoe HHom N um mcofiuwpfiamcmppomm>m Hmflucmuom prw3np< N mNO MQOHONOHQHOOOO N pamx 0.0m N.mm N.m m.w: H.Hm N.OH Om = O.HO I O.ONH H.ON 0.0 N.OO O.NO cnmz OOONO HHO O.HH O.O O.O O.ON N.O O.N am e 0.0N I H.ONH 0.0m 0.0 I 0.0N 0.0m cmoz OOOOOOO O.Om O.O 0.0 N.Om 0.0 N.OH mm = O.ON I N.NmH O.NN N.ON O.NO 0.0m new: ONNOHHO O.NO O.OO H.O O.HO O.OH O.ON OOOO e O.OO I O.ONH N.OH O.O O.OO HO new: nHOmzm N.OO N.NNH Nm O.ONI 0.00 mm HHOO HOW O.HO I O ONH Om O.N O.OO Om Nmsne Hmens NO I HNH OO O.O mm Nm mxonnmmw O.NN I ONH Om N.OO 0.00H NO OONH O.OOHI N.HOH NN O.m N.OO ON OnoNHOOOO N.OO I N.HOH mm O.OO N.HOH OO OONHEHOOOOOHV OO I O.OOH Om O.NO O.NHH ON nnONmOOenO"u N.OHHI N.HmH OH N.ONI N.HOH NO NOOOOOONO O.OHHI 0.00H ON O.OOI N.HOH mm NmnncnnOMw O.NO I N.HOH Om H.OO N.OO Om nnOHOONOOlo N.OO I ONH ON H.OO 0.00 Om women: HO I NmH NO Om OO Nm oNNoHHO OO I OOH HN mO OOH NN OHN O.HN I H.OHH OH 0.0N O.NO HO OOOOHOO N.OHHI N.NmH OH O.NNI N.OO Om dmdmcnm O.Om I O.HO OH N.ONI O.OOH mO Omnmnm HOE O.ONHI O.OOH O.NHI N.HO OO mm I N.NmH N.ON 0.00 NO SE EE EE SE SE soapmpm Om ONO Om ONO ONNO HHNNO gene: A.O.OOOOVII.O OHOOO 66 precipitation, adjusted potential evapotranspiration, and potential soil moisture storage from all other stations in each region were calculated to illustrate the station's representativeness within each region (Table 7). Mean monthly precipitation (Ppt), adjusted potential evapotranspiration (APE), and soil moisture storage (St) for the representative stations in each test site for 1970- 1977 were also calculated (Table 8). When reviewing Table 8 note that in all instances Stl + Ppt2 - APE2 = St2 except where St is equal to or larger than 100 mm (maximum storage level used) and, in the last portion, when using the mean values. The mean values are from a times series 1970-1977 and reflect both positive and negative deviation from the mean. Yield Equation The yield equation proposed for this study was: Y = lel+b2X2+b3X3+U where: Y = yield in kilograms per hectare bl-b3 = regression coefficients X1 = millimeters of potential soil moisture by months of the growing season X2 = percentage of category acreage of given soil in wheat producing region X = acreage of rain fed wheat within each study area U = error term It became apparent after running cross tabulations of the soils and land cover/use and comparing them with wheat acreage 67 .Ho>OH Oosopflwcoo om. map pm OCOOONOHO OHOCOOHchmHm mosaw> cmoz* *Hm.l NO.I ma. mom. ma. mo. *mm. mo.l :0. *m3. mH.I no. Oxommmm OH.I HO.I OH. *NO. NO.I OO.I NH. NO.I OO. HO. OO. HO.I ONNOHHO OO.I OO. OO. *N.N OO.I HN. OO. OO.I OH. HO.I OO. OH. OHOmzm Om ONO ONN Om ONO ONN Om ONO ONN Om ONO ONN .HHLQO noun: Npmsmpmm prscmw .AmmmHImmmHV meOOOm opzpmaoz Hfiow Hmfipcmpom pun COHONOHQOQOOOOQ IneO HOHOOOOON OOOOOOOO .OOHOOOHNHOONN .npcoz NO neoHOOOO NOONO OONHN OOOOOHom on» now omcoammm cofiwom cmoz opp SOLO moocomommfim mepcoohOmII.N OHQNB O.OOI HNH O.ON N.OOI OO O.ON N.ONI NO N.OH O NHI ON H.O NNOH O.NI HNH O.HO N.ON OO N.OO OO O.NN N.OO O.NN O.NO O.OH ONOH O.OOHI N.OOH O.N N.ONI O.OO O.O O.OH O.OO O.HO O.N ON O.OH ONOH N.OOI 0.0HH N.OO 0.00 N.ONH N.ONH O.HO NO 0.00 0.0N O.NH 0.00 ONOH N.OHHI O.NOH N.ON O.NOHI O.OOH O.H O.OOI O.OO OH N.ON O.NH N.NO ONOH O.NN N.OOH NN OOH OO OO OOH N.OH O.OO OOH O N.OO NNOH O.OI O.OHH O.NOH O.OOI N.ONH O.OO O.OOI O.OO NN H.OI Nm O.OH HNOH O.ONHI H.OOH 0.0 O.OHHI 0.00H H.Om m.mOI O.NO O.O O.I OO H.ON ONOH oxmnnnm H.ONI NOH O.HO O.NOI O.HOH O.OO OH OO H.Nm O.ON N.ON O.NO NNOH O.N O OHH N.OO 0.00 OO 0.00 OOH O.HN 0.00 H.OO N.ON 0.00 ONOH O.ONHI O.ONH O.NO OI O.NHH O.N OOH OO O.OO O.OO N.OH H.OO OWMH O H O.ONHI N.HOH H.ON O.O N.HOH O.OO O.NO O.OO N.ON OOH N.O O.NO ONOH 0.0 O.OOH O.NN O.NN N.HOH ON OOH H.N N.HO OOH O O.HO NNOH O.NHI O.OOH H.OO O.OOI N.HOH O.OO O. N.NO O.NO O.ON O.NO O.ON HNOH O.OOHI N.OOH 0.0 N.OOI N.ONH O.OO O.ONI 0.0N N.OH O.OO O.NO H.ON ONOH ONNOHHO OOI N.NOH N.NO O.ONI O.OO H.HO O.OH H.OO O.ON H.OO O.OO N.OO NNOH NOI H.OHH ON H.OO H.OHH O.OO O.ON O.OO N.OO O.NH O.NO H.HO ONOH O.OOHI m.ONH 0.0H O.OI 0.0HH 0.0m O.NN H.OO 0.0HH O O.NO 0.0N ONOH O.ONI O.HOH O.NO O.O N.ONH O.OO OOH N.OO O.ONH OOH O.ON O.OOH ONOH O.OOHI N.ONH O.OH N.OOI 0.00 O.OO N.OHI N.ON N.OH O.OO 0.00 N.OO ONOH H.OOI NOH O.OO O.NO O.OO H.OO OOH O.OO O.OO OOH 0.00 O.NO NNOH O.HN O.NO 0.00H O.OHI 0.00H H.Nm 0.00 0.00 OO O.N ON N.OO HNOH N.OOHI O.NOH N.NH O.HH O.OHH OHH O.NH N.HO O.OO N.OO N.OO 0.00 ONOH OHOozm QOOV OOOV OOOV QOOV oeev OeeV NeeO Heec AeeO Aeeo AOOO Heeo Om ONO ONN Om ONO ONN Om ONO ONN Om ONO ONN NOON O OOHOOOO HHNNO seem: OONOOOON OONOOOO .OOHO pnme OonO OH OoHOOOm N>HOOOOonONNON meo NON NNOHIONOH NoN OOOHO> OOONOOO ONOOnHoz HHoO OOO .OOHOONHNOOONOONneO HancmpoN OOOOOOOO .OOHOOOHNHOONN NHOOOoz :szII.O OHOOO 69 OH.ON OO.ON ON.OO OH.OO NN.OH NO.Nm ON.NO NO.HN OO.OO OO.OO OO.NH Nm.Om NO e HO.ONI ON.OOH OO.OO ON.OI ON.OOH O.OO OH.NO O.NO OO.HO OO.OO ON.ON NO.NO ONO: nOONO H NH.OO OO.ON ON.OO OO.OO NN.ON OH.NO ON.ON NO.NN ON.ON HO.OO HO.OH ON.ON NO e HO.NOI HO.OOH OO.OO OO.ONI OH.OOH HO.OO ON.N OO.NO OH.OO OO.ON OO.ON OO.HO ONO: OOOOOOO OO.ON OH.ON HO.OO OO.OO NO.OH OO.OO OO.OO NN.NN ON.NN O.OO OO.HH HO.OH on e OO.HOI ON.OOH OH.NO OO.H OO.OOH H0.00 OO.NO OO.NO HO.OO O0.00 0.0H O.OO Omoz ONNOHHO OO.NN NH.ON OH.HO NO.NO HO.OH 0.0N OO.NO ON.OH OO.OO NO.NO OO.OH ON.OO am e OO.OOI OO.OOH OO.HO OO.O ON.NOH OH.OO HH.HO HO.NO ON.NN ON.OO OO.OO HH.NN ONO: OHOOOO ANNOHIONOHV ONNV ceeV eNeV ONNO ONNV ONNV ONeV cNeV eeeV AeeV flees Neec Om ONO ONN Om ONO ONN Om ONO ONN Om ONO ONN NOON O OOHOOOO chHQOOO flog N.Hmzhflmh grab. UOSCH ...EOOII . w manna. 70 estimates published by the Syrian government that it is necessary to break the soils down by area percentage. Refer to Table A for an example. The independent variables are presented for each station as they were entered into the regression model, therefore, the first independent variable represents the best one independent variable model to pre- dict yield, the first and second independent variable repre- sent the best two variable model, etc. The results of the multiple stepwise regression equations are illustrated in Table 9. The computer software used to determine the regres— sion coefficients in Table 9 was the Statistical Package for the Social Sciences.1 The initial null hypothesis was rejected in each in- stance and the alternative accepted. The Durbin-Watson test for the Swedia and Hasseke sites indicated zero autocorrela- tion, while the Alleppo station had a negative value for the d statistic. A negative autocorrelation would indicate that the dependent variable was smaller than the actual value. This would indicate that the error terms were negatively correlated (Neter and Wasseman, 1978). The Durbin-Watson statistics were plotted at the minimum sample size for each equation and no autocorrelation was indicated. In an attempt to determine what the explanation power and significance level would be for the dependent variable yield, available independent variables were loaded into the lNie, N.H. Statistical Package for the Social Sciences. New York: McGraw-Hill, 1975. 71 .O OHOOONNO OOHN m OOO .N .H OOHOOO Op NONoN OHHOOOOO .owONoum ONSOOOoE HHOO Hmfipcmpom N owONOOm** .EOUOONO mo OONwOON.m.Q cum MOOOOHOOOOIONO OOCOHOHNOOOO coammoprN mo NONNO ONOOcmum NAaV .m.m OOOCOHOONOOOO COHOOONMONNQ OHO>OH OO:OOHNH:mHmNO mpmzoa ONOOOQOHQOO OOOOH36500O OOOOHOON £OH£3 OsHm> m .QOOOOQHENOOOO OHQHOHSE mo OQOHOHOOOOONmm* mm.mamz mfi.mmwm pamohoucH OIO OO.O OO.O HN.mI OOO. NN. HHNNO ONONoOO .m OIN HO.N O0.0H ON.OHI NOO. ON. NNOOOOO OOONOOO .N ON.H OIH OO.OH N0.0 O0.0 OOO. NN. NNOONOON OOONOOO .H Oxmwmwm O0.00N HO.ONNI paooNOch OIN OO.O NN.NH NO.OI OOO. NO. OOONH x HHoO .N HO.N OIH OO.OH OH.OH OO.OI OHO. OO. OoNOO OOONoOO .H ONNOHHO m.m:N :.mam pamohmpsH OIN OO.O OO.O OH.O OOO. HN. noNOz OOONOOO .N OO.H OIH OH.O OO.H ON.O ONO. OO. OOHHNNO OOONOOO .H OHOOOO OOOOOO .N.N N NOV .O.O O a O OOOHOOHNO> ONOONONOONH ICHQNSQ < m .NOO OONHN OHHOO ONO OOONOOOIINNOHIONOH NoN OOOOOOO ONO .OQQOHHO .mficozm Now mOmeHpmm UHOHO amonz coamwopwom OHQHOHSEII.m OHQOE 72 equation. The equation read: Wheat yield=blSt(Jan) + b2St(Feb) + b St(Mar) + buSt(Ap) 3 + b Ppt(Jan) + b6Ppt(Feb) + b Ppt(Mar) + b8Ppt(Ap) 5 7 + b9APE(Jan) + blOAPE(Feb) + bllAPE(Mar) + bl2APE(Ap) + b X (soils) 13 13°°°°33X33 Where: Wheat yield = kilograms per hectare St = soil moisture storage Ppt = precipitation APE = adjusted potential evapotranspiration bl—b33 = regression coefficients Xl3-X33 = percentage of each producing soil found in each study set (Tables 2-A) The results are shown in Table 10. The second run allowed the rejection of HO. Consequently, the alternative was accepted, H Actual wheat yield for Syria 1' can be predicted with this model at a 90% confidence level. The variable with the highest explanatory power remained the same for each test site for both runs. Soils data failed to play a significant role in explaining wheat yield for any of the three test sites. The yield equa- tion for the Hasseke site recorded soils as the sixth variable adding 6% to the explanatory power of the model, but at 7A.5 confidence level, which indicates it is a nonsignificant vari- able. The Durbin-Watson test for the Swedia and Hasseke sites indicated zero autocorrelation. While the same test recorded a negative value for Alleppo on the first run, a positive value was recorded for the second run. In this instance the .meNoum ONSpwHoE HHom HOHOCOOOQ N OwONOOm** .EOUOONM mo OONwOON.m.Q can moapmfipmpmlmum OOCOHOOOOOOO COHOOONwON mo NONNO ONOUGOOONAGV .m.m MOOCOHOHNNOOO coammONwONNn OHO>OH OocmOHchwHwNO MNOSOQ ONOO Imcquxm OOOOHSESOOO OOOOHMON LOHSS O5HN> m .EOOOOCHENOOOO OHQHOHSE mo OCOHOHOOOOON m* 73 N ON.OOOH m.momH uaooNOch OIO OO.O NO.m ON.HI OOO. HO. :oHOOOHNHOONN HHNNO .O OIO O0.0 O0.0 OH.O OOO. OO. .O.N OOOOOOOO HHNNO .m mIm om.oa mm.m OO.NHI OHO. Ow. .m.m OOOOSOOO ONOSNQOO .m OO.N OIH OO.OH OO.O NH.O OOO. ON. NOHOOOHNHOONN NNOONOON .H oxmmmwm mm.wm mm.wmw pamoNOch HIO NO.NOHH NNO. HO.OI ONO. OOO. .O.N OOOOOOOO NNOONOON .O mim 0.0M mom. Nm.mHI mmo. mm. .m.m Ompm5m©< Ohmscmh .m OIO NH.O NNO. ON.O OOO. HO. OOHOOOHNHOONN OoNOO .O OIO OO.O OOO. N0.0H NNO. NO. OOHOOOHNHOONN NNOONOON .O OIN HO.N OON. NH.ONI HmO. ON. OOHOOOHNHOONN NNOOOOO .N ON. OIH om.oa mwm. NO.HI mHo. mm. noNOz meNoum .H ONNOHHO mfi.mzz mm.OONHI OQOoNOOCH OIO HH.O OH.O ON.NH OOO. NO. .O.N OOOOOOOO NNOONOON .O OIO NN.OH NO.N NO.OI OHO. OO. OOHOOOHNHOONN OoNOz .m mlm mm.OH ON.m wN.Hm moo. Om. .m.m UmpmSnOO HHNQO .m HO.H OIH OH.O NO.N HO.OH ONO. OO. HHNNO OOONOOO .H OHOOOO OOOOOz .N.O N HOV .0.0 O a NO *nOHOOHNO> OOOOOONOOOH ICHQNSQ < .csm OcoommIImOHnOHNO> HHOIINNOHIQNOH Non Oxmmmmm paw .OQQOHHO .Ofipozm NON OOOOEHpmm OHOHO pmmsz coameNwOm OHQHOHSZII.OH OHQOB 7A positive autocorrelation indicated that the predicted value of the dependent variable was larger than the actual value. Once again the plot of the Durbin—Watson test indicated no autocorrelation problems were present for any of the sites. Final Yield Equation Upon closer review of the variables, a number of strong interrelationships among the independent variables (multi- collinearity) were present. Where high correlations (370%) were present, the explanatory power of each independent vari- able was checked against the dependent variable. The inde- pendent variable with the highest explanatory power was main- tained and the other eliminated. This process was repeated for each test site. Those independent variables remaining were reloaded into the equation and run for a third time. The revised equation based on the previously described ana- lysis was run for each test site. Swedia Wheat yield=blSt(Jan) + b2St(Feb) + b3(Mar) + buPpt(Jan) + b Ppt(Feb) + b6Ppt(Mar) 5 + b APE(Jan) + b8APE(Mar) + b APE(Ap) 7 9 Where: Wheat yield = kilograms per hectare bl-b9 = regression coefficients Alleppo Wheat yield=blSt(Feb) + b28t(Mar) + b Ppt(Feb) + buPpt(Mar) + bSPpt(Ap) APE(Mar) + b8APE(Ap) 3 + b6APE(Jan) + b7 Where: Wheat yield = kilograms per hectare bl-b8 = regression coefficients 75 Hasseke Wheat yield = blet(Feb) + b2Ppt(Ap) + b3APE(Mar) + buAPE(Ap) Where: Wheat yield kilograms per hectare bl-bA regression coefficients. The results are shown in Table 11. Once the multicollinearity was adjusted for each equation, the positioning of the independent variables changed appreciably. The best one term predictor became precipitation in April for Swedia, potential soil moisture storage in March for Alleppo, and precipitation in February for Hasseke. The primary reason for the difference in the placement of the predictor variables is explained by the fairly wide variations in the meteorological phenomena be- tween sites (Table 6). The reduction of the multicollin- earity and the repositioning of independent variables, how- ever, did not change the R2 estimates significantly. The HO: was again rejected and the alternative accepted for each test site. The minimum confidence levels were .911 for Swedia, .916 for Alleppo, and .922 for Hasseke. The Durbin—Watson statistic for Alleppo was positive by .07, while the statistic for Swedia and Hasseke were both in the zero region of the scale. The plot of the Durbin-Watson statistic failed to identify autocorrelation for any of the sites. Further comparisons of the models for each region showed that only two variables were common to all three models. 76 .meNOOO ONSOOHoE HHow Hwfipcouog N OmmNOOm** .EOOOONO No OONwOON.m.Q ocm MOHOOHOOOOIONO OOCOHOHOOOOO coameNwON No NONNO .m.m OOOCOOOHMNOOO COHmmONwoNNO OHO>OH OoQOOHmfichmuo ONOSOQ ONOOOCOHQ IxO OOO HSE:oom OOOOHMON £OH£3 Ozam> O .COOOOCHENOOOO OHQHOHSE mo OGOHOHNOOOON m* ONOOCOOONA m O0.00N Om.HONI uaooNOOCH OIO N0.0 N0.0 OO.O- ONO. OO. .O.N OOOOOOOO OoNOz .O OIO O0.0 OH.O O0.0 OOO. NO. .O.N OOOOOOOO HHNNO .m OIN OO.OH OO.N NO.O OHO. OO. OOHOOOHNHOONN HHNNO .N mw.m wIH mm.ma ww.m H0.0H moo. on. soapmpfiafiomhm ONOSLQOm .H OOOOOOO m©.mHN m.momm| pQOoNoucH NIO ON.HH ON.O N0.0 OOO. NO. OOHOOOHNHOONN NNOONOON .O OIO ON.NH HN.O OO.OHI NOO. OO. .O.N OOOOOOOO NNOOOOO .O Olm mo.HH wo.m O©.m Hmo. mm. COOOOOHQHOONN HHNOO .m mum ON.O NH.O O0.0N OOO. OO. .O.N OOOOSOUO :onz .N OO.H OIH O0.0H OO.H NO.H OHO. OO. OoNOz OOONOOO .H OQmOHHO mm.m©m w.mmeI pawopmpsH HIO ON.ON O0.0 O0.00 OOO. OO. .O.N OOOOOOOO OoNOz .O OIO OO.N OO.H OH.O NOO. HO. .O.N OOOOOOOO HHNNO .O OIm mm.m ON.O OH.OmI Omo. ow. ONOSNQOO OwONOpm .m OIN O0.0H mm.O O0.00 OHO. Hm. **£oNOz meNOOm .m OO.N OIH OH.O ONO. ON.O NOO. HO. OOHOOOHNHOONN HHNNO .H OHOOOO compmz .m.Q O ADV .m.m n O mm *mOHOOHNm> OCOUCOQOUCH IOHONOO < .csm ONHQBIIOOHOOHNO> OOOOOHOwIINNmHIONmH Non Oxommmm pew .OOQOHHO .mfipozm NON mOmefipmm OHOHO pawn: COHOOONwOm OHQOOHSZII.HH OHOOB 77 These variables were the precipitation in April and the ad- justed potential evapotranspiration during March. To test the significance of the R2 using these variables, an analysis of variance was run for all regions together. The first run compared April precipitation against wheat yield for all three test sites combined. The results were: 2 Mean Significance R_ DNF. Swine F cfi‘F April Precipitation .20 1 665502.75 (explained) 5.56 .028 119A89.87 (residual) The second run compared March adjusted P.E. against wheat yield again for all three test sites combined. The results were: 2 Nbai Sigujimrme BL. D.F. Square F: of F March Adjusted P.E. .25 1 A2l7l7.50 (explained) 3.61 .0A5 11665A.52 (residual) The third run compared both of the previous independent variables against wheat yield with the following results: 2 2 Significance .I'LAPLF. orF April Precipitation .20 .20 5.56 .028 March Adjusted P.E. .25 .05 3.61 .0A5 The F values were highly significant, thus indicating a direct relationship between wheat yield and these variables. The best one term explanatory variable, based on this ana— lysis, was April precipitation with March adjusted P.E. only adding .05 to the overall R2 value. Under these conditions, 1970-1977 data for all three test sites would be of doubtful use in calculating the March adjusted P.E. 78 Further analysis with combinations of variables and re— gions indicated that the most significant estimates of wheat yield were obtained with equations reported in Table 11. Consequently, Table 11 reflects the models and responses that were used in the following loss function analysis. Loss Function The final regression equations used for each test site have been used to estimate wheat yield at P:.90 confidence interval. However, data used to establish these yield esti- mates were not collected with a 100% precision (none ever is). In this instance, the margin of variable precision error accepted was i 10% at a .90 confidence interval. The difference between a i 10% sampling error versus : 10% pre- cision error is that sampling determines the representative- ness of the composition of the population while precision refers to the representativeness or continual recurrence, through Observation, of a unit within the population based on numerous runs. As a result, a sample size of n with a i 10% sampling error may also have a i 10% data precision error. As a result, the decision maker basing his decision on the .90 confidence interval without taking into considera- tion the effects of the variable precision of the data could be taking a considerable risk. In order to quantify what the potential monetary risk would be for each test site using mean value variable precision data, a quadratic loss function was used. The quadratic loss function is used to determine the Optimal estimator of a central value based on past mean 79 values. In this instance the quadratic form was used to calculate the monetary loss incurred with a decision through a given data precision error. The quadratic loss function used was based on work by Wonnacott and Wonnacott (1972) using Bayesian Analysis. Assumptions As noted previously (page 11), five assumptions had to be made before the potential monetary losses could be cal— culated: 1. Acreage and yield statistics reported by the Syrian government were at least 95% precise (an estimate based on review of variations in reported statistics). 2. Syrian cultural and/or farming practices would not change yield characteristics from one study site to the next. 3. Pre-determined precision levels assigned to each variable were representative of that variable (based on Chapter III). A. Syrian recorded moisture and temperature values were 90% precise (refer to Ward, 1967, catchment errors). 5. Adjusted evapotranspiration corrections (Penman values) calculated for each station were 92% precise (refer to McGuinness and Bordre, 1972, p. 15 for comparisons of Penman and lysimeter values). Refer to Chapter I, Assumptions, for a discussion of the appropriateness of these assumptions to the study. 8O Actions and States The Syrian Department of Agriculture's (SDA) utility function was assumed to be the "determination of potential wheat expansion areas to increase wheat revenue with the highest potential return and the least risk." To assess this utility function, gated: four potential actions were investi- Expand into 100% of the available rangeland Expand into 75% of the available rangeland Expand into 50% of the available rangeland Expand into 25% of the available rangeland The SDA's choices were assessed under five states of variable probability of occurring for each test site, which reflect yield. site. Bl potential reductions in key variables and wheat They were: Precipitation, potential evapotranspiration, soil moisture and yield equal the mean reported values Precipitation, potential soil moisture, and yield of the mean values Precipitation, potential soil moisture, and yield of the mean values Precipitation, potential soil moisture, and yield of the mean values Precipitation, potential soil moisture, and yield of the mean values Quadratic loss functions were The maximum potential gross evapotranspiration, are reduced to 95% evapotranspiration, are reduced to 90% evapotranspiration, are reduced to 85% evapotranspiration, are reduced to 80% calculated for each test gain (MPG) to be derived 81 by wheat expansion for each State by each Action was calcu- lated. The reported yield was multiplied by .95 to reflect the potential observation error made by the Syrian reporting services. The .95 was based on the variation in yield re- ported in the Syrian Statistical Abstracts, 1978. The $ (U.S.) reflects the 1977 government supported local wheat price in Syria--61 Syrian piastors per kilogram of wheat = .15 U.S. MPG = [Reported Yield (.95)] - Expansion Area $ in U.S. per kilogram of wheat The maximum potential dollar loss (MPDL) was calculated as: MPDL = [l—(Precision Xl - Precision X2 - Precision X3 Precision XA - Precision X5)2] - [Observation error (.95) . R2 (%)] - Expansion Area - $ in U.S. per kilo- gram of wheat The R2 (%) reflects the potential error from using the yield model to estimate potential wheat yield using only the vari- ables reported in Table 11 for each test site. The minimum potential gross gain (MPGG) was calculated as: MPGG = MPG - MPDL Examples for a 100% expansion into the available rangeland are provided at the end of Tables 12, 13, and 1A. Loss Function Comparisons Between Sites When State 1 Action 1 was compared between Swedia and Alleppo, the MPGG for Alleppo was found to be A1% larger than 82 AcoOOOONxO OOOHV ONO.ONO.NO OONzIONz AconcOaxO OOOHV OOO.HOO.NHO I OH..OOO.OOO.OOO.NN. OOH..OOO.OOO.OOOO.OON.IHO OH..OOO.OOO.HOOOO.O HONNONO.VIHO OH..OOO.OOO.HOm. HNANO..NO.V.HNO..OO.V.HNO..OO.V.OO.VIHO AcoHOcONxO OOOHV OOH.ONO.ONO OOH..OOO.OOO.HOO.VOOOO .AOO.VOOOH. AOONzV chc mwopo HmHucmpom ESEHCHZ .o 13sz wmoq NOHHOQ HmHucouom ESEwaz .p HO.HVHO :oOOoO .HO.HVHO OOOOO "OHNEOOO HOO.OOO.H OON.NH0.0 OOO.NHO.OI HON.ON0.0 HNOAOOOAO OOHAOMHAN OHN.OON.OH NONaNOO.OH O OOO.OOO.O OOO.OOH.OH OOO.HON.OH NOH.ONO.ON NOO. V O ON0.000.H OHN.OON.O ONOOOONmO OHN.OO0.0 HOOAOHO.O NOO.ONO.N OOO.NOO.HH ONO.OON.OH O OHO.NHO.O HOH.ONO.OH NNO.OON.OH OOO.OOO.HN NOO. V O NHH.OOO.H OOO.OOO.O NHO.NOO.O IOOH.ONN.OI ONO.OOO.O OOO.ONO.O ONO.OHH«NH OOO.OOHAOH O NON.OON.O NOO.HOO.HH OOO.NOH.NH OOH.ONO.NN HOO. V O OOH.OON.H OHN.ONO.O ONH.OOO.O NNO.NOH.N OomOOON.O NHO.NNOAO ONOAOONANH OOO.OOOANH N OOH.OOO.O NON.OOO.NH OOH.NOH.OH OOO.OOH.ON HOO. V O OOONz OON.NOO.H OO0.00N.O OOO.OOO.O ONO.ONO.N OOONz NNN.OOO.O OOO.ONO.O OHO.OOO.OH OOO.HOO.NH H OONz HOO.NOO.O OOO.OON.NH NON.NOH.OH OOH.ONO.ON NO.HV m HON.VOO HOO.VOO HON.VNO NO.HVHO OOOOO OOOOOO .cHwo wmoac HwHucouom ESEHCHZ new Ammoq NOHHOQ HwHucmuom Ezefixmz .cfimo Hmecouom 535 Ime2 mauzocm OpHm umoe meozm an» no mHmOHOCO COHuocsm mmoq OHumNcwsaII.mH mHnt 83 ONO..HOO.VNOOO.HNHOO..NO..OO..NO..ANO..OO.V*HO AOOHOOONOO OOOHV NOO.ONO.OHO I OONOION: I :HOO OOONO HOHOOOOON eschHz .o HOOOOOONOO ROOHV NOO.OOO.HNO OH..OOH.NOO.HOO.OO. OH..OOH.NOO.HOONNAOOO.VIH OH..OOH.NOO. AeoHOcONxO OOOHV OON.HOO.HOO I HOONzV HOONOV wmoq NOHHOQ HwHucouom Esewaz .p OOH..OOH.NOO.AOO.VNOOO I HONzV OHOO HOOOOOOON essOsz .O HO.HVHO OOHOoO .AO.HVHO OOOOO ”OHNEOOO OO0.0NH.N OOO.OOO.O HOOOONO.O OOO.HON.OI OOOONHN.O NHO.OOO.O ONOAOOOANH OOO.OOOAOH O NOO.NOO.O OOO.OON.NH ONO.NNH.OH NO0.000.0N NOO. V O NON.HHO.N OOO.NN0.0 ON0.000.0 NOH.OON.O HNWAOOOAO OOHOHOO.O OHN.HOO.OH OONANNO.NH O OOO.HON.O ONN.OOO.OH OOO.ONO.ON NOO.NOH.NN HOO. V O OON.NOO.N OOO.OOO.O HON.HOO.N HOO.OON.O OOH.OON.O OON.OOO.O NOOINON.OH NOO.OON.OH O OOO.HOH.N OON.NOO.OH OOH.ONO.HN OOO.OON.ON HOO. V O OON.OOO.N NHO.OOO.O ONO.OON.N OOOHNOO.OH NOOONOO.O OOO.OHO.OH HOOAONO.OH OON.OOOION N OOO.OOO.N HHO.HOH.OH NHN.NNN.NN ONO.OOO.OO NOO. V m oOON: NOH.OHN.N OOO.OOO.O OOO.NOH.O NOO»ONO.OH OHON: OON.HNN.O HNO.NOOIOH HONIOHOOOH NO0.000.HN H OONz NNO.OOO.N OOO.OOO.OH HON.HNO.ON OON.HOO.HO HO.HV m NON.VOO NOO.VOO NON.VNO NO.HVHO OOOOO eoHsoO cho OOONU HmHucmuom EOEOCHZ new .mmoq NOHHOQ Hmecmoom Ezefixwz .cho HOHOCOOON EseHsz wcfizocm OpHm pme OQQOHH< Opp mo mOmOHOCO coHuocsm mmoq OHOONONSOII.MH OHQNB 8A AconcONxO OOOHV NOH.OHO.ONO I OH..OON.ON0.0HOOO.Hm AQLZIOLZ AcoanOaxO OOOHV OOO.OOO.NNO OH..OON.ONO.HOO.OO. AOOOOO.VIHH H..OON.ONO. OOO..HOO.VNNOO.ONHNO..NO..OO..OO.VIHO I mnoq NOHHOO HOOOOOOON EOstOz .o AcoHOOONxO OOOHV OOO.OHO.OOO I OOH..OON.ONO.HOO.VNNON HOONzV :Hmo mmoao HmHucouom ESEHCHZ .o HOONzV Acmzv CHOU Hmfiucouom Essdxmz .w Ao.HVH< cofiuo< .Ao.HVHm mpmum "OHQmem ONO.OOO.O HOOMOOm.O HOO.OON.OH NOO.NHO.ON HOOIONOAO NN0.000.0 OOO.OOO.OH OOO.OHO.NH O HOO.NOO.O OOO.OON.OH OOO.OOH.ON ONN.OOO.NO AOO. V O OOO.OON.O OHO.OHO.OH NNO.ONO.OH IIOOO.NOO.ON OOO.OON.O OON.OH0.0 ONO.ONN.OH OOOAOOO.OH O OOO.OOO.O OOH.OO0.0H NON.NOO.ON OOO.ONO.OO NOO. V m OOO.OHO.O OON.HOO.HH OOO.NOO«OH NOO.OOO.NN OHOAOOO.O NON.ONOAOH OOO.OHOIOH ONO.OOOION O OH0.000.0H OOO.HHH.HN OOO.OOO.HO NOO.NNN.NO NOO. V m NOO.NNO.O ONO.OOO.HH OHO.NOO.NH NOO.OON.ON NOwaOH0.0 OOH.OOO.OH NONAOOO.OH OOOAONN.HN N OOO.HOH.HH OO0.00N.NN NOO.ONO.OO NON.NO0.00 NOO. V O oOON: ONN.ONH.O HOO.NON.NH NNOMOOO.OH NOH.OH0.0N OHON: ONO.OOOAW OOH.OOHAHH ONN.OON.OH OOO.OOO.NN H OON: OOO.ONN.HH OON.OOO.ON OOO.OOH.OO OOO.OHO.OO HO.HV O HON.VOO HOO.VOO HON.VNO HO.HVHO OOOOO cfimo OOONo Hwfipcmuom ESEHCH: cam .mmoq NOHHOQ HmHucmuom Esewaz .cfimo HOHOCOOON Ezefixwz wcfisonm mpr oxmwmm: vs» no mHmmHmc< :oHuocsm mmoq OHOONOOSGII.OH OHQOB 85 Swedia. The Alleppo site, however, is 27% larger than the Swedia. To compare how much of the MPGG difference was due to the overall precision levels, .28 for Swedia and .A7 for Alleppo, and how much was due to area, the Swedia MPDL was calculated using the area and yield values for Alleppo. The MPGG for Swedia became $16,A85,521 or 67% less MPGG than the Alleppo area, thus, a 19% difference in overall precision could result in a potential loss of $8,028,581. The same procedures were used to compare Swedia versus Alleppo and Alleppo versus Hasseke. The results were: 1. Swedia, using Alleppo data = $9,636,371 MPGG or 11.A% difference, or $1,2A0,296 for a A% difference in over- all precision. 2. Alleppo, using Hasseke data = $18,175,959 MPGG or 26% difference, or $6,339,1A3 for a 15% difference in over- all precision. Probability of Moisture Occurrence To better establish the risk in commercial value the decision maker would face in taking any one action, the prob- ability of moisture occurring at each State was introduced. The MPGG values were multiplied by the probability of mois- ture occurrence. The resultant value represents the adjusted minimum potential net gain to be expected from any future Action based on a 20 year record. The procedure is illustra- ted for each test site in Tables 15, 16, and 17. Normally, 35 years of data is considered the minimum to estimate these variations (refer to Ward, 1967). However, in this instance 86 .AON.VOO + HOO.VOO + AOVOO + HOH.VOO + AOH.VNO + HOO.VHO I OONOOOO .NOHOQESONOO HOOOO NON OONO ONOHOONOSO On 0» Ooesmmw ONo3 mpoHNOO Ommsp ONHNSU Omoso0NO mQONO one .NNSOON OOHoNOEEOO HOHONOOOQ o ucomONQoN Op 0 OUHO> O NO>Hw mm; mm ONO mepcmo INOQ wasp pcmmONaoN Op OOOOO mmz Om Opmum .ONOOON NOON cm a NON osam> NONE HOONO>O On» NO mow swap OOOH ONOz mosaw> OON OOONOOON map No Ommxx .moumN owcmnoxm OOOOOONO NNOH so Ommmn aONOHHOU .m.: CH ONO OOSHO> OOONOQON HO m NO>Hm mm: mm ONO mepcoo INOQ mHsp pNOmONQON Op OOUOO was Om Opmpm .ONOOON NOOO om m Non OSHO> came HHONO>O ONp No Rom swap mmoa ONOB mmzam> pam OOONOOON map No Ommm* .wmpmN omcmnoxm HOHOONNO NNOO so comma .mNOOHOU .m.: 2H ONO mosfim> OOpNOQON HHOO OOONOO.OON.H ONN.NOO.O NOO.HON.O OOO.OHO.N I OONz o o o o onwm xmmm O O O O NOO. VOO O O O O O HOO. VON O O O O O AOO. VON O ONO.OON HO0.0HO NOO.ONN OON.OOO.H AOO. VNO OH ONOO.OOO.H OOO.HOO.N ONO.OOO.O NOH.NOO.O NO.HVHO OO HON.VOO NOO.VOOIIIIIOON.VNO NO.HVHO OOOOO OoNoNNOooO NOHpo< pamm .NOHpOO ONO Opmpm comm Op OpOm OQQOOHO ONp NON oocoNNSOoo ONSmeoz No OpHOOmeONm OQp Op OOpOSOUO Amwsz mcfimw mmONU HOHpNOpom ESEHNHEII.OH OHOOE 88 .AOOOO + “OO.OOO + HOH.OOO + AOH.Omm + AOH.OOO + Hmm.OHm u OOO2*** .EOHpQESmcoo awooa pom mace pzmHoa%m3m mm on cmESmmw who: mGOHme mmmnp wcHasc Umosvopa maopo mne .CHSpmH amaommEEoo Hmflpsmpoa o ucmmmmamh on o m5am> m cm>Hw mm: wm vsm mmeCmo ILmQ wasp usmwmhamp 0p Umcvm mm: mm madam .Upoomm Ham» om w pom mSHm> cams HHmpm>o map mo &om can» mmmH mhmz mmSHm> pam Umomoomh map mo Nom** .mmpmm mwcmcoxm HmHOHmmo wmmfi co cmmmn “mamaaoc .m.D CH mpm mmsam> umphogmh HH<$ ***mmm.OOO.O OOO.OOOHO HOO.OmH.mH OOO.OOH.OH u OOOz O O O O HOVOm **Om HmH.mOm mOm.OOO mm:.mmp OOO.OOO HOO. vmm m OOO.Omm NOO.HOOnH Omm.mmm.H mOO.mOOnm HmO. OOm OH omm.Hmm OOH.MOH«H OOO.OmO.H Omm.OOm.m HOO. Omm OH Omm.mOm OOO.OOH.H HOO.OOO.H mmmnmmm.m Hmm. Omm OH *NOO.mOH.m mmH.OOm.O :Hm.mmO.O mON.OOm.O HO.HOHm mm Hmm.VO< AOm.var||||kmO.vm< HO.HOH< mpmpm mocmpn:ooo 20Hpo< pama .QOHpo< 6cm mudpm 50mm mp mpfim mxmmwmm on» pom mesmhhdooo musOmHoz no OOHHHOOOOHO map On OmpszO< Hoomzv mchO mmopO Hmecmuom EzeHcstu.OH mHOme 89 the available 20 year cycle (1958—1977) was used to illus- trate the procedure. Monetary Risk The previous three tables (15-17) would allow a Syrian decision maker to determine the percentage chance (potential risk) of no commercial crop associated with each Action. Based on the probability of a wheat crop versus no commercial crop for Swedia, the chance or risk of no MPGG was 25% (B6) with a potential average yearly return of $5,458,580 (100%). The risk of no commercial crop in the Alleppo region was 35% with a potential average yearly return of $7,015,450. The 10% greater risk of no commercial crop in the Alleppo site represented a $1,556,870 greater potential of an average yearly return, or $31,137,U00 over 20 years (total value not dis- counted). The difference in risk of no commercial crop be— tween the Swedia site and the Hasseke site was 5% for the 20 year period. The 5% additional risk of no commercial crop, however, carried a $10,721,386 greater potential of an average yearly return or $21u,u27,72o based on the 20 year record (total value not discounted). Based on this information alone the Syrian decision maker would conclude that the least risk of no commercial crop would be the Swedia site, while the largest MPGG would come from the Hasseke site. Yet, the decision maker would be basing this decision on a yield estimate with an overall model precision 90 of 28% for Swedia.2 When overall model precision levels (Tables 12—1u) were multiplied by the percentages of Ppt occurrence (Tables 15— 17), the overall confidence in the estimate became .28 (.75) = 21% for Swedia, .32 (.65) = 20.8% for Alleppo, and .u3 (.70) = 32.9% for Hasseke (Wonnacott and Wonnacott, 1972, Bayesian analysis). Expanding into the Swedia test site presented the least risk in the overall estimate with the highest MPGG. Expansion Site Expansion into the Hasseke site best fulfills the common utility function of the Syrian Agriculture Sector, "To pro- vide an estimate of the utility associated with identifying potential wheat expansion areas to increase wheat revenue with the highest potential return and the least risk." The Syrian farmers' common utility function, "To provide an esti- mate of the utility associated with identifying wheat expan- sion areas to increase wheat revenue with the highest return and the least risk," would also best be satisfied by acquiring land in the Hasseke site. The decision maker must also answer which Action(s) should be taken. To determine this, the cost of production, harvest, and the analysis must be considered with respect to MPGG and the yearly Agricultural Ministry's budget. 2Model Precision 0904.90..92)°(.90-.92)-(.9o-.92)]2 (5211776)2 .28 91 Cost Versus MPGG Production costs and returns per 100 kilograms for the 1976-77 rainfed wheat crop in northeastern Syria were: Variable Costs $ 6.75 (U.S.) Family Labor (planting and harvest) __;35 Total Cost $ 7.20 Value of Wheat (supported price) $15;gg Net Earnings $ 7.80 Sixty percent of the variable costs were accounted for in seed and power costs (animal and tractor). Hired labor and ferti- lizers accounted for another 20% and 20% was spread over new equipment, repairs, etc. The mean yearly wheat yield for the Hasseke site adjusted by potential reporting error was 501 kilograms of wheat per hectare. The total costs per hectare became $36.07, while the total net earnings were $39.08. The total costs of conducting the analysis of the study sites was based on a breakdown of actual CRIES expenditures for the remote sensing, computer, digitizing, materials, and personnel during the Syrian project (1978-1980). The con- densed budget, as it would reflect this project, was: Salaries and Wages Investigator (100%) $ 30,000 Interpretation (100%) 25,000 Computer Programmer (75%) 18,750 Cartographer (50%) 12,500 Secretarial Support (75%) 9,000 Subtotal $ 95,250 92 Fringe Benefits (18%) $ 17,145 Subtotal $112,395 Travel, Transportation and Per Diem Two, four-week TDY's for two (air $ 9,600 fare, P.D., misc.) Materials Landsat imagery (from Italy), diazo $ 6,000 film, equipment rental, mapping supplies and miscellaneous Total Direct Cost/Year $127,995 Indirect Cost (21%) $ 26,879 Total Adjusted Project Cost $15u,87u The total cost of the analysis per hectare for all three test sites combined was: $15U,87u/l,307,200 hectares = $ .12 per hectare The adjusted minimum potential net gain (MPNG) per hectare in the Hasseke site was $7.80 - .12 = $7.68 or 51.2% of the government supported price. To determine what the actual MPNG for each State and Action would be, the total costs of production were subtracted from the MPGG reported in Table 17 (Table 18). By subtracting the total MPNG values in Table 18 from the MPGG values in Table 17, the total cost for each Action excluding the expenditures during periods when only subsis- tence crops were recorded, was determined: A1 = 7,766,385; A2 = 5,82U,787; A3 centage return on expenditures was 8%. = 3,883,19u; A” = 1,9u1,598. The per— 93 Table l8.--Minimum Potential Net Gains (MPNG) Minus the Variable and Analysis Costs for the Hasseke Site. Action State Al(l.0) A2(.757_"‘K3(.50) Au(.25) 81(1.0) u,u61,748 3,3u6,311 2,230,87u 1,115,u37 B2( .95) 1,211,0u6 908,28u 605,523 302,761 B3( .90) 1,1u7,306 860,u79 573,653 286,836 Bu( .85) 1,083,567 812,675 5A1,783 270,892 B5( .80) 509,91u 382,u35 25u,957 127,u78 *Total MPNG = 8,u13,581 6,310,18u u,206,790 2,103,395 *Example: State values used in the equation are the MPGG values from Table 17. Action 1 Value of Net Earnings per 100 kilograms of wheat $7,80/15 = .52; MPNG = .52 (8,580,285) + .52 (2,328,935) + .52 (2,206,359) + .52 (2,083,783) + .52 (980,60u) + .52 (0) 94 Action Versus Budget The question of which Action the decision maker should take became one of how large the agricultural sector's yearly budget would be. If the Syrian government provides money for both the total variable costs of planting and harvesting, plus a guaranteed wheat price, the government must have re- serves large enough to pay both the MPG and the MPGG. Based on the analysis, the minimum reserves needed are reported as MPGG values in Table 18 and the maximum would be the MPG values reported in Table 14. At $15 per kilogram, Action 1 would require risking $16,179,966 (MPG) to $46,913,408 (MPGG); Action 2, $12,134,971 to $35,185,056; Action 3, $8,089,984 to $23,956,704; and Action 4, $40,044,993 to $16,728,352. However, if the govern— ment just opened the land and let the farmers bear the risk for the total expenditures, the government would only be re— sponsible for the guaranteed wheat price. The minimum re- serves required are reported as MPNG values in Table 18, while the maximum reserves needed were calculated by multi- plying the MPG values in Table 14 by the percentage of net return per kilogram of wheat (.52). Action 1 would require a minimum of $8,413,581 to $24,394,972; Action 2, $6,310,184 to $18,296,229; Action 3, $4,206,790 to $12,197,486; and Action 4, $2,103,395 to $6,098,743. Obviously, a careful analysis of world market prices for wheat would have to accompany the yearly setting of the government's supported price for wheat. 95 Information Value The information value was measured as the difference be— tween the total MPGG (Table 14) and the total adjusted MPNG (Table 18) or the potential loss from overestimating the net gains from a given Action. The information value for each Action, based on the 20 year period, was calculated by sub- tracting the potential returns from the costs of production and analysis (Table 19). Under each Action the decision maker would have antici- pated a 50% higher return if the variability in the precision levels were not considered. By utilizing the adjusted MPNG values based on a 20 year probability of occurrence, a future plan of action could be determined with a more realistic po- tential return value. The Syrian government could use this procedure to re- gionalize the country's best potential agricultural areas. By grouping crops such as wheat, barley, lentils, etc., yield equations and quadratic loss functions could be run to deter- mine where the highest yield potentials could be expected. Minimum potential net gain values could be derived for each crop grouping. The MPNG values could then be used to deter- mine the minimum farm acreage allotment needed to produce the median or subsistence income. By running different simula- tions of this procedure, those regions offering the highest MPNG could be prioritized for development. Those regions having marginal MPNG values for subsistence incomes could be prioritized with the highest ranking going to those regions 96 .mtoHHoo .m.: :H omppoaoa modam> Ha Hmp09* mum.amu.m osam> coapw590McH 6H60 662 HoHoeooom eseonz ooomsno< Omz.mHH.H oHoO HoHpoooom eoeonz Hoooe oom.OMH.m OOm.mOm.m mmm.moe.m OOO.OOO.O mee.OHH mam.mmm eHO.OOm mOO.mOO mO:.OmH 5mm.zmm mmzwmwm :Hm.mom m Omm.mzm OOO.NOO OOO.HmO Ome.msm HOO. v m HOm.OOm mOO.mOO Oom.mee mOO.OOO NOO.OON MOOiHOm OOO.OHO OomimOO.H O mOm.OHm OO~.OMO.H OOH.mmm.H mem.mOO.m Hmm. v m HOO.OOO OOH.ONm HOH.OOO HON.OOO.H OOO.OON mmo.mem OOO.OOO oom.OOH4H m www.mzm mee.OOO.H OOO.OOO.H OOm.mOH.m HOO. V m OHOJONN .Omm.mmm Ommmmmm OOO.OOH.H HOO.mOm mmw4mOO OON.OOO OOOdHHme m OOm.OOm HOO.OmH.H HOH.Ome.H mmm.eHm.m Hmm. V m mmH.OHm.H O:O.OMO.O oOm.emO.m OOOMOOO.O OewwOmm.m HHmioem.m OOO.HOO4O H ONH.OON.O OOO.MOO.O OON.Omm.O HO.HV m . O . m . m . H Hmm O < HOm V a Hm» v a HO HO a oooom coHoo< coamcmoxm pmmnz oumocHHoQ Op mpmo coamfiompm .opam mxommmm 0:» CH mmoh< oHomHLm> wcfimb ozam> COHmeLOMCHuI.mH oHnme 97 needing further study with the potential of returning study costs. The break down of regions using this procedure could also be the basis for developing new agricultural tax asses- ments based on the MPNG or the MPGG values. Needs in re- gional marketing and transportation networks could also be compared to the various simulations of crop groups and MPNG values. However, caution should be used in applying this procedure beyond the regional level it was designed for without re-evaluating the precision of the independent vari- ables needed to make the yield estimate. Soils, for example, should be reassessed at each level to determine their sig- nificance to the overall yield equations. CHAPTER V SUMMARY, CONCLUSIONS AND RECOMMENDATIONS FOR FUTURE RESEARCH The purpose of this research was to investigate the value of information derived from using variable precision data to delineate potential wheat expansion areas in Syria. To illustrate the value of the information derived, three objectives were formulated. The first objective was to identify potential areas suitable for wheat production within three Syrian test sites using remote sensing techniques, water balance equations, and the most current soil data available. An expansion area was defined as an uncropped area (rangeland) capable of sup- porting a wheat crop. A multiple stepwise regression analy- sis was used for each test site. Soil moisture availability, precipitation, potential evapotranspiration, and the percen- tage of each soil within the crop producing area were used as independent variables. Wheat yield was the dependent variable. In every instance the soils data, when considered only by area composition, failed to play a significant role in the explanation of wheat yield. The expansion areas within the test sites were then ranked according to their potential wheat production. The mean yield values from each test site were multiplied by the 1977 government supported price for wheat per kilogram to determine the potential net gains from wheat expansion. 98 99 Each variable used in the regression analysis has a potential measurement error. Consequently, it was not feasible to anticipate that the potential net gains obtained from multiplying the mean yield and the dollar value would reflect the actual gains to be derived over time. To measure what effect the variability in the precision levels among the independent variables would be, a quadratic loss function analysis was used. The second objective was to establish the monetary risks of making wheat yield estimates using variable precision data with a quadratic loss function. Before the potential mone- tary losses could be calculated, five assumptions had to be made. The assumptions made in this study were considered to be appropriate at the macro level at which they were used, but the representativeness of the precision levels could vary considerably at a micro level. The Syrian Department of Agriculture's (SDA) utility function was the "determination of potential wheat expansion areas to increase wheat revenue with the highest potential return and the least risk." To assess the SDA's utility function, four potential Actions were considered: A1 = 100% = 75% expansion, A expansion, A = 50% expansion, and A4 = 2 3 25% expansion. These four Actions were tested by five States reflecting the effects of potential reductions in the mean values of the independent variables: B1 = no reduction, B2 = 5% reduction, B3 = 10% reduction, Bu = 15% reduction, and B5 = 20% reduction (i.e., for precipitation B1 = 40 mm, B2 = 38 mm, B3 = 36 mm, B4 = 34 mm, and B5 = 32 mm). 100 The maximum potential gross gain, maximum potential net loss and the minimum potential gross gain were calculated for each Action and State. To better establish the monetary risks the Syrian decision maker would face over time, the available 20 year precipitation record was used to establish an adjusted minimum potential net gain value that represented the MPGG to be expected from any future Action based on the available record. Based on this information, the test site with least risk of no commercial crop and the highest MPGG was Hasseke. The Hasseke site also best fulfilled the Syrian Agricultural Sector's and the potential farmer's utility functions. The third objective was to compute the monetary gains or losses associated with expansion into the test site with the highest return using 1977 dollar values. To determine what the potential gains might be, it was first necessary to es- tablish what the costs and returns for local rainfed wheat were. Included in the costs were the variable cost, family labor, and the cost of conducting this study. The value of the wheat reflected the 1977 government supported price. The actual MPNG for each State and Action was calculated by subtracting the total costs from the adjusted MPGG (re- flecting the probability of occurrence). The potential percentage return on each dollar spent by the Syrian Agri- cultural Sector for wheat production in the Hasseke test site was 8%. 101 Information Value Information value was established by measuring the dif- ference between the total MPG minus total costs and the total adjusted MPNG minus total costs according to the probability of occurrence. The information value equaled the potential loss from overestimating the net gains from a given Action. If the MPG was overestimated, the size of the farm allotments could be underestimated, thus, not deriving enough income to keep the farmers on the land. This would also violate the farmers' common utility function. Recommendations for Future Research How reliable the information should be and what the con- sequences are of using variable precision data to identify crop expansion areas and potential yields pose difficult questions to a decision maker. Ideally the decision maker should specify the precision level for data required for the project prior to beginning a study. By identifying a combined precision level, he has specified how large of a risk he is willing to take. This type of risk can then be quantified in- to monetary gains or losses. However, the major problem in identifying crop expansion areas lies in the need for time series data to establish trends. The measurement devices set up 20 or more years ago and the resultant data do not allow a present day decision maker the luxury of being able to specify precision levels needed and the resultant risk he is willing to take. 102 The procedure described in this work provides a frame- work for establishing precision levels and the determination of monetary risks from a given decision. By calculating minimum potential net gains using a quadratic loss function for several Actions and States, the decision maker can assess several alternatives at one time. For example, for crop groupings that have particularly high national priority or high international market value, the Syrian government sup- ported prices could be set high to induce local farmers to produce those crops. By calculating corresponding yields and MPNG values for each crop grouping (prior to growing periods) and corresponding them to the optimum yield regions, government crop support prices could be set. If dollar re— turns are not sufficient (too much risk), the decision maker can increase the MPNG by prioritizing the regions and col- lecting more primary data. Regions prioritized for increased crop production but indicating too much risk using regional level data could have the risk reduced through further primary data collection at a more refined level. Areas with the most fertile, well drained soils could be identified through fieldwork and given a high priority. Next, areas adjacent to rivers or streams could be mapped and soil bores taken to determine the feasi- bility of irrigation. Once the region's area had been re— duced to those sections with the highest potential, MPNG values could be recalculated for several crop groupings to determine the optimum return based on the established utility 103 function. These MPNG values would reflect higher precision levels for moisture availability and moisture quantities in the prioritized areas. To assess how effective this procedure actually is, it would be necessary to track the MPNG over time and compare it to the estimated MPNG. It is recommended that in similar future projects where multiphase (reconnaissance to regional to detailed) studies will be completed, that this procedure be used. Identification of independent variables for yield estimates and the required precision levels needed at each planning level would be beneficial not only to Syrian planners but also to all decision makers throughout the Middle East. Future studies should give specific attention to the handling of soils data and its overall value as a variable in the yield equation. More attention should be given to rankingtfluatexture of the soil as an independent variable. Comparisons to illustrate how detailed the soil mapping unit descriptions must be to be economically feasible would be extremely beneficial to international studies in underdeveloped nations, as soil mapping is extremely time consuming and costly at the semi—detailed and detailed levels. Similar comparisons need to be made concerning the type and density of meteorolo- gical stations required to provide data at an economically feasible precision level at each study level. APPENDICES 104 .mmmfi «.m.m .mcopom ocm .H.w .mmmccfizcoz mo xmoz so oommm mo. Ho. mo. 00. ma. 0H. NH. ma. NH. no. mo. mo. mo. opfimzanhonfi HH. Ho. mo. mo. NH. mm. :m. mm. OH. OH. :0. mo. mo. :mEcom HH. mo. mo. mo. ma. :m. mm. mm. mH. mo. mo. :0. no. oHUUOLOImocmHm NH. OO. OO. OO. OH. OH. mm. mm. Om. HH. OO. OO. mO. pooosHmHH coo: m m m M MN m .m. a m a m m moozpoz HmownHoem :mECom pew .oaooflholhocmam .opfimzapCHOSB on» new mwcfiowom HmpoEHmmH szpo< hp :oHHMHHchmppoam>m HMHpcopom zHHwQ owopo>< no wcomemQEoo manpcoz < XHszmm< APPENDIX B Syrian Map and Map File Codes for the Soils Map Numeric Code Legend Symbol Used in the on the l:500,000 Description Computer Map File Display Maps l AXRc H/LS Lithic Rhodoxeralfs and associated soils from limestone on hilly topo- graphy. 2 AXRc M/LS Lithic Rhodoxeralfs and associated soils from limestone on mountains. 3 DOAh H/LS Lithic Camborthids and associated soils from limestone on hilly topo- graphy. 4 DOAh HD/LS Lithic Camborthids and associated soils from limestone on maturely dissected plains. 5 DOAh R/LS Lithic Camborthids and associated soils from limestone on rolling plains. 6 DOAh S/LS Lithic Camborthids and associated soils from limestone on steeply sloping hills and escarpments. 7 DOAh U/LS Lithic Camborthids and associated soils from limestone on undulating plains. 8 DOGa H/TS Typic Gypsiorthids and associated soils from weakly consolidated sedi— mentary rocks on hilly topography. 105 106 Appendix B (cont'd.) Legend Symbol on the 1:500,000 Display Maps Numeric Code Used in the Computer Map File Description 9 DOGa HD/TS Typic Gypsiorthids and 10 11 12 13 l4 15 l6 DOGa DOGa DOGa DOGb DOGb DOGd DOGd L/TS R/TS U/TS L/TS U/TS L/TS R/TS associated soils from weakly consolidated sedi- mentary rocks on maturely dissected plains. Typic Gypsiorthids and associated soils from weakly consolidated sedi- mentary rocks on level plains. Typic Gypsiorthids and associated soils from weakly consolidated sedi- mentary rocks on rolling plains. Typic Gypsiorthids and associated soils from weakly consolidated sedi- mentary rocks on undulating plains. Calcic Gypsiorthids and associated soils from weakly consolidated sedi- mentary rocks on level plains. Calcic Gypsiorthids and associated soils from weakly consolidated sedi- mentary rocks on undulating plains. Petrogypsic Gypsiorthids and associated soils from weakly consolidated sedi- mentary rocks on level plains. Petrogypsic Gypsiorthids and associated soils from weakly consolidated sedi— mentary rocks on rolling plains. 107 Appendix B (cont'd.) Numeric Code Used in the Computer Map File Legend Symbol on the l:500,000 Display Maps Description l7 DOGd U/TS Petrogypsic Gypsiorthids 18 19 2O 21 22 23 24 25 DOLa DOLa DOLa DOLe DOLe DOLe DOPa DOPa R/LS U/B U/LS L/LS R/LS U/LS L/LS U/LS and associated soils from weakly consolidated sedi- mentary rocks on undulating plains. Typic Calciorthids and associated soils from limestone on rolling plains. Typic Calciorthids and associated soils from basalt on undulating plains. Typic Calciorthids and associated soils from limestone on undulating plains. Lithic Calciorthids and associated soils from limestone on level plains. Lithic Calciorthids and associated soils from limestone on rolling plains. Lithic Calciorthids and associated soils from limestone on undulating plains. Typic Paleorthids and associated soils from limestone on level plains. Typic Paleorthids and associated soils from limestone on undulating plains. 108 Appendix B (cont'd.) Numeric Code Used in the Computer Map File Legend Symbol on the 1:500,000 Display Maps Description 26 27 28 29 3O 31 32 33 34 35 EFHg EFXa EOHa EOHa EOHa EOHc EOHc EOHc EOHc EOHc L/A L/A L/L L/U U/U H/B H/LS HD/LS L/B R/B Xeric Torrifluvents and associated soils from alluvium on level plains. Typic Xerofluvents and associated soils from alluvium on level plains. Typic Torriorthents and associated soils from loess on level to gently sloping plains. Typic Torriorthents and associated soils from un- consolidated materials on level to undulating plains. Typic Torriorthents and associated soils from un- consolidated materials on undulating plains. Lithic Torriorthents and associated soils from basalt on hilly topography. Lithic Torriorthents and associated soils from limestone on hilly topo- graphy. Lithic Torriorthents and associated soils from limestone on maturely dissected plains. Lithic Torriorthents and associated soils from basalt on level to un- dulating plains. Lithic Torriorthents and associated soils from basalt on rolling plains. 109 Appendix B (cont'd.) Numeric Code Legend Symbol Used in the on the l:500,000 Description Computer Map File Display Maps 36 EOHc U/B Lithic Torriorthents and associated soils from basalt on rolling plains. 37 EOHk L/U Xeric Torriorthents and associated soils from un- consolidated materials on level plains. 38 EOHk R/U Xeric Torriorthents and associated soils from un- consolidated materials on rolling plains. 39 EOHk U/U Xeric Torriorthents and associated soils from un- consolidated materials on undulating plains. 40 EOXa L/U Typic Xerorthents and associated soils from unconsolidated materials on level plains. 41 EOXd H/CH Lithic Xerorthents and associated soils from marl on hilly topography. 42 EOXd L/LS Lithic Xerorthents and associated soils from limestone on level plains. 43 EOXd R/B Lithic Xerorthents and associated soils from basalt on rolling plains. 44 EOXd S/B Lithic Xerorthents and associated soils from basalt on steep hills. 45 EOXv L/LS Xerorthents, lithic ver- tic phase and associated soils from limestone on level to undulating plains. 110 Appendix B (cont'd.) Numeric Code Used in the Computer Map File Legend Symbol on the 1:500,000 Display Maps Description 46 47 48 49 50 51 52 53 54 55 EOXv EOXv EOXv EOXV EOXv IAHa IAHb IAHb IAHb IASa R/LS U/LS R/B H/LS U/B L/U L/A L/CH L/D L/A Xerorthents, lithic ver- tic phase and associated soils from limestone on undulating plains. Xerorthents, lithic ver- tic phase and associated soils from limestone on undulating plains. Xerorthents, lithic ver- tic phase and associated soils from basalt on rolling topography. Xerorthents, lithic ver- tic phase and associated soils from limestone on hilly topography. Xerorthents, lithic ver- tic phase and associated soils from basalt on un- dulating plains. Typic Haplaquepts and associated soils from un- consolidated materials on level plains. Aeric Haplaquepts and associated soils from alluvium on level plains. Aeric Haplaquepts and associated soils from marl on level plains. Aeric Haplaquepts and associated soils from colluvium on level plains. Typic Haplaquepts and associated soils from alluvium on level plains. Appendix B (cont'd.) 111 Numeric Code Used in the Computer Map File Legend Symbol on the 1:500,000 Display Maps Description 56 57 58 59 60 61 62 63 64 65 IASb IOXa IOXa IOXh IOXh IOXh IOXh IOXh IOXh IOXk L/A L/U R/U H/TS R/LS S/B S/CH S/LS U/LS L/U Aeric Haplaquepts and associated soils from alluvium on level plains. Typic Xerochrepts and associated soils from un— consolidated materials on level plains. Typic Xerochrepts and associated soils from un- consolidated materials on rolling plains. Lithic Xerochrepts and associated soils from weakly consolidated materials on hilly topo- graphy. Lithic Xerochrepts and associated soils from limestone on rolling plains. Lithic Xerochrepts and associated soils from basalt on steep hills. Lithic Xerochrepts and associated soils from marl on steep hills. Lithic Xerochrepts and associated soils from limestone on steep hills. Lithic Xerochrepts and associated soils from limestone on undulating plains. Vertic Xerochrepts and associated soils from un- consolidated materials on level plains. 112 Appendix B (cont'd.) Numeric Code Used in the Computer Map File Legend Symbol on the l:500,000 Display Maps Description 66 67 68 69 70 71 72 73 74 75 IOXk IOXk VXCa VXCa VXCa VXCa VXCa VXPa VXPa VXPa R/U U/U L/LS L/U R/B U/B U/LS L/SD R/B U/B Vertic Xerochrepts and associated soils from unconsolidated materials on rolling plains. Vertic Xerochrepts and associated soils from unconsolidated materials on undulating plains. Typic Chromoxererts and associated soils from limestone on level plains. Typic Chromoxererts and associated soils from unconsolidated materials on level plains. Typic Chromoxererts and associated soils from basalt on rolling plains. Typic Chromoxererts and associated soils from basalt on undulating plains. Typic Chromoxererts and associated soils from limestone on undulating plains. Typic Pelloxererts and associated soils from calcareous sandstone on level plains. Typic Pelloxererts and associated soils from basalt on rolling plains. Typic Pelloxererts and associated soils from basalt on undulating plains. 113 Appendix B (cont'd.) Numeric Code Legend Symbol Used in the on the l:500,000 Description Computer Map File Display Map 76 DOLe HD/LS Lithic Calciorthids and associated soils from limestone on maturely dissected plains. 77 EOXd U/B Lithic Xerorthents and associated soils from basalt on undulating plains. 78 EOHc S/LS Lithic Torriorthents and associated soils from limestone on steep hills. 79 VXCa L/B Typic Chromoxererts and associated soils from basalt on level plains. 80 EOXd R/U Lithic Xerorthents and associated soils from basalt on rolling topo- graphy. 81 DOPa HD/LS Typic Paleorthids and associated soils from limestone on maturely dissected plains. BIBLIOGRAPHY BIBLIOGRAPHY Ackerson, K., Miller L. and Atchley, A. Land Resources. Vol. II of Syria Agricultural Sector Assessment. Edited by W.A. Faught. 5 Vols. Damascus, Syria, 1980. Adnam, D. "An Evaluation of Remote Sensing With Respect to Crop Acreage Estimation in Canada." Report presented to the Crops Section Agriculture Division Statics Canada. Ottawa, Canada, January 1975. Arrow, K.J. "Economic Welfare and the Allocation of Re- sources for Invention." The Rate and Direction of In- ventive Activity: Economic and Social Factors. National Bureau of Economic Research. Princeton University Press. 1962. Aspects of Theory of Risk Bearing. Helsinki: Aca- demic Bookstore. 1965. Essays in the Theory of Risk-Bearing. Chicago: Markum Publishing Co. 1971. Barry, R.G. "Evaporation and Transpiration." Climate in Review. Edited by Geoffrey McBoyle, Boston: Houghton Mifflin Co. 1973. Bates, D.G. "The Role of State in Peasant-Nomad Mutualism." AnthrOpological Quarterly. Vol. 44:3 (1978), pp. 109-131. Bauer, M.E. and Cipra, J.E. Identification of Agricultural Crops by Computer Processing of ERTS MSS Data. LARS Re- port, West Lafayette, Indiana. 1973. Baumgardner, M. "Remotely Sensed Multispectral Scanner Data-—A Tool for Soil Surveys." A seminar presented at Michigan State University, November 13, 1980. Baumgardner, M.F., Anderson, J.A. and LARS Staff. Mapping Soils, CrOps and Rangelands by Machine Analysis of Multi— Temporal ERTS-l Data. LARS Report, West Lafayette, Indiana. 1974. Blalock, H. Social Statistics. New York: McGraw-Hill. 1972. 114 115 Carter, D. Three Water Balance Maps of Southwest Asia. Vol. XI. Centerton: Laboratory of Climatology. 1958. . Applications of the Water Budget in Physical Geo- graphy. Vol. XXVI. Elmer: Thornthwaite Associates Laboratory of Climatology. 1973. Chappelle, D.E. "Precision Required in Allowable Cut Deter- minations on the National Forests: An Inquiry into the Nature of the ApprOpriate Loss Function." Paper sub- mitted to the Intermountain Forest and Range Experiment Station, East Lansing, Michigan. February 1975. "Quantitative Analysis in a Qualitative World: Modeling Forestry Systems to Improve Decision-Making." Workshop at the Society of American Foresters National Convention, Cleveland, Ohio. September 26-30, 1971. "How Much is Information Worth?" Proceedings: Re- source Data Management Symposium. West Lafayette, In- diana. 1976. Clawson, M., Landsbert, H.H. and Alexander, L.T. The Agri- cultural Potential of the Middle East. New York: Ameri- can Elsevier Publishing Co. 1971. Clough, D.J. Remote Sensing and Economic Value of Errors in Wheat Forecasting. Report to the Canada Centre for Remote Sensing, Ottawa, Canada. 1974. Cochran, W.G. Sampling Techniques. New York: John Wiley and Sons, Inc. 1963. Sampling Techniques. 2d ed. New York: John Wiley and Sons. 1977. Committee on Remote Sensing. Resource Sensing from Space: Prospects for Developing Countries. Washington, D.C.: National Academy of Sciences. 1977. CRIES. Honduras Land Resource Base Report. East Lansing, Michigan. November 1980. Deming, W.E. Sample Design in Business Research. New York: John Wiley and Sons, Inc. 1960. Deschamps, R. "Risk Aversion and Demand Function." Eco- nometrica. 41(3), May 1973. Dummeier, E.F., Heflebower, R.B. and Norman, T. Economics With ApplicatiOns to Agriculture. New York: McGraw- Hill. 1950. 116 Duncan, G.T. "A Matrix Measure of Multivariate Local Risk Aversion." Econometrica. 45(4), May 1977. Eagleman, J.R. "An Experimentally Derived Model for Actual Evapotranspiration." Agricultural Meteorology, Vol. 8, June, 1971, pp. 385—394. Economics and Statistics Service. "1981 AgRISTARS, Domestic Crops and Land Cover Project." U.S.D.A. Temporary Direc— tive, Washington, D. C., January 28, 1981. Eisgruber, L.M. Potential Benefits of Remote Sensing Theo- retical Framework and an Empirical Estimate. Report to the Laboratory for Applications of Remote Sensing, Purdue University, West Lafayette, Indiana. 1972. Fisher, W.D. Clustering and Aggregation in Economics. Bal- timore: The Johns Hopkins Press. 1969. Frazee, C.J., Myers, V.I. and Westin, F.C. "Density Slicing Techniques for Soil Survey." Proceedings: Soil Science Society American. 1972. Graaff, J. De V. Theoretical Welfare Economics. Cambridge University Press. 1967. Hamburg, M. Statistical Analysis for Decision Making. New York: Harcourt, Brace and World. 1970. Hamilton, D.A. Precision Requirements for Some Information in Timber Management Decisions. Unpublished Dissertation, Department of Forestry, Ames, Iowa: Iowa State University. 1970. Hanoch, G. "Risk Aversion and Consumer Preferences." Eco- nometrica, 45(2), March 1977. Hanuschak, G.A. "Precision of CrOp-Area Estimates." Pro- ceedings of the Thirteenth International Symposium on Remote Sensing of Environment. Ann Arbor, Michigan. 1979. Hargreaves, G.H. World Water for Agriculture. Report for AID Contract, Salt Lake City, Utah, January 1979. Hayami, Y. and Peterson, W. "Social Returns to Public In- formation Services. The Case of Statistical Reporting of U.S. Farm Commodities." American Economic Review. April 1972. Hossain, A. and Chandburg, M.U. "Utilization of Remote Sen- sing Technology in Natural Resources Development in Bangladesh." Remote Sensing of Earth Resources. 1975. Howe, C.W. Natural Resource Economics. New York: John Wiley and Sons. 1979. 117 Kahneman, D. and Tversky, A. "Prospect Theory: An Analysis of Decision Under Risk." Econometrica. 47(2), March 1979. Keeney, R.L. "Risk Independence and Multiattributed Utility Functions." Econometrica. 41(1), January 1973. Kendall, M.G. and Buckland, W.R. A Dictionary of Statistical Terms. New York: Hafner Publishing Co. 1960. Kibler, W.E., Thome, P.G., Potter, T.D. and Doyle, F.S. AgRISTARS: Management/Organization Plan. Report to the U.S.D.A., NASA, U.S.D.C. and U.S.D.I., Washington, D.C., January 1980. AgRISTARS: Technical Program Plan. Report to the U.S.D.A., NASA, U.S.D.C. and U.S.D.I., Washington, D.C., January 1980. King, K.M. "Evaporation from Land Surface." Proceedings: Hydrology Symposium No. 2, Evaporation. 1961. Lee, R. Forest Microclimatology. New York: Columbia Uni- versity Press. 1978. MacDonald, R.B. and Hall, F.G. "Global Crop Forecasting." Science (May 1980), pp. 670, 679. Magnusson, G. Production Under Risk: A Theoretical Study. Sweden: Almquist and Wiksells Boktrysckeri AB Uppsala. 1969. McGuinness, J.L. and Bordne, E.F. A Comparison of Lysimeter- Derived Potential Evapotranspiration with Computed Values. U.S. Department of Agriculture Tech. Bu11., No. 1452, Washington, D. C., 1972. Ministry of Social Affairs and Labor, Syria. Statistical Abstract, 1978. Damascus, Syria. 1978. Mokma, D. Personal communication. January 1982. Myers, V., Moore, D., DeVries, M. and Worcester, B. "Remote Sensing for Monitoring Resources for Development and Con- servation of Desert and Semi-Desert Areas." Proceedings of the American Society of Photogrammetry. 1978. NASA. Proceedings of Technical Sessions: The LACIE Sym- posium. Houston, Texas, October 1978. Neter, J. and Wasserman, W. Applied Linear Statistical Models. Homewood: Richard D. Irwin, Inc. 1974. Nix, H.A. and Fitzpatrick. "An Index of CrOp Water Stress Related to Wheat and Grain Sorghum Yields." Agricul— tural Meteorology, Vol. 6:12 (1968), pp. 321-337. 118 Park, A.B. Program Plan for Developing the Capability of Forecasting Crop Production. Report to FAO, Rome, Italy. 1975. Pollack, R.A. "The Risk Independence Axiom." Econometrica. 41(1), January 1973. Pratt, J.W. "Risk Aversion in the Small and in the Large." Econometrica. 32(1-2), January-April 1964. Riemenschneider, C.H. "Economic Structure, Price Discovery Mechanisms and the Informational Content and Nature of USDA Prices." Paper presented at AAEA-USDA Agricultural and Rural Data Workshop, May 4-6, 1977. Richardson, A.J., Wiegand, C.L., Torline, R.J. and Gautreaux, M.R. "Landsat Agricultural Land-Use Survey." Photogram— metric Engineering and Remote Sensing, Vol. 43:2 (1977), pp. 207-216. Safley, J. "Comparison of Methods for Estimation of Drought Intensity and the Relationships of Drought to Yields of Agronomic Crops in Tennessee." Unpublished Ph.D. disser- tation, University of Tennessee. 1974. Swagler, R. Caveat Emptor. Lexington, Massachusetts: D.C. Heath. 1975. Swalm, R.O. "Utility Theory--Insights into Risk Taking." Concepts and Applications of Modern Decision Models. Ed. Tummalu, V.M. Rod Henshaw, R.G., East Lansing: MSU Busi- ness Studies. 1976. Syrian Ministry of Defense. Climatic Atlas of Syria. Damas— cus. 1977. Terjung, W.H. "Climatology for Geographers." Annals of American Assoc. of Geographers. Vol. 51 (1961), pp. 345- 355. Theil, H. Optimal Decision Rules for Government and Indus- try. Amsterdam: Holland Publishing Co. 1964. Thomas, R. and Hag, C. "Two Phase Sampling for Wheat Acreage Estimation." Machine Processing of Remotely Sensed Data Symposium. 1977. Thornthwaite, C.W. "An Approach Toward a Rational Classi- fication of Climate." Geographical Review. 1948. Thornthwaite, C.W. and Mather, J.R. Estimating Soil Trac- tionability from Climatic Data. Vol. VII. Centerton: Laboratory of Climatology. 1951. 119 U.S. Department of Agriculture. Soil Taxonomy, A Basic System for Soil Classification for Making and Inter- preting Soil Surveys. Washington, D. C.: U.S. Depart— ment of Agriculture. 1975. VanLiere, W.J., et a1. Soil Map of Syria, 1:500,000. 1968. Ward, R.C. Principals of Hydrology. New York: McGraw- Hill. 1967. Weismiller, R.A., Persinger, I.D. and Montgomery, O.L. "Soil Inventory from Digital Analysis of Satellite Scanner and Onagraphic Data." Soil Science Society of America. 1977. Westin, F.C. and Frazee, C.J. "Landsat Data, Its Use in a Soil Survey Program." Soil Science Society of America Proceedings. 1978. Willmot, C.J. WATBUG: A Fortran IV Algorithm Calculating the Climatic Water Budget. Report to the C.W. Thornthwaite Laboratory of Climatology, Vol. 30, Elmer, New Jersey. 1977. Witter, S. and Hill—Rowley, R. The Development of a Genera— lized Land Cover/Use Map of Syria Through Visual Inter— pgetation of Landsat Imagery. A report to the Syrian Agricultural Sector Assessment Project, East Lansing, Michigan. 1979. Witter, S., Schultink, G. and Lusch, D. Land Cover/Use Inventory of Landsat Imagery. Report to SIstema De In- ventorio y Evaluation De RECUNSOS Agricola, East Lansing, Michigan. 1980. Wonnacott, T. and Wonnacott, R. Introductory Statistics for Business and Economics. New York: John Wiley and Sons, Inc. 1972. Worcester, B.K. and Moore, D. "Delineation of Soil Land— scapes in the Sudd Region of Sudan on Landsat Imagery." Brookings: Remote Sensing Institute. 1978. Worcester, B.K. and Dalstead, K.J. "Applicability of Land- sat Imagery for Mapping Soils in Arid and Semi-Arid Regions." Remote Sensing of Earth Resources. Edited by F. Shahrokhi and T. Paludan. Tullahoma: U.T. Space Institute. 1980. Yao, A.Y.M. "The R Index for Plant Water Requirement." Agricultural Meteorology, Vol. 6, December, 1968, pp, 259- 273. T 71777777 If! (11171111111111ll 10740 8274