EVALUATION OF PELLET GROUP SURVEYS FOR ESTIMATING DEER POPULATIONS IN MICHIGAN By Lawrence Atwell Ryel A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Zoology 1971 P L E A S E NO TE : S o m e pa ge s m a y have i n d i sti nc t print. F i l m e d as received. U n i v e r s i t y M i c r o f i l m s , A Xe ro x Educ ati on C o m p a n y ABSTRACT EVALUATION OF PELLET GROUP SURVEYS FOR ESTIMATING DEER POPULATIONS IN MICHIGAN BY Lawrence Atwell Ryel Deer pellet group surveys are used and recommended by wildlife bio logists as a deer census technique throughout the northern and western deer ranges. The objective of this study was to evaluate the technique as used in Michigan. The appraisal was based on an examination of data produced by the considerable amount of research and survey work done in Michigan from 1953 through 1969. Use of pellet group counts as a census method is dependent upon sev eral assumptions: deer defecate at a rather constant frequency, pellet groups persist long enough to be counted, groups can be found and counted accurately, a deposition period can be delineated, and groups found can be aged relative to the deposition period. Calculations of deer numbers from pellet group counts utilize the algorithm: mean pellet groups ^ deer population = 1 ^ per plot pjlqt size size of area deposition period X defecation rate For a given survey, plot size, area, deposition period, and defecation rate have been generally considered fixed. Defecation rates were studied using penned deer fed several diets at the Cusino Wildlife Research Station in the Upper Peninsula and at the Houghton Lake Wildlife Research Station in northern Lower Michigan. Studies were also undertaken at both stations to determine longevity of pellet groups in several types of overstory and ground cover. Fifteen experimental pellet group surveys were conducted on two Lawrence Atwell Ryel large enclosures with known numbers of deer, the 1.8 square mile George Reserve in southern Michigan and the 1 square mile Cusino enclosure. Systematic sampling designs were employed for the most part. Extensive operational pellet group surveys were carried out in the Upper Peninsula and northern Lower Michigan since the early 1950's. Since 1959, both regions have been surveyed annually. Two-stage, strati fied random sampling with optimum allocation was used. was based on expected over-winter deer populations. units were geographic sections (square miles). Stratification The first stage Second stage sampling units (courses) consisted of eight rectangular 1/50-acre plots, 72.6 feet by 12 feet, spaced 5 chains apart in a straight line. Courses were lo cated within sections by restricted randomization procedures. Average over-winter population levels, obtained directly from pellet group counts, were adjusted to provide estimates of the spring and previous fall herds by considering proportional contributions from deer known to have been removed, i.e. legal harvest and over-winter mortality. Various aspects of these surveys were analyzed. Studies of defecation rates yielded the following means per deer per day: adult bucks 15.61, adult does 12.89, buck fawns 14.74, and doe fawns 11.89. Indices for a given survey should be weighted by herd composition. Groups of deer fed various winter diets showed little difference in defe cation rates within sex and age classes. The available data are less than decisive concerning whether there is a decrease in defecation rates in mid-winter. No data are available for the period from May through December. All Michigan work has been based on a fall-to-spring accumulation of pellet groups. The autumnal fall of leaves has been used to provide a Lawrence Atwell Ryel beginning reference point and the mean date of the spring survey the ter minus . Leaf fall dates for 11 years were relatively uniform within north ern game districts; the range in most cases was less than 2 weeks. The average time interval in the spring required to search all sam ple plots within a game district was 24 days. Adjusting all pellet group counts to a common spring date had little effect on estimates. Making accurate counts was found to be the most important phase of the technique, including the finding of all groups present through dili gent searching, and ascertaining the age of groups found relative to the deposition period. The problem of missing groups was resolved to a large extent by using two experienced men per crew, each checking the other's work. Studies of pellet weathering revealed nearly all groups deposited during fall and winter were extant in the spring. In addition, many older groups were also visible, up to 5 years in some instances. For survey purposes, groups dropped during the established deposition period are termed "new," while those dropped prior to the period are "old." All groups found on top of recent leaf litter should be considered new regardless of appearance. Correctly aging groups on sites devoid of fallen leaves appears to be a serious problem. groups from sample plots in Attempts to remove pellet the fall were not successful. A group should be counted only if its midpoint falls within the plot. If possible, pellet group counts should be avoided on bright, sunny days, which makes groups difficult to see in the contrasting light, or immedi ately following a rain, which tends to make old groups appear fresh. Results of 10 experimental pellet group surveys on the George Reserve showed a very poor fit with known populations. Available evidence pointed Lawrence Atwell Ryel to mistakes in aging pellet groups as the major problem. Five surveys on the Cusino enclosure revealed a much better agreement. Frequency distributions of pellet group counts on the Cusino enclo sure, the George Reserve, and three or four extensive areas in the northern Lower Peninsula were satisfactorily described by fitted negative binomial distributions. The relationship between the log of the means and the log of the variances, however, provided a simpler way of estima ting standard deviations for sample allocation purposes. Variances used included only estimates of sampling error in counting pellet groups. Additional contributions due to adding variance components for defecation rate and deposition period were found to be small enough to be safely dis regarded. Accuracy of extensive surveys could not be properly appraised since real herd sizes were unknown. fair agreement. Comparison with other indices indicated a Evidently the technique is more accurate under Upper Peninsula weather conditions and cover types. For estimates of equal precision, stratified sampling resulted in considerable savings in man power when compared to simple random sampling. Stratification was not accurate enough, however, to warrant annual selection of new samples. No differences were found in pellet group counts by position on course lines. Analysis of time records and comparisons of within course vari ability indicate optimal strategy would be to reduce the number of plots per course to five and increase the number of courses. ACKNOWLEDGMENTS Grateful acknowledgments are made to my graduate committee, Drs. William E. Cooper, Chairman, George J. Wallace, John L, Gill, and George A. Petrides for their guidance during the course of my studies and in the preparation of this manuscript. Dr. Cooper deserves special credit for supplying the necessary impetus for me to complete my studies. Special thanks are due to Dr. Walter H. Conley and especially to Dr. Charles E. Cress for graciously consenting to fill in for Dr. Petrides and Dr. Gill respectively, both of whom were out of the country when I completed my write-up. I am grateful to the Michigan Department of Natural Resources for the opportunity to conduct much of this study as a component of my regu lar job. My supervisor, Dr. David H. Jenkins, Chief of the Research and Development Division, offered continual encouragement and supplied me with invaluable information on the E. S. George Reserve and its deer herd. My colleagues Carl L. Bennett, Jr. and Louis J. Hawn also contri buted valuable suggestions and assistance. This study was conducted as part of my work as leader of Federal Aid in Wildlife Restoration Project W-96-R. Special thanks are due the late Dr. Phillip J. Clark, Department of Zoology at Michigan State, and my former supervisor, Dr. L. Lee Eberhardt for their council in the initial stages of this study. I would like to thank Dr. Francis C. Evans, Department of Zoology and Associate Director of the E. S. George Reserve, and Dr. Warren W. Chase, formerly professor of Wildlife Management, both at the University of Michigan, for allowing me to carry out experimental pellet group ii surveys on the Reserve and for providing me with information on the deer harvests there. Drs. Archibald B. Cowan and Dale R. McCullough also kindly supplied me with deer harvest and population data from the Reserve. Literally hundreds of individuals from the DNR and the U. S. Forest Service participated in collecting data, and without their help this study would not have been possible. Benjamin C. Jenkins, the DNR’s northern game habitat specialist, and Herman F. Olson, formerly Regional wildlife specialist, U. S. Forest Service, Milwaukee, Wisconsin, were responsible for initiating me in the use of the pellet group survey technique. I am especially grateful to wildlife research biologists Robert C. VanEtten, Louis J. Verme, Edward J. Mikula, and Rodney R. Smith who carried out research studies at the Cusino Wildlife Research Station; and to Arlow P. Boyce, Elsworth M. Harger, and Herbert E. Johnson who conducted similar studies at the Houghton Lake Wildlife Research Station. Dr. M. L. "George" Dudzinski, Commonwealth Scientific and Industrial Research Organization, Canberra, Australia, supplied me with a computer program for fitting the negative binomial distribution to frequency data. Dennis Burck, DNR computer programmer, modified this for use on the Bur roughs 5500 computer. I will be forever indebted to Mrs. H. Rex Caster, Ms. Marilyn L. Moss, and Mrs. Douglas L. Meister for their "beyond the call of duty" efforts to read my writing, correct my spelling, and type the many versions of the manuscript. Dean B. Armstrong did his usual excellent job in draft ing the figures. Especial appreciation is made to Mrs. Bernadette McCarthy Henderson, Department of Zoology, for her constant inspiration and encouragement during my years at Michigan State. iii Finally, I wish to thank my wife Rosie, daughter Gale, and son Ron for their support, patience, and understanding while I spent my nights and weekends engaged in studies and writing. iv TABLE OF CONTENTS Page I. INTRODUCTION.................................. O b j e c t i v e s .......... . ................................. .................................. Deer census techniques Michigan deer census methods ........................... Scope of the present study ................... . . . . . II. THE PELLET GROUP TECHNIQUE .................................... Terminology.................... Historical b a c k ground .................. Basic c p n c e p t s ............................................ Additional applications of fecal studies ............... III. 1 1 4 7 14 14 14 15 17 GENERAL M E T H O D O L O G Y ............................................. 19 Determination of defecationrates ........................ Weathering of pellet groups............................... Experimental field trials ............................... Extensive surveys ........................................ IV. 1 DETERMINATION OF DEFECATION RATES ............................ 19 22 23 27 33 B a c k g r o u n d .................................................33 Previous studies ........................................ 33 C o p r o p h a g y ................ ................................37 Results of Michigan studies ............................... 37 Sex. and age differences . ................................ 38 Differences‘between diets ............................... 42 Differences between times of year ....................... 44 Corral studies .......................................... 51 V. DEPOSITION P E R I O D ............ 53 Leaf Fall D a t e s .............................................. 53 R e s u l t s .....................................................54 D i s c u s s i o n .................................................57 Dates of Spring S u r v e y s ...................................... 59 R e s u l t s ..................................................... 60 v TABLE OF CONTENTS - Continued VI. Page COUNTING AND AGING PELLET G R O U P S ................................ 70 Weathering of Pellet Groups ............................... 70 Pre-leaf fall g r o u p s ...................................... 70 Post-leaf fall g r o u p s ...................................... 71 Spring and summer groups ............................... 81 81 Determination of pellet a g e s .................... Clearing plots and marking g r o u p s ................ 83 D i s c u s s i o n .................................................84 Counting Errors ............................................ 87 R e s u l t s ............................... 87 The "Deming m e t h o d " ........................................ 96 D i s c u s s i o n .................................................98 Fecal Identification......................................... 100 VII. EXPERIMENTAL FIELD TRIALS .................................... 103 George Reserve .............................................. 103 Description of the area .........................103 The deer h e r d ............................................. 104 R e s u l t s ....................................................110 120 Cusino E n c l o s u r e ............ Description of the a r e a ................................... 120 The deer h e r d ............................................. 122 R e s u l t s ....................................................123 D i s c u s s i o n ................................................ . 125 VIII. EXTENSIVE SURVEYS ............................................ 130 Deer Population Estimates ................................. 130 Distribution of Pellet Groups ............................. 134 First Stage S a m p l e s ......................................... 135 Sample allocation ........................................ 136 Accuracy of stratification ............................. 137 Sample sizes and effort ................................. 146 Effect of stratification ............................... 150 Second Stage Samples ........................................ 151 General f o r m a t ........................... 151 Plot location b i a s e s ..................................... 152 Plot size and s h a p e ............................... 153 Optimum number of plots per c o u r s e ...................... 156 IX. SUMMARY AND C O N C L U S I O N S ....................................... 159 APPENDIX A - THE MICHIGAN DEER R E S O U R C E .............................165 APPENDIX B - SUPPLEMENTAL T A B L E S ..................................... 168 BIBLIOGRAPHY ........................................................ vi 225 LIST OF TABLES Table 1. ige Comparison of pellet group size and of defecation rates in Cusino deer, late February, 1954 ..................... 49 2. Summary of leaf fall dates by game districts .............. 55 3. Simple linear correlation matrix of leaf fall dates by game d i s t r i c t s ............................................ 56 4. Distribution of major cover types in northern Michigan 1966 58 5. Dates of pellet group surveys by game district ............ 61 6. Frequency distribution of courses run by date, 1962 northern surveys .......................................... 64 Ranks of the mean-date plots were searched by strata for selected game district surveys ........................... 66 Extended median tests of survey dates among strata for selected game district surveys ........................... 67 Comparison of survey means from original records with those computed from counts adjusted to a common spring survey date for selected game district surveys ................. 69 Study of weathering of deer pellet groups at Houghton Lake 1954-55 .............................................. 72 Proportion of post-leaf fall deer pellet groups visible in succeeding years in the Cusino enclosure 1953-57 . . . 76 New and old pellet groups found in the spring with and without "removal" in the fall on surveys in districts 7 and 8 . . ............................................ 85 Results of rechecking plots in the George Reserve following the 1955 pellet group survey ............................. 89 Summary of counting errors on the 1956 pellet group survey in the Cusino enclosure ................................. 90 Summary of counting errors on the 1958 pellet group survey in the George Reserve ................................... 91 7. 8. 9. 10. 11. 12. 13. 14. 15. vii L.._____ LIST OF TABLES - Continued 16. Page Comparison of original counts with later rechecks*on the 1966 northern s u r v e y s ............................... 95 17. Known deer populations for the George Reserve based on aging m e t h o d .............................................. 107 18. Comparison of deer drive counts with population estimates based on the aging method for the George R e s e r v e ........ 109 19. Average composition of the overwinter deer herd on the George Reserve ............................................ 20. Summary of results of deer pellet group surveys on the George Reserve ............................................ 109 Ill 21. Cover type areas on the George Reserve based on pellet group s u r v e y s ............................................ 112 22. Proportion of plots counted during deer pellet group surveys on the George R e s e r v e ........................... 114 23. Effect of estimating groups on uncounted plots for deer pellet group surveys on the George Reserve .............. 114 24. Observed old and new pellet groups by cover types on the George R e s e r v e ........................................116 25. Intercorrelations of weather factors, results of pellet group surveys and deer populations on the George Reserve 118 26. Mean new groups per plot by cover types from deer pellet group surveys on the George R e s e r v e ..................... 121 27. Summary of results of the deer pellet group surveys in the Cusino enclosure ............................. 124 28. Statistical analysis of the 1962 deer pellet group survey in district 7 ....................... .............. 131 29. Corrections for deer removals for the 1962 deer pellet group survey in district 7 ............................... 132 30. Comparison of actual results with stratification for deer pellet group surveys in game districts 31. Ranks of chi-square values for goodness of fit tests comparing actual allocations with optimum allocations viii 141 . . 145 LIST OF TABLES - Continued 32. Page Effort required for deer pellet group surveys, 1960-1969 147 33. Number of deer pellet group courses required for various levels of precision based on results of 1960 game district surveys ................. . . . . . . . 149 34. Comparison of stratified random sampling with simulated simple random sampling for selected game district s u r v e y s .....................................................149 35. Friedman's tests on the rank of pellet group abundance by position on course l i n e s ................................ 154 36. Estimates of the optimum number of plots per course based on 1958 deer pellet group s u r v e y s ..........................158 37. Chronology of deer pellet group surveys in Michigan 38. Defecation rates for adult bucks in Michigan feeding trials 172 39. Defecation rates for adult females in Michigan feeding t r i a l s .................................................... 173 40. Defecation rates for buck fawns in Michigan feeding trials 175 41. Defecation rates for doe fawns in Michigan feeding trials 176 42. ANOVA table, defecation rates for three sex and age classes of deer on a hardwood-conifer diet, Cusino 1953 178 ANOVA table, defecation rates for two sex and age classes of deer on six diets for six dates between January and April, Cusino 1954 .................. 178 ANOVA table, defecation rates for four sex and age classes of deer on the control diet, Cusino 1959 ................. 179 ANOVA table, defecation rates for three sex and age classes of deer on the swamp-conifer diet, Cusino 1959 ........... 179 ANOVA table, defecation rates for adult does on the control and swamp-conifer diets, Cusino 1959 ..................... 180 ANOVA table, defecation rates for buck fawns on the control and swamp-conifer diets, Cusino 1959 ..................... 180 43. 44. 45. 46. 47. ix . . . . 168 LIST OF 48. TABLES - Continued Page ANOVA table, defecation rates for doe fawns on the control and swamp-conifer diets, Cusino 1959 ..................... 181 49. Deposition periods for pellet group surveys................. 182 50. Performance of individuals in counting pellet George Reserve, 1958 .................. 185 51. 52. groups, Performance of individuals in counting pellet groups, George Reserve, 1959 ...................................... 185 Performance of individuals in counting pellet groups, George Reserve, 1960 ...................................... 186 53. Performance of individuals in counting pellet groups, George Reserve, 1 9 6 1 ........................................ 186 54. Performance of individuals in counting pellet groups, George Reserve, 1962 ...................................... 187 Performance of individuals in counting pellet groups, George Reserve, 1963 .................................... 187 Comparison of original counts with later rechecks, northern Lower Peninsula survey, 1956 188 Comparison of original counts with later rechecks, northern Lower Peninsula surveys, 1957 ................... 189 Comparison of original counts with later rechecks, Upper Peninsula surveys, 1957 190 55. 56. 57. 58. 59. Calculation of deer populations for George Reserve based on removal and aging m e t h o d ......................... 191 60. Comparisons of counts of deer pellet groups with fitted negative binomial distributions for the George Reserve. . 205 Comparisons of counts of deer pellet groups with fitted negative binomial distributions for the Cusino enclosure 209 61. 62. Deer population estimates and antlered buck harvests for game d i s t r i c t s .............................................. 211 63. Deer population estimates and antlered buck harvests for game d i s t r i c t s .............................................. 216 x LIST OF TABLES - Continued 64. 65. 66. Page Comparisons of counts of deer pellet groups with fitted negative binomial distributions for northern surveys in 1953 ................................. 218 Comparison of optimum with actual allocation for deer pellet group surveys in game districts, 1959 to 1964 . . . 220 Comparison of optimum with actual allocation for deer pellet group surveys in game districts, 1965 to 1969 . . . 223 xi LIST OF ILLUSTRATIONS Figure Page 1. Locations of deer pellet group surveys, 1953 to 1957 . . . . 8 2. Locations of deer pellet group surveys, 1958 ............... 9 3. Locations of deer pellet group surveys, 1959-1964 10 4. Locations of deer pellet group surveys, 1965-1969 11 . ......... 43 5. Defecation rates for Michigan white-tailed deer 6. Diagramatic representation of a hypothetical Michigan deer h e r d ...................................................45 7. Disappearance of pellet groups over time in several overstory cover types. Cusino Wildlife Research Station 74 Model of pellet group counting; data are from the 1956 Cusino enclosure survey ................................. 79 9. Location of pellet group courses 1959-1964 surveys ......... 138 10. Location of pellet group courses 1965-1969 surveys ......... 139 8. xii I. INTRODUCTION Objectives The ultimate goal of this and related studies is the proper manage ment of the white-tailed deer (Odocoileus virginianus) in Michigan. The short-term objective is to evaluate pellet group surveys for censusing deer under Michigan conditions. At present, this technique is the one most universally recommended by wildlife biologists throughout the north ern and western deer ranges, and has been used and studied in Michigan for a number of years. The Michigan deer herd is a valuable renewable resource (Appendix A) and its perpetuation is an important activity of the Department of Natural Resources.* Deer census techniques Petrides (1968) has defined conservation as "... a philosophy which advocates the use of natural resources on a sustained and permanent basis in the public interest." Moreover, as Lauckhart (1953) states "... renewable natural resources are managed for maximum harvest and game is no exception. The success of a game management pro gram is measured by the game in the hunter's bag and fish in the fisherman's basket." One of the major problems of wildlife managers is determining the size of the populations they are attempting to manage. The magnitude of their problems varies inversely with the closeness of the harvest to *0n November 15, 1968, the name of the Michigan Department of Conserva tion was changed by law to the Michigan Department of Natural Resources. 1 sustained level of the populations involved. Carhart (1946) stated it quite simply when.he said, "The first step in a management plan for deer is to secure a good estimate of the number of deer on a given range." This basically is the problem considered here. While it may be the first step, it is also perhaps the most difficult. Population estimators have been the concern of wildlife workers since the inception of the modern profession in the 1930*s and the in formed naturalists and gamekeepers that preceded them. It is safe to say that present day professional wildlifers are still looking for better methods to use with nearly every game species, and the same is true of zoologists working with wild populations of many kinds of animals. The difficulties of wildlife census may not be obvious to the un informed. Only in the case of species in imminent danger of extinction and certain gregarious forms, is it possible to see an entire population (Dasmann 1964, Gilbert 1967). Getting reliable information, at reason able costs, on the abundance of a game species distributed over millions of acres each year is not an easy task (Ruhl 1932). wildlife requires that some form of sampling be used. Censusing most The real problem is the difficulty of getting accurate counts of the animals on selected sample plots, not the lack of statistical methods (Dice 1941). Ancestors of white-tailed deer have been identified in some North American Miocene strata, and Pleistocene forms have been found which are almost identical with present species of Odocoileus (Scott 1937). Thus, deer have survived for several million years in North America and con tinue to be widespread occurring even in areas with high human popula tions. The survival of deer is primarily due to the continued presence of suitable food and cover, their behavioral and physiological res ponses to seasonally inclement weather, their reproductive capacity, their ability to adapt to changing conditions, and their skill at de tecting and evading predators, including man. Several of the factors responsible for long-time survival work at cross-purposes to attempts at census. At best, deer are difficult to see and behavior patterns cause unexpected paradoxes. In'northern Michigan, deer are most visible at spring break-up when the population is at its yearly low, and least visible 3 months later when the popula tion is at its yearly high after fawns are dropped. Through the years, a great number of methods have been tried for censusing deer with varying degrees of success. The volume of papers in the literature is an expression of the need and concern for adequate methods. Rasmussen and Doman (1943) in a review article listed 30 references on deer census, Taylor (1947) cited 39 papers describing "new" techniques in hoofed animals including census and determination of population trends, and Hazzard (1958) gives 199 references on big game census methods. McCain and Taylor (1956) summarized methods which have been tried with mule deer (Odocoileus hemionus) to determine popu lation levels: (1) direct counts made on foot, horseback, motor vehicles, and aircraft; (2) indirect counts of tracks, beds, trails, rubbed trees, shed antlers, dung, and plants eaten; (3) hunter kill; and (4) trapping and tagging. The Great Lakes Deer Group (1964) has outlined deer cen sus methods used in Michigan, Wisconsin, Minnesota, and Ontario and pointed out areas needing further research. Jenkins and Marchinton (1969) present an up-to-date comparison of several census techniques used for white-tailed deer in the Southeast. Alexander (1958) indicates that any wildlife management technique, if it is to be considered practical, must be workable by field men, and he lists several desirable characteristics: "1. 2. 3. 4. 5. 6. It must have ease of application, requiring only a reason able amount of experience. It must be capable of obtaining data or giving accomplish ments in quantity in a short period. It should not require a large personnel. It must be reasonably inexpensive; requiring limited equipment. It should provide data or results relatively free from the influence of the individual operators. It should furnish data of a type that permits fast and easy analysis and interpretation, or give distinct results, readily measured and evaluated." Other desirable characteristics for census methods might be added to his list: 7. 8. 9. 10. It It It It should be possible to do it at any time of the year. should work equally well on all sizes of areas. should be based on valid sampling theory. should produce accurate estimates. Michigan deer census methods Previous to the establishment of the Michigan Conservation Depart ment in 1921 and particularly the founding of a separate Game Division in 1928, information on Michigan deer numbers is known only from reports by early explorers, trappers, missionaries, surveyors, lumbermen, set tlers, hunters, and early naturalists and railroad shipping records. Later, around the turn of the century, there were a few zoological sur veys, some authorized by the State legislature and some sponsored by the University of Michigan. During 1930 and 1931, attempts were made by the Game Division to obtain an estimate of Michigan's deer herd through the aid of Conserva tion Officers, hunters, lumbermen and timber cruisers. Estimates of 82,200 deer in the Upper Peninsula and 35,000 in the northern Lower Pen insula were arrived at which seemed logical at the time (Bartlett 1945), although later evidence disclosed that these were serious underestimates. Systematic records originated about 1932 with the establishment of periodic reports from each Conservation Officer, and subsequently other Department personnel, on the number of deer seen per hour in the field. This covered the period from July 1 to the end of October (including also the first 11 days of November from 1932 to 1934). Later (1965) counts were extended to include May and June and still later (1968), April was also encompassed. Essentially these are roadside counts made in conjunction with other duties. Such records probably reflect long term trends of the deer herd, however; they are influenced by weather, changes in personnel, interest of workers toward deer, amount of night patrol, increasingly better road systems, faster automobiles, etc. On the other hand, they tend to be made by the same workers in the same areas year after year and are based on thousands of hours of observations. Rather large fluctuations in deer seen may occur within a given county from one year to the next, but data for blocks of counties tend to show smaller fluctuations. The first methods to give actual population levels were census drives on specific areas conducted with the aid of the Civilian Conser vation Corps camp program from 1935 to 1940 (Bartlett 1950). Properly handled census drives are quite accurate for small areas, but manpower and terrain limit their usefulness. needed to do a good Roughly, 100 men per section are job. Results of the early census drives in 1935 were surprising. They revealed that numbers of deer were considerably greater than realized. Total figures were calculated by averaging the results from all drives made in a particular year. Census areas were believed to be representa tive, but evidently this was not true. Only certain types of areas can be properly driven. Retrospection indicates that the expanded totals were probably too high. Since 1940 only a few areas were driven annu ally, mostly for public relation purposes. Michigan biologists felt the need for better information about deer populations because of adverse public reaction following the first large scale antlerless deer harvest in the fall of 1952 (Bennett, Ryel, and Hawn 1966). The pellet group survey seemed to be the best available technique and surveys in four areas of northern Lower Michigan were carri ed out the following spring in cooperation with the U. S. Forest Service. Surveys quickly became routine annual operations, gradually increasing in extent and consuming considerable time and effort. The U. S. Forest Ser vice had made pellet group counts on the Ottawa National Forest (western Upper Peninsula) as an index to relative deer use of certain timber types from 1950 to 1952 (Olson 1952). However, the 1953 surveys marked the first use of the technique here for actual population estimates. Since 1959 the entire northern two-thirds of the state has been surveyed each year, as well as several other areas. Eberhardt (1960) developed some methods of estimating the relative abundance of Michigan deer based on sex and age data from hunter-shot deer, embryo counts from car-killed does, estimates of legal harvest, sight records, and the number of deer killed on the highways. Neverthe less, he considered pellet group counts as the most accurate method and used them as a standard for comparison with his other estimators. Re cently, Croon et al. (1968) and McCullough, Olson, and Queal (1969) have experimented with infrared scanning for deer census in Michigan, but as yet this technique cannot detect deer through green foliage nor discrimi nate between animals in the size range from foxes to horses. Scope of the present study Normally, considerable research precedes use of a technique; how ever, only limited study evidently had been done anywhere before 1953 on pellet group surveys. Neff (1968), in a review paper, lists but four papers earlier than this date. I suppose that technically, research preceded censusapplications in Michigan, but the lead time was less than a week. Studies on defeca tion rates were begun at the Cusino Wildlife Research Station, near Shingleton, on April 1, 1953, just prior to the beginning of extensive field surveys in the northern Lower Peninsula. searched in Lake County on April 8. The first plots were On April 29, an experimental sur vey was carried out on the E. S. GeorgeReserve, nearPinckney, census the confined deer herd there. to Since 1953 Michigan workers have probably carried on more research on the technique and conducted more surveys than any other state, yet most of the accumulated data have never been analyzed. Only two tech nical papers (Eberhardt and VanEtten 1956, VanEtten and Bennett 1965) have been published. Research studies at Cusino were done under Federal Aid to Wildlife Restoration Project W-70-R. Those at Houghton Lake were carried out under Projects W-63-R and W-95-R. Planning and computational phases of extensive surveys were included in the work of Project W-96-R. Pro gress summaries were included in the various unpublished Quarterly and Annual Reports submitted for these Projects. In addition, annual reports giving population estimates from operational surveys were written. Figures 1 to 4 show the locations of the various areas involved in these studies and surveys. Table 37 lists the various surveys and cites 8 SPRING 1953 SPRING 1954 Cusino Enclosure C usino* Enclosure Mio Ranger District Entire Northern Lower Peninsula Houghton La ke Area La ke County Lake County George Reserve George Reserve SPRING 1956 SPRING 1957 Entire U pper Peninsula CENTRA!. * EflST Area 6 ,7 ,8 E n tire N o rth e rn Lower Peninsula E n tire N o rth e rn Low er P en in sula Area 2 (Includes most o f Lake County) County George R e s e rv e * Figure 1. Locations of deer pellet group surveys, 1953-1957. 9 Alger - Delta Schoolcraft Cusino \ Counties Enclosure Oickinson County Area CKAIKVDII Roscommon- € Oscoda Crawford I Counties r-M io Ranger [ D is tric t Clare-Gladwin Counties OCUNA U X 11AC George Reserve Figure 2. Locations of deer pellet group surveys, 1958. 10 D istrict I Ottawa National forest D is tric t 3 D is tric t 4 Hiawatha National F orest, West Hiowotho Notional Forest, East D is tric t 2 D istrict 5 Mio Ranger D istrict Hiawatha National Forest Deer M gt. Demonstration n Area '* District 7 f Huron Section D istrict 6 of the on-Manistee National Forest Manistee Section of the Huron-M anistee Notional Forest D is tric t 9 D is tric t 8 TUSCOtA SAQMAW GIATWT GEHIHC OTTAWA CUHTOH Unit IAW T Q A I SMAWAUR B a rry -A lle g a n IATON •AN luU N JACKSON George R eserve TT.JOSIW Figure 3. MUSbUI lIN Aw el UONtOt Locations of deer pellet group surveys, 1959-1964. 11 District I Ottawa National Forest D istrict 3 Hiawatha National Forest .West D istrict 4 Hiawatha Notional Forest, East Drummond Island D is tric t 2 D is tric t 5 Beaver Island V MIMA D is tric t 7 D is tric t 6 ^ ¥*** Huron Section of the Huron-Manistee National Forest Manistee Section of the Huron - Manistee N ational Forest tujcwa 1ANIIAC OTTAWA SAINT a A l UACQMS CMIIAMD VAN IUI2N KAIAUAZOO si. jour* Figure 4. JACKSON IIA N O t KlltSCVUi IIN A W II MONK* Locations of deer pellet group surveys, 1965-1969. 12 descriptive reports which give survey results. In 1966, the Michigan Conservation Commission contracted the Research Triangle Institute of Durham North Carolina, to audit the Department's procedures for estimating deer populations and deer kill (Research Triangle Institute 1966). Their personnel reviewed sampling and computational procedures as well as the estimates given in available reports. No new analyses were conducted. As an employee of the Michigan Department of Natural Resources, I have been involved with pellet group surveys every year since their in ception in the Lower Peninsula in 1953. I participated in the first experimental survey on the George Reserve that year and in operational surveys in northern Lower Michigan from 1953 through 1957. became the technical supervisor of all such surveys. In 1958 I This included sam ple design, selection of courses, conducting briefing sessions, editing returned survey cards, and preparing and distributing survey results within the time structure necessary for setting annual regulations. Carl L. Bennett, Jr. assisted me from 1964 to 1969. I conducted experi mental surveys on the George Reserve from 1958 to 1963. As a biometrician with the Game Division and later the Research and Development Division, I also instigated the 1958 Cusino enclosure experimental survey and new research studies on defecation rates in 1959. The present study will appraise the pellet group technique as it has been used in Michigan, suggest modifications for future surveys and point out areas needing future research. Unfortunately, time overlaps and feedback between research and application make it difficult to develop a logical and even presentation. The data base encompasses all research and survey work done through 1969. Original records were 13 examined whenever possible. I also did an extensive survey of the literature and pertinent findings from other studies are cited. attempt will be made here to delve Into deer herd dynamics. No II. THE PELLET GROUP TECHNIQUE Terminology All living animals excrete wastes resulting from digestive and meta bolic processes. In the Cervidae and some other ruminants, each defeca tion occurs as a group of discrete, elongated or rounded objects (Murie 1954). A large number of terms have been used to refer to mammalian def ecations: feces, fecal matter, dung, spoor, sign, droppings, manure, scat, and many colloquialisms. Understandably, wildlife biologists have come to refer to defecations of the deer family, and those of other animals whose defecations are similar, by the euphemism "pellet groups." Each of the small components is called a "pellet." Pellets are also used to refer to defecations of the Leporidae (hares and rabbits), the Erethizontidae (porcupines), and others which ordinarily do not deposit clusters of pellets at one time. Sample counts of deer defecations have commonly been called pellet group surveys. Historical background Seton (1925) urged the study of feces as a means of learning about the habits, food and whereabouts of mammal species. His observation that, "Normally, the white-tailed deer evacuates every hour of its active life, which may mean 20 or more times a day" may have planted the seed for a new census method. Later, several workers used pellet counts in studies of rabbit popu lations (Taylor 1930, Vorhies and Taylor 1933, Hendrickson 1936, and MacLulich 1937). Ruhl (1932) suggested use of pellet counts for big game species and even anticipated that there would be difficulties in de termining the age of pellets and establishing a deposition period. 14 15 The Interstate Deer Herd Committee (1946) gives credit to the Cooperative Wildlife Research Units for first using deer pellet group counts in 1938. Bennett, English, and McCain (1940) also carried out surveys that year in Pennsylvania. wide acceptance. Since then the technique has received At present it is perhaps the most used and recommended deer census method in the northern and western states. Neff (1968) lists 57 references on the pellet group method, mostly concerned with mule or white-tailed deer, but in addition he refers to elk (Cervus canadensis). moose (Alces alces), Barbary sheep (Ammotragus lervia), domestic sheep (Ovis aries) , and beef cattle (Bos taurus). Uses of fecal counts to gauge animal abundance have ranged from small mammals in Wisconsin (Emlen et al. 1957) to African elephants (Loxodonta africana) (Wing and Buss 1970) and from red deer (Cervus elaphus) in Russia (Yurgenson 1963) to red deer, fallow deer (Dama dama), sheep, chamois (Rupicapra rupicapra). brush tailed possum (Trichosurus vulpecula), domestic pig (Sus scrofa), domes tic goat (Capra hircus), and wild (European) rabbit (Orvctolagus cuniculus) in New Zealand (Riney 1957). Basic concepts Basically the pellet group system represents a systematic applica tion of the experienced hunter's method of reading "sign" to gauge the abundance of game (Bennqtt et al. 1940). The idea is that the quantity of pellet groups which are found is directly related to the number of deer that are present. a number of factors. Quantifying this relationship, however, involves Taylor and Williams (1956) observed that the number of pellets of the European rabbit seen on the ground at any given time depend upon: (1) the number of rabbits present (2) the average rate at which rabbits have been producting pellets over a period prior to the counting 16 (3) the rate at which pellets disappear due to such factors as weather, bacteria, stock, etc. (4) the efficiency of the observer. It is a trivial change to substitute deer for rabbits. In this regard, Michigan pellet group surveys have been predicated on several assumptions: (1) (2) (3) (4) (5) Deer defecate at a rather constant frequency Pellet groups persist long enough to be counted Pellet groups can be found and counted in the field An explicit deposition period can be delineated The age of pellet groups, which are present, can be extablished relative to the deposition period The current paper essentially investigates the validity of these assump tions. Since complete counts of anything which is very numerous or which occurs over a large area are virtually impossible, some form of sampling is necessary for pellet group surveys. Actually, only in the case where deer are confined in small pens for short periods can complete counts be made. Even here sampling may be the preferred approach. As Deming (1960) points out, a sample is not a last resort, but is usually the best way to do a job. Michigan studies have always employed carefully defined plots of known size which are distributed in some fashion over the area of inter est. Except for certain research studies, the ultimate sampling unit has been 1/50-acre plots. Sampling designs have varied among the several surveys and will be discussed under the appropriate section. The deposition period in Michigan work has been bounded by the autumnal fall of leaves and the mean date of the survey. Surveys are carried out in the spring soon after the snow disappears and in as short a time span as possible. The calculation of deer numbers from pellet group utilizes the algorithm: deer = population mean pellet groups X 1 X size of area per plot_________ plot size_______________ deposition period X defecation rate For a given survey, plot size, area, deposition period, and defe cation rate have been considered fixed, the only variable being the mean number of groups per plot. Actually both deposition period and defeca tion rate are also subject to variation and their importance will be discussed in later chapters. Additional applications of fecal studies Study of fecal material can also provide other sorts of information besides measuring abundance. Although these other uses have not been employed extensively in Michigan, a brief account is given below as an aid to planning possible multipurpose surveys in the future. Adams (1957) proposed using pellets to study food habits of herbi vores and later several workers applied this technique to deer, for example: Lay (1965), Zyznar and Urness (1969), and Segelquist, Ward, and Leonard (1969). Short and Remmenga (1965) reported that fecal cellu lose content seemed to have predictive value in estimating range forage consumed. They suggest that samples of pellet groups could be collected during population surveys for chemical assay to estimate plant tissue eaten from the range. Petrides, Golley, and Brisbin (1969) discuss deter mination of food energy flow in large herbivores and point out that food energy not utilized by the animal becomes defecated wastes. Severinghaus and Cheatum (1956) report that the winter range of deer was studied by putting red and blue dye into soybean molasses cakes which colored the droppings. Kindel (1960) tested a number of dyes for use in marking ru minant feces to trace animal movements. (1967) fed and injected 65 Nellis, Jenkins and Marshall into rabbits, opossums, foxes, and bobcats (species names not given). tectable for over a year. When injected, the radioactive zinc was de When fed to fox, opossum, and bobcat, it was detectable in the feces for about a month. McCaffery and Creed (1969) and Shafer and Liscinsky (1968) used the numbers of pellet groups in vari ous cover types as an indication of relative deer use. Samuel and Trainer (1969) analyzed pellets picked up on regular pellet group surveys in Wisconsin for information on the prevalence and distribution of several endoparasites of deer. Finally, fecal examination might be useful to ob tain information about age composition. Pellets of young fawns are obvi ously smaller than those of adults, but to date this approach has been little studied, nor have any workers reported sex differences in deer droppings as Bailey (1956) concluded are present in wild turkeys (Meleagris gallopavo). III. GENERAL METHODOLOGY The following sections explain the sources of the data which are evaluated in later chapters. Determination of defecation rates Rates of daily pellet group deposition were investigated with penned animals at the Cusino and Houghton Lake Wildlife Research Sta tions during 1953, 1954, and 1959. At Cusino, defecation rates of one adult male, two adult females and one female fawn were studied for and early May of 1953. a total of 73 deer-days in April The deer were fed a hardwood-conifer diet made up largely of sugar maple (Acer saccharum) and northern white cedar (Thu.ja occidentalis) which simulated late winter conditions for this area. Other species - red maple (Acer rubrum), hemlock (Tsuga canadensis), as pen (both quaking, Populus tremuloides, and bigtooth, grandidentata) , and willow (Salix sp.) - were also fed, but consumed in low quantities. Pellet groups were counted and removed every 24 to 72 hours depending on weather conditions. All counts were converted to rates per 24 hours. Much more extensive work at Cusino was carried out in 1954. pens of three deer each, half adults and half fawns, were studied. but one deer, an adult buck, were females. Twelve All A pen of adults and a pen of fawns constituted the samples for each of six different diets of natural browse. Eight counts on each pen were made between January and April. Each count was made approximately 24 hours following a snowfall of suffi cient depth - about 4 inches - to obliterate signs of all previous deposi tion. On February 27, counts of pellets were made for selected groups in each pen. The control diet consisted of nine browse species - white cedar, 19 20 balsam fir (Abies balsamea), hemlock, red maple, sugar maple, yellow birch (Betula lutea) , paper birch (IJ. papyrifera), black ash (Fraxinus nigra), and red-osier dogwood (Cornus stolonifera) - fed in unlimited quantities, simulating a white cedar deer yard in excellent condition. The swamp conifer diet simulated a medium condition deer yard where white cedar is moderately abundant. It included nine species of which tamarack (Larix laricina), balsam fir, paper birch, red maple, and the moderate amounts of white cedar supplied, were the species consumed the most. The swamp hardwood diet consisted of eight deciduous species commonly found around the periphery of deer yards and simulated a browsed-out white cedar deer yard. Red maple, paper birch, aspen, and American elm (Ulmus americana) comprised the bulk of the browse eaten. The hemlock-hardwood diet consisted of the 10 major species found in the northern hardwoods type. Red maple, aspen, American elm, along with hemlock, the only conifer included, were preferred. The fire suc cession diet consisted of jack pine (Pinus banksiana), red maple, black cherry (Prunus serotina) , aspen, paper birch, and five shrubs - species commonly found profusely sprouting or seeding following fires. The final diet was mixed conifer - upland hardwood consiting of white pine (Pinus strobus), balsam fir, white spruce (Picea glauca), red maple, paper birch, aspen, yellow birch, and sugar maple. At Houghton Lake in' 1954, seven counts were made on four deer on 3 successive days in late May. Information on the sex, age, or diet of these deer is not available. In 1959 at the Cusino Station, 20 deer were fed one of two diets. Varying numbers of counts were made on each deer within the overall period from January 7 to April 24. Both diets control and swamp conifer, 21 were essentially the same as those in 1954. kept In a pen. Generally three deer were Prior to pellet counts, each deer was moved individually to a special pen for 2 days and then returned. ed at either 24 or 48-hour intervals. Pellet groups were count Longer counts were converted to 24-hour rates and each received the same weight as one 24-hour count. Studies at Houghton Lake in 1959 involved 11 deer in seven pens on three different diets. Diets were composed of one of three test species plus a combination of five to seven other browse plants that are associ ated with it. Test species constituted 50 per cent of the diet and the remaining 50 per cent consisted of approximately equal amounts of the other species. Adjustments were made in amounts fed to insure that associated species made up no more than 50 per cent of the food consumed. The diets used were sweet fern (Comptonia peregrina), aspen, and willow. Counts were made from January 28 to March 5. Pens were cleared of pellet groups prior to each rate determination and counts were made 24 hours later. In some instances, fresh snowfall was used instead of clearing. Also in 1959, a study was carried out in a 6-acre enclosure near Houghton Lake. The cover type was a typical white cedar swamp associa tion and was used by deer as a wintering area (Dead Stream deeryard). Inmate crews had enclosed an area 6 chains by 10 chains with a 7 1/2-foot fence in the winter of 1951-52 for use in a cedar topping study (Harger, 1954). Four deer - adult buck, adult doe, buck fawn, and doe fawn - were released into the area January 7, 1959 and were removed in the spring. Following transfer of the deer, biologists counted pellet groups on 60 transects 396 feet by 3 feet from June 23 to 25. completely across the enclosure. with replacement. Transects ran Those counted were selected at random 22 Weathering of pellet groups Studies were undertaken at both Cusino and Houghton Lake Wildlife Research Stations to determine the longevity of pellet groups in several situations. In addition, researchers hoped to develop some guidelines with which to judge the age of pellet groups. In the square-mile Cusino enclosure, observations were made of the rate of decomposition and vegetative coverage of naturally deposited pellet groups at varying intervals up to 5 years. Recently dropped groups in various cover types were marked with numbered wooden stakes. A detailed description of each group and its surroundings was recorded at every visit. Summer groups were observed more closely the first year to study their rate of decomposition. All groups visit were examined the following spring or early extantat the last summerto note their appearance at the usual time of pellet group surveys. The first groups were marked during the summer of 1953, and ulti mately 318 groups were considered during the 5-year study period. Groups deposited in summer or early autumn (pre-leaf fall) numbered 123, while post-leaf fall examples totaled 195. Observations on many groups were terminated as soon as the groups became completely covered by vegetation, but others were continued to investigate the durability of individual pellets in the duff. At Houghton Lake, workers utilized a different study technique. Pellet groups were collected from penned deer and cover types. placedin selected On September 1, 1953 three groups were setout in each of five types - oak (Quercus sp. ), jack pine, open grass and sedge, aspen, and conifer swamp. Three examinations were made at weekly intervals. At the end of 24 days the study was replicated with a new series of groups. 23 Sixty more groups were distributed February 25 and March 7 and 10, 1954-30 in an upland aspen site and 30 in a swamp conifer stand. About one-half of the groups placed in each location were collected from deer fed a jack pine-oak diet and the remainder a balsam fir-northern white cedar diet. All groups were re-examined once on May 16, 1955. Experimental field trials Experimental pellet group surveys were conducted on two large deer enclosures. One of these, the Edwin S. George Reserve near Pinckney, covering 1.8 square miles, is owned by the University of Michigan and operated as an outdoor laboratory for ecological studies. The other, a square mile enclosure at the Cusino Wildlife Research Station near Shingleton, is used for deer population and range research by the Mich igan Department of Natural Resources. Ten surveys (1953-56, 1958-63) were carried out at the George Reserve and five (1953-56, 1958) at the Cusino enclosure. Herd size is quite accurately known on each area and this enables comparisons between results of pellet group surveys and ac tual populations. With areas in this size range, it is impractical to select plots wholly at random. Travel must be primarily by foot. Hence, to facil itate locating plots, systematic sampling designs were used for the most part. On the George Reserve the sampling plan consisted of establishing plots at uniform distances on lines which crossed the area. In Univer sity research and in this study, sample plots were located with the aid of aerial photographs on which 5-chain intervals have been inscribed in a grid pattern. During the first 3 years, plots were systematically lo cated on every other north-south grid line with a 4-chain spacing between 24 plots. Every fifth plot was marked by a permanent stake with a desig nated number. The plots were circular and 1/50-acre in size (16.7 feet in diameter). A new sample was selected in 1956 in a similar fashion, except that plots were not staked. In addition, plots were changed to a rectangular format (12 feet x 72.6 feet). From 1958 to 1963 a stratified design was used primarily to spread samples across the whole area as outlined by O'Toole (1964). was divided into six east-west belts 20 chains deep. The Reserve Two east-west tran sect lines were then selected at random within each belt (the northern most belt being very small, had only one line in some years). placed 4 chains apart on each line, beginning at a randomly selected starting distance (up to 5 chains) from the main boundary fence. Plots were central road or the Plots were the same as in 1956, but each was deflected 45° to the right of the direction of travel. Major cover-type boundaries in the Reserve are orientated in a north to south fashion, hence, east to west lines cut across cover types. Total plots numbered about 280 in every year except 1956 when about 240 were set up. hand compass. All plots were spaced by pacing and lines run using a Generally the work was done in 1 or 2 days depending on the number of men involved. days. Usually the survey required 11 to 15 man- For the most part, two-man crews were used, and personnel were nearly all regular Game Division biologists. The major deviation was in 1956 when one Game Division biologist and five University of Michigan students completed the counts in 2 days. Until 1955, crews simply aided each other in counting groups. Seem ingly low counts in 1955, however, prompted a recheck of 13 plots by two men 19 days after the original survey. Beginning in 1958 a concurrent 25 recheck system was used. Each man searched one-half of the plot, then rechecked his partner's counts on the other half. Metal disks were often used to mark counted groups so recheckers could readily determine which groups were missed on the first count. Leaf fall dates were determined by the local District Game biol ogist. Oaks were especially troublesome on the Reserve, since their leaves fall intermittently from fall through spring. Equally disturb ing were the large areas of marsh and upland grass where leaves were not present as a guide to pellet group age. In such instances, teams simply had to judge the age of the groups present. At briefing sessions held the first morning of each survey, crews were provided with some guidelines to use in aging pellet groups and were given "pep talks" on the importance of diligent effort. On the final six surveys, both old and new groups were tallied, and cover types for each plot were recorded. In the Cusino enclosure a 10 chain x 10 chain grid has been estab lished for study purposes. line - a total of 153. Numbered posts occur every 5 chains in each The lines are partially brushed out and marked with paint on trees between posts. In 1953, three circular 1/50-acre plots were placed in conjunction with each post. One used the post as plot center, another was staked 40 feet due west and the third 40 feet due east. In 1954, 1955, and 1956 the central plot was eliminated because of potential bias due to human disturbance in using the lines as travel lanes. In all cases, estimates were based on the sum of the two or three plots at each post. The 1956 Cusino survey was easily the most painstaking ever carried out in Michigan. In the fall of 1955, each plot was searched two or more times and the groups found were recorded. External characteristics and 26 the exact location on the ground were entered on charts for each group so that it could be identified later without disturbing or marking the groups in any manner. Checks were spaced to cover the leaf fall period and until snow cover prevented further examination. The following spring, as soon as the snow cover was gone, two experienced game biol ogists not connected with the previous fall’s survey, each surveyed about half of the plots. metal disks. They marked all pellet groups found with Different colored disks were used to indicate whether they felt each group was deposited prior to or after leaf fall. Finally, all plots were resurveyed using original plot records and the disks present to arrive at the best possible count. In 1958 a stratified random sampling design was used. Each 2 1/2- acre block (5 chains x 5 chains) was assigned to one of four deer density classes, based on the results of previous surveys. These blocks were the primary sampling units and 100 were selected at random within strata. The number selected per stratum was an optimum allocation based on pre vious survey data. Each selected block was sub-sampled by a north-south line of three 1/50-acre rectangular plots. line was selected at random. The starting point for each The sum of ..the three plots constituted the data analyzed. With the exception of 1956, old groups were not tallied by crews. In 1953 and 1954, crews removed groups from the plots after counting. Manpower for the enclosure's surveys consisted of biologists from the research station, usually assisted by inmates. survey required about 4 to 5 crew-days to complete; a crew consisting of two men. courses were run by two men, a biologist and an In 1958 all In most years, the inmate, each checking the work of the other in the usual manner. 27 Population estimates were calculated as outlined in Chapter II. Weighted defecation rates for each area were developed from herd compo sition data. Chapter IV explains the origin of these rates. No unbiased variance estimators are available for systematic samp ling, however, Yates (1960) gives formulae for some approximations (usually inflated). Kish (1965) points out that a systematic sample, in many cases, can be accepted for practical purposes as a good approxi mation of random sampling. He indicates, though, this practice may also produce variances which are high because of induced stratification effects. For ease of computation I chose to compute means and variances for the 1953 to 1956 surveys on both areas as if they were simple random samples. Later surveys were analyzed as stratified random samples. Counts for individual 1/50-acre plots on the George Reserve entered the computa tions, while the sum of the two or three plots at each location consti tuted the Cusino data. Only estimates of sampling error in counting pellet groups were considered. No attempt was made to include variance components due to deposition rate or deposition period which were con sidered fixed insofar as population estimates were concerned. 2 standard errors of the mean were calculated for all surveys. Limits of Since t Q^- 2 in the range of degrees of freedom encountered here, such limits are approximately 95 per cent confidence intervals. Cochran (1963) indicates that the normal approximation is generally satisfactory even with markedly skewed distributions unless very exact statements of limits are needed. gated for pellet counts. Validity of this assumption was investi Negative binomial distributions were fitted to these data by the method of maximum likelihood (Bliss 1953). Extensive surveys Procedures for extensive surveys have gradually evolved over a number 28 of years to a rather efficient operation. They are presented in some detail for the benefit of other workers who may wish to adopt them for similar surveys. The 1953 surveys utilized a systematic design wherein five sections (square miles) were sampled in each full geographic township. Sections were arbitrarily selected in approximately a "five of spades" pattern, favoring accessibility and public ownership. Each section was in turn sub-sampled by a "course" of five circular, 1/50-acre plots located 8 chains apart in a cardinal direction. by means of hand compass and pacing. Plot locations were determined Course lines turned at right angles if formidable barriers (lakes, rivers, military bases, etc.) were encoun tered. Plot centers were marked by numbered wooden stakes. To facilitate rapid location and relocation, all course lines were begun from driveable roads where possible. Workers were instructed to check maps and aerial photos of selected sections, beginning from the NW corner and proceeding clockwise around the perimeter of the section until a driveable road was located. On-site inspection of roads was often necessary to determine actual driveability. If no useable road was pres ent, east-west or north-south roads crossing the section were selected in turn. Lacking suitable roads, workers drove as close as they could to the section and walked the shortest distance to the section line. For each course, a random number from one to nine was selected to deter mine how far to drive (tenths of a mile) along the access road before starting the course line. Plot 1 was 8 chains from the starting point. Courses were searched by state or federal game biologists working alone. Plots on two of the units were also searched in 1954. Based on the results of the 1953 surveys, a two-stage, stratified random sampling plan was developed in 1954 and used with modification 29 henceforth. Sections (square miles) constituted the primary stage. Local game biologists assigned each section to one of five strata of expected deer abundance during the over-winter period: stratum I, colored red on the stratification maps, over 50 deer/square mile; stratum II, colored yellow, 35 to 50: stratum III, colored brown, 20 to 35, stra tum IV, colored blue, 5 to 20; and stratum V, uncolored, 0 to 5. tional sections were included if at least half was present. sample sizes were set by available manpower. Frac Overall Dispensing of the primary stage sample over the various strata was by optimum allocation (Cochran 1963). Sections within each stratum were picked using tables of random numbers (Rand Corporation 1955). The second stage sample in each se lected section consisted of eight 1/50-acre plots arranged in a straight line 5 chains apart. Location of the course line within each section followed the 1953 practices, except Plot 1 was 5 chains from the starting point. Beginning with the 1956 surveys, plots were changed to the rectang ular shape with the center line set at 45° to the right of the line of travel. Midpoints of each end of the plot were marked with wooden stakes. A plastic clothesline was stretched between the stakes while counts were made. An additional feature was the introduction of a system of recheck ing one-fifth of the courses, selected at random, by research biologists. Beginning in 1959 two-man crews were used and the concurrent re check system was employed. The first plot was randomly located 0 to 5 chains from the starting point. The nine northern game districts formed separate sampling units with 60 courses in each. Additional courses were established within the Huron-Manistee National Forest boundary and in some game districts to provide more precise estimates. Except in two dis tricts, plots were established in the spring of 1959 and resurveyed the 30 following 5 years. All plots in Districts 7 and 8 werestaked in Septem ber 1958 and cleared of all pellet groups. Game districts in the northern Lower Peninsula were reduced from five to four in 1965. New stratification maps were drawn up and new samples were selected in all districts. to IV were modified as follows: Population levels for Strata I Stratum I 35 deer/square mile, Stratum II 25 to 35, Stratum III 15 to 25, and Stratum IV 5 to 15. sample size was kept at 60 courses per district. The basic Plots were staked in the spring of 1965 and searched each year through 1969. A major change involved starting from a randomly selected corner to initiate location procedures rather than always beginning at the northwest. new instructions allowed no offsetting of course lines. Furthermore, Plots falling in permanent bodies of water were assumed to have zero groups. Plots ac cessible to deer, but not able to be searched by crews, were assigned the average of the plots which were counted. If none of theeight plots on a course could be reached, it was dropped from the survey. Examples of troublesome sites included Air Force bases, flooded marshes and river basins, freshly plowed fields, sheep pastures, and private land where the owner refused access. IBM mark-sense cards were used to record field data on the 1953 and 1954 surveys. Problems with erasures and bent and water-soaked cards prompted a change to Royal-McBee keysort cards for all subsequent sur veys. Recently, Patton and Casner (1970) have reported good results with IBM Port-A-Punch cards but these have not been tried in Michigan. All original records are on file in Lansing. Game biologists residing in various parts of the state were in structed to keep annual records of local leaf fall conditions. Each year 31 note is made of the date when most (80 to 90 per cent) of the leaves have fallen. For a given survey unit, the deposition period was consid ered to extend from the leaf fall date to the mean date that survey courses were searched. Means and variances were computed from course totals only, which is the appropriate analysis when the main objective is the estimation of a total. This is true even when, as here, the second stage sample is systematic (Cochran 1963, Eberhardt 1963, Kish 1965). Because of the generally small sampling fractions used, finite population corrections were ignored. Confidence limits of 2 standard errors as per cent of the mean (approximately 95 per cent) were calculated. Average over-winter populations were computed as given in Chapter II. Conversion of these estimates to fall (pre-hunting season) and spring (pre-fawning) populations was accomplished by taking into account deer killed during hunting seasons and other mortality during the deposi tion period. Losses were removed instantaneously on dates believed to be weighted averages. The third and fourth days of the hunting season were used for the legal harvest in the northern Lower Peninsula and Upper Peninsula respectively. Other fall and early winter losses were removed December 1 and late winter and spring losses on March 1 in both peninsulas. Mail surveys to samples of deer license purchasers are conducted each year by the Department of Natural Resources to estimate the legal kill (Eberhardt and Murray 1960, Bennett et al. 1966, Ryel 1970). Be cause final sales figures for hunting licenses are not available when pellet group surveys are completed, previous results utilized preliminary estimates of legal harvests. Final kill data have been used here to pre pare revised population estimates for all surveys. 32 Estimates of losses other than legal kill were obtained from strati fied random sample surveys carried out in northern Michigan during some years. These involved crews searching sample plots of ground and deter mining cause and approximate time of death for all carcasses found (Whitlock and Eberhardt 1956, Ryel and Bennett 1962). In later surveys, dead deer searches were combined with pellet groups surveys to reduce manpower requirements. Northern biologists were asked to estimate losses in their districts for those years when surveys were not made. Not in cluded in mortality estimates are deer shot illegally and removed during the pellet deposition period. Hence, actual fall populations are slight ly higher and spring populations slightly lower than the estimates given. Using computations similar to those carried out by Eberhardt (1960), an average herd composition following the hunting season was obtained. These are expected to be close to the averages for the entire over-winter period, since the bulk of hunting season losses usually occurs within a month of the leaf fall date. The Upper Peninsula herd averaged about 15 per cent adult bucks, 49 per cent adult does, 19 per cent buck fawns, and 17 per cent doe fawns. A weighted pellet deposition rate of 13.47 groups per deer-day was used for estimates (see Chapter IV for the source of these defecation rates). Similarly, the average northern Lower Peninsula composition was 9 per cent adult bucks, 52 per cent adult does, 22 per cent buck fawns and 17 per cent doe fawns, resulting in a weighted depo sition rate of 13.37. Original population estimates were calculated using a standard rate of 12.7 statewide. based on revised defecation rates. All estimates given here are IV. DETEBMINATION OF DEFECATION SATES The assumption that deer defecate a fixed number of groups per deer per day is an integral part of the technique whether pellet group counts are used as an index of relative abundance or as a population estimator. Background The deer is a ruminant. Feeding studies by Mautz (1969) revealed that, in general, food is consumed on a regular and frequent basis and materials are passed from the rumen at a regular rate. He reported that in two adult males the passage time of natural diets averaged 30.1 and 40.6 hours through the entire alimentary tract. Zyznar and Urness (1969) found in studies of both mule and white-tailed deer that normal time lapse from feeding to evacuation, as indicated by presence of basic funchsin dye, was about 36 hours for a number of forage species. About half of the dry weight of natural deer foods is defecated. For example, Mautz and Petrides (1967) found digestibility of the dry matter in northern white cedar to be 52.45 per cent. In later studies, Mautz (1969) found digestibility coefficients of dry matter to be 50.70, 49.42 and 54.17 per cent in quaking aspen, bluegrass (Poa pratensis), and staghorn sumac (Rhus typhina) respectively. Ullrey et al. (1967) found corresponding coefficients for northern white cedar and jack pine to be about 41 per cent each. Previous studies Essentially two ways have been used to determine the defecation rate of deer. One involves controlled studies of penned deer; the other, counts made in large enclosures with free-ranging deer. There are ad vantages and disadvantages to each system. For example, the use of penned deer enables one to study individuals 33 34 on specific diets and make periodic counts covering known time intervals. The disadvantages are: (1) confined deer often trample and scatter pel lets, (2) penned animals are probably less active than free-ranging deer, (3) the diet is restricted to the foods which are provided, (4) it is ex pensive to house and feed individual deer, and (5) only relatively docile individuals can be used, which ordinarily means few males. In large corrals deer can be more active, can select their own food (within limits), and are less liable to trample or scatter pellet groups. Furthermore, the simulation is closer to that of an actual field survey. That is, factors which may affect disappearance of pellet groups in the wild may also be at work in large range pens. The disadvantages are: (1) it is difficult to make complete counts and sampling may be necessary, (2) it is usually possible to make counts only once at the end of the yard ing season, (3) if multiple deer are used, possible sex and age differences cannot be studied, (4) the food consumed is unknown, and (5) deer may die or be removed without the knowledge of the investigator. In spite of the widespread use of the technique, all early census estimates were based on a single study of mule deer defecation rates (Rasmussen and Doman 1943). Michigan studies, begun in 1953 and dis cussed in detail here, evidently were only the second such study and to date apparently the only studies dealing with white-tailed deer. The first available data for determining defecation rates, were ob tained by Rasmussen and Doman (1943) in 1941. Although not actually given in the paper, rates can be calculated from the information presented. They worked with mule deer in a 741-acre fenced area in Utah. An aver age of 172 animals, both fawns and adults, was present during the study period. Counts were made on 123 circular 1/100-acre plots in August, September and October; apparently the pellets present were removed during 35 each count. able. The weather was hot and dry with little green forage avail The determination of known deer-days of use was based on all deer present. McCain (1948) computed an overall average of 12.7 groups per deer-day from these data, the so-called "McCain Index." Rogers, Julander and Robinette (1958), speculating on the Rasmussen and Doman study, sug gested that the defecation rate, as derived, was probably low since fawns born in late June and early July would have defecated less while receiv ing milk than those feeding strictly on forage. In addition, some of the small fawn groups may have been missed on the August counts. Still later, Julander, Ferguson and Dealy (1963) computed an overall mean of 12.6 from these same data, giving the August mean as 11.1, September as 13.1, and October as 13.5. Dasmann and Taber (1955) studied a population of black-tailed deer (0. h. columbianus) on a 400-acre chaparral area in California. Actual populations on the area were unknown but estimated from four census meth ods, including pellet group counts. Five pellet counts were made over a period of 18 months on 50 circular, 100-square foot plots, located at random. At each visit, pellet groups were counted and removed. They concluded that defecation rate must have varied with diet and gave tenta tive defecation rates as follows: April-June on sprouting brush - 10.0, July-October on dormant brush - 13.0, and November-March on green herbaceious feed - 17.0. Rogers et al. (1958) studied confined mule deer during five winters in Colorado. A series of deer-tight, contiguous pastures varying from 90 to 190 acres and stocked with from 3 to 40 deer were used. Adults made up 60 per cent of the deer involved, but sex composition is not given. After removing the deer in the spring, pellet groups were counted on 36 sample plots placed along line transects. The average defecation rate was 15.21 ± 6.97 per cent (95 per cent confidence limits). was 12.82 to 20.52. The range Smith (1964) points out, however, that they used the wrong weighing factor - pellet groups counted instead of deer-days in their computations and concludes that 15.01 is a better estimate. Smith (1964) studied defecation rates of mule deer over a 12-year period beginning in 1947, using both individually penned animals and paddocks stocked with known numbers of deer. Winter defecation rates for penned animals on natural forage were 16.9 for fawns, 13.0 for year lings, and 10.6 for older deer. His analyses showed the higher defeca tion rate for fawns to be significantly different than those observed in older animals. Data are not given by sex. For all deer in paddocks, the over-winter average was 13.2 groups per day. In Arizona, Neff (1964) reported the defecations of one adult mule deer (sex unknown) on dormant browse for two periods of 7 and 7 1/2-days in February to be 13.0 and 13.9 groups per day. McKean (1965) studying mule deer in Colorado, counted pellet groups deposited in five small paddocks for a 20-day period in each of three successive winters. The paddocks were established on pine-juniper range and deer ate natural food. The overall defecation rate was determined to be 13.2, but he noted that rates were usually higher in the lightly stocked paddocks than those with relatively more deer. Neither the sex nor the age of the animals used is given. Hines (1963) studied the defecation rate of 15 black-tailed deer in Oregon in a 340-acre enclosure from December to May. Deer fed on natural foods and dropped an average of 18.5 pellet groups per deer-day. 37 Coprophagy It is well established that members of the Lagomorpha frequently reingest their own fecal pellets (coprophagy). For example, Meyers (1955) has observed it in the European rabbit in Australia, Hamilton (1955) in the swamp rabbit (Sylvilagus polustus) in Florida, Geis (1957) in the cottontail (Sylvilagus floridanus) in Michigan, and Bookhout (1959) in the snowshoe hare (Lepus americanus) also in Michigan. Bailey (1969) found about 30 per cent of the pellets produced by penned cottontails were of the soft type which are normally reingested. dence, however, that deer reingest their own pellets. There is no evi Being ruminants, they would presumably not be able to extract much additional sustenance from fecal matter. Short, Medin and Anderson (1965) show that deer have a relatively small rumen coupled with a relatively high basal metabolic rate. Since the digestion of fibrous foods requires a long retention time, the small dimensions of the deer rumen would seem to preclude ex tensive fiber digestion. Hence, I conclude that coprophagy is not a factor in deer pellet group surveys. Results of Michigan studies Results of penned studies will be reported here. Surveys on the Cusino enclosure and the George Reserve will be discussed in detail in Chapter VII. Tables 38 to 41 present the pertinent results of the penned studies by sex and age. Results for all determinations were adjusted to a deer- day basis (one deer for 24 hours) and all analyses were made on the ad justed data. Included are records for all deer or pens of deer of the same type which were deemed "good counts" by the personnel present, a total of 263 determinations involving 60 deer. Not listed in these 38 tables are (1) the 1954 Houghton Lake data because the kind of deer used was not reported, (2) one pen with an adult buck and two adult does in the 1954 Cusino study, (3) records for six adult does at Cusino in 1959 which were injured in transfer between pens or became ill during the trials, and (4) one pen with a fawn of each sex in the 1959 Houghton Lake study. Before proceeding into a more detailed discussion, I would like to point out some of the weaknesses inherent in the data. Obviously, the deer involved were not a random sample of any natural population, nor even representative in terms of sex or age. pens or diets at random. Deer were not assigned to Counts were made at irregular intervals based on weather, work schedules, and convenience of the workers. Deer diets were constructed by workers to simulate natural diets, but were not based on observations of the diet of wild deer. My hope is that, none theless, the data are similar to those that might have been obtained by more appropriate methods. No better information seems to exist. Con sequently, with reservations, I have carried out some statistical compu tations as if the implied assumtions were satisfied. The 1954 Houghton Lake study produced an average of only 11.79 pel let groups per deer-day. Results of other studies will be discussed in four sections, which consider differences between sex and age, differ ences between diets, differences in time of year, and result of a large corral study. Sex and age differences 1 have used a simple breakdown into four sex and age categories since exact ages were not available for many of the adult deer used and sample sizes were small for males. Fawns are defined here as animals 39 less than one year old and, hence, adults are those one year and older. The samples are heavily weighted toward female deer, mainly because females are easier to handle in pens. The evidence of sex and age differences is rather conflicting. An analysis of variance (two level, nested, mixed model with unequal sample sizes) on the 1953 Cusino data, Table 42, indicates no difference be tween kind of deer, but among individuals was significant at the P level. The average rate per individual was 13.32. Regrettably, in this analysis the average coefficients of the same variance components (EMS column) for kind of deer do not correspond at difference levels because of unequal sample sizes. Hence, no exact F test can be performed. Satterthwaite's approximation (Sokal and Rohlf 1969) can be used, but preliminary computations indicate significance would not be reached at the .05 level and I did not perform all of the necessary mathematics. For the 1954 Cusino data, Eberhardt and VanEtten (1956) indicated highly significant differences (.01 level) between ages and also between diets. I performed a somewhat more complex analysis (three way, Model I ANOVA without replications) treating dates as a third variable. second order interaction was used in significance testing. the added effect on a| due to the AXBXC interaction is zero. The This assumes Signifi cance was obtained between age (.01 level) and diet (.01 level), and in addition between the diets x dates interaction (.05 level), Table 43. A significant interaction, however, raises some question about the value of testing the diets and dates main effects. per day on every diet. to 13.24 for adults. Fawns averaged less groups Overall fawn deposition rates were 12.10 compared Since only one deer in this study was a male, basi cally these data indicate that adult females average higher than female 40 fawns. Absence of significant interaction between ages and diets indi cates that diets did not affect daily rates of fawns and adults differ ently. A re-examination of original records and reports reveals some dis turbing aspects about the conduct of this latter study. Replacement deer were added to various pens to maintain the quota of three deer per pen. Actually, in only four out of the 12 pens were the same three deer present for all six counts. In seven pens, one deer was substituted and in another pen two extras were used. final six were reported. Eight counts were made, but only the The first two were conducted with inmate labor and were deemed inaccurate. Obviously, caution should be exercised in interpreting the results of these trials. For the 1959 Cusino studies, defecation rates for eight deer on the control diet (two in each of the four sex-age groups) were compared. Analysis of variance (two-level, nested, mixed model with unequal sample sizes), Table 44, showed no difference among the kinds of deer, but a difference within kind of deer (P<.01). deer was 14.30. Mean defecation rate for all A similar analysis revealed similar results for six deer on the swamp-conifer diet, two adult does, one buck fawn and three doe fawns (Table 45). Here the average rate was 12.38. deer on the same diets are available for testing. No other groups of As in the 1953 Cusino study described above, unequal sample sizes in both cases preclude an exact F test of the effect of kind of deer. Again preliminary compta- tions indicated significance would not be reached at the .05 level even using closer approximations. Only one of the above studies reveals significant differences in defecation rate between sex and/or age categories. This is not too surprising in light of the high variability noted in individual deer and the small sample sizes available for males. Smith (1964), working with captive mule deer in Utah, reported that fawns exhibited higher rates of defecation than adults. This is the direct opposite of 1964 Cusino data. Unfortunately, Smith did not report his data by sex. All but one of the adult deer and all of the fawns at Cusino were females. Because of the homeostatic state of the rumen, 1 would expect that the range of possible defecation rates would be quite limited. Actual mean rates for individual animals reported above ranged from 8.80 to 21.38, and no significant differences in defecation rates were demon strated between sex and age classes. Pending further study, particularly of males, I computed pooled means for each sex-age group regardless of diet. Counts made on groups of two or three deer per pen were treated as if they came from a single deer. Means and variances, weighted by degrees of freedom, were obtained in the usual fashion (Dixon and Massey 1957) n v + „ Y + ... 4- (n^ - 1) Sj2 + (n2 - 1) s22 + ... + (i^ - 1) sfc2 n l + n 2 + *" * + nk “ k where nQ is the adjusted mean sample size (Snedecor 1956) The resultant means and 95 per cent confidence intervals are adult bucks 15.61 ± 2.03 adult does 12.89 ± 1.60 buck fawns 14.74 ± 3.04 doe fawns 11.89 ± 1.79 42 Tables 38 to 41 and Figure 5 present these data In some detail. The averages obtained seem different enough to warrant computation of weighted defecation rates for field surveys using some measure of herd composition. For the weighted defecation rates used to estimate Upper Peninsula (13.47) and northern Lower Peninsula (13.37) deer populations, confidence limits of 2 standard errors of the mean were 6.59 per cent and 6.97 per cent respectively. Using a variance approximation suggested by Deming (1964), I examined the effect on the confidence limits for population estimates due to adding a component of variation for defecation rates to that based on counting pellet groups. additional contribution. Results indicated only a small For example, the revised confidence limits for the 1969 survey in District 6 were calculated to be ±39.57 per cent (2 standard errors of the mean) compared to ±38.96 per cent when the defe cation rate was considered constant. Differences between diets The analysis of the 1954 Cusino studies (Table 43), as mentioned above, discloses a difference between diets (P<.01). Duncan's Multiple Range test (Steel and Torrie 1960) indicated that three diets, hemlockhardwood (mean 11.80), fire succession (mean 12.21), and mixed coniferupland hardwoods (mean'12.36), differed (less) from the control diet (mean 14.06), at the .05 level, but not among themselves. These three diets are not a typical winter diet for a large proportion of Michigan deer. The commonest diets, swamp conifer (mean 12.76), and swamp hard woods (mean 12.83), showed no significant difference from the control or each other. Analysis of the 1959 Cusino data (two level, nested, mixed model ANOVAS with unequal sample sizes), Tables 46, 47, and 48, also shows no Mean 15.61 ADULT BUCKS _____________________________________ m -------------- • - Mean 14.74 BUCK FAWNS • • • Mean 12.89 V ADULT DOES Mean 11.89 DOE FAWNS 1 1 0 1 1 2 W ' ’• 1 4 1 1 6 1 1 8 1 1 10 VJpW • 1 1 9" "W 9 1 12 PELLET GROUPS PER DAY Figure 5. Defecation rates for Michigan white-tailed deer. 9 1 14 1 1 16 1 1 18 1 I. 20 1 .... J 22 difference between control and swamp-coriifer diets for groups of four adult does, three buck fawns and five doe fawns. Again unequal sample sizes prevent exact F tests between diets in each instance; however, it is obvious that significance would not be obtained at the .05 level by more precise methods. In the two female groups, differences were indi cated between deer within diets (P<.05 and P<.01 respectively). None of the trials at Houghton Lake lend themselves to such analyses. I conclude from the available evidence that for deer feeding on what we believe to be the usual winter fare of wild deer, differences in diet are not a major factor influencing defecation rates. Evidently, individ ual differences in deer are more important than diet. Differences between times of year While virtually all studies of pellet deposition rate in Michigan have been conducted during late winter and early spring (see Tables 38 to 41), the mid-point of the period from leaf fall to mean survey date is mid or late January in both peninsulas. Actually the deposition period normally stretches over 6 months in the Upper Peninsula and slight ly less in the northern Lower Peninsula, Figure 6. Cowan and Long (1962) observed that adult male white-tailed deer in their studies voluntarily decreased food consumption and lost weight starting abruptly at the onset of rut, and continuing throughout the winter. In the discussion following a paper by R. L. Cowan (1962), I. McT. Cowan noted that some of his captive black-tailed deer had gone 60 days during the rut "without eating an ounce of food." Penned deer fed low nutritional diets undergo considerable weight loss and many lowdiet fawns die when their weight drop exceeds about 30 per cent of their fall levels (Verme 1967). FAWN M O R TALITY POACHING p o p u l a t io n ACCIDENTS L E G A L HARVEST C R IP P L IN G LOSS I STARVATION PREDATION POACHING ACCIDENTS -p> Ln BREEDING 1 0 % Bucks 4 5 % Does 5 0 % Does 3 5 % fe w * 4 0 % Fawns deer BASK; % MAY J JUN I JUL I AUG ( SEP OCT NOV f DEC I JAN | FEB j MAR t A P R | U --------------------- P E L L E T GROUP D EPO SITIO N P E R IO D ------------------* | Figure 6 . Diagramatic representation of a hypothetical Michigan deer herd. 46 Furthermore, it has long been known that deer, even oh high nutri tional diets fed ad libitum, undergo reduced food consuption and weight loss in mid-winter. Normally the decline starts in November and reaches a low point in February or March after which there is an increase (Davenport 1939, McBeath 1941, French et al. 1955, Magruder et al. 1957, McEwen et al. 1957, Silver and Colovos 1957, Wood, Cowan, and Norden 1962, Ozoga and Verme 1970). This has even been observed in Louisiana where cold winters are not a factor (Fowler, Newsom, and Short 1968). There is some evidence that there is an initial decrease in metabolism coincident with change to winter coat in September and that further re duction is due to onset of the rut in October or November and later still due to shorter day lengths (Silver and Colovos 1957, Silver et al. 1969, Silver 1969). Hoffman and Robinson (1966) reported a corresponding re duced function of various endocrine glands during January and February. The concern is that there may be a corresponding decline in pellet group deposition during mid-winter when deposition studies have been carried out. If true, computed rates would be smaller than the true average rate. Linear regression lines, defecation rates (Y) versus dates (X), were fitted to all records included in Tables 38 to 41 (except the pen of adult bucks with only two observations). Of the 35 records, only three entries, two pens of does and one doe fawn disclosed significant (P<.05, P<.05 and P<.01 respectively) negative slopes while one pen of does showed a significant (P<.05) positive slope. studied for defecation rates died of malnutrition. Three of the deer None of these showed evidence of a progressive decline in pellet deposition. Overall, about two-thirds (22) of the 35 had negative slopes and this strongly suggests 47 at least a slight decline for most deer In mld-wlnter. As noted above, analysis of the 1954 Cusino study revealed a sig nificant diet x date interaction. This' suggests there was some change in defecation rate with date in one or more of the diets. One problem is that many of these records involve multiple deer per pen and many extend over a fairly short range of time. Nonetheless, if we look at only individual deer studied for relatively long periods, we obtain much the same results. From the 1959 investigations at Cusino and Houghton Lake, there are 13 individual deer, two adult bucks, four adult does, three buck fawns, and four doe fawns upon which counts were made in portions of three or more months. In only two instances, a doe (observation 12, Table 39) and a doe fawn (observation 12, Table 41), were slopes of linear regression lines fitted to these data significantly different (P<.01 and P<.05) from zero. However, none of the 13 have negative slopes Including these two. Intuitively one would expect that reduced intake would result in reduced fecal formation. The real question, though, is whether the rate differs, not the volume. Conflicting data are reported in the liter ature. Mautz (1969) studied the mean retention time of food in experi mental white-tailed deer using 51-chromium as a tracer. He progres sively decreased the food intake for one deer while leaving the others at ad libitum levels of feeding. He concluded that the rate of food passage was not affected significantly by the amounts of food eaten. Ullrey et al. (1964), however, noted food passage in their experimental white-tailed deer was generally slower amounts of food were consumed. than normal when very small Smith (1964) found an increase of defeca tion rate in mule deer with increased food consumption when rates are 48 plotted against consumption on a per hundredweight basis. Rogers et al. (1958) quote E. A. Hewitt of Iowa State to the effect that for livestock: "There is a very definite correlation with a decrease or increase in the amount of forage consumed upon the defecation rate . . . cattle usually defecate 10 to 24 times daily - more frequently if the food contains more water as in the case of succulent pasture." Some crude data are available from both the 1954 and 1959 Cusino investigations which can be used to compare food ingested with defecation rate. All involved pens with two to four deer. Average browse consump tion was reported for 3-week periods for the various pens. This can be compared with three to six pellet rate determinations per pen. unfortu nately, two determinations were commonly made in the same browse period. Data for three of the 12 pens in 1954 showed significant correlations. Adults on fire succession and fawns on hemlock-hardwood showed a signi ficant positive correlation (P<.05) of food consumption versus pellet deposition rate, but fawns on the control diet showed a significant negative correlation (P<.01). In 1959, none of five pens, with three deer in each, showed a significant correlation. Biologists at Cusino sampled the groups present in each of 12 pens and counted the number of pellets per group on February 27, 1954. Table 1 shows a comparison of the mean pellets per group with the mean defe cation rate per deer made on February 26. Data for neither age group reveals a significant correlation between the two variables. There is evidence in the literature that switching from woody browse to lush new leafy foods results in increased defecation rates. Longhurst (1954) reported that domestic sheep in California placed on green grass after feeding on grain, stubble, and hay increased defeca tion rates from a mean of 13.26 to 15.50 per animal day. He also cites Table 1 Comparison of Pellet Group Size and of Defecation Rates in Cusino Deer, Late February, 1954 Diet Control Swamp Conifer Swamp Hardwoods Hemlock Hardwoods Fire Succession Mixed ConifersUpland Hardwoods 69.0 64.3 Correlation Coefficient r Adult Does Mean pellets per group 73.6 71.3 94.6 ] 67.0 .27' Mean groups per deer 15.0 13.6 13.3 l 12.0 12.0 13.0 62.3 67.3 52.0 67.0 63.0 Doe Fawns Mean pellets per group 49.0 -.05' Mean groups per deer 14.0 12.0 14.0 ll.'O 12.3 10.0 ^ n e adult buck and two adult does were included in this pen, all other pens held three females each. 2This is a measure of the intensity of association between the two sets of means. P .05 for 4 degrees of freedom, r = .75. 50 unpublished data from A. D. Smith that summer rates are higher than winter rates in penned mule deer. From studies of both mule and white tailed deer, Zyznar and Urness (1969) conclude that coarse browse ap parently passes through the digestive tract more slowly than concentrates and herbaceous foods. Mautz (1969) reported passage rate was inversely correlated with food particle size. Studies by Short, Medin, and Anderson (1966) seem to explain the mechanism involved. They found in mule deer that during the summer when the animals are feeding on succulent herbs, forbs, grasses and leaves, there is a greater rumen fermentation rate and a greater turnover of rumen contents. What these studies seem to indicate is that defecation rates are fairly consistent in mid-winter, but with evidence for a slight decline. During this period, deer are subsisting on dormant, woody browse. Con sumption of herbaceous foods apparently results in somewhat higher defecation rates. Normally this should not be a problem in the spring because surveys are usually completed before green-up. In the fall, however, deer are still able to obtain at least some herbaceous material. Hence, one might expect deposition rates then to be higher than winter, but there seems to be no way of showing this with the information at hand. I feel strongly that defecation rates need to be studied for the period from mid-October to the end of December. The diets used should simulate the mix of plant species utilized by the majority of wild deer at this time. Such studies, however, appear to be unusually difficult. Determination and collection of the proper foods for penned studies would be next to impossible. In spite of their inherent problems, the best approach might be to use rather large corrals. 51 Corral studies To date, Michigan has only carried out one pellet rate determina tion which involved the use of a large "corral." This was the 1959 Houghton Lake study. The four deer stocked in the 6 acres represent a deer herd of 427 per square mile. While this seems excessive, many natural populations in Michigan have exceeded this during the winter yarding period. On the average, deer yards in northern Lower Michigan occupy less than 10 per cent of the summer range (Bartlett 1950). When this study area was fenced off during the winter of 1951-52, we drove 28 deer out of the in terior before the fourth side was put in place. Nevertheless, when biologists examined the area in May the adult buck had already died. Herbert Johnson, biologist at Houghton Lake, who discovered the dead animal estimated the time of its death as late March. However, since we do not know exactly when the buck died, it greatly re duces the usefulness of the study. The adult doe was removed May 27 and the two fawns not until June 8 . Biologists making the counts had difficulty distinguishing separate groups of pellets, particularly in runways where they were often piled on top of one another. Best estimates were that a total of 2,883 groups were present on the 60 transects. This provides an estimate of 10,571 ± 531.72 (95 per cent confidence interval) for the 6 acres. If we assume the adult buck died on March 31, then the total deer days becomes 527 and the average pellet groups per deer day = 10,571 * 527 => 20.06. Our other experience leads us to be surprised at this high rate. One complicating factor was that observers were not concerned about distinguishing old groups, since the area had been deer free for 7 years. 52 We assumed that none of the groups deposited prior to building the fence would still be extant. However, as VanEtten and Bennett (1965) reported, some groups may persist at least as long as 5 years. Normally these would be classified as "old" by field men during the regular surveys. The author visited the enclosure a number of times in the years immedi ately following its construction and was impressed by the persistence of pellet groups present. We must admit, therefore, that the possibility exists that some old groups were counted, but it is difficult to believe these constituted a very high proportion. A potentially more serious problem was that the deer were not re moved in March or April when the normal yarding period ends. Further more, in most years the average date of the extensive deer pellet group surveys is mid-April in this area. As mentioned above, evidence suggests that switching from dormant, woody browse to lush, new growth results in increased defecation. Very likely these deer had been feeding on leafy foliage for at least 1 1 / 2 months. Hence, I cannot view the results of this trial as having much significance in relation to pellet group surveys. V. DEPOSITION PERIOD In the Rocky Mountain states, mule deer characteristically migrate between rather discrete winter and summer ranges. On a range used by deer only in the winter, workers need merely to distinguish the recent winter's pellet groups from those of the previous winter and note the time of arrival and departure of deer (McCain and Taylor 1956). In Michigan, however, the difference between summer and winter deer range is frequently ambiguous. Although deer tend to be more restrictive in their choice of winter cover, they commonly use many of the same sites throughout the year. All Michigan work has been based on a fall-to-spring accumulation of pellet groups. Ideally, this period should be bounded by instantan eous beginning and ending points, such that each pellet group encountered could be easily assigned either within or without the period. Limita tions in technology and manpower have generally precluded the ideal situation to date. Hence, less definite limits have been utilized. The autumnal fall of leaves has been used to provide a beginning reference point and the mean date of the spring survey the other bound. LEAF FALL DATES The premise is that the fallen leaves will blanket the ground and cover up all pellet groups deposited earlier. In northern Michigan, leaves of most deciduous trees and shrubs fall during a fairly short period. Often various combinations of hard rains, winds, and heavy, wet snows will bring down most leaves in one or two days. Leaves of some oaks and beech (Fagus grandifolia), however, may persist until late winter or early spring. Cropland, pasture, wild openings, marshes, and conifer stands are also potential problem areas. 53 54 While no specific studies have been carried out to explore varia tions in leaf fall dates, reported dates were studied to provide some indication of variability and potential problems. Results Records of leaf fall dates for 11 years (1959 to 1969) disclose that they have been relatively uniform within each game district (Tables 2 and 49). In most cases the range is about 2 weeks, the maximum being 27 days in old District 8 . Standard errors range from .55 to 3.60 days. In every case, except old District 8 and old District 9, the range in leaf fall dates is less than 10 per cent of the average deposition period. A simple linear correlation matrix (Table 3) of dates among the several districts suggests that their leaf fall dates show only a slight tending to covary. Significant correlations (P<.05) were obtained in only 7 out of 54 instances: new District 6 versus Districts 4 and 3, old District 6 versus Districts 1, 3, and 4, and District 3 versus Dis tricts 1 and 4. Relative lack of correlation reflects vegetational and climatic differences as well as observer differences. Leaves tend to drop progressively later from north to south, but the influence of the Great Lakes precludes a regular progression. Furthermore, no regular procedures are followed by field men to determining leaf fall dates. Leaf fall dates are the result of impressions tempered by past experience and perhaps unduly weighted by conditions in the areas where each biol ogist lives and where he is working in mid-October. Since game districts cover several counties, leaf fall may vary somewhat from one part to another. An average date for the district is selected. At times airplane flights provide a better perspective, but such trips are generally coin cidental in regard to leaf fall. 55 Table 2 Summary of Leaf Fall Dates by Game Districts District Number of years Mean Standard error of the mean Range 1 11 Oct. 12 1.45 Oct. 2 ■- Oct. 18 2 11 Oct. 14 1.22 Oct. 8 •- Oct. 21 3 11 Oct. 14 1.51 Oct. 8 •- Oct. 23 4 11 Oct. 24 1.51 Oct. 17 - Oct. 31 5 11 Oct. 19 1.22 Oct. 11 - Oct. 25 Old 6 6 Oct. 23 2.03 Oct. 16 - Oct. 27 Old 7 6 Oct. 23 2.02 Oct. 18 - Oct. 31 Old 8 6 Oct. 25 3.60 Oct. 11 - Nov. 6 Old 9 6 Oct. 20 2.79 Oct. 16 - Nov. 3 New 6 5 Oct. 24 2.20 Oct. 22 - Nov. 3 New 7 5 Oct. 20 1.86 Oct. 15 - Oct. 25 New 8 5 Oct. 24 .55 Oct. 22 - Oct. 25 I Table 3 Simple Linear Correlation Matrix of Leaf Fall Dates by Game Districts 1 District 1 1.00 2 3 4 5 Old 6 New 6 Old 7 New 7 Old 8 New 8 Old 9 * = P<.05 ** = P<.01 2 .50 1.00 3 Old New New 7 New 6 Old 7 Old 6 8 8 .88* .56 -.27 .37 .41 -.40 -.11 .67 .17 -.38 -.06 .15 -.80 -.51 -.01 .86* .93* -.22 .44 .42 -.11 -.23 .20 .91* .94* .33 .71 .52 .34 -.01 .75 .52 .60 -.05 -.09 -.21 4 5 .72* .57 .09 .56 .23 .26 1.00 .78** 1.00 1.00 -.06 1.00 .17 1.00 .67 .56 1.00 Old 9 .08 .25 .68 .62 1.00 .64 1.00 .75 1.00 1.00 in on 57 Leaf fall dates do not enter the population estimator (Chapter II) per se, but only as a part of the deposition period. A variance esti mate for deposition period can be obtained from the variability in the length of the yearly periods given in Table 49. Adding such a component of variation, however, has very little influence on confidence limits. As an example, approximate confidence limits (Deming 1964) for the 1969 survey in District 6 were calculated to be ±39.20 per cent (2 standard errors of the mean). This can be compared to +38.96 per cent when both deposition period and defecation rate were assumed constant and only counts of pellet groups were considered random variables. Taking into account variance components for all three of these elements, limits of ±39.82 per cent were calculated. The additional contributions seem small enough to be safely disregarded. Hence, confidence limits report ed here include only estimates of the sampling errors due to counting pellet groups. Discussion In my experience, where fallen leaves do not cover the ground, survey crews use matted vegetation as an equivalent reference point. Adequate snow cover to flatten ground plants, however, may not occur until some time after leaf fall, particularly in the Lower Peninsula. In general, cover types in the Upper Peninsula seem more favorable to pellet group surveys, than in the northern Lower Peninsula (Table 4). The area covered by oak and cropland in the Upper Peninsula totals less than 5 per cent, while in the northern Lower Peninsula it comprises over one-quarter of the land area. Similarly, hardwood stands, except oak- hickory, cover nearly 60 per cent of the Upper Peninsula but only about one-third of the northern Lower Peninsula. Pine occupies less than 10 58 Table 4 Distribution of Major Cover Types in Northern Michigan 19661 Northern Upper Peninsula Lower Peninsula Acres Prop. Acres Prop. 93,700 .0088 43,000 .0038 Red pine 234,900 .0222 325,100 .0285 Jack pine 412,300 .0390 472,100 .0415 82,000 .0072 White pine Scotch pine White spruce balsam fir 872,600 .0824 193,000 .0169 Black spruce 398,700 .0377 29,400 .0026 Tamarack 115,200 .0109 40,300 .0035 Nor. white cedar 843,800 .0797 333,800 .0293 3 ,429,500 .3240 1,313,400 .1153 92,400 .0087 1,296,100 .1138 586,100 .0554 630,500 .0554 1 ,792,000 .1693 2,061,300 .1810 Paper birch 218,800 .0207 174,000 .0153 Unproductive and reserved 374,600 .0354 57,700 .0051 Cropland 246,500 .0233 1,169,400 .1027 142,200 .0134 572,400 .0503 Other 731,100 .0691 2,593.900 .2278 TOTAL 10 ,584,400 1.0000 11,387,400 1.0000 Maple-beech-b irch Oak-hickory Ash-elm-cottonwood Aspen Pasture and range * *After Chase, Pfeifer and Spencer, Jr. 1970. 59 per cent in both regions. The only "troublesome" types where the Upper Peninsula leads, are the other conifers which total 21 per cent there compared to 5 per cent in the northern Lower Peninsula. Leaf fall and pellet group deposition both occur on a continuum over time, while matting of fallen leaves and herbaceous plants by snow cover are essentially discontinuous events. Workers agree that the near ideal situation occurs on northern hard wood sites in the Upper Peninsula. Here snow normally arrives in Novem ber and persists until spring breakup. short period and is complete. Leaf fall occurs during a fairly Deep snow compresses ground cover, re sulting in a flat table-like appearance upon which overwintering pellets are conspicuously displayed in the spring. The early fall snow cover prevents deep frost penetration, and runoff from spring snow melt is rapidly absorbed. While it appears that field biologists are providing fairly accurate leaf fall dates, the real problem is to decide whether a particular group was dropped before or after the selected date. In computations the as sumption is made that all pellet groups deposited on the plots during the selected deposition period will persist, be found, and be identified properly on the spring survey. Furthermore, that any groups found which were dropped prior to the leaf fall date will be classified as "old" and will not enter into population estimates. The veracity of these assump tions will be considered in Chapter VI. DATES OF SPRING SURVEYS The dates that plots are searched in the spring are known with cer tainty. Ideally, counts are carried out as soon as possible after the snow disappears. Research studies on small study units present no real 60 problems In this respect. a few days. Ordinarily they are completed In a matter of With extensive operational surveys, however, many circum stances arise which interfere with this hoped-for situation. Survey periods have ranged up to one and one-half months. The commonest problem is that the spring thaw may proceed gradually with patches of snow remaining on north-facing slopes and in swamps for several weeks. Then too, the opposite situation - an early breakup with no precipitaion - may result in many workers being involved with forest fire control. Other hazards to efficient operation include spring snows, impassable roads, floods, and rains. A prolonged survey period may also mean new herbaceous growth will have started before counts are completed. Besides, the possible effect on pellet deposition, growing vegatation disturbs the leaf mat and makes finding and aging pellet groupsmore dif ficult . Logically, extended counting periods might cause serious biases. For example, if high deer population areas were searched later than low population areas, overestimates would result. Basically, our concern is with the extensive northern surveys. Records for the 94 game district surveys run from 1959 to 1969 were examined to determine sample of 11 ranges of survey periods. of these were analyzed in Records from arandom more detail to determinethe effects of long survey periods. Results Table 5 provides an overview of survey dates. The overall average length of survey periods is 24.04 days, varying from as little as 8 days for District 5 in 1965 to 45 days for District 8 in 1962. In 72 (77 per cent) instances, median and mean survey dates were within one day of Table 5 Dates of Pellet Group Surveys by Game District District 1 Range 2 Days between Mean date Median date Range 3 Days between Mean date Median date Range 4 Days between Mean date Median date Range 5 Days between Mean date Median date Range 61 Days between Mean date Median date Range Days between Mean date Median date 1960 1959 Apr 23- Apr 25May 12 May 24 20 30 May 4 May 8 May 4 May 5 Apr 20- Apr 25May 6 May 27 17 33 Apr 28 May 4 Apr 29 May 3 Apr 23- Apr 25May 11 May 24 19 30 May 2 May 8 May 2 May 4 Apr 21- Apr 25May 11 May 11 21 17 May 2 May 4 May 5 May 2 Apr 25- Apr 19May 12 Apr 29 17 11 Apr 30 Apr 26 Apr 30 Apr 26 Apr 20- Apr 20Apr 30 May 2 11 13 Apr 25 Apr 26 Apr 26 Apr 26 1961 Apr 27May 19 23 May 9 May 10 Apr 18May 5 18 Apr 27 Apr 28 Apr 14May 12 29 Apr 28 Apr 26 Apr 14May 9 26 Apr 25 Apr 26 Apr 7Apr 30 24 Apr 21 Apr 21 Apr 4May 5 32 Apr 16 Apr 18 1962 Apr 25May 17 23 May 7 May 8 Apr 26May 15 1963 1964 Apr 4- Apr 24May 6 May 20 33 27 Apr 23 May 5 Apr 24 May 4 Apr 5- Apr 20Apr 17 May 6 20 13 17 May 6 Apr 11 Apr 25 May 7 Apr 11 Apr 25 Apr 27- Apr 3- Apr 13May 17 May 7 May 8 21 35 26 May 8 Apr 16 Apr 25 May 8 Apr 17 Apr 27 Apr 25- Apr 16- Apr 21May 16 May 10 May 18 22 25 28 May 7 Apr 22 Apr 29 May 7 Apr 23 Apr 28 Apr 30- Apr 10- Apr 20May 14 Apr 29 May 5 15 20 16 May 5 Apr 18 Apr 25 May 4 Apr 18 Apr 24 Apr 23- Apr 8- Apr 5May 15 May 1 May 1 23 24 27 Apr 29 Apr 13 Apr 18 Apr 26 Apr 11 ..A?r I?.. 1965 May 3May 25 23 May 15 May 17 Apr 28May 25 28 May 11 May 12 Apr 27May 20 24 May 11 May 12 May 1May 20 1966 Apr 28May 25 28 May 12 May 12 Apr 20May 13 24 May 7 May 8 Apr 18May 26 39 May 7 May 7 Apr 16May 13 20 28 May 10 May 4 May 11 May 4 May 3- Mar 29May 10 May 11 8 44 May 6 Apr 21 May 6 Apr 20 Apr 22- Apr 18May 15 May 11 24 24 Apr 30 Apr 25 Apr 30 Apr 22 1967 Apr 24May 17 24 May 4 May 3 Apr 17May 5 19 Apr 26 Apr 26 Apr 19May 18 30 May 7 May 9 Apr 19May 15 27 May 2 May 3 Apr 10Apr 29 1968 Apr 10May 6 27 Apr 25 Apr 29 Apr 9Apr 24 16 Apr 16 Apr 17 Apr 2May 9 38 Apr 16 Apr 17 Apr 16May 7 22 Apr 24 Apr 24 Apr 1Apr 19 20 19 Apr 19 Apr 9 Apr 19 Apr 10 Apr 17- Apr 2Apr 27 May 2 11 31 Apr 21 Apr 17 Apr 20 Apr 16 1969 Apr 23May 16 24 May 5 May 6 Apr 23May 9 17 May 3 May 5 Apr 28May 13 16 May 4 May 6 Apr 23May 15 23 May 6 May 7 Apr 14May 19 36 Apr 26 Apr 26 Apr 10May 5 26 Apr 21 Apr 23 Table 5 Dis trict 7l Range 81 Days between Mean date Median date Range 91 Days between Mean date Median date Range Days between Mean date Median date 1959 Apr 10May 7 28 Apr 24 Apr 23 Apr 7May 7 31 Apr 18 Apr 17 Apr 8Apr 22 15 Apr 14 Apr 14 1960 Apr 20May 12 23 Apr 27 Apr 27 Apr 12Apr 22 1961 Apr 11May 8 28 Apr 15 Apr 13 Apr 4Apr 19 11 16 Apr 19 Apr 10 Apr 19 Apr 11 Apr 14- Apr 10May 12 Apr 26 29 17 Apr 23 Apr 15 Apr 23 Apr 14 (cont'd.) 1962 1963 Apr 19- Apr 8May 8 May 13 20 36 Apr 28 Apr 18 Apr 27 Apr 17 Apr 3- Apr 2May 17 May 1 45 30 Apr 11 Apr 10 Apr 6 Apr 9 Apr 9- Apr 6May 9 Apr 25 31 20 Apr 17 Apr 19 Apr 16 Apr 17 ^ a m e District boundaries were changed in 1965. 1964 Apr 13Apr 26 14 Apr 17 Apr 16 Mar 20May 1 43 Apr 11 Apr 13 Mar 24Apr 15 23 Apr 6 Apr 9 1965 1966 Apr 19- Mar 28May 10 May 5 22 39 Apr 29 Apr 18 Apr 29 Apr 19 Apr 20- Mar 28May 3 Apr 28 32 14 Apr 25 Apr 10 Apr 26 Apr 18 1967 Apr 10May 3 24 Apr 15 Apr 13 Apr 10May 3 24 Apr 18 Apr 18 1968 Mar 27Apr 22 27 Apr 4 Apr 2 Apr 1Apr 30 30 Apr 9 Apr 9 1969 Apr 14May 9 26 Apr 23 Apr 23 Apr 7Apr 22 16 Apr 14 Apr 14 63 each other suggesting symmetrical frequency distributions although not "bell-shaped." The reason for this involves the finite character of the manpower resources. As an example, Table 6 presents frequency distri butions of dates for 1962 surveys. Ranks of mean survey dates by strata for the 11 surveys are given in Table 7. Significance of these stratum ranks was tested using Friedman’s test (Campbell 1967). In order to do these computations, data for stratum V and the 1962 District 5 survey were dropped and ranks recast accordingly. Results indicated there was not sufficient evidence (a = .05) to reject the null hypothesis of no difference between strata, .05
.05 level (,05
n£n 2 unless n^ = n 2 = n£ = n£ = 0 nj >0 and n 2>0 unless n^ = n 2 = n£ = n£ = 0 In terms of application then, the first man must find at least one group on his half-plot unless there is really none there, otherwise the estimate will be undefined or zero. Furthermore, the second m a n ’s adjustments must always add to the first, otherwise the estimated total will be less than those actually found. Finally, the recheckers must not find enough additional groups to make the product of the two additional counts equal to or greater than the product of the original counts, other wise the estimate will be undefined or negative. Hendrick suggests that some of these difficulties can be overcome by lumping data from several plots for the same crew, but this masks the real or potential differences of finding groups in various cover types, and complicates estimating procedures. Discussion The relative success of these surveys has depended in very large part on the work habits, motivation, and skill of the game biologists involved. On the northern survey, they set the example for their crew members and are responsible for resolving differences in numbers and group age. Obvious as it is, this point was difficult to get across at briefing sessions. Many seemed to feel that statistical "mumbo-jumbo" could salvage a poor job. Some personnel may have genuine eyesight problems, but in most cases missed groups can be attributed to haste and lack of concentration. Ideally, counts should be made on bright, overcast days rather than under clear skies. Intense sunlight causes confusing patterns of light and shadow which makes pellet groups difficult to see. Recheckers should traverse their partner's half-plot in the opposite direction on sunshiny days. On the other hand, counts during or following rain should also be avoided. Robinette, Ferguson, and Gashwiler (1958) point out fresh and older groups are often indistinguishable when wet. Riney (1957) recommends waiting at least 2 or 3 days following a rain. Winter deer concentrations in deer yards, tree cutting operations, and heavily used runways can result in high pellet density. result when survey plots fall on such sites. Problems Not only is it difficult to sort out the different groups, but merely counting a hundred or more groups is very tedious. Often two or more groups will be intermixed and careful examination will be necessary to properly gauge the number of groups actually present. Size, shape, texture, and color can be assessed to differentiate individual groups with the reservation that pellets frequently have a slightly different hue where they have been in contact with the ground. Neff (1968) and his co-workers in Arizona concluded that adjacent groups of very similar mule deer pellets should be counted separately unless they were definitely connected by scattered pellets. Otherwise, observers who tend to be "lumpers" might make serious under counts . Another aspect of counting pellet groups concerns the fact that all groups deposited by deer are not in the form of neat and distinct piles of pellets. Riney noted that under zoo conditions, 26.5 per cent of red deer and 17.5 per cent of fallow deer pellet groups were "stringers." Robinette et al. (1958) found 8 per cent of 1,485 mule deer groups were "strewn-out." Unfortunately, such data are not available in Michigan, but there is ample empirical evidence that scattered groups occur here frequently in the wild. counts. Hence, the size of the plot may affect the Batcheler (1971) suggests that in general, the larger the plot the more likely that scattered pellet assemblages will be identified properly as groups. Every group should have a chance of being counted. Robinette et al. (1958) found the mean length of 185 strewn-out mule deer groups to be 14.5 100 feet. If plots were "too small" none could contain half of such a group. A circular .001 acre plot, for example, could contain half of a group 14.5 feet long only if the group was strewn across the center. plot size of 12' x 72.6' should handle most groups properly. Michigan's Notwith standing, on future surveys I recommend adopting the suggestion of Robinette and his co-workers to count a group only if its midpoint (cen ter of gravity) falls within the plot. On the other side of the coin, Michigan workers were instructed to count even one pellet as a group if after careful examination it could not be assigned to a nearby group. This procedure was based on the practice of Bennett et al. (1940). More recent studies of mule deer droppings by Smith (1964) and Neff (1968) and both mule and white-tailed deer by Hart (1958), recommend that a larger number of pellets be used. Neff rather arbitrarily set 30 as the minimum for his Arizona surveys while Hart used five in South Dakota. Under Michigan conditions I feel that 10 pellets would be appropriate as an interim threshold for use in future surveys, pending receipt of needed research findings. FECAL IDENTIFICATION Murie (1954) points out the great similarity between the droppings of many hoofed animals and the difficulty of identifying pellet groups to species. In Michigan, we have occasional problems distinguishing the feces of porcupine (Erethizon dorsatum). snowshoe hare, and cottontail rabbit from deer; but more troublesome are elk and domestic sheep. Pellets left by the first three animals named can usually be ac curately identified, and field men are instructed in recognizing them. Both rabbit and hare droppings tend to be round, rather than oblong, 101 and flattened from side to side rather than circular In cross-section. Porcupine pellets are usually longer than deer, of a rougher texture, often with a longitudinal groove, and usually scattered under a tree rather than in distinct groups. lying among the pellets. Frequently, cut twigs will be found At times, large piles of porcupine droppings are found at the base of a hollow tree or fallen log which an individual animal has used for a den. Elk occur in numbers only in one area of the state, centered in the Pigeon River State Forest near Gaylord. Adult elk pellets are notice ably larger in diameter than deer and normally groups contain more pel lets. Overlaps occur, however, between the droppings of small elk and large deer. Field biologists who make pellet group counts in the elk country are familiar with the droppings of both species and decide on the ground to which species questionable groups belong. Experimental elk pellet group surveys undertaken in the major elk range from 1963 to 1967 indicated about two to eight times as many deer as elk pellet groups present in the areas searched. Domestic sheep pellet groups seem to be virtually impossible to distinguish visually from deer, and plots falling in known sheep pas tures are not counted. Howard (1967) working on western ranges where both mule deer and pronghorn (Antilocapra americana) occurred developed a simple chemical method of distinguishing pellet groups of the two species. He determined mean pH values of 6.07 for mule deer and 7.61 for pronghorn and found no overlap. Nagy and Gilbert (1968) extended this technique to mule deer and domestic sheep and found a significant difference in pH between the two, averaging 7.31 for sheep and 5.72 and 6.46 for mule deer from two different ranges. We have not carried out similar investigations in Michigan for elk, sheep or deer pellets, nor am I aware of any such work done with white tailed deer. be researched. This is an area offering considerable promise and should VII. EXPERIMENTAL FIELD TRIALS Michigan is indeed fortunate and unique in having two large deer enclosures available for deer research - the Edwin S. George Reserve in southern Lower Michigan and the Cusino enclosure in the Upper Penin sula. The two areas are quite dissimilar ecologically, geographically, and climatologically. Neff (1968), in his literature review of the pellet group technique, reported that only Michigan and Georgia had tested pellet group counts on areas with known populations. Experimental pellet group surveys on such fenced areas allow researchers to "look up the answer in the back of the book." Georgia studies (Downing et al. 1965) were aborted when it became apparent insects were rapidly destroying pellets. similar trials is not unexpected. Lack of The cost of enclosing and maintain ing even one square mile with a deer-proof fence, and the difficulty of keeping tabs on the true size of the enclosed herd are formidable ob stacles. GEORGE RESERVE Description of the area The George Reserve is located in southern Michigan, in Livingston County, 3 miles west of Pinckney. The following historical notes and description of the area is largely from Jenkins (1964). This property was purchased in 1927 as a country estate by Colonel Edwin S. George, a Detroit industrialist. 12 separate farms. Prior to this date the area had been occupied by Since 1927 there has been no cultivation, no lunber- ing, no grazing by domestic stock, no artificial feeding, no fires, and no shrub or tree planting except one small planting of white pine and red pine (Pinus resinosa). For the most part, only red pine survived. 103 104 The terrain ranges from rolling to rough. Glacial action produced steep- sided eskers, moraines, kettle holes, swamps, and marshes. of loam, loamy sands, sandy loams, peats, and muck. Soils consist About one-third of the area consists of open grassland, about one-third Is In upland hard woods (largely oak, Quercus velutlna. £. rubra, and (J.. alba: hickory, Carya glabra. _C. ovata; and aspen, both quaking and blgtooth), about onetenth each Is In marsh and wooded swamp (mostly tamarack, poison sumac, Rhus v e m i x . and a variety of other shrubs) and the remainder in ponds, bogs, lowland shrubs, and the pine plantation. It became clear early in this study that the actual size of the Reserve was other than the published reports would indicate: Hickie (1937) gave "2 sections"; O'Roke and Hamerstrom (1948) "about 1,200 acres"; and Tody (1949) "1,268 acres." Several crude attempts by me to ascertain the true size stimulated Jenkins (1964) to determine that the area inside the fence is 1,146 acres with another 122 acres outsidethe fence. report I have used 1,146 as the true acreage. In this The deer herd Colonel George enjoyed seeing deer and consequently enclosed the area with a nearly deer-proof 7-foot fence and stocked it in 1928 with two bucks and four does. The deer came from Grand Island in Lake Superior near Munising, Michigan. After six breeding seasons, a deer census drive count indicated a population of 162 in December, 1933, including two found dead prior to the drive (Hickie 1937). Jenkins (1964) believes the true population at this time may have been even higher. Deer drive count3 have been conducted with University personnel and students essentially annually since 1933 (Chase and Jenkins 1962). Details about the herd and its population dynamics can be found in 105 Hickie (1937), Kelker (1947), O'Roke and Hamerstrom (1948), Hamerstrom and C a m b u m (1950), Chase and Jenkins (1962), Jenkins (1964), and Haney (1969). Since 1952, lower jaws from all deer shot or found dead have been collected. Although many of the participants in the annual drive counts had some doubts about their accuracy, it was not until Jenkins (1964) review ed original drive records and analyzed ages from deer jaws, that actual evidence of discrepancies come to light. Dr. Jenkins and Dr. L. Dale Fay of the Game Pathology Laboratory, Department of Natural Resources, aged all collected deer jaws from 1953 through the 1962-63 harvest, using tooth replacement and wear criteria (Ryel et al. 1960). Since that time, Dr. Dale R. McCullough (Professor in the Department of Wildlife and Fish eries) and his students have continued this task. Jenkins (1964) described his "aging method" of determining deer popu lations on the Reserve as follows: "If the age of each deer is known the actual population during earlier years can then be reconstructed. To illustrate: If a deer was 3 1/2 years old when killed in December 1957, it must have been present as a 21/2-year-old in 1956; as a 1 1/2-yearold in 1955; and as a fawn in 1954." This is essentially the "virtual population estimation procedure" developed by Fry (1949) for fisheries studies. The chief problem with the method is that it requires a wait of several years before we can be confident that most of the animals alive in a given year have died. For example, the last pellet group survey on the Reserve was carried out in 1963 and I felt it necessary to wait at least 5 years before determining the population for that year. Actually, after 1965-66, harvests produced only five deer which were alive in the spring of 1963 or before, two each in 1966-67 and 1967-68 and one in 1968-69. Jenkins indicated that 106 during the first 11 years in which jaws were collected, only 22 deer over 4 1/2 years old were killed. The accuracy of aging deer by the tooth nique varies inversely with age (Ryel et al. replacement and wear tech 1961). Inrecent years, a newer and more accurate technique of using annuli in the cementum has been developed (Low and Cowan 1963, Ransom 1966, Gilbert 1966). All jaws from the reserve older than 2 1/2 years were aged again by Bromley (1968) using this latter technique and population estimates were adjusted ac cordingly. Ages for about 20 deer were changed. In addition, my data differ slightly from those of Jenkins because of inclusion of several deer that turned up in a re-examination of old records by Dr. McCullough. Deer populations constructed from the aging method are shown in Table 17 for the period covered by pellet group surveys. More detailed informa tion on the construction of yearly populations is given in Table 59. The fence surrounding the Reserve prior to renovation in 1963 was only 7 feet high with a 12-inch barbed wire overhang on both sides of the top. This cannot be considered completely deer-proof and some deer were observed to jump the fence on deer drives. Jenkins' (1964) analyses, however, indicated fence jumping must have occurred only rarely and per haps nearly always during the drives when such instances were recorded. A few deer were also observed jumping back inside the fence during or following deer drives. I have arbitrarily considered all animals leav ing the Reserve and all unaged deer which were found dead to be 1 1/2 years old (median age of the herd). Deer found dead were assumed to have died on the median date of the known harvests. All deer which jumped the fence were assumed not to have returned. University of Michigan personnel feel that the efficiency of the Table 17 Known Deer Populations for the George Reserve Based on Aging Method Year of survey Deposition period Deer on area at leaf fall Deer days Deer killed after leaf fall Subtract deer days Net total deer days Average deer per sq. mile 179 87 15,573 14 1,718 13,855 43.23 1954 182 118 21,476 42 5,175 16,301 50.02 1955 181 123 22,263 46 4,930 17,333 53.48 1956 162 125 20,250 38 4,878 15,372 52.99 1958 167 108 18,036 58 6,017 12,019 40.19 1959 165 84 13,860 17 2,126 11,734 39.72 1960 163 101 16,463 51 5,925 10,538 36.10 1961 161 87 14,007 33 3,475 10,532 36.53 1962 144 94 13,536 49 5,078 8,458 32.80 1963 153 66 10,098 13 1,570 8,528 31.13 107 1953 108 drive counts has been much Increased In later years. This Is due to both the use of more participants and a better disciplined operation. A comparison between drive counts and aging estimates, Table 18, supports this feeling. Nevertheless, it is difficult to account for apparent over estimates in drive counts which occurred in three years, particularly the difference of 16 deer in 1963. At first I was tempted to blame this on the fence renovation project in 1963 which might have allowed some deer to escape that year. An undocumented loss of deer in 1963 would also af fect the reconstructed populations for previous years as well, although to a lesser extent. Inspection of the data could easily lead one to be lieve that such a loss occurred. University of Michigan workers, however reported that the method used to repair and extend the height of the fence did not result in a break at any time. Furthermore, the normal behavior of deer in regard to their home ranges would seem to rule out any mass exodus even if it were possible. losses would be through poaching. of local people ran strongly Another probability for deer In the 1961-63 period the feelings against the deer harvest operations on the Reserve and this resentment may have resulted in illegal shooting and re moval of deer from the Reserve. Indeed a few such instances are actually known but there is no definitive evidence that any large deer losses from poaching occurred on the Reserve during or about 1963. During the study period, over-winter deer populations averaged 41.6 per square mile with a range from about 31 to 53. Table 19 shows the average composition of the over-winter herds together with the expected deposition rates. In pellet survey computations a weighted defecation rate of 13.79 was used. Reference to Table 17 points up the potential problems which could 109 Table 18 Comparison of Deer Drive Counts with Population Estimates Based on the Aging Method for the George Reserve Year 1952-53 1953-54 1954-55 1955-56 1956-57 1957-58 1958-59 1959-60 1960-61 1961-62 1962-63 Date of drive count Jan. Jan. Dec. Nov. Nov. Dec. Dec. Dec. Dec. Dec. Dec. Deer counted on drive2 10 9 11 12 101 14 6 5 10 9 1 54 73 91 91 132 110 74 96 73 88 82 Deer at time of drive from aging method 78 97 119 125 153 108 84 99 79 84 66 Difference 24 24 2.8 34 21 - 2 10 3 6 - 4 -16 1Two drives were ' held, results for only the first are given here. zData from Jenkins (1964). Table 19 Average Composition of the Overwinter Deer Herd on the George Reserve Proportion of deer days Buck Adult fawns does Doe fawns Expected defecation rate Year Adult bucks 1952-53 1953-54 1954-55 1955-56 1957-58 1958-59 1959-60 1960-61 1961-62 1962-63 .3874 .2257 .2446 .1740 .2891 .1962 .2585 .2249 .2120 .2703 .1289 .1977 .3149 .2194 .2163 .3068 .2740 .2602 .2454 .1250 .2849 .3170 .3127 .3415 .3451 .2658 .3509 .3060 .3502 .3369 .1988 .2596 .1278 .2651 .1495 .2312 .1166 .2089 .1924 .2678 13.98 13.61 14.01 13.50 13.93 13.76 13.98 13.77 13.73 13.59 Average .2483 .2288 .3211 .2018 13.79 110 result if the extent and timing of mortality is not considered in con structing the average over-winter populations. Similarly, attempts to determine accurate fall or spring population levels from pellet group sur veys would be virtually impossible without similar knowledge. This fact, however, seems to have been overlooked by other workers and for several years these adjustments were unique to Michigan studies. Results Eberhardt and VanEtten (1956) reported the gross results of the first three surveys and the author (Ryel 1959a) discussed the first six in more detail. Summaries of the 1960, 1961, and 1962 surveys were fiven in Ryel 1960, 1961, and 1962 respectively. All of these papers, by necessity, compared survey results with deer population estimates based on drive counts, which, as we have seen, were quite inaccurate in some years. While pellet group survey estimates compare favorably with drivecount populations for four of the first six surveys (Ryel 1959a) they fit very poorly with the more accurate population data for these and subsequent years, Table 20. The relationship between known and estimated deer population estimates is, in fact inverse. The simple linear corre lation coefficient is r = -.62 which does not quite reach significance at the .05 level (.6319 for 8 degrees of freedom). In only three years do the estimated and known populations differ by less than 10 deer per square mile. In addition to revisions in the size of the known deer herd, some changes were also made in survey records. Approximately a quarter of the Reserve is in cover types which are commonly flooded or soggy in the spring - bogs, marsh, lowland shrubs, and tamarack swamp (Table 21). Table 20 Summary of Results of Deer Pellet Group Surveys on the George Reserve Year of survey Deposition period New groups found Number of 1/50 acre plots Estimated deer per sq. mile2 Known deer per sq. mile Estimated population compared to known Apparent deposition rate 1953 179 6141 277 28.7 ±20.7% 43.2 - 14.5 9.14 1954 182 558 278 25.6 ±23.7% 50.0 - 24.4 7.01 1955 original 181 343 278 15.8 ±23.3% 53.5 - 37.7 4.07 1955 recheck 200 32.5s 53.5 - 21.0 8.39 1956 162 806 241 47.9 ±21.8% 53.0 - 5.1 12.51 1958 167 677 279 33.5 ±22.2% 40.2 - 6.7 11.49 1959 165 602 277 30.5 ±21.0% 39.7 - 9.2 10.61 1960 163 1,001 280 50.4 ±23.0% 36.1 + 14.3 19.26 1961 161 1,115 284 55.5 ±21.6% 36.5 + 19.0 20.94 1962 144 842 284 49.8 ±22.4% 32.8 + 17.0 20.95 1963 153 1,430 287 75.1 ±15.2% 31.1 + 44.0 33.28 XA11 groups were counted in 1953. 2Limits of 2 standard errors of mean, approximately 95 per cent confidence limits. 3Based on simple expansion from the results of a recheck of 13 plots. Table 21 Cover Type Areas on the George Reserve Based on Pellet Group Surveys1 Upland grass Pine 158.5 390.6 8.2 435.6 1,146.2 21.1 124.4 365.0 8.5 424.1 1,145.9 37.4 53.4 87.3 419.4 8.7 434.8 1,146.0 155.6 19.5 30.5 130.5 385.1 0 412.1 1,146.0 31.3 117.4 27.5 31.6 117.4 405.5 4.2 411.1 1,146.0 1958 18.8 153.2 4.2 64.8 57.0 373.9 4.0 470.1 1,146.0 6-year average 25.1 110.6 34.2 37.1 112.6 390.4 4.2 431.8 1,146.0 80.2 68.8 45.8 527.1 298.0 1,146.0 455.0 396.5 1,146.0 Lowland shrubs Year Pond Marsh 1963 23.0 65.1 44.7 20.5 1962 35.1 96.4 71.3 1961 29.5 75.5 1960 12.7 1959 2 3 Tamarack swamp Bog 126.1 Y" 5.7 114.6 19.5 154. 7 Mixed hardwood Total j XA11 ,areas in acres. O 2From O'Roke and Hamerstrom (1948), but their data was converted to 1,146 acres. 3From Tody (1949), but his data converted to 1,146 acres. 113 Permanent ponds occupy another 2 per cent. During the first three sur veys, workers generally made no effort to count groups In such areas even though deer may have used them during the over-winter period. Uncounted plots were simply ignored and estimates computed using the results from those plots which were searched. Living organisms, however, are usually not distributed at random, thus, if deer frequented wet areas more or less than the average of the other sites, computations might be consider ably in error. Hence, in 1956 and 1958, crews were asked to make an extra effort to count wet plots. This strategy succeeded in reducing the proportion of uncounted plots about one-third by 1958, Table 22. Begin ning in 1959 we instructed crews to attempt counts in all areas except, of course, permanent ponds. On the last five surveys the number of un counted plots was negligible. For the purpose of computing estimates, I assigned values for the uncounted plots. recorded as zero. lized. In all years, plots falling in permanent ponds were For the first three surveys the same sample was uti Because of varying weather conditions, plots may have been counted 1, 2, or 3 years or not at all. If a wet plot had been counted at least 1 year, it provided a basis of estimating the missing year or years through linear regression methods. Plots not counted in any of the first 3 years and on all subsequent surveys (except ponds), were assigned the average of the "wet" plots for that year which were counted. Cover types were not routinely recorded until 1958; but on earlier surveys re corders usually stated the reasons why certain plots were not counted. The net results of these manipulations actually had little effect on calculated populations, the largest difference being less than two deer per square mile, Table 23. 114 Table 22 Proportion of Plots Counted During Deer Pellet Group Surveys on the George Reserve Plots counted Plots not counted Number Per cent Total plots 1953 239 38 13.7 277 1954 237 41 14.7 278 1955 241 37 13.3 278 1956 223 18 7.5 241 1958 254 25 9.0 279 1959 275 2 .7 277 1960 275 5 1.8 280 1961 284 0 0.0 284 1962 284 0 0.0 284 1963 283 4 1.4 287 Table 23 Effect of Estimating Groups on Uncounted Plots for Deer Pellet Group Surveys on the George Reserve Unadjusted Adjusted 1953 28.8 ±23.7%1 28.7 ±20.7% 1954 26.1 ±27.2% 25.6 ±23.7% 1955 15.0 ±28.1% 15.8 ±23.3% 1956 49.4 ±23.0% 47.9 ±21.8% 1958 35.4 ±22.6% 33.5 ±22.2% 1Deer per square mile with limits of ±2 standard errors of the mean, approximately 95 per cent confidence limits. 115 One of the more striking aspects of Table 20 is the rather progres sive Increase in apparent defecation rate. This is computed by setting the estimation equation equal to the known herd size and solving for the defecation rate necessary to produce it. Population estimates derived from the first six surveys were less than the known populations while the last four were higher. The simple linear correlation between year of the survey and apparent defecation rate, r = .87, is significantly different from 0 at the .01 level. Further study of survey data reveals a reverse progression in the proportion of pellet groups found which were classi fied as old, Table 24. The simple linear correlation coefficient between apparent defecation rate and the proportion of old groups for the 6 years when old groups were tallied is r = -.94 for 8 degrees of freedom, which is also highly significant (.01 level). Assuming an average of 60 per cent new groups greatly increases the accuracy of the 1960 to 1963 surveys and brings all six within 50 per cent of the known populations. This sug gests that the survey inaccuracies were caused by non-random errors, in particular, mistakes in aging groups. The studies described earlier noted that weathering of pellet groups varied with cover types, site conditions, and weather factors. I examined available weather data which might affect longevity and/or appearance of pellets. Multiple linear regression methods were used to provide an em pirical prediction model for the known populations and to assess the relationships among the variables. In other words, given the pellet sur vey results, can gross environmental factors be used to produce a linear equation which predicts the known population? The factors considered included: Y = known deer population/square mile Table 24 Observed Old and New Pellet Groups by Cover Types on the George Reserve Year Marsh Bog Lowland shrubs Tamarack swamp Upland grass Pine Mixed hardwoods Total Prop. new groups 42 0 42 10 0 10 48 3 51 195 2 197 376 39 415 48 2 50 711 15 726 1,430 61 1,491 95.9% 1962 New Old Total 46 3 49 1 0 1 107 2 109 142 0 142 279 67 346 6 0 6 261 11 272 842 83 925 91.0% 1961 New Old Total 75 1 76 78 0 78 17 3 20 18 2 20 771 118 889 1 0 1 155 21 176 1,115 145 1,260 88.5% 1960 New Old Total 104 4 108 23 1 24 16 0 16 89 1 90 548 170 718 221 34 255 1,001 210 1,211 82.7% 1959 New Old Total 58 20 78 13 3 16 82 72 154 212 67 279 13 0 13 217 29 246 602 191 793 75.9% 1958 New Old Total 36 0 36 1 0 1 14 0 14 198 161 359 0 0 0 409 26 435 677 187 864 78.4% 7 0 7 . 19 0 19 116 1963 New Old Total x^ = estimated deer population/square mile X 2 = degree days June to September (sums of negative departures of average daily temperatures from 65° F) = degree days November to March x^ = days 1 inch or more snow on ground x,. = days 2 inches or more snow on ground Xg = days with trace or more rain April to September x^ = days with measurable rain April to September Climatological data were obtained from the General Motors Proving Grounds near Milford (U. S. Department of Commerce 1952 to 1963). This, the closest regular Weather Bureau Station, is located about 18 miles north east of the Reserve in Livingston County. variables are shown in Table 25. •Intercorrelations of these None of the variables is significantly correlated with either the known or estimated deer populations, although, as noted before, they approach significance with each other at the .05 level. In addition, degree days June to September approach a signif icant correlation with the estimated deer population, r = .60 (.05 level = .63 for 8 degrees of freedom). Using stepwise regression procedures (Draper and Smith 1967), the "best" regression equation for predicting the known population level was: Y = -.2028X-L - .0083X3 - .6449X 7 + 139.9527 and the ANOVA table is as follows: source degrees of freedom sum of squares mean square F 4.65 due to regression 3 415.4159 138.4716 about regression 6 178.7941 29.7990 Total 9 594.2090 2 The calculated value of R is 415.4149 = .6991 which means this model 118 Table 25 Intercorrelations of Weather Factors, Results of Pellet Group Surveys and Deer Populations on the George Reserve1 X1 x2 1.00 x2 x3 x4 x5 .05 -.46 -.19 -.52 .46 .55 -.09 -.02 .81 .71 -.32 -.29 1.00 .95 .23 .20 1.00 .32 .24 1.00 .83 -.21 .60 .41 1.00 -.13 -.43 -.50 1.00 .34 1.00 x4 x5 x6 1.00 x7 ^ e e text for explanation of variables. P .05 ~ *63, P .01 = *76* x7 -.13 -.62 x3 x6 00 Y X1 • to Y Significance levels are: 119 accounts for about 70 per cent of the variation about the mean. The cal culated F is not quite significant at the .05 level, F(3, 6, .05) = 4.76. The above calculations simply provide a means of explaining the way the several factors are related mathematically. In no way do they pre tend to furnish an explanation of the observed phenomena. They have, however, partially satisfied my feeling that weather variations could have influenced pellet group appearance enough to cause survey crew mem bers to make mistakes in aging many of the groups found. The results of fitting the negative binomial distribution to the George Reserve data are shown in Table 60. Chi-square goodness-of-fit tests indicate a satisfactory fit in every year except 1961. Such com putations assume simple random samples, while in reality sampling was systematic and for some years stratified. However, following the reason ing of Bowden et al. (1969), I considered these data to be similar to counts which might have been obtained from randomly located plots. Habi tats were irregularly located and deer movements in relation to cover induces non-regular location of pellet groups. Locations of plots were pre-selected and independent of observer bias. The stratification used was to insure a more uniform area coverage and was not related to rela tive deer abundance. The negative binomial is a two parameter distribution - the mean (x) and a positive exponent (k). The value of k has been used by some ecol ogists as a measure of the "clumping" of populations. Relative clumping and values of k tend to vary inversely (Southwood 1966). _ of the negative binomial is equal to x + x2 . The variance Hence, if k was constant over the years, the mean and variance would depend only on the mean. On the Reserve, values for k ranged from .2758 to .7121 for the 10 years' 120 data. A x2 test for a single pooled k value, .4119, was made according to Bliss (1953) and was rejected, .005>p. This suggests that different patterns of deer distribution may have occurred over the years. Associ ation between ranks of k and degree days November through March and with the number of days with 1 Inch or more of snow on the ground from Novem ber through April (U. S. Department of Commerce 1952 to 1963) was meas ured with Kendall's coefficient of rank correlation (Campbell 1967). Both associations were positive and significant, P = .046 and .002
P>.005) while the other three surveys did not.
Values for k varied from .4361 to .6623 and were well within the range
obtained from the George Reserve surveys, even though the Cusino data
represents the sum of three (1953) or two (1954-56) 1/50 acre plots
rather than one.
A x2 test
for a single pooled value of k, .5119 (Bliss
1953), was not rejected at the .05 level (.10>P>.05), suggesting rather
more consistent over-winter deer concentrations than on the George Re
serve.
This seems logical in view of the white-tailed deer's behavior
patterns during the relatively severe winter weather here (Ozoga 1968).
DISCUSSION
The fifteen surveys reported here were carried out primarily to
field test the pellet group technique against contained deer herds of
known size.
The record, as noted above, was not outstanding, partic
ularly on the George Reserve.
Previous chapters have discussed several
potential sources of trouble in pellet group surveys, but there is no
completely satisfactory explanation for these discrepancies.
One hypothesis is that defecation rates of free-ranging deer are
126
really higher than the results of penned studies would Indicate.
Support
for this viewpoint is provided by the 1959 Houghton Lake study of deer
in a 6-acre enclosure described earlier, and from the fact that the "best"
surveys on both the Cusino enclosure and the George Reserve yielded over
estimates of the known populations.
The wide fluctuations in apparent
defecation rates in the George Reserve and Cusino studies (Tables 20 and
27), however, suggest we are not dealing with a consistent bias in this
component.
Still I cannot completely rule out faulty defecation rates
based on the available evidence and further research is indicated.
An alternative hypothesis would implicate aging and counting errors
as being the major causes of trouble.
Some evidence has been presented
for the George Reserve which intimates weather factors may influence
appearance of some pellets enough to confuse workers.
The 1955 surveys
supply additional support for interaction between counting pellet groups
and a widespread phenomenon such as weather.
All surveys conducted that
year resulted in serious underestimates, including the only underestimate
of the five Cusino surveys.
Results of recounts following the 1955
George Reserve survey have been given previously.
Preliminary inspection
of the northern Lower Peninsula survey data revealed counts seemingly so
low that analyses were never completed.
The somewhat better performance
of the Cusino enclosure surveys compared to those on the George Reserve
seems related to the better compaction of the leaf mat and herbaceous
vegetation by the deep accumulation of snow at Cusino, thereby facilitat
ing the finding and aging of pellet groups in the spring.
Furthermore,
the presence of large grassy openings and the abundance of oaks on the
Reserve added to the difficulties of finding and properly aging pellet
groups.
127
An additional problem introduced by these field surveys involves
chance sampling errors.
With pellet group surveys we are interested
chiefly in obtaining estimates of the mean with appropriate confidence
intervals.
Pellet count data have been treated as if they approximated
random samples from normal populations.
Confidence limits of 2 standard
errors of the mean (about 95 per cent) were computed to compare esti
mated with known populations.
As discussed in Chapters IV and V, compo
nents of variation due to determination of defecation rates and deposition
period are evidently small enough to be regarded as negligible and were
not included.
Later work has indicated pellet groups are far from normally
distributed and can be described rather well by several discrete conta
gious distributions, notably the negative binomial.
The rationale involved here is that most living organisms are not
distributed at random, although at times the sampling designs used may
not provide adequate data with which to show statistically significant
departures from randomness.
When natural populations are sampled by
quadrats, the resulting sample distribution is nearly always asymmetri
cal with an excess of samples with low values and a few samples with very
high numbers.
Of course, one cannot assume exact correspondence to theo
retical distributions based on samples taken from unknown populations.
The best we can conclude is that the field data can be represented
"adequately" by a theorietical distribution.
In the negative binomial, the smaller the value of k, the greater
the aggregation of the objects being counted.
Conversely, when k is
"large" (about 8 or more) the distribution approaches the Poisson and
hence is virtually random (Southwood 1966).
Although k may be Influenced
by the size of the sampling unit (Cole 1946), both Bowden et al.
(1969)
128
and McConnell and Smith (1970) found "small" k values similar to the
George Reserve and Cusino data using several different plot sizes for
counting mule deer pellet groups.
Thompson (1965) puts forth the idea that simple frequency of occur
rence (presence or absence) could be used to estimate pellet group den
sity, providing a value for k can be established.
Such a system elimin
ates the need for tedious complete enumeration on each plot.
McConnell
and Smith (1970), however, caution that frequency counts are much less
efficient than total counts.
Several times the number of plots may be
needed to produce estimates with the same precision.
Furthermore, they
report that serious biases may occur in such a system.
An obvious pro
blem would be the regular occurrence in our surveys of a few plots with
extreme counts which often have a considerable influence on the mean.
Because of extreme skewness and kurtosis for distributions with k