Ml ) “WIN ‘\ ( _—_‘_ ___———— ‘ Ml Ml ‘ [:3 WW EVALUATEQH QF THE F3 ENCEPAL FiHG:N§CKEB FHEAgAHT FGFLEL ATTGN ENQECES EN MECH {GAR Hunts Em» the Degree of M S 17 If! ”Mffiufl Igtrrvfzt‘ 9n” l‘iaCflLuéI‘ uafiif‘: ht? EiibJZLY Gaie Charies Jamsen i968 _ 1 w lfl$11u11\fl[ufl1lllm1flflflll4\|\ “WW L [:1 ‘ ABSTRACT EVALUATION OF THE PRINCIPAL RING-NECKED PHEASANT POPULATION INDICES IN MICHIGAN by Gale Charles Jamsen The objective of the study was to evaluate whether the principal indices of pheasant abundance used by the Michigan Department of Conservation were providing useful information. For the time period 1956 to 1965, the crowing-cock and rural mail carrier brood survey were related to mail survey kill estimates by linear regression equations. Since information on fall pheasant populations is of most concern to game managers, prediction equations were constructed to provide kill estimates from the crowing-cock and rural mail carrier brood survey results. As expected, the crowing-cock index was an imprecise tool in predicting the 1966 fall cock harvest. However, in four out of five study areas, predictions were possible with wide confidence limits. Approximately 20 per cent of the crowing-cock routes provided data suitable for use in the prediction equations. The rural mail carrier brood survey results were highly correlated with the legal harvest. Linear regression equa- tions predicted the 1966 kill with a high degree of accuracy. Gale Charles Jamsen Kill estimates for three of the five study areas deviated less than five per cent from the mail survey estimates. The rural mail carrier brood survey provides excellent informa— tion for making management decisions on regulation of the harvest. Farm area as a basis for computing kill densities was not as satisfactory as the total study area when kill densi- ties were correlated with the crowing-cock and brood indices. Mail survey results appear to be precisely estimating the legal harvest of cocks. Considerable eVidence exists that a constant proportion of cocks are being harvested in each study area. Thus, kill estimates are probably reflect- ing the magnitude of fall pheasant populations. This index is the most valuable because the results can be easily con- verted into density values. EVALUATION OF THE PRINCIPAL RING-NECKED PHEASANT POPULATION INDICES IN MICHIGAN BY Gale Charles Jamsen A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE Department of Fisheries and Wildlife 1968 t I") _ ‘ - I l 6. ACKNOWLEDGMENTS The cooperation from members of the Game and the Research and Development Divisions of the Michigan Department of Conservation has been most helpful in the formulation and execution of this study. Editing by F. W. Stuewer and L. A. Ryel was especially appreciated. Drs. W. E. Cooper, Department of Zoology, and G. A. Petrides, Department of Fisheries and Wildlife, deserve special thanks for their assistance in see- ing this study to completion. The efforts of Drs. L. W. Gysel and R. H. Baker, Department of Fisheries and Wildlife, also are appreciated. ii TABLE OF CONTENTS Page INTRODUCTION . . . . . . . . . . . . . . . . . . . . . 1 METHODS . . . . . . . . . . . . . . . . . . . . . . . 3 DISCUSSION AND RESULTS . . . . . . . . . . . . . . . . 7 Crowing-cock survey . . . . . . . . . . . . . . . 7 Relation of crowing to cock abundance. . . . 8 Causes of variation in crowing results . . . 9 Relation of crowing-cock index to kill estimates . . . . . . . . . . . . . . . 15 Rural mail carrier brood surVey . . . . . . . . . 25 Mail survey kill estimates. . . . . . . . . . . . 50 Advantages over previous surveys . . . . . . 50 Relation of fall populations to kill estimates . . . . . . . . . . . . . . . 50 CONCLUSIONS . . . . . . . . . . . . . . . . . . . . . 59 SUMMARY . . . . . . . . . . . . . . . . . . . . . . . 42 LITERATURE CITED . . . . . . . . . . . . . . . . . . . 44 APPENDIX . . . . . . . . . . . . . . . . . . . . . . . 46 iii LIST OF TABLES TABLE Page 1. Characteristics of study areas and surveys in the primary pheasant range of Michigan. . . . . . 5 2. Analysis of variance for counts by experienced participants in a crowing-cock survey (random model), Sanilac County, Michigan, June 1, 1961. . 10 5. Single station counts of cock pheasant calls com- paring the 40-minute period before sunrise with the 40-minute period after sunrise, Ingham County, Michigan. . . . . . . . . . . . . . . . . 12 4. Effect of daily timing of the crowing-cock survey on the results, Michigan primary pheasant range . 12 5. Pheasant sex ratios from Conservation Department late winter roadside surveys, Michigan primary pheasant range. . . . . . . . . . . . . . . . . . 16 6. Pheasant sex and age ratios and mortality data, Rose Lake Wildlife Research Station, East Lansing, Michigan, 1956-1965 . . . . . . . . . . . . . . . 18 7. Correlations of total crows per crowing-cock route (X1) and the respective county (X2) and area (X3) kill per square mile, Michigan primary pheasant range, 1956-1965 . . . . . . . . . . . . 20 8. Pheasant kill estimates in 1966 with 95 per cent confidence limits, Michigan primary pheasant range . . . . . . . . . . . . . . . . . . . . . . 21 9. Per cent deviation of preseason estimates of 1966 pheasant kill from mail survey estimates, Michigan primary pheasant range . . . . . . . . . 22 10. Correlations of kill per square mile of pheasant range (X1) with broods per 1000 miles (X2) and broods per ten carrier days (X3), Michigan primary pheasant range, 1956-1965 . . . . . . . . 25 iv LIST OF TABLES - Continued TABLE 11. 12. 15. 14. 15. 16. 17. 18. 19. Covariance analysis for homogeneity of regres— sions concerning broods observed per 1000 miles and population level as cocks killed per square mile of study area, Michigan primary pheasant range, 1956-1965. . . . . . . . . . . . . . . . . Correlations of total crows per crowing-cock route (x1) and kill per square mile of pheasant habitat as total area (X2) and farmland (X3). Michigan primary pheasant range, 1956-1965. . . . Correlations of pheasant broods observed per 1000 miles (X1) and kill per square mile of pheasant habitat as total area (X2) and farmland (X3), Michigan primary pheasant range, 1956-1965. . . . Michigan pheasant hunting results from mail sur- Vey estimates . . . . . . . . . . . . . . . . . . Per cent of pheasant cocks harvested. Based on a postseason survey by rural mail carriers in Michigan's primary pheasant range . . . . . . . . Male pheasant age ratios derived from wings and feet collected by cooperating Michigan hunters. . Pheasant hunter numbers in Michigan's primary pheasant range, 1956-1965 . . . . . . . . . . . . Regression equations for predicting area kill A (hundreds of cock pheasants) per square mile (Y) from total of calls per crowing-cock route (X). Michigan primary pheasant range, 1956-1965. . . . Regression equations for predicting area kill (hundreds of cock pheasants) per square mile (Y) from broods observed per 1000 miles (X) of travel by rural mail carriers, Michigan primary pheasant range, 1956—1965. . . . . . . . . . . . . . . . . Page 27 29 29 52 54 55 58 47 48 FIGURE 1. 2. LIST OF FIGURES Page >Study areas in Michigan's primary pheasant range. . . . . . . . . . . . . . . . . . . . . 4 Typical relationship between kill densities and crowing-cock and rural mail carrier brood survey results, Area V, Michigan primary pheasant range, 1956-1965. . . . . . . . . . . 49 vi INTRODUCTION The ring-necked pheasant was introduced into Michigan in 1895 with the first season opening in 1925. Since that time, the pheasant has become the state's major game bird. The harvest peaked in 1944 when an estimated 1,401,076 cock pheasants were legally taken. The average annual bag for the state over the past 50 years has been just under one million cocks. During the period of this study 1956- 1965, the pheasant season opened on October 20th and termi- nated on November 10th. The bag limits have been two cocks per day and eight for the season. Management of upland game species such as the ring- necked pheasant requires periodic pOpulation assessments during key periods in the life cycle. The Michigan Depart- ment of Conservation does this by initiating numerous sur- veys throughout the Michigan pheasant range. The three most important surveys are conducted during the Spring breeding season, in mid-summer when nearly all broods have been hatched, and after the hunting season for estimates of the harvest. In all three surveys only a segment of the total pheasant population is sampled. In the spring an index to breeding male abundance is obtained from the crowing-cock survey. This survey is essentially the same as that developed by Kimball (1949). The mechanics of this survey entail the travel during early morning of twenty mile routes on rural roads. Male pheasant calls or crows are counted in a tw0dminute period at one mile intervals. -The summer survey enumerates broods. This results in an index that reflects the success of the breeding season. Approximately 600 rural mail carriers in Michigan's pheasant range cooperate with the Conservation Department by counting pheasant broods along their routes. Usually this is done during a two week period in early August. The third major index of the pheasant population is the kill estimates obtained from a mail survey of approximately one per cent of Michigan's small game hunters. This sample survey is conducted early in the year following the hunting season (October 20th-November 10th). The objectives of this thesis are to evaluate whether the indices are providing useful information concerning the pheasant population and whether the assumptions necessary for precision in each index are being satisfied. METHODS ‘ Data for this study were supplied by the Game and the Research and Development Divisions of the Michigan Department of Conservation. Original report forms completed by partici- pants in the annual crowing-cock survey as well as Game Division reports summarizing this survey were made available. In addition, experimental data which had only been summarized were made available for analysis so various assumptions of the crowing-cock survey could be tested. The data from the rural mail carrier brood survey were primarily derived from Game Division Reports. The mail survey sample of small game hunters provided pheasant hunter numbers and pheasants killed in each county plus information on success and effort. Most of these data are from Research and Development Division reports and annual survey records. The geographical areas (Fig. 1) used in this study are the same as those used by MacMullan (1960). They comprise approximately 72 per cent of Michigan's primary pheasant range (Table 1). These areas were selected primarily for convenience in analyzing a large amount of data. In many cases the data collected were only available on a geographical study unit basis since the original records were destroyed. Also, the study areas permitted contrasts between areas with .owamu ucwmmmam mumEHum w.swmwnofiz :H among Suzum .H gunman uuym<§md ' ,a wmn<:<.:4 zo»z_4u a. m , . l . 24.¢ow mmm dfi mm mam.wm mmcmm mumEanm mam.mam mmd mm mm mom.na Hmuoa www.mm ww N m >mm.a Aumwzv > mampom oma m m wam.¢ Aummznuzomv >H www.mm OOH m m m>m.m Aamuucmuv HHH www.mm om m m oma.m AummmSDDOmv HH 0mm.m> how. u w ham.¢ Anesnav H mnmwuumo mumfluumo mmusou mmaucsoo mmHHE mmu¢ Hana >9 HHME mo maa3ouu mo mumswm mcm>HnU mmaflz mumnesz mo Hmnfisz HmQEDZ .smmflnoflz mo mmcmu ucmmmmnm wumfiflum ms» CH m>m>nsm Ucm mmmum wpsum mo moflumfiumuumnmzo .H magma relatively high densities and those with medium and low densities. Reasonably complete records in the ten year period, 1956-1965, were available for analysis. About 70 per cent of the 56 crowing-cock routes were run continuously during the entire period and were suitable for analysis. In the following analyses the mail survey kill esti- mates serve as the standard for testing the other two indices. Pheasant populations available to the hunter are of the most interest and importance to the managers of this game Species. Further, it is assumed the mail survey re- sults are unbiased estimates of a constant proportion of the fall cock population. The indices of male breeders and broods produced were then analyzed to determine their relative value in predicting the size of this fall pOpu- lation. The three indices will be evaluated in the order they appear chronologically in the pheasant life cycle. A dis- cussion of the assumptions underlying the results of each survey will be followed by remarks on how well it is meeting its objectives. DISCUSSION AND RESULTS Growing-cock survey The crowing-cock survey has been widely used as an index to breeding pheasant males. However, this segment of the spring breeding population is of little importance due to the polygamous nature of the species. An index of breed- ing pheasant hens would be more desirable as a clue to potential productivity for a given year. Accurate estimates of sex ratios would provide an adjustment factor for trans- forming the crowing-cock index into an index of breeding hens. Since accurate spring sex ratios are difficult to obtain (Stokes,_1954:82-88), in many cases the crowing-cock index becomes the breeding index. If the sex ratio is relatively constant over the portion of the pheasant range that is of interest, the crowing-cock index might suffice as an indicator of spring breeder abundance. An even better use of the crowing-cock index would be as a predictor of the fall population through its correlation with the legal harvest. The chain of events required for this to happen appear quite remote, since adult cocks usually compose only 10 per cent or less of the fall harvest. For the spring crowing-cock index to serve as a good predictor of the fall population, certain assumptions must be met. They are: 1. Growing is directly related to cock population size. 2. Crowing-cock survey results produce unbiased in- dices of breeding cock pOpulations. 5. The ratio of hens to cocks in the spring pheasant population is constant. 4. A constant percentage of hens hatch broods of a constant size. 5. Spring-to-fall adult and juvenile mortality remains fairly static over the years. 6. Hunting kill is a constant percentage of the fall cock population and the kill estimates have rather narrow confidence limits. The only assumptions where sufficient data are available for analysis are the first three and the last. The last one will be considered under the discussion of kill estimates; the intermediate assumptions will be tested by correlating the crowing index and the legal harvest. A discussion of the first three assumptions follows. Relation of growing to cock abundance -- Research in Illinois in 1947 and 1949 (Robertson, 1958:59) suggests that there is a linear relation between the number of calls per two-minute listening period and known cock pOpulations. Gates (1966) also found a linear relation (r=0.96) between his crowing intensity index and known cock populations in Wisconsin. After carefully examining the data of Gates (1966), the corre- lation of calls per two-minute period and the square of the known cock pOpulation appeared to be the more meaningful com- parison. This yielded a correlation coefficient (r) of 0.99. A correlation this high is Open to suSpicion as being too good. More information is needed before the nature of the relation is adequately described. However, it does lend sup- port to the logical assumption that crowing and cock numbers are related. Causes of‘variation in crowing results -- Kimball (1949) who developed the crowing-cock survey states, "The accuracy of the crowing count and the success with which it can be used are dependent largely upon the following factors: 1. Variation in ability of the individuals conducting the survey to hear cock calls. 2. Daily trend and duration of maximum cock crowing. 5. Seasonal trend and duration of maximum cock crowing. 4. Uniformity of results. 5. Effect of variable factors, such as weather and cover, upon the count." Carney and Petrides (1957) found close agreement among experienced participants in a crowing-cock survey. Partici- pants without previous experience failed to agree with each other or with experienced participants. The Michigan Depart- ment of Conservation in 1961 conducted an experimental crowing- cock survey in Sanilac County with experienced biologists. The data were only superficially examined at that time.‘ The replicated counts suggested a tw0dway analysis of variance as being the best means of testing the null hypothesis that there were no differences between individuals in their ability to count crowing pheasants (Table 2). These experienced 10 Table 2. Analysis of variance for counts by experienced participants in a crowing-cock survey (random model), Sanilac County, Michigan, June 1, 1961. Degrees of Sum of Mean Source freedom squares square F Participants 6 521.000 55.500 1.681 Stations 5 1524.857 441.619 Interaction 18 572.810 51.825 5.808* Error 56 468.000 8.557 Total 85 2686.667 *Significant (P < 0.05) workers, evidently, also do not differ significantly in their ability to record calls. Since Michigan has primarily used experienced personnel during the time period of this study, changes in participants have probably contributed little to the variation in results. There is little disagreement that cocks begin crowing at a high level from 40 to 50 minutes before sunrise and continue crowing intensely until some time after sunrise. Kimball (1949) suggested that the daily intensity of crowing was fairly constant from 40 minutes before sunrise to 50 minutes after sunrise. Gates (1966) reviewed this matter, however, and demonstrated that in Iowa and South Dakota crow- ing reached a peak shortly after counting began and declined rapidly within one hour thereafter. In Wisconsin he found that the decline was moderate and that low-density stations 11 had a significantly greater average rate of decrease than did high-density stations. Kozicky (1952) used a circular crowing route of ten miles where stops one and two also served as nine and ten, reSpectively. His comparison of the same stations at different times also yielded a significant difference. Analysis (Table 5) by the "t“ test of two-minute counts collected in 1961 and 1962 from Ingham County indicated that crowing declined significantly (P < 0.05) in three out of four cases when 40-minute periods before and after sunrise were compared. To learn how this bias might affect field data, random selection and analysis (Table 4) of five crowing routes was accomplished for both 1961 and 1965 to discover if the daily time period of the survey influenced the results. These routes had been cruised forward on one day and in rer verse order a day or so later. Two out of five crowing routes each year had mean counts per stop that were significantly different from'each other. The data were tested by the Wilcoxon matched-pairs signed-ranks test (Siegel, 1956:75-85). Two routes, close to being significant, were tested further by the more powerful paired "t" test and were then found to be significantly different. The results of this analysis are not conclusive. The effect of day—to-day variation may be accounting for the differences. This effect could be elimi- nated by performing a similar trial with equally skilled participants on the same day. 12 Table 5. Single station counts of cock pheasant calls compar- ing the 40-minute period before sunrise with the 40-minute period after sunrise, Ingham County, Michigan. _ _‘- i — ‘ - Mean number of cock-calls per two-minute period Degrees of Date Before sunrise After sunrise freedom t 5-4-61 12.2 10.0 16 1.55 5-15-61 20.6 11.8 17 6.57* 5-21-62** 7.9 5.4 -- -- *Significant (p < 0.05) **The Mann‘Whitney U test (n1=8, n2=9, U=1.5) was used because the variances of the samples were not homogeneous [P(U=1.5) < 0.05]. Table 4. Effect of daily timing of the crowing-cock survey on the results, Michigan primary pheasant range. _ 1 n L Mean number of calls per stog Crowing-cock route Forward ”Backward 1961 Clinton* 5.55 5.55 Eaton 15.85 14.45 Huron W * 9.50 10.15 Lenawee NE 10.50 14.15 Washtenaw 5.40 7.20 1965 Clinton* 9.00 8.80 Hillsdale 5.70 4.15 St. Clair N 14.80 -15.75 Lenawee W* 15.25 7.40 Tuscola NW 5.75 6.55 * Mean values significantly different (P < 0.05). 15 In Wisconsin, seasonal peaks of crowing occurred approxi- mately the same time (April 25-May 15) during 1959 to 1964 (Gates, 1966). Differences in spring weather failed to in- fluence the peak period. Michigan's peak period in 1951 and 1952 occurred during the month of May (Blouch, 1952). The crowing-cock survey has been conducted during this month since 1952 and the Wisconsin results lend support to the validity of this practice. Gates (1966) suggests that the chief problem in using the crowing count as a population index is the day-to-day variability of the results. Counts varying from 7.5 to 20 per cent of the mean have been reported. Under restrictive weather conditions, Gates (1966) obtained results that varied less than 10 per cent from the mean at the peak of crowing. No similar data on this tOpic exists for Michigan. A way of testing whether the variability from day-to-day is signifi- cant would be by means of a two-Way analysis of variance. Replicated counts at each station of a segment of a crowing route could be recorded over a number of days. Comparisons could be made between days and if careful weather data are collected, their effect on the results could be analyzed. Kimball (1949) dismisses cover and weather as factors influencing the counting of calls when restrictions on wind velocity are met. Wind velocity and cloud cover were vari- ables considered especially important by Gates (1966). He suggested that crowing counts be made only on calm and clear mornings. 14 Other factors such as the effect of the Spring sex ratio on crowing behavior, the relation of crowing intensity to population density, the adequacy of the two-minute counting period, and the prOportion of non-crowing cocks could be sources of variation. From Gates (1966) one can deduce that none of these factors has a significant influence on most crowing tallies. He found that crowing counts in areas of high pheasant density could lead to pOpulation overestimates because of mutual stimulation. Assessing this effect outside of an intensive study area would be extremely difficult. The two-minute count is long enough for some cocks to be recorded twice, but from evidence collected by Nelson §£“_l. (1962) and Gates (1966) it should only be important at densi- ties higher than those found in Michigan. -Burger (1966) discovered from 5 to 8 minute counts that 71 to 92 per cent of the cocks he detected would have been recorded during a two-minute count. For this study it is assumed that the percentage of cocks calling in a two-minute period is constant. In summary, one can see that the accuracy of the final result of a crowing-cock survey is affected by many sources of variation. The most serious appears to be day-to-day variation due to the fact that the routes are run only once. Assuming all the sources of variation could be eliminated or accurately measured, one would still have an index that would be difficult to interpret for there is no standard or gauge that it can be measured against. Unfortunately, most of 15 the sources of variation cannot be easily measured. There- fore, the data analyzed here are assumed to be providing an unbiased estimate of cock abundance for each study area. Its acceptance or rejection as useful information occurs when it is judged against some acceptable standard established at an earlier or later time. Relation of crowing-cock index to kill estimates -- Since the pheasant is a polygamous species, the Spring sex ratio should be constant over the years for each study area or it should be precisely determined so the crowing-cock index can be adjusted to represent the Spring breeding population of females. Smith and Gallizioli (1965) discovered a linear relationship between spring call counts and hunting success of the monogomous Gambel quail. This permitted them to make good predictions of hunting success by means of linear regression equations for three study areas in Arizona. Sex ratios of breeding pheasants are difficult to obtain due to their behavior. As cocks come into breeding condition in March and April, their conSpicuous behavior strongly biases sex ratios determined by roadside counts (Wagner §t_gl,, 1965: 27). Thus, late winter sex ratios (Table 5) have been examined as the best substitute for Spring sex ratios. Roadside sur- veys by Conservation Department personnel during February and March provide this data. Stokes (1954:82-88) in his Pelee Island study clearly points out the difficulties in ob- taining meaningful sex ratios. Factors that bias the results 16 .AUHmE u 20 .mnmsnnmm n ma .mumHmEoocH no mHQMHHm>mcD mmmfi on HOHHQ mumnm m.ma $.m a.o m.o o.m n.0N m.N N.ma >.H >.m >.m m.m m.m 0.0a > >.a m.fi o.H N.o O.H m.a m.H m.fi m.o o.fi o.a «.fi m.m N.N >H m.m N.H o.H m.o N.¢ >.m m.¢ 0.4 ¢.H m.m o.m w.¢ m.N h.¢ HHH m.a N.w H.N m.m o.m >.m m.¢ N.m m.¢ m.m H.> m.w m.n m.> HH $.w m.m >.N m.m o.¢ m.m ¢.m m.w m.¢ N.¢ m.m m.m o.Na.m.mH H z m z m S m S. m z m z m S m. mwud .IIIIIIII llllllll IIIIIIII. llllllll pllllLHI mmmfi «mmfi mmmfi Noma Hmma ommfi mmmmd Axuoo “ma mamm%\0Humu xmm .mmsmn Dammmwsm mumfiflum cmmflnoflz .m>m>usm mpflmpmou HmDSAB mumH ucwfiuummmn SOHum>ummsoo_Eoum moaumn xmm ugmmmmnm .m magma 17 were investigated. They are method of observation, weather (snow cover), and the time of the winter when counts are made. Stokes' weekly estimates of the sex ratio of a given popu- lation were highly variable. The Rose Lake Wildlife Research Station in Central Michigan is another location where spring sex ratios have been carefully determined. The use of these sex ratios to estimate productivity available to the hunter yields values (Table 6) for 1961 and 1962 that appear to exceed biological potential (Stokes, 1954:87). With this background, the data (Table 5) from the late winter survey were analyzed by short-cut multiple comparisons (Kurtz 33 31., 1965). The February and March data were treated separately. Differences between years for either month could not be detected. However, Area V had a signifi- cantly different mean sex ratio (P < 0.05) than Areas III and IV in February. In March, Areas I and V were significantly different from Area IV (P < 0.05). Next the February and March data were compared by means of the Wilcoxon matched- pairs signed-ranks test (Siegel, 1956:75-85). It was believed the February results would yield sex ratios with a higher proportion of hens due to the effect of snow cover (MacMullan, 1960:129-145). This was the case, P(z=1.70)=0.04. .It is not my intent to select one month or the other as the more repre- sentative of the "true“ Spring sex ratio. ‘Rather, it is to Show that there is a fair amount of variability in sex ratio determinations from roadside surveys. This is eSpecially true 18 .Anmuemmav wmxoum mo ponumE wsu >9 Umusmfioo .muuommn mumucsn Eoum Umusmfioom .mHQMHHm>MSD mumnu Q ma Hudw.m mm SN.H OO.H m.¢ cmmE UmemHmBSD Hm «umm.m am om.H mm. N.¢ mmmd ON II Sm II mm. 0 ¢mmfi. ma Humo.m 5m NN.a NH.a o.m mmmfl ma Hubm.w mm ed.d ao.a m.m Nmmfi Ha Humm.w mm ma.a mo.a m.¢ Hmmfi mm Hum>.H mm m¢.d OH.H >.> Coma ma Humm.d mm m¢.a mm. H.> mmma Hm HuOd.m mm HN.H mm. N.¢ mmma ma aumm.m Hm om.a oa.H m.m wmmfi, ma dumo.m mm mm.a mm. N.¢ mmma Ammn Hmuou Ammamamm Haax may DwmumEHumm mpm>ummno mcflumm Hmmw. mo ucmu ummv Danna "masow SH mmaflcm>zfl qcflggsn mo >mp umufim mmoH mcflammfluu. mo ucmo mom Axoou Hod mgmnv oaumm xmm .mmmdlmmma .smmHSUHE .mSHmcmq pmmm .coaumum zunmmmmm.mMHHUHfl3 mxmq wmom .mump muflamuuoa paw moaumu mmm paw xmm ucmmmmnm .m magma 19 in Area V where very small samples of birds are observed in some years. From this analysis it was concluded that an assumption of a constant sex ratio for each study area from 1956 to 1965 would be reasonable and the crowing-cock index could be directly correlated with the kill. First, each crowing route was correlated with its respective county kill (Table 7). Biases in the kill estimates at the county level could be severe due to small sample Size and an urban inflation factor (MacMullan, 1960:58). .Approximately 25 per cent of the crow- ing route indices were Significantly correlated (P < 0.05) 'with their county kill. Then the crowing route indices were correlated with the study area kill figures (Table 7) and similar results occurred. Linear regression equations were prepared using the crowing-cock indices and area kills that were linearly re- lated. Crowing-cock index values for 1966 were utilized in these equations and the predictions compared with mail survey estimates (Tables 8 and 9). In many of the cases the mail survey estimates were outside of the very wide confidence intervals of the regression estimates. Thus, they were not especially useful. Area I was difficult to interpret. A rapidly declining pOpulation in this area required predic- tions to be made from beyond the range of the data. Crowing- cock index values for 1966 and 1967 suggest that a logarithmic transformation of the crowing index might improve its relation 20 Table 7. Correlations of total crows per crowing-cock route (X1) and their respective county (X2) and area (X3) kill per Square mile, Michigan primary pheasant range, 1956-1965. Crowing route rX1X2 rxlxa Area I Bay 0.88* 0.89* Huron W 0.59 0.80* Huron N 0.71* 0.77* Sanilac E 0.65 0.65 Sanilac W 0.11 0.19 St. Clair N 0.49 0.55 Tuscola NW 0.68* 0.66 Area II Lenawee SE —0.15 0.54 Lenawee W 0.09 -0.20 Lenawee NE 0.54 0.55 Monroe N 0.01 -0.05 Monroe S -0.58 0.21 Area III Clinton 0.26 0.58 Eaton 0.55 0.55 Ingham 0.55 0.69* Ionia 0.69* 0.70* Shiawassee 0.25 0.21 Area IV Barry 0.55 0.44 Berrien 0.55 0.25 Branch 0.17 0.24 Calhoun 0.49 0.20 Cass 0.88* 0.67* Kalamazoo —0.29 0.15 .Area V Ottawa 0.89* 0.95* Allegan 0.58 0.45 *Significant (P < 0.05) 21 Table 8. Pheasant kill estimates in 1966 with 95% confidence limits, Michigan primary pheasant range. Brood survey Crowing-cock Broods per Broods per 10 Mail Area survey 1000 miles carrier days survey I 190,400:20%a 87,500f67% 72,100:118% 89,250 175,600t58%b 151,900:49%C II d 176,60019% 181,40016% 142,080 III 80,100127%: 68,900t25% 67,900:25% 67,760 88,700t18% IV 66,100i19%g 61,200t15% 65,700:14% 65,840 v 81,700f5%h 68,500f10% 72,000110% 74,850 a Bay County crowing route. U‘ Huron County North crowing route. 0 Huron County West crowing route. 9.. Crowing-cock survey results not significantly correlated with the kill. eIngham County crowing route. ronia County crowing route. 9Cass County crowing route. hOttawa County crowing route. 22 Table 9. Per cent deviation of preseason estimates of 1966 pheasant kill from mail survey estimates, Michigan primary pheasant range. W Brood survey, Crowing-cock Broods per Broods per 10 Area survey 1000 miles carrier days + 94.5% + 70.2%C II d +19.5% +21.7% III +18.2%: +1.7% +0.2% +50.9% IV +5.5%9 —4.1% -O.2% v +9.27%h -9.6% -5.8% aBay County crowing route. bHuron County North crowing route. cHuron County West crowing route. dCrowing-cock survey results not significantly correlated with the kill. eIngham County crowing route. ronia County crowing route. 9Cass County crowing route. hOttawa County crowing route. 25 with the kill density and be of some utility for making future predictions. The crowing index and area kill in study area II have very little relation to each other. A relatively stable kill for the period is matched by fluctuating cock pOpulations. Cocks appear to be underharvested compared to Areas I and V. This is probably explained by the fact that much of the land in this area is unavailable to hunters because it includes large metropolitan areas (e.g. Detroit) and landowners are reluctant to permit hunters to use their land (Zorb, 1959). Klonglan and Kozicky (1955) found that crowing routes in areas of low pheasant population density produced data that was much more variable than in areas of high densities. This may be an explanation for the narrow confidence limits in 1966 kill predictions for the high pheasant density range (Area V) as opposed to the fairly wide confidence limits found in the low to medium pheasant density range represented by Areas III and IV (Table 8). Rural mail carrier brood survey The rural mail carrier brood survey usually is carried out during the first two weeks of August. Cooperating mail- men, around 600, travel approximately 400,000 miles throughout Michigan's primary pheasant range and enumerate pheasant broods, hens, and cocks. To evaluate this survey, broods were correlated with the hunting kill. MacMullan (1960) demonstrated that a high correlation existed between these two quantities. He also analyzed the factors that contributed 24 to this situation prior to 1957. In investigating the rela- tion during the time period covered in this study, kill figures on a density basis were used. Brood survey results were converted to broods per unit of distance traveled and per unit of time to test which set of data provides the best kill predictions. Brood data for the analysis in this study were utilized as broods per thousand miles since this implies a sampling unit in terms of area. The Conservation Department presents its data as broods per ten carrier days in the belief that a unit of time in the field as a sample will yield as good an index as a unit of distance traveled (MacMullan, 1960). This point was tested by regressing kill per square mile of each study area upon the respective number of broods per thousand miles and ten carrier days. Two sets of regression equations were prepared. Data from the 1966 brood survey were introduced into the equations and the resulting kill estimates were compared with the 1966 mail survey results (Tables 8 and 9). Broods per ten carrier days yield closer estimates to the mail survey in Areas III to V, but for practical purposes the estimates provided by broods per thousand miles as the independent variable are satisfactory (Table 10). In Area I the 1966 brood index value was con- siderably lower than the range of the previous ten years. This causes one to be very hesitant in attempting to place much importance in the estimates, although they are 25 surprisingly close to the mail survey estimate. In addition, it Should be noted that a recent decline in pheasant numbers over the primary pheasant range requires that 1966 estimates be made from slightly beyond the range of data for Area II and III when using broods per 1000 miles as a base and in Area III when using broods per ten carrier days. Table 10. Correlations of kill per square mile of pheasant range (X1) with broods per 1000 miles (X2) and broods per ten carrier days (X3), Michigan primary pheasant range, 1956-1965. = Area rxixa rXixs I 0.95*“ 0.88* II 0.75* 0.80* III 0.88* 0.88* IV 0.68* 0.67* V 0.90* 0.87* *Significant (P < 0.05). At this time it is appropriate to look into the reasons why such a good linear relation exists between the brood survey results and the kill density. It appears the most important factors contributing to an accurate brood index are large sample size, a random sample of broods appearing on highways, and the timing of the survey. With nearly 600 mailmen participating annually, the primary pheasant range is intensively surveyed. Also, although the time of day 26 (9:00 a.m. to 5:00 p.m.) is far from ideal for observing broods it is constant year after year. To optimize the survey to the time period when adult pheasants are most active would require beginning the survey at sunrise, which is ob- viously impractical. Dalke (1957) observed that adult pheasants are most active during the hour following sunrise and little feeding occurred during the middle of the day in either summer or winter. Experience through participation in the survey for many years also adds to its reliability. Annual turnover among rural mail carriers is only about five per cent. Finally, the weather and crop phenology are moni- tored via climatological reports and crop weather bulletins. This with the aid of hatching curves and the judgment of the pheasant specialist insure that the brood survey will be conducted each year at a time when a maximum number of broods are available for observation. To learn if changes in the brood index caused uniform changes in the kill density throughout the range, Slopes and levels of the individual area regressions were tested (Freese, 1964). The slopes were not significantly different but the levels were different (Table 11). Thus, the use of one re- gression equation utilizing all the data would probably err in predicting an average kill density for all the study areas. While working on this phase of the study and observing the difference in the precision of the kill estimates for the different study areas, the idea that the total land area 27 Table 11. Covariance analysis for homogeneity of regression concerning broods observed per 1000 miles and population level as cocks killed per square mile of study area, Michigan primary pheasant range, 1956-1965. fl:— —_ -> L Degrees of Sum of Mean Source freedom squares square F Area I 8 547.022 Area II 8 151.088 Area III 8 121.916 Area IV 8 52.481 Area V 8 129.627 Total 40 782.154 19.555 Difference 4 110.499 27.625 1.415 (Slopes) Pooled regres- 44 892.655 20.287 sion Difference 4 5768.976 942.244 46.446* (levels) Single regression 48 4661.609 * Significant (P < 0.01). 28 was not suitable as pheasant habitat suggested itself. Since the pheasant is closely associated with farmland and has a diet composed mostly of corn and grains, it seems logical that farm area (U. S. Bureau of the Census, 1959 and 1964) rather than the total study area would be more suitable as the basis for computing pheasant denisities. This hypothesis was tested by correlating crowing-cock results and the kill per square mile of farmland. .Correlations equal to or slightly lower than those obtained from the total area kill densities were computed except for Area IV where the corre— lation decreased to where it was not longer significant (Table 12). In testing the hypothesis by correlating broods per thousand miles with kill per square mile of farmland, only in Areas I and III were correlations as good as the correlations of broods per thousand miles and kill per square mile of study area (Table 15). It is not clear why farm area does not serve as well as the total area in defining pheasant habitat. The good correlation between the brood index and kill densities for the study areas serve to strengthen the conten- tion by MacMullan (1960) that: 1. Late summer mortality and brood size are relatively constant from year to year. 2. The brood survey has been appropriately timed. 5. The precision and accuracy of the brood survey and computed kill are quite high. 29 Table 12. Correlations of total crows per crowing-cock route (X1) and kill per square mile of pheasant habitat as total area (X2) and farmland (X3), Michigan primary pheasant range, 1956-1965. p I Area Crowing route FX1X2 rxlxs I Bay 0.89* 0.88* I Huron W 0.80* 0.80* I Huron N 0.77* 0.77* III Ingham 0.69* 0.67* III Ionia 0.70* 0.69* IV ‘ Cass 0.67* 0.56 V Ottawa 0.95* 0.92* * Significant (P < 0.05). Table 15. Correlations of pheasant broods observed per 1000 miles (X1) and kill per square mile of pheasant habitat as total area (X2) and farmland (X3). Michigan primary pheasant range, 1956-1965. M II II Area rX1X2 rxlxa I 0.95* 0.95* II 0.75* 0.48' III 0.88* 0.87* IV 0.68* 0.54 V 0.90* 0.80* * Significant (P < 0.05). 50 Mail survey kill estimates Advantages over previous surveys -- The mail survey of approximately a one per cent sample of Michigan small game hunters was initiated in 1955 to obtain a kill estimate for pheasants. Prior to that time, kill estimates were depend- ent upon an unenforced "compulsory" return (totaling only 10 per cent in 1950) of report forms provided with the hunting license. The sampling system described by Blouch (1956) and Eberhardt and Murray (1960) has the advantage of being representative of the hunting population. .A nearly complete response, about 90 per cent, from a systematic sample with a random start avoids the possibility that unsuccessful hunters may be less inclined to report as may have been the case in the previous system. Also, all areas of the state appear in the sample. In the first sample survey it was found that neither failure to hunt nor poor success was a primary reason for neglecting to reply (Blouch, 1956). Relation ofifallgpopulations to kill estimates -- The mail survey kill estimates are assumed to be unbiased esti- mates of the actual kill for the reasons cited above. Biases which are not practical to detect in a post season sampling method, however, could influence the estimate. They include: 1. Memory bias in which the hunter does not remember what he Shot. 2. Prestige bias in which a hunter consciously or unconsciously exaggerates his kill. 51 5. Party bias in which two hunters report shooting the same game. No tests have been made and no proof exists that these biases affected kill estimates in Michigan. -Confidence limits for the kill estimates were computed (Cochran, 1965:26-27) for the 1956 to 1965 time period (Table 14). They are sur- prisingly uniform. The only thorough test of this sampling procedure is to run two random samples of the same Size under the same conditions. In 1962 two sample surveys of approximately the same size were conducted for the purpose of detecting whether the length of time between the end of the season and the time of the survey had an effect on the estimates. The early survey produced the smaller kill esti- mate (Table-14). The total kill estimate from the later survey fell outside of the 95 per cent confidence limits of the early survey estimate. This indicates the samples 'were from different populations. .This may be the case, but judgment should probably be reserved until the hypothesis of no difference in sample estimates can be tested. The confidence limits on the estimates are computed as if each mail survey was a simple random sample. .The systematic samples appear to fit the model where the population is in "random" order (Cochran, 1962:225). The skewed frequency distribution of pheasants per hunter presented itself as a problem in determining confidence limits. However, Cochran (1965:26) states that it is sufficient for the means of sample estimates to be normally distributed if unbiased total 52 Table 14. Michigan pheasant hunting results from mail survey estimates. Kill with 95% confidence Small game Year limits Hunters Days hunted licensesb 1956 1,125,680t4.2% 549,550 2,421,200 758,086 1957 1,257,570:4.0% 515,960 2,559,290 691,072 1958 1,181,54014.0% 528,780 2,469,140 700,710 1959 914,600:4.8% 454,080 2,098,910 657,071 1960 974,490:4.5% 469,570 2,267,540 647,989 1961 846,470a 455,260 2,085,180 627,514 1962 956,66014.5% 486,810 2,168,540 658,945 1965 720,15014.4% 421,480 2,018,450 616,845 1964 857,590t4.2% 444,070 2,066,605 654,950 1965 685,77014.6% 455,440 2,155,420 652,246 aData unavailable for computing confidence limits. Tabulated from license dealer sales. estimates with confidence limits are to be made. The data over the years indicate this condition is probably satisfied. It has been hypothesized that the kill densities are good indices of fall pheasant populations. For this to be the case there must be some indication of the percentage of hens removed from the population by hunters and a constant percentage of cocks must be harvested annually for a given area. No information is available on the illegal hen kill in Michigan. It is not thought to fluctuate widely or to be a 55 limiting factor to the pheasant population (V. S. Janson, Personal communication). The per cent of cocks harvested annually from each study area was computed (Table 15). The basis for the computations was the sex ratio data from the rural mail carrier post season survey. .Also, a 1:1 preseason sex ratio and a neglig- ible loss of hens during the hunting season was assumed. Small samples of pheasants as well as snowfall over parts of the range in early December are the chief weaknesses of the post season survey. Small samples often result in nearly equal numbers of hens and cocks being observed, while snow on the ground tends to inflate the ratio of hens to cocks actually present in the population. .The results of such biases are poor estimates of the per cent harvest as evidenced in Area IV. The data in Table 15 were analyzed by the Short-cut method of Kurtz g§.al. (1965). Partial results follow: fi’ M _ ==== Years 1965 1961 1965 19621960 1959 1957 1956 19641958 Means 421.9 49.0 5;.8 55.9 56.0 61.5 62.0 62.1 65.3: 74.2 The per cent harvest in 1958 appears inflated. Snow cover during the survey period might explain the unusually high per- centage of cocks taken. The low percentage of cocks harvested in 1965 is probably due to the adverse weather conditions on the opening day. Further analysis determined that each study area had a per cent harvest of cocks significantly different 54 mwu6numm mo conumfi >9 Umunmeoo umm>umn ucmo Hmm .Afl0m6v .UmEDmmm 06umu xmm sommmmmum 6u6m a.mw 6.00 0.00 0.00 0.06 0.00 0.60 N.¢5 0.m0 6.N0 £002 6.00 5.00 5.50 0.05 6.65 5.05 0.00 0.60 0.00 0.00 0.00 > 0.0m 6.0m 0.00 0.00 0.0m 0.0 6.0m 0.00 0.60 5.06 6.0m >6 0.56 0.0m 6.66 0.50 0.00 N.6¢ 0.50 0.00 N.N5 0.00 0.00 HHH 0.00 0.00 0.00 6.56 0.00 0.00 0.00 0.00 5.05 5.50 0.00 HH 0.05 0.00 6.00 6.65 5.00 5.05 0.05 6.65 0.00 0.05 5.05 H :80: 0006 6006 0006 N006 6006 0006 0006 0006 5006 0006 mmu¢ .mmcmn ucmmmmnm mumfi6um m.c006£o62 c6 mHmHHHmo 66MB awash ha mm>usm GOmmmm umom m :0 UmmMm 0.00umm>nms mxuoo ucmwmmnm 00 Damn Hmm .06 06989 55 from the other areas. In conclusion, the per cent of cocks harvested in each study area appears to be relatively constant. Age ratios in the kill also reflect the differences between low and high density areas (Table 16). These figures derived from wings and feet collected early in the)season are probably biased towards a higher number of juveniles in the kill than appear in the population (Eberhardt and Blouch, 1955). Specifically, the age ratio data points to the contribution of broods to the fall kill. Thus, if the juveniles taken annually provide the bulk of the cocks har- vested, the brood survey and the mail survey are essentially measuring the same quantity. Table 16. Male pheasant age ratios derived from wings and feet collected by cooperating Michigan hunters. Juvenileymales per adult male Mean per cent of Area 1959 1961 1962a Mean juveniles in kill I 24.2 16.2 55.5 24.6 96 II 16.9 21.4 10.4 17.5 95 III 10.7 11.9 8.6 10.4 91 IV 6.7 10.7 b 8.7 90 V 19.5 19.7 .10.? 16.6 94 aCollections on a statewide basis discontinued after 1962. Data for 1956-1958 and 1960 is unavailable. bNo adults in the sample. 56 A number of factors affect the proportion of cocks shot. They work chiefly on the effort (hunter-days) expended in hunting pheasants. If the population densities within a study area remain relatively stable, then constant annual effort should produce a fairly constant percentage of cocks harvested annually. Factors influencing effort are: 1. Hunting season length. 2. Changes in bag limits. 5. Variation in hunting conditions. 4. Changes in hunter numbers. During the time period (1956-1965) studied, the season length remained constant at 22 days. Increasing the season length or decreasing it by a few days probably would not have had much of an effect on the total harvest. Most of the kill and effort occurs during the Opening day and the first weekend of the season. During the 1955 Michigan season, 76 per cent of the cocks were harvested during the first week with 51 per cent taken on opening day (Blouch, 1956). .Also 66 per cent of the total effort (gun-hours) was expended dur- ing the first week of that year. The bag limits of two cocks per day and eight for the season have remained constant from 1956 to 1965 and would exert a stabilizing influence on the total kill. Variations in hunting conditions are difficult to assess. The timing of corn-picking, weather, and conditions of natural vegetation influence the kill. Some of these factors 57 are compensatory. Unfortunately, there is no good way of measuring their effect. MacMullan (1960) was not able to de- tect any measurable effect of these conditions on the kill. The opportunity does exist for assessing the effect of the weather on the opening day of the 1965 season. A heavy rain fell over most of the range on that day. The expected results of this occurrence would be a reduced harvest for 1965. Table 15 indicates that the per cent of cocks harvested was down appreciably for all the study areas in that year. Thus, unusual weather conditions during the first few days of the hunting season Should be considered by anyone con- cerned with pheasant population changes. MacMullan (1960) in his study of Michigan pheasant popu- lations uses small game licenses to indicate there was little change in pheasant hunter numbers through 1956. Better data are now available (Table 14). Small game licenses, pheasant hunters, and days hunted by pheasant hunters in the state indicate that annual changes as high as 10 per cent are un— usual. In the study areas (Table 17) this was also true. In examining the mail survey sample data it was noted that close to 40 per cent of the pheasant hunters did not kill any pheasants. Also, approximately 50 per cent of the pheasant hunters harvested about 75 per cent of the pheasants. So, losses of pheasant hunters as high as 50 per cent from one year to the next would probably not have an appreciable effect on the total harvest if it can be assumed that the 58 Table 17. Pheasant hunter numbers in Michigan's primary pheasant range, 1956-1965. :— 3 Mean hunters Coefficient per square of Area Range Mean mile variation I 86,622—125,600 105,704 22.6 15.5% II 87,551-102,465 96,029 50.8 5.5% III 41,905-57,909 49,400 14.6 11.5% IV 52,509-45,497 58,674 8.6 10.8% v 52,552-45,874 57,288 19.7 12.7% majority of the unsuccessful and probably the least skillful of the hunters would drop from the hunting ranks in such a situation. .Since the changes in hunter numbers or effort from 1956 to 1965 has been slight, the assumption of a rela- tively constant effort in this time period seems warranted. In summary, the kill estimates are providing precise estimates of the legal harvest. They are also indicating fall population levels when the above mentioned assumptions are satisfied. Considerable evidence exists that these assumptions are being met. This permits estimates to be made of the total fall population. This is the only time in the pheasant life cycle when relatively good estimates of total numbers are possible. Finally, the kill estimates serve as a standard for evaluating the other indices of abundance. 1. CONCLUSIONS The crowing-cock index is a very imprecise indicator of the fall cock pheasant population in Michigan. Only 20 per cent of the crowing-cock routes in the study areas provided data useful in generating linear regression equations for predicting the fall kill. A logical action would be to abandon this survey or to limit it to an experimental basis in the hope that it can be refined to yield better information. The rural mail carrier brood survey is serving as an ex- cellent source of information for management decisions concerning utilization of the species. Its chief import- ance appears to be in the area of public relations. When pheasant numbers are low, reports on brood abundance and its relation to the fall harvest enable the casual hunter to decide whether he will participate. These re- ports also ease the disappointment of hunters who are not very successful. Since the cost of this survey is minimal, its continued existence appears justified. The mail survey kill figures are the only precise esti- mates of cock pheasant populations and certainly the most valuable in recording historical trends. Experimentation 59 40 with sampling methods could possibly yield more precise estimates. The extensive surveys (late winter, hunter success, and the rural mail carrier postseason survey) complementing the three major surveys are supplying essentially the same information about pheasant population levels and appear to be of marginal value. All the surveys presently indi- cate that the southeast (Area II) and west (Area V) portions of the primary pheasant range are areas of high density. The "thumb" (Area I) and central (Area 111) portion are of medium density while the southwest (Area IV) has a low pheasant density. Why this distribution occurs is still unknown although high pheasant densities appear to be associated with lake-bed clay soils (MacMullan, 1960:154). Surveys of pheasant populations are noted for their imprecision. Carefully thought out experimental designs using aids such as the digital computer appear to be an approach most likely to produce the precision necessary to a better understanding of pheasant population dynamics. Management of a species implies manipulation of the Species for the benefit of man. Planning for optimum utilization of a game Species requires changing of the daily and season bag limits when evidence for a change is present. 41 Also, it may mean harvesting both sexes of a game species. Game birds that are not readily identified by sex do not appear to be harmed in maintaining their population level when they are hunted. Since the biology of the pheasant is probably known better than any other game bird and is nearly as well understood as that of the white-tailed deer, experimental hen seasons Should play the same role as the antlerless deer harvest in utilizing this valuable wildlife resource. Thus, extensive and intensive surveys can become valuable in detecting changes in the population due to management decisions. Direct management of the type mentioned above has not been practiced in Michigan. This is probably because the Conservation Department lacks discretionary authority to manage game species. Convincing state legislators of the soundness of bio- logical principles is difficult. This leads to a con- servative stance in direct manipulation of Species such as the pheasant whose population Size fluctuates quite widely. SUMMARY The three principal indices of pheasant abundance in Michigan's primary pheasant range were evaluated for the 1956 to 1965 time period. The mail survey kill estimates were established as the standard in evaluating the crowing- cock and the rural mail carrier brood survey. AS expected, the crowing-cock index was an imprecise tool in predicting the 1966 fall cock harvest with linear regression equations. However, in four out of five study areas, predictions were possible with wide confidence limits. Approximately 20 per cent of the crowing-cock routes pro- vided data suitable for use in the prediction equations. The rural mail carrier brood survey results were highly correlated with the legal harvest. Linear regression equa- tions predicted the 1966 kill with a high degree of accuracy. Kill estimates for three of the five study areas deviated less than five per cent from the mail survey estimate. The remaining two deviated 10 and 20 per cent from the mail sur- vey estimates. The brood index is equally valuable in terms of broods per unit of time or broods per unit of distance traveled by mailmen. The rural mail carrier brood survey provides excellent information for making management decisions on regulation of the harvest. 42 45 Farm area as a basis for computing kill densities was not as satisfactory as the total study area when kill densi- ties were correlated with the crowing-cock and brood in- dices. Mail surveys of small game hunters appear to estimate closely the legal harvest of cocks. Considerable evidence exists that a constant proportion of cocks are being har- vested in each study area. Thus, kill estimates are prob— ably reflecting the magnitude of fall pheasant populations. This index is the most valuable because the results can be easily converted into density values. LITERATURE CITED Blouch, Ralph I. 1952. The 1952 Pheasant breeding season. Michigan Dept..Conserv., Game Div. Rept. 1165. 9 pp. . 1956. Results of a survey of pheasant hunters by mail. Papers Michigan Acad. Sci., Arts, and ' Letters 41:99-107. Burger, George V. 1966. Observations on aggressive behavior of male ring-necked pheasants in Wisconsin. J. Wildl. Mgmt. 50(1):57-64. Carney,.Samuel M., and George A. Petrides. 1957. Analysis of variation among participants in pheasant cock- crowing censuses. J. Wildl. Mgmt. 21(4):592-597. Cochran, William G. 1965. Sampling techniques. John Wiley & Sons, Inc., New York and London. xvii * 415 pp. Dalke, Paul L. 1957. Food habits of adult pheasants in Michigan based on crop analysis method. Ecology. 18(2):199—215. Eberhardt, Lee, and Ralph I. Blouch. 1955. Analysis of pheasant age ratios. Trans. N. Am. Wildl. Conf. 20:557-567. , and R. M. Murray. 1960. Estimating the kill of game animals by licensed hunters. Proc. Soc. Statist. Sect., 120th Annual Meeting Am. Statist. Assoc. pp. 182-188. Freese, Frank. 1964. Linear regression methods for forest research. U. S. Dept. Agriculture, U. S. Forest Service Research Paper FPL 17. .156 pp. Gates, John M. .1966. Crowing counts as indices to cock pheasant populations in Wisconsin. J. Wildl. Mgmt. 50(4):755-744. Kimball, James W. 1949. The crowing count pheasant census. J. Wildl. Mgmt. 15(1):101-120. Kozicky, Edward L. 1952. -Variations in two Spring indices of male ring—necked pheasant populations. J. Wildl. Mgmt. 16(4):429-457. 44 45 Kurtz, T. E., R. F. Link, J. W. Tukey, and D. L. Wallace. 1965. Shortcut multiple comparisons for balanced single and double classifications: part 1, results. Technometrics 7(2):95-161. MacMullan, Ralph A. 1960. Michigan pheasant populations. Michigan Dept. Conserv., Game Div. Rept. 2277. xiii + 169 pp. Nelson, Robert D., Irven O. Buss, and Gary A. Baines. 1962. Daily and seasonal crowing frequency of ring-necked pheasants. J. Wildl. Mgmt. 26(5):269-272. Petrides, George A. 1954. Estimating the percentage kill in ring-necked pheasants and other game species. J. Wildl. Mgmt. 18(5):294-297. Robertson, William B., Jr. 1948. Investigations of ring- necked pheasants in Illinois. Illinois Dept. Conserv., Tech. Bull. 1. 158 pp. Siegel, Sidney. 1956. Nonparametric statistics for the behavioral sciences. McGraw-Hill Book Company, Inc., New York, Toronto, and London. xvii + 512 pp. Smith, Ronald H. and Steve Gallizioli. 1965. Predicting hunter success by means of a spring call count of Gambel quail. J. Wildl. Mgmt. 29(4):806-815. Stokes, Allen W. [1954.] Population studies of the ring- necked pheasant on Pelee Island, Ontario. Ontario Dept. Lands and Forests Tech. Bull., Wildl. Serv. 4. 154 pp. U. S. Bureau of the Census. 1961. U. S. Census of Agri- culture:1959. U. S. Government Printing Office, Washington. xxv + 255 pp. . 1966. U. S. Census of Agriculture:1964 (prelimi- nary report). U. 8. Government Printing Office, Washington. Wagner, Frederic H., C. D. Besadny, and Cyril Kabat. 1965. Population ecology and management of Wisconsin pheasants. Wisconsin Conserv. Dept. Tech. Bull. 54. 168 pp. Zorb, Gordon L. 1959. Hunter permission survey, pheasant seasons, 1958 and 1959. Michigan Dept. Conserv., Game Div. Rept. 2250. .12 pp. APPEND IX 46 47 Table 18. Regression equations for predicting area kill A (hundreds of cock pheasants) per square mile (Y) from total of calls per crowing-cock route (x), Michigan primary pheasant range, 1956-1965. Area Crowing-cock route Equation X I Bay Y 52.019 + 0.187x I Huron W T 50.454 + 0.112x I Huron N § 51.804 + 0.165x III Ingham § 15.027 + 0.045x III Ionia § 25.451 + 0.067x IV Case Y 9.299 + 0.148x v Ottawa § 21.665 + 0.090x Table 19. 48 .Regression equations for predicting area kill A (hundreds of cock pheasants) per square mile (Y) from broods observed perJIXX)miles (x) of travel by rural mail carriers, Michigan primary pheasant range, 1956-1965. Area Equation I § = :5.095 + 2.787x II T = 26.661 + 2.584x III § = 10.194 + 1.858x IV 9 = 6.992 + 1.825x v 9 = 21.215 + 1.972x 49 0 H g 60-- w u w :3 U‘ U) 3? 4.0-» m H H -a K A Y = 21.665 + 0.090x 0 1L l I 0 200 400 600 Total calls per crowing-cock route 0 ,_I -a E 0 60‘- H m :3 O" m r H 0 C1+40» H H -a M A Y = 21.215 + 1.972x 0 I 4. 0 10 20 Broods per thousand miles ’Figure 2. Typical relationship between kill densities and crowing-cock and rural mail carrier brood survey results, Area V, Michigan primary pheasant range, 1956-1965. ‘ll1111111111111l“