Black-Shouldered Kite - Small Mammal - Vegetation Relationships in Northwestern California

Black-shouldered kite - Small mammal - Vegetation Relationships in Northwestern California

Jeffrey R. Dunk

A Thesis Presented to The Faculty of Humboldt State University

In Partial Fulfillment of the Requirements for the Degree Master of Science

February 1992 BLACK-SHOULDERED KITE - SMALL MAMMAL - VEGETATION

RELATIONSHIPS IN NORTHWESTERN CALIFORNIA

Jeffrey R. Dunk

Approved by the Master's Thesis Committee

Robert J. Cooper, Chairman

David W. Kitchen

Timothy E. Lawlor

Director, Natural Resources Graduate Program Date

92/W-235/02/14 Natural Resources Graduate Program Number

Approved by the Dean of Graduate Studies

Susan H. Bicknell ABSTRACT

This thesis is written in 2 sections to divide the major themes of the study. The second section is partially based on the work presented in the first. I studied the relationships between populations of both California voles (Microtus californicus) and Western harvest mice (Reithrodontomys megalotis) and vegetation structure and composition in northwestern California. Both species exhibited annual fluctuations in abundance during the 19 months of the study. In general, M. californicus populations were largest in areas with lower cover height and larger percent brown grass cover. This finding is contrary to the majority of Microtus spp. studies that have examined population - vegetation relationships. The discrepancy is likely due to the fact that all areas that I trapped had fairly abundant cover (no areas were grazed). The increase of M. californicus populations (from the annual low) coincided with grass seeds being more available (i.e., dropping off of grasses). To my knowledge I recorded the largest annually fluctuating M. californicus population ever found (range = 0 - 1414 ind/ha, Jolly-Seber estimate). I propose that resource limitation is a primary factor in M. californicus population fluctuations. R. megalotis populations fluctuated synchronously iii iv with those of M. californicus, but at much reduced numbers. Increasing populations of R. megalotis were associated with an increased availability of grass seeds. The largest populations of R. megalotis were associated with relatively low cover height of vegetation. Each month I developed discriminant function analysis models to predict both species of small mammal abundances using vegetation (structure and composition) variables. Mean percent correct classification was 70.03% and 88.2% for M. californicus and R. megalotis, respectively (n = 19). I also studied the relationship between black- shouldered kite (Elanus caeruleus) territory size and both prey abundance and competitor abundance in northwestern California. Kite territory size ranged from 1.6 - 21.5 ha (n = 26). Estimated mean number of M. californicus per territory was 1483 (SE = 163, n = 25). Competitor abundance (i.e. total raptor abundance) ranged from 4.8 - 31.0 individuals/km2 and was strongly correlated with abundance of M. californicus. Both estimated prey abundance and competitor abundance were negatively correlated with kite territory size. After developing a multiple regression model using both variables, partial correlation analysis revealed that once the effects of prey abundance were statistically V controlled, competitor abundance continued to be significantly correlated with kite territory size. When the effects of competitor abundance were statistically controlled, prey abundance was no longer significantly correlated with kite territory size. I conclude that kite territory size is proximately regulated by competitor abundance and ultimately regulated by prey abundance because M. californicus abundance regulates numbers of raptors. ACKNOWLEDGMENTS

I am indebted to so many people for their help, support, and encouragement prior to and throughout this entire experience that I am not quite sure where to begin. I will start by thanking my parents, Bill and Barbara Dunk, for their support (v and $) from the first mention of this endeavor. They continued this support even after they found out that scientists have to pay to have their work published. I thank Grandma Lou, Uncle Mike and Aunt Pat, and Bill and Marilyn Gillaspy for introducing me to natural areas at a young age and showing me how to enjoy them. It is very unlikely that I would be in this field had it not been for them.

I thank G. Monroe of California Department of Fish and Game for allowing me access to state property to conduct my study. I also thank H. and P. Hunt for their thoughtful consideration of my project during their activities on the study area.

Dr. James R. Koplin (deceased) first introduced me to the system and animals that I studied. For those that knew him, his influence on me will be obvious. My friends TallChief A. Comet, Douglas G. Leslie, and I had many discussions on black-shouldered kites before this project began. Their support, enthusiasm, and ideas are largely responsible for my enthusiasm and my thesis topic.

vi vii I thank my friends S. Beatty, J. Browning, K. Bradley, T. Burgess, D. Call, T. Comet, R. Cooper, S. Dunk, L. Ellis, R. Grosz, B. and D. Kristan, D. Leahy, D. Leslie, R. Lightfoot, R. Long, G. Roemer, S. Steinberg, L. Tisue, C. Verhey , and N. Weisman for providing help and companionship in the field. Thanks also to A. Franklin for letting me sub-permit under his banding permit. I thank my committee members Drs. T. E. Lawlor and D. W. Kitchen for their many helpful comments on drafts of my thesis. I would also like to thank K. Moon and R. Brown for their assistance whenever it was asked. Sabra Steinberg helped during all phases of my graduate life. I thank her for her support, encouragement, companionship, and tolerance (of my pre-dawn ventures into the field each month) throughout. I am indebted to my major professor, Dr. Robert J. Cooper, for his encouragement, friendship, guidance, and support in every phase of this project. He went well beyond his "duties" as a major professor. Partial funding for this project came from 2 Humboldt State University Research and Creative Activity grants to Dr. Cooper. Lastly, I would like to thank the Wildlife Department at HSU for attracting such high caliber students. It was while interacting with these people that my greatest educational experiences occurred. TABLE OF CONTENTS Page

ABSTRACT . . . . . . . . . • . iii ACKNOWLEDGMENTS . . . . . . . • • vi LIST OF TABLES . . . . . . . . . ix LIST OF FIGURES . . . . . . . • . xi GENERAL INTRODUCTION • • • . • • • • 1 STUDY AREA . . . . . . . . . . 2 SMALL MAMMAL - VEGETATION RELATIONSHIPS . . . 4 INTRODUCTION . . . . . . . . . 4 METHODS . . . . . . . . . . • 9 Small mammal abundance . . . . . . 9 Vegetation characteristics . . . . . . 11 RESULTS . . . . . . . . . • • 13 Microtus californicus . . . . . . . 13 Reithrodontomys megalotis . . . . . . 33 DISCUSSION . . . . . . . . . . 41 Microtus californicus . . . . . . . 41 Reithrodontomvs megalotis . . . . . . 48 TERRITORY SIZE REGULATION IN BLACK-SHOULDERED KITES. 50 INTRODUCTION . . . . . . . . . 50 METHODS . . . . . . . . . . . 53 RESULTS . . . . . . . . . • • 58 DISCUSSION • • • • • • • • . • 63 LITERATURE CITED . . . . . . . . . 69 viii LIST OF TABLES Table Page 1 Comparison of 4 methods of estimating California vole (Microtus californicus) populations. All estimates are rounded to the nearest whole number. Numbers in parentheses are standard errors. (na) means that the calculation could not be made. Fay Slough Wildlife Area, Eureka, California, June 1989 - December 1990. . . . . . 15 2 Sex ratios of California voles (Microtus californicus) (percent female) caught on the Fay Slough Wildlife area, Eureka, California. June 1989 - December 1990. Total number of individuals (male and female) sexed in parentheses. Dashes (-) represent no individuals caught or sex not determined. . 19 3 Jolly-Seber estimates of monthly survival probabilities (and SE) for Microtus californicus caught on the Fay Slough Wildlife Area, Eureka, California, from June 1989 - December 1990. Dashes (-) represent parameters not estimated by the model. Blanks represent a grid not being trapped that month. . . . . . . . . 21 4 Four-group stepwise discriminant analysis of Microtus californicus abundance in ungrazed grassland habitat. Only variables entered into the model are represented. Numbers presented are structure coefficients for the first two discriminant functions (DF1 and DF2). Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. . 29 5 Evaluation of Stepwise discriminant analysis models ability to predict Microtus californicus abundance based on monthly vegetation characteristics in small mammal grids. Observations were considered correctly classified if they were classified into the grid they came from or a grid with similar M. californicus abundance (similar = within 5 individuals). Fay Slough Wildlife Area, Eureka, California, June 1989 - December 1990. . . . . . . . . . 33 ix •

x 6 Stepwise-regression analysis results using Microtus californicus abundance (naive) as the dependent variable and vegetation parameters as independent variablesa. Analyses included increase phases in abundance of M. californicus, decrease phases, and all data regardless of population phase. Only variables entered into models are presented. Results where no variables were entered into models are not included. . . . . . . 34 7 Stepwise-regression analysis results using Reithrodontomys megalotis abundance (naive) as the dependent variable and vegetation parameters as independent variables. Analyses included increase phases in abundance of R. megalotis, decrease phases, and all data regardless of population phase. Only variables entered into models are presented. Results where no variables were entered into models are not included. . . • • • • 38 8 Evaluation of Stepwise discriminant analysis models ability to predict Reithrodontomys megalotis abundance based on monthly vegetation characteristics in small mammal grids. Observations were considered correctly classified if they were classified into the grid they came from or a grid with similar R. megalotis abundance (similar = within 5 individuals). Fay Slough Wildlife Area, Eureka, California, June 1989 - December 1990. • • • 40 9 Monthly raptor species abundances at the Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. . 59 10 Stepwise-regression analysis and partial correlation analysis results using black- shouldered kite territory size as the dependent variable and competitor abundance and prey abundance as independent variables. For each pair of analyses, the first model presented is the best. Following those, models where entry of independent variables was changed are presented. * next to variable names means that this variable was entered into the model first. . 61 LIST OF FIGURES

Figure Page 1 Location of study area, Fay Slough Wildlife Area, in northwestern California. . . 3 2 Number of individual California voles (Microtus californicus) caught within each trapping grid each month, on the Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. . . . . 14 3 Mean percent green grass cover within each small mammal trapping grid. Bars represent 1 SE. Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. . . 22 4 Mean percent brown grass cover within each small mammal trapping grid. Bars represent 1 SE. Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. . . . . . . 23 5 Mean percent green herbaceous cover within each small mammal trapping grid. Bars represent 1 SE. Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. . 24 6 Mean percent brown herbaceous cover within each small mammal trapping grid. Bars represent 1 SE. Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. . 25 7 Mean percent grass seed cover within each small mammal trapping grid. Bars represent 1 SE. Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. . . . 26 8 Mean cover height (cm) within each small mammal trapping grid. Bars represent 1 SE. Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. . . 27 9 Number of individual western harvest mice (Reithrodontomys megalotis) caught within each trapping grid, each month, on the Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. . 36

xi GENERAL INTRODUCTION

This thesis is written in 2 sections in order to separate the 2 major themes of the study and make the manuscript easier to read. Each section is treated as a discrete study although the second section is dependant on the first. Although each section is treated discretely, after reading both, a single, larger, study is obvious. The first section, small mammal - vegetation relationships, examines the relationship of small mammal populations, within a macrohabitat, to vegetation structure and composition. I developed models to predict small mammal abundance based on vegetation structure and composition and to distinguish between habitats of differing qualities. Section 2, territory size regulation in black- shouldered kites, examines the separated and combined influences that competitor abundance and prey abundance had on territory size of kites. I developed models to predict kite territory size based on prey and competitor abundance and discuss their relative importance in determining kite territory size. Section 2 draws considerably from the small mammal and vegetation work presented in section 1. The same study area was used for all aspects of the field work, so a single study area description is provided.

1 STUDY AREA

The study was conducted on the Fay Slough Wildlife Area (Fig. 1) in Eureka, California. The 135 ha. area was used for cattle grazing prior to 1987 when it was acquired by California Department of Fish and Game. It was composed primarily of ungrazed pasture with small patches of alders (Alnus spp.) and blackberry bushes (Rubus spp.). Alders were planted along slough borders in 1988, and provided important perching locations for raptors. A number of freshwater and brackish sloughs dissect the area. The topography is flat with elevations ranging from -0.5 m to 2 m. The climate in the area is maritime, with cool summers and mild winters. Summers are characterized by foggy nights and mornings. Precipitation was greatest between November and April though rain occurred during most months of this study. The growing season is generally between April and July during which time grasses develop, then drop, their seeds. Vegetation composition and structure was relatively constant throughout the study though limited (rotational) grazing occurred on approximately 35 ha beginning in Dec 1989. Primary plant species included Italian ryegrass (Lolium multiflorum), perennial ryegrass (L. perenne), velvet grass (Holcus lanatus), and rose clover (Medicago spp.).

2 3

Figure 1. Location of study area. Fay Slough Wildlife Area, in northwestern California. SMALL MAMMAL - VEGETATION RELATIONSHIPS INTRODUCTION Microtine rodents have received an enormous research effort during the past 50 years in both Europe and North America. Much of the work has focused on understanding the ecology of these rodents in order to control their well- documented cycles (primarily peaks) of abundance. The idea of managing habitat to encourage microtine populations has received little attention. Tamarin (1985) reviewed much of the work done in North America to date. Much of the work on California voles (Microtus californicus) has focused on population regulation, primarily by means of analyzing demographic (Krebs 1966, Ostfeld et al. 1985) and behavioral (Ostfeld 1985, Heske et al. 1988) features or a combination of both (Hestbeck 1986). Pearson (1966, 1971) suggested that M. californicus populations, especially population lows, are regulated by predation. The view that predation causes microtine cycles has received little support in North America (Boonstra 1977).

Habitat requirements of all microtine species are difficult to generalize because studies comprise many areas (Getz 1985). Many authors have noted the positive relationship between microtines and vegetative cover (Eadie 1953, Mossman 1955, Birney et al. 1976, and others). It 4 5 appears that a threshold level of vegetative cover is necessary for microtine populations to increase (Birney et al. 1976). The positive effects of vegetative cover seem to be decreased risk of predation (Klatt and Getz 1987) and increased food quantity (Lack 1954) and quality (Pitelka 1964). The relative importance of each of those factors in contributing to habitat quality is largely unknown and probably varies depending on the microtine species in question. Koplin and Hoffman (1968) and Klatt and Getz (1987) each showed how 2 sympatric Microtus species used available habitat differently. Those findings exemplify the need for site- and species-specific habitat relationships if microtine-habitat relationships are to be understood. Other factors that may potentially confound identification of habitat characteristics important to microtine species include lack of: 1) quantification of vegetation composition, structure, or both during a study, relying instead upon previous non-specific site descriptions presented in the literature; 2) information on the abundance of other small mammal species caught during a study; and 3) information on the abundance and types of predators in an area. Although there is no consensus as to the importance of food quantity and/or quality as factors limiting the density of natural populations (Batzli 1988), results from many studies suggest that they are key extrinsic factors 6 influencing population size and cyclic patterns of microtine rodents. Batzli (1986) experimentally assessed the affects of different diets on reproduction of M. californicus and concluded that mineral deficiencies caused declines in reproduction. Bergeron and Jodoin (1987) found that protein and caloric contents in plants decreased during summer months and that food preference of M. pennsylvanicus was closely linked to protein content and low levels of digestive inhibitors in plants. Also, several authors have found that enclosed Microtus populations provided with supplemental food reached higher densities and exhibited different cyclic patterns than populations without supplemental food (Taitt and Krebs 1981, Ford and Pitelka 1984, Desy et al. 1990). Other population parameters have also been found to be affected by food supply. Ford and Pitelka (1984) reported that survivorship, breeding rate, and rate of population change of M. californicus were significantly correlated with vegetative production. Ostfeld (1986) showed that female M. californicus home ranges decreased in size when supplied with supplemental food, while male home range sizes were a function of access to females (indirectly related to food abundance). Similarly, Jones (1990) found territory size of female M. pennsylvanicus was inversely correlated to available forage and concluded that forage quantity is a factor in controlling population density. 7 Most attempts to explain vole cycles have been based on the assumption that voles interact with a relatively constant plant food supply (Hornfeldt et al. 1986). Yet the above studies suggest that plant production rhythms may play a role as possible triggers to vole cycles. If the quantity and quality of plant food changes over the course of a vole cycle, then vegetative characteristics might serve as good predictors of vole population parameters. Relative to research on microtine rodents, western harvest mice (Reithrodontomys megalotis) have received little attention. Of the studies that mention R. megalotis, few were initiated to address specific questions about their ecology. R. megalotis habitat is generally thought to be in areas of dense, tall grass (Ford 1977, Abramsky 1978, Abramsky et al. 1979). Blaustein (1981) reported asynchronous fluctuations in abundance of R. megalotis between 4 areas and was unable to predict abundance within areas based on abundances from the previous year. Blaustein (1981), Heske et al. (1984), and Johnson and Gaines (1988) all reported that R. megalotis underwent annual cycles of abundance, and that R. megalotis abundance was negatively affected by high abundance microtine populations, suggesting possible competition between the 2 species. Conversely, Pitcher and Keller (1979) found the largest populations of M. montanus had no measurable affect on R. megalotis. 8 However, Heske et al. (1984) attributed those results to relatively low (peak) abundance of M. montanus (ca. 140 voles/ha). In this study I provide further information on the relationships between M. californicus and R. megalotis populations and vegetation structure and composition in a coastal northwestern California grassland. I tested whether either vole or harvest mouse abundance could be predicted with vegetation variables. Specifically, I tested the following null hypotheses: 1) There is no difference in M. californicus abundances among 4 ungrazed grassland areas; 2) M. californicus abundance is not related to vegetation structure and composition; 3) There is no difference in R. megalotis abundance among 4 ungrazed grassland areas; and 4) R. megalotis abundance is not related to vegetation structure and composition or M. californicus abundance. METHODS Small Mammal Abundance Small mammals were trapped monthly from June 1989 through December 1990. Four randomly selected permanent locations were used to sample small mammal populations. One of the initial trap grid locations was disturbed after the fifth month of the study, so another grid location was randomly chosen. Small mammals were trapped in Sherman live-traps spaced by 5 m in an 8 x 5 pattern (40 traps per grid). Each trap was locked open with no food or bedding for 24 - 36 hours prior to being set. Traps with fecal material and food in them were wiped clean. All traps were then set and furnished with bird-seed and polyester pillow stuffing. Traps were checked within 1 hr of sunrise and prior to sunset for 4 consecutive days. Traps were then shut and left in the grid until the following month. Each trap location was uniquely numbered. The following data were recorded at each capture location: 1) species, 2) sex (of microtines), 3) weight (g), 4) trap number, 5) date, and 6) time of capture (AM or PM). Individuals of each species were uniquely toe-clipped for identification. All individuals were weighed with a spring scale to the nearest g. Captured individuals were released within 1 m of the capture location. I used the number of individuals (of each species)

9 10 caught per month as a naive estimate of abundance that is directly comparable between months regardless of the number of individuals caught. I expressed density (naive) as number of individuals caught each month (in each grid) divided by grid area (700 m2). Minimum number known to be alive (MNKA) (Krebs 1966) is also widely used, so I used it to calculate density of M. californicus, facilitating comparison with most Microtus studies. MNKA assumes, for trapping occasions 1 through k, that individuals caught and marked in a grid prior to time t, not caught at time t, but subsequently caught at some time after time t, were in the grid at time t (for this study, t = 1, 2, ..., k refer to months and not trapping occasions within a month). For example, if an individual was caught in September and not caught again until January, it was assumed present in the grid during all months between September and January. I also estimated vole populations using the Jolly-Seber estimate (Jolly 1965) which uses individual capture histories between months (i.e. regardless of whether an individual was caught 1 or 8 times during a month, it was scored as a 1 for that month). I estimated population sizes monthly survival probabilities of voles within each grid with model A of program JOLLY (Brownie et al. 1986). Lastly, I estimated vole populations using the Petersen method by using the first 2 trapping days as the initial marking period and the last 2 as the recapture period. 11 Vegetation Characteristics Vegetation was sampled in each small mammal trapping grid within 1 week post-trapping each month (except October 1989 when no vegetation sampling was done). Vegetation characteristics were measured next to each trap location in 0.5 m2 circular plots. The side of the trap next to which vegetation plots were placed was randomly determined each month for each grid. On each plot, I measured percentage of the plot covered by: 1) green grass (Gramineae), 2) brown grass, 3) green herbaceous vegetation, 4) brown herbaceous vegetation, 5) seed heads on grasses (this measure was not exclusive of grass cover), 6) green rush (Cyperaceae), and 7) brown rush (parameters 6 and 7 were measured starting in Dec 1989). All measures were visually estimated and classified into 1 of 6 cover classes: 0, 1-20%, 21-40%, 41- 600, 61-80%, or 81-100%. I also measured cover height of vegetation in each plot. Cover height was measured at each plots' center by pushing a meter stick through the vegetation, flush to the ground, and recording the height (cm) at which the meter stick and plot frame intersected. When grids were partially flooded, plots that fell over open water were not measured, but recorded as flooded, assuming that open water was unsuitable habitat for small mammals. Monthly vegetation characteristics were analyzed using Stepwise Discriminant Analysis (Dixon 1985) to find the vegetative characteristics that most influenced 12 abundance of voles and harvest mice at any point in time. Vegetation parameters were the independent variables, while vole and harvest mouse abundances were dependent variables. Each grid was considered a distinct population; vegetation plots within grids were the observations to be classified. Model accuracy was assessed by examining correct classification rates (i.e. the percentage of vegetation plots correctly classified to the grid they came from). Grids were combined into one population if they had similar abundance of voles or harvest mice in a particular month. Small mammal abundances (naive) were considered similar if they were within 5 individuals of one another. It would be unreasonable to consider populations as different if they only differed by a few individuals. I felt that differences of 5 or fewer individuals would alleviate differences due to random variations in capture probabilities. Vegetation characteristics influencing vole and harvest mouse abundances within each trapping grid were analyzed using stepwise-regression analysis. These analyses were done for population increase phases, population decrease phases, and with all data regardless of population phase. Population increase phases included the last month of low abundance through peak abundance periods. Population decrease phases started at the month following peak abundance and continued through the first month of lowest abundance. RESULTS Microtus californicus Vole populations fluctuated annually in all grids (Fig. 2). Populations increased from June 1989 through September 1989, then remained relatively constant until March - April 1990 when all populations declined (no individuals were captured during April 1990). Populations increased from May 1990 through September 1990 and again remained relatively stable through December 1990 (Fig. 2). Though vole populations in all grids followed the same general pattern of abundance, there were large differences in the magnitude of fluctuations (Fig. 2). Vole density ranged from 0 - 914, 0 - 971, 0 - 1243, and 0 - 1414 individuals ha-1 based on number of individuals captured, MNKA, Petersen, and Jolly-Seber estimates, respectively (all estimates were determined by dividing the population estimate by 700 m2). Of the 4 population estimators used, number of individuals caught was always the smallest while the Petersen and Jolly-Seber estimates were the largest (Table 1). Mean difference between number of individuals caught and MKNA was 1.89 individuals. In general, sex ratios changed from a predominantly female bias during population increases to male bias during peak abundance and population decline phases (Table 2). Grid 5 did not follow this pattern as strongly, being female

13 14

Figure 2. Number of individual California voles (Microtus californicus) caught within each trapping grid each month, on the Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. 15

Table 1. Comparison of 4 methods of estimating California vole (Microtus californicus) populations. All estimates are rounded to the nearest whole number. Numbers in parentheses are standard errors. (na) means that the calculation could not be made. Fay Slough Wildlife Area, Eureka, California, June 1989 - December 1990.

Estimator # caughta MNKAb c P J-Sd Month

Jun 1989 Grid 1 0 0 0 na Grid 2 13 13 17 (4.9) na Grid 3 1 1 0 na Grid 4 1 1 0 na

Jul 1989 Grid 1 0 0 0 0 Grid 2 15 15 36 (59.9) 16 (na) Grid 3 5 5 5 (4.9) 5 (na) Grid 4 19 19 22 (13.2) 20 (na)

Aug 1989 Grid 1 8 8 0 9 (na) Grid 2 28 31 36 (20.1) 58 (43.8) Grid 3 4 5 0 5 (1.6) Grid 4 33 34 34 (5.7) 37 (3.1)

Sep 1989 Grid 1 10 10 11 (3.5) 10 (na) Grid 2 64 68 83 (8.6) 87 (14.7) Grid 3 14 16 16 (3.9) 27 (15.1) Grid 4 39 45 38 (4.1) 49 (6.1)

Oct 1989 Grid 1 10 10 7 (4.5) na Grid 2 34 44 37 (5.4) 51 (5.4) Grid 3 8 13 11 (16.4) 23 (14.2) Grid 4 33 42 37 (5.6) 47 (4.7)

Nov 1989 Grid 2 56 62 99 (24.2) 74 (9.1) Grid 3 9 13 12 (19.4) 21 (10.0) Grid 4 37 47 42 (6.7) 55 (4.6) 16

Estimator # caughta bMNKA PC J-Sd Month

Dec 1989 Grid 2 39 46 54 (12.6) 47 (4.9) Grid 3 12 14 12 (5.8) 14 (3.1) Grid 4 44 58 47 (6.2) 77 (9.7) Grid 5 8 8 9 (4.3) na

Jan 1990 Grid 2 33 33 40 (7.2) 30 (na) Grid 3 13 14 11 (4.8) 15 (2.5) Grid 4 42 55 46 (5.6) 68 (8.3) Grid 5 32 33 45 (15.6) 38 (9.0)

Feb 1990 Grid 2 19 19 23 (5.7) 19 (na) Grid 3 15 16 36 (59.9) 24 (12.2) Grid 4 39 43 46 (7.6) 64 (11.6) Grid 5 22 24 48 (39.2) 44 (15.7)

Mar 1990 Grid 2 0 0 0 0 Grid 3 5 5 5 (2.5) 5 (na) Grid 4 28 28 28 (9.3) 28 (na) Grid 5 24 24 32 (9.8) 25 (na)

Apr 1990 Grid 2 0 0 0 0 Grid 3 0 0 0 0 Grid 4 0 0 0 0 Grid 5 0 1 0 0

May 1990 Grid 2 0 0 0 0 Grid 3 3 3 2 (2.5) 3 (na) Grid 4 0 0 0 0 Grid 5 10 11 10 (1.9) 16 (na) 17

b Estimator # caughta MNKA c P J-Sd Month

Jun 1990 Grid 2 0 0 0 0 Grid 3 3 3 4 (na) 2 (na) Grid 4 0 0 0 0 Grid 5 17 19 18 (3.3) 24 (4.8)

Jul 1990 Grid 2 6 6 12 (na) 6 (na) Grid 3 6 8 7 (3.1) 20 (na) Grid 4 9 10 15 (22.4) 10 (na) Grid 5 15 19 16 (2.7) 24 (4.9)

Aug 1990 Grid 2 17 18 21 (10.4) 20 (6.1) Grid 3 7 11 7 (1.3) 21 (12.9) Grid 4 14 14 18 (7.8) 18 (3.5) Grid 5 10 14 11 (2.5) 18 (4.1)

Sep 1990 Grid 2 41 41 44 (6.3) 39 (na) Grid 3 8 11 8 (2.2) 11 (3.3) Grid 4 22 22 23 (5.4) 22 (na) Grid 5 13 14 12 (4.6) 14 (1.3)

Oct 1990 Grid 2 30 31 35 (8.2) 34 (5.4) Grid 3 12 13 12 (na) 13 (3.3) Grid 4 28 29 29 (3.5) 29 (1.7) Grid 5 18 19 19 (5.0) 19 (1.6) Nov 1990 Grid 2 19 19 22 (6.5) 20 (1.5) Grid 3 11 12 11 (4.8) 18 (8.0) Grid 4 43 47 55 (5.4) 55 (8.1) Grid 5 24 25 28 (7.2) 23 (1.5) 18 Table 1. Comparison of 4 methods of estimating California vole (Microtus californicus) populations. All estimates are rounded to the nearest whole number. Numbers in parentheses are standard errors. (na) means that the calculation could not be made. Fay Slough Wildlife Area, Eureka, California, June 1989 - December 1990. (continued)

Estimator # caughta MNKAb Pc J-Sd Month Dec 1990 Grid 2 18 18 21 (5.7) na Grid 3 11 11 11 (2.4) na Grid 4 33 33 34 (2.6) na Grid 5 22 22 24 (2.6) na

a Number of individuals caught within a grid b Minimum number known to be alive (after Krebs 1966) cPetersen estimate d Jolly-Seber estimate 19

Table 2. Sex ratios of California voles (Microtus, californicus) (percent female) caught on the Fay Slough Wildlife area, Eureka, California. June 1989 - December 1990. Total number of individuals (male and female) sexed in parentheses. Dashes (-) represent no individuals caught or sex not determined.

# Grid 1 2 3 4 5 Month 1989 June - - - - July - 53(13) 60(5) 71(14) August 37(8) 54(28) 75(4) 47(32) September 30(10) 57(60) 46(13) 58(31) October 33(9) 38(34) 43(7) 58(31) November 43(54) 50(6) 45(31) December 34(35) 55(11) 42(38) 14(7) 1990 January 25(32) 50(12) 37(38) 28(25) February 42(17) 80(15) 64(39) 70(20) March - 100(5) 44(27) 70(23) April - - - - May - 0(3) - 40(10) June - 100(1) - 47(17) July 50(6) 33(6) 67(9) 60(15) August 59(17) 57(7) 57(14) 60(10) September 60(40) 75(8) 59(22) 58(12) October 53(30) 58(12) 57(28) 44(18) November 42(19) 55(11) 41(41) 41(24) December 39(18) 64(11) 41(32) 50(20) 20 biased prior to the population crash. Estimated survival probabilities were generally synchronous with abundance on each grid (Table 3). Estimated survival probabilities were > 0.8 in all grids during population highs of both years (Table 3). Precision of estimates (standard errors) of survival probability were significantly, inversely, correlated with number of individuals caught (r = -0.84, P < 0.001, n = 46) and ranged from 0.08 - 0.53 (excluding one month with a sample of 1) (Table 3). All vegetation characteristics fluctuated annually within each grid (Figs. 3 - 8). Within-grid variation in all vegetation parameters was relatively small and constant at any point in time (Figs. 3 - 8). Green grass was most abundant in April and May 1990 and then declined rapidly through September (Fig. 3). Brown grass abundance (Fig. 4) was inversely correlated to green grass abundance in all grids (r = -0.982, -0.987, -0.996, and -0.747 for grids 2 - 5, respectively, P < 0.001 for all). Herb abundance was low throughout the study, reaching a maximum of 11.8% cover during August 1990 in grid 2 (Figs. 5 and 6). Green herbs were most abundant from May - September (Fig. 5); after that few persisted. Brown herbs were most abundant from October - March (Fig. 6). Grass seeds (on grasses) were most abundant during June and July, declined rapidly from August through October, and reached lowest abundance during Table 3. Jolly-Seber estimates of monthly survival probabilities (and SE) for Microtus californicus caught on the Fay Slough Wildlife Area, Eureka, California, from June 1989 - Dec 1990. Dashes (-) represent parameters not estimated by the model. Blanks represent a grid not being trapped that month.

Grid # 1 2 3 4 5 Month June - - 1.000 (1.00) - July - 0.294 (0.16) 0.650 (0.26) 0.626 (0.12) August 0.570 (0.19) 0.745 (0.12) 1.071 (0.43) 0.707 (0.10) September - 0.591 (0.08) 0.917 (0.53) 0.768 (0.09) October - 0.873 (0.12) 0.400 (0.26) 0.798 (0.09) November 0.533 (0.09) 0.521 (0.18) 0.918 (0.09) December 0.486 (0.08) 0.838 (0.18) 0.858 (0.11) 1.036 (0.22) January 0.233 (-) 0.625 (0.35) 0.881 (0.18) 0.659 (0.24) February - 0.200 (0.10) 0.297 (-) 0.238 (0.09) March - 0.200 (-) - 0.044 (-) April - - - - May - 0.333 (0.27) - 0.822 (0.17) June - 1.400 (0.48) - 0.737 (0.17) July 0.846 (0.27) 1.000 (0.46) 0.667 (0.19) 0.698 (0.20) August 0.706 (0.11) 0.586 (0.26) 0.571 (0.13) 0.572 (0.17) September 0.518 (0.11) 0.781 (0.27) 0.781 (0.10) 0.798 (0.13) October 0.422 (0.10) 0.533 (0.24) 0.808 (0.12) 0.704 (0.12) November - - - December - - - - 2 1 22

Figure 3. Mean percent green grass cover within each small mammal trapping grid. Bars represent 1 SE. Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. 23

Figure 4. Mean percent brown grass cover within each small mammal trapping grid. Bars represent 1 SE. Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. 24

Figure 5. Mean percent green herbaceous cover within each small mammal trapping grid. Bars represent 1 SE. Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. 25

Figure 6. Mean percent brown herbaceous cover within each small mammal trapping grid. Bars represent 1 SE. Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. 26

Figure 7. Mean percent grass seed cover within each small mammal trapping grid. Bars represent 1 SE. Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. 27

Figure 8. Mean cover height (cm) within each small mammal trapping grid. Bars represent 1 SE. Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. 28 March and April (Fig. 7). On all grids cover was highest during June of both 1989 and 1990 and lowest during January (Fig. 8). Mean cover height increase (for grids 2 - 5) between May and June (1990) was 9.9 cm. From the 1989 vole population increase through February 1990, grids with larger percentage brown grass and lower cover height were usually associated with larger vole populations (Table 4, Figs. 2, 4, and 8). No voles were caught in grid 2 during March 1990, at which time grids with a larger percentage brown grass cover or a larger percent brown rush cover had larger vole populations. No voles were caught during April 1990 in any grid (Fig 2). Grids 3 (3 individuals) and 5 (10 individuals) were the first grids to begin to show increased populations, in May, from the annual population low (Fig. 2). High cover height and large percentage green rush distinguished grid 5 from all other grids during May and June 1990, respectively. Voles on grids 2 and 4 began to increase from the annual population low in July 1990 (Fig. 2). During this time the largest vole population was associated with low percentage brown grass cover (Table 4, Figs. 2 and 4). During August 1990, large vole populations were associated with areas that had a large percentage brown grass cover, low cover height, and a small percentage of brown herbs. In September 1990, the largest vole population (grid 2) was associated with large percentage brown grass cover, low cover height, and large 29 Table 4. Four-group stepwise discriminant analysis of Microtus californicus abundance in ungrazed grassland habitat. Only variables entered into the model are represented. Numbers presented are structure coefficients for the first two discriminant functions (DF1 and DF2). Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. 30 Table 4. Four-group stepwise discriminant analysis of Microtus californicus abundance in ungrazed grassland habitat. Only variables entered into the model are represented. Numbers presented are structure coefficients for the first two discriminant functions (DF1 and DF2). Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. (continued)

Variable DF1 DF2 March 1990 Brown Grass 0.257 0.533 Seed 0.081 -0.150 Brown Rush -0.458 -0.836 April 1990 Cover Height -0.659 -0.678 Seed 0.801 -0.429 Brown Rush -0.307 -0.881 May 1990 Green Grass. 0.259 0.915 Seed 0.825 -0.378 Cover Height -0.559 -0.577 June 1990 Brown Grass 0.392 0.349 Green Herb 0.013 0.153 Cover Height -0.585 -0.605 Seed 0.504 -0.478 Green Rush -0.866 0.368 July 1990 Brown Grass 0.141 0.990 Seed -0.066 0.581 August 1990 Brown Grass 0.859 -0.506 Brown Herb 0.149 0.614 Seed 0.548 -0.293 September 1990 Brown Grass 0.892 -0.416 Green Herb 0.080 0.665 Cover Height -0.407 -0.749 31 Table 4. Four-group stepwise discriminant analysis of Microtus californicus abundance in ungrazed grassland habitat. Only variables entered into the model are represented. Numbers presented are structure coefficients for the first two discriminant functions (DF1 and DF2). Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. (continued)

Variable DF1 DF2 October 1990 Brown Grass 0.812 -0.584 Green Herb 0.100 0.523 Seed -0.041 -0.348 Cover Height -0.674 -0.517 November 1990 Brown Grass 0.814 -0.381 Green Herb 0.257 0.119 Seed 0.153 -0.067 Brown Rush -0.708 -0.516 December 1990 Green Grass -0.120 0.992 Cover Height -0.815 -0.443 Brown Rush -0.777 -0.457 32 percentage green herb cover. Medium abundance vole populations (grid 4) were associated with large percentage brown grass and low cover height (but higher than grid 2). During October 1990 large vole populations (grids 2 and 4) were associated with large percentage brown grass cover and low cover height. During Nov 1990 the largest vole population (grid 4) was associated with large percentage brown grass and green herb cover and small percentage seed cover. Medium abundance vole populations (grids 2 and 5) were associated with either characteristics similar to grid 4 or with large percentage brown rush cover and large percentage seed cover. During December 1990 the largest vole population (grid 4) was associated with medium cover height and large percentage brown grass cover. Medium abundance populations (grids 2 and 5) were associated with either medium cover height and medium percentage brown grass cover or high cover height, medium percentage brown grass cover, and large percentage brown rush cover. Discriminant models were relatively effective at correctly predicting vole abundance each month (Table 5). Mean correct classification was 70.03%, whereas 25% would be expected if classification rates were random. Vegetation characteristics explained a significant amount of variation in vole abundance during all population phases in grids that supported large populations (Table 6). Grids that had larger vole populations increased in 33 Table 5. Evaluation of the ability of Stepwise discriminant analysis models to predict Microtus californicus abundance based on monthly vegetation characteristics in small mammal grids. Observations were considered correctly classified if they were classified into the grid they came from or a grid with similar M. californicus abundance (similar = within 5 individuals). Fay Slough Wildlife Area, Eureka, California, June 1989 - December 1990.

Percent classified Mean % classified to grid with similar Month to original grid abundance

June 1989 50.0 77.50 July 1989 58.1 71.25 August 1989 63.1 81.25 September 1989 66.9 77.68 October 1989 N/A N/A November 1989 70.0 81.67 December 1989 59.4 82.50 January 1990 56.3 58.13 February 1990 53.8 56.40 March 1990 61.3 69.38 April 1990 58.7 100 May 1990 52.5 55.00 June 1990 61.3 87.50 July 1990 54.7 84.38 August 1990 56.3 75.63 September 1990 58.7 68.75 October 1990 64.8 75.62 November 1990 61.6 64.18 December 1990 62.5 63.75 Table 6. Stepwise-regression analysis results using Microtus californicus abundance (naive) as the dependent variable and vegetation parameters as independent variablesa. Analysis include increase phases in abundance of M. californicus, decrease phases, and all data regardless of population phase. Only variables entered into models are presented. Results where no variables were entered into models are not included.

Adjusted Grid/phase Model P Intercept b (variable) r2 n

GRID 2 Increase <0.05 147.89 -1.215 (seed), -3.331 (CH) 0.723 7 Decrease <0.0025 33.70 4.062 (seed), -2.745 (CH) 0.768 8 All Data <0.01 -10.05 0.657 (BG) 0.352 17 GRID 3 Increase <0.001 11.93 -1.366 (GH), 0.097 (seed) 0.682 14 GRID 4 Increase <0.001 79.38 -2.054 (CH) 0.894 11

Decrease <0.05 -5.87 0.741 (BG), -61.531 (GH), 0.996 5 1.536 (seed All Data <0.001 -8.22 19.248 (BH), 0.348 (BG) 0.781 17 a CH = Cover Height, GH = Green Herb, BG = Brown Grass, BH = Brown Herb. 34

35 abundance when cover height decreased or when cover height and seed cover decreased (Table 6). Grid 3 had a much smaller population of voles than all other grids during most months, and voles increased in abundance when green herb cover decreased and seed cover increased (Table 6). Partitioning population phases within each grid improved the performance of regression models (r2) over using the entire data set (Table 6). Reithrodontomvs megalotis Harvest mice populations fluctuated annually in grids 2 - 5 (Fig. 9). No individuals were caught in grid 1. Grid 2 was the first grid in which harvest mice were captured, with the population fluctuating from 0 to 4 individuals from July through October 1989. After that time the population increased to a high of 11 individuals during November 1989. The first harvest mice were caught in grid 4 during October 1989. After that the population followed the same general pattern as on grid 2, but reached 21 individuals in November 1989 (Fig. 9). These "early" harvest mice populations were associated with a large percentage brown grass cover and low cover height (Figs. 4 and 8). During these increases in abundance, between September and November 1989, grass seed cover (on grasses) declined from 39.8% to 11% and 56% to 24.5% on grids 2 and 4, respectively. Increasing populations were significantly correlated with brown grass cover (inversely) and brown herb cover, while decreasing 36

Figure 9. Number of individual western harvest mice (Reithrodontomys megalotis) caught within each trapping grid, each month on the Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990. 37 populations were significantly correlated with brown herb cover on grid 2 (Table 7). The harvest mouse population on grid 3 generally increased from December 1989 through April 1990 then declined to 0 individuals in June 1990. Increases in populations on grid 3 were significantly, inversely, correlated with brown grass cover, but no significant correlations between vegetation parameters and harvest mouse abundance were found during decline phases or when all data from grid 3 were used (Table 7). Increasing populations on grid 4 were significantly, inversely, correlated with brown grass cover and seed cover, while decreasing populations were significantly correlated with seed cover and green herb cover (inversely). The populations in grid 5 followed the same general pattern as grid 3. Increases in harvest mouse populations on grid 5 were significantly, inversely, correlated with seed cover (Table 7). Harvest mouse populations on grids 3 and 5 were increasing at the same time that populations on grids 2 and 4 were decreasing in abundance. These "late" populations were associated with a large percentage grass cover and a high cover height relative to grids 2 and 4. Although cover height was higher in grids 3 and 5 at this time, it was lower than it had been in these grids during preceding months (Fig. 8). The second increase in harvest mouse populations occurred on all grids between September and November 1990 (Fig. 9). Harvest mouse populations peaked in either November or December of 1990 Table 7. Stepwise-regression analysis results using Reithrodontomys megalotis abundance (naive) as the dependent variable and vegetation parameters as independent variablesa. Analysis include increase phases in abundance of R. megalotis, decrease phases, and all data regardless of population phase. Only variables entered into models are presented. Results where no variables were entered into models are not included.

Adjusted Grid/phase Model P Intercept b (variable) r2 n

GRID 2 Increase <0.025 51.36 -0.840 (BG), 1.559 (BH) 0.641 9 Decrease <0.025 -0.03 1.140 (BH) 0.532 8 All data <0.05 1.02 1.390 (BH) 0.234 17 GRID 3 Increase <0.025 6.21 -0.106 (BG) 0.380 14 All data <0.025 6.81 -0.125 (seed) 0.300 17 GRID 4 Increase <0.025 63.69 -0.691 (BG), -0.184 (seed) 0.884 6 Decrease <0.05 0.94 0.898 (seed), -6.408 (GH) 0.754 7 All data <0.025 -0.295 6.582 (BH) 0.280 17 GRID 5 Increase <0.05 13.70 -0.444 (seed) 0.411 10 All data <0.0025 13.35 -0.386 (seed) 0.585 17 a BG = brown grass, BH = brown herb, GH = green herb. 38

39 in all grids. These population increases occurred within 2 months of a drastic decline in percentage cover of grass seeds on all grids (mean % decrease between August and October 1990 = 31.5, SE = 4.76) (Fig 7). Stepwise-discriminant analysis models accurately predicted harvest mouse abundance based on vegetation characteristics each month (mean % correctly classified = 88.2) except during April 1990 (Table 8). All classification rates were much greater than expected by chance (25%). Generally, only weak positive correlations between vole and harvest mouse abundances were found (r = 0.32, 0.06, and 0.01, for grids 2, 3, and 5 respectively, P > 0.05 for all grids. A significant positive correlation existed in grid 4 (r = 0.55, P = 0.016). 40 Table 8. Evaluation of the ability of stepwise discriminant analysis models to predict Reithrodontomys megalotis abundance based on monthly vegetation characteristics in small mammal grids. Observations were considered correctly classified if they were classified into the grid they came from or a grid with similar R. megalotis abundance (similar = within 5 individuals). Fay Slough Wildlife Area, Eureka, California, June 1989 - December 1990.

Percent classified Mean % classified to grid with similar Month to original grid abundance

June 1989 100 100 July 1989 100 100 August 1989 100 100 September 1989 100 100 October 1989 100 100 November 1989 70.8 70.8 December 1989 59.4 76.3 January 1990 56.3 84.4 February 1990 53.8 91.9 March 1990 61.3 80.6 April 1990 58.7 62.5 May 1990 52.5 83.1 June 1990 100 100 July 1990 100 100 August 1990 56.3 79.4 September 1990 100 100 October 1990 64.8 90.0 November 1990 61.6 82.5 December 1990 62.5 73.8 DISCUSSION Microtus californicus Lack (1954) suggested that rodent population cycles were a result of rodents increasing in numbers until their food supply was depleted, leading to starvation, population declines, plant recovery, and subsequent rodent population increases. My findings are consistent with the hypothesis of resource limitation as a primary factor in regulating vole populations. Stepwise-regression models, using vegetation characteristics, explained 76.8% and 99.6% of the variation in abundance of M. californicus during population declines on grids 2 and 4 respectively (Table 6). Seed cover was the common feature to both models and was positively related to decreasing populations of M. californicus (Table 6). Though I did not experimentally manipulate vegetation characteristics, the natural phenology of the vegetation allowed analysis within and among grids. Although methods differed from those in this study, many studies have focused on the effect(s) of food (Taitt and Krebs 1981, 1983, Ford and Pitelka 1984, Desy et al. 1990) and cover (Birney et al. 1976, Taitt and Krebs 1983) on microtine populations and/or behavior. Each of those studies reported increased microtine densities in food supplemented areas and/or areas with greater cover. My study supports the assumption as well. There was a

41 42 correspondence of the vole population increase and decrease phases to the availability of seeds (on all trapping grids) during both years; some of these correlations were highly significant (Table 6). It is important to note that percentage seed cover measurements were estimates of seed abundance on grasses. This is significant because during the decline in percentage seed cover, through summer, seeds were becoming more available to small mammals. Batzli and Pitelka (1970) reported that seeds formed 70% of the diets of M. californicus. I interpret the initial low seed cover periods (approx. October - December) as being high seed availability periods for small mammals, whereas the later low seed cover periods (January - April) were likely periods when few seeds were available to small mammals because they had either germinated or were consumed. Small mammal populations on all grids declined dramatically between February and April 1990. Increased cover, per se, had no detectable positive effect on population size within grids; in fact, it was significantly negatively correlated with population size in grids 2 and 4 (Table 6). Birney et al. (1976) noted that it is what cover provides (i.e. not cover alone) that is important to cycling Microtus populations. I found the largest population in the grid that had the lowest cover height, which supports their interpretation. The fact that trapping grids with lower cover consistently had larger vole 43 populations is in contrast to the findings of Birney et al. (1976) (see their Fig. 3), but consistent with the hypothesis of food quantity/quality being important in regulating populations. Ford and Pitelka (1984) found significant positive correlations between peak standing crop of vegetation and both proportion of females breeding and rate of population change. The seemingly contrasting findings of Ford and Pitelka (1984) and mine are probably due to differences in vegetation height between the 2 study areas. The vegetation in their study area was much shorter than in my study area. Their peak standing crop of vegetation was < 10 cm (measured with a point frame) while only 1 of the 72 monthly cover height estimates was < 10 cm during my study. Furthermore, cover height is an estimate of the height of 100% cover (horizontally) whereas point frames measure the height that the point touches vegetation, so it is likely that the lowest standing crop of vegetation in my study was still greater than the peak standing crop in Ford and Pitelka's (1984). In this study, it is likely that vegetation cover was never below the minimum or "threshold" level (Birney et al. 1976) that M. californicus probably require.

Taitt and Krebs (1985) reviewed much of the work done on Microtus population dynamics and cycles in North America. They estimated that of 106 yrs of data, 59% involved annually fluctuating Microtus populations while 41% 44 were parts of cycles. Furthermore, Taitt and Krebs (1985) reported pronounced declines in abundance during spring in all populations that fluctuated annually. My findings followed these general "rules". The drastic nature of vole cycles in this study were somewhat surprising. Ostfeld and Tamarin (1986) suggested that in relatively aseasonal environments (i.e. those with small variations in temperature and precipitation) voles should cycle less dramatically or not at all because food resources would probably not fluctuate as they do in strongly seasonal environments. Support for this hypothesis comes from studies by Ostfeld and Klosterman (1986) and Krohne (1982); both studied M. californicus along the north- central California coast and found populations to be stable relative to other areas. Ostfeld and Tamarin (1986) suggested that voles perceive those habitats as aseasonal. The north-central coast areas are much less seasonal than those where most M. californicus work has been done (near San Francisco Bay), but more seasonal than my study area. Ostfeld and Tamarin (1986) reported that the presence of perennial green vegetation is critical to support breeding in M. californicus and that summer fog (which is characteristic in coastal California) is sufficient to support grass growth. Voles on my study area fluctuated annually and all populations exhibited a severe decline during spring. In fact, these are among the most dramatic 45 fluctuations in abundance ever reported for the species (see Taitt and Krebs 1985). Though abundance between grids varied considerably within months, the peak density that I found (1414 individuals/ha) is the highest ever reported for an annually fluctuating, natural, M. californicus population. Thus, my findings do not support Ostfeld and Tamarins' (1986) hypothesis. Instead, I suggest that areas with aseasonal climates can still have seasonal variability in vegetation structure and composition. Plants respond to both temperature and photoperiod, thus both should be considered when defining seasonality. It is generally assumed that microtine populations with female-biased sex ratios will be larger than those with male-biased sex ratios because of their potentially larger reproductive output (Keller 1985). Lidicker (1973) reported female-biased sex ratios during population increases followed by an equal sex ratio in M. californicus. Populations on 3 of my trapping grids similarly exhibited female-biased sex ratios during the population increases, with no consistent pattern between grids during the remainder of the cycle. Ostfeld et al. (1985) never found male M. californicus to constitute > 50% of populations (in 4 trapping grids). They also found that grids in higher quality habitat had a more pronounced female-biased sex ratio. Contrary to those findings, grid 3 in my study had 46 the fewest voles and a consistently higher female-biased sex ratio than the other 3 grids. Furthermore, grids 2 and 4 (the 2 "highest quality" grids) had male-biased sex ratios in 6 of 14 and 7 of 15 months when animals were caught, respectively. My results suggest that within-grid sex ratios may not be good predictors of grid quality. Local populations can increase from immigration, reproduction, or both. Early phase populations in my study increased initially from immigration, then (apparently) from local reproduction. I presumed that subsequent immigration, emigration, mortality, and reproduction contributed to the changing sex ratios. Many researchers have reported increased survival during increasing and peak population phases of Microtus spp. (Krebs 1966, Krebs et al. 1973, and others). Similarly, I found populations in all grids (regardless of absolute abundance) to have increased survival probabilities during the population increase and stable phases. It appears that resource abundance/availability helped to improve survivorship during population increase and stable phases. I interpreted decreased survivorship during the population decline phase to be a function of large populations decreasing the quantity/quality of critical vegetation resources. Andersson and Jonasson (1986) were unable to accept a food-quantity shortage hypothesis to explain population 47 declines of Clethrionomvs rufocanus or M. oeconomus in Sweden. They also reported a cessation of reproduction in voles when biomass of green herbaceous vegetation was at its peak, and suggested that predation could have been responsible for population declines (though predation was not quantified). My findings were similar, with population declines and lows during periods when green grass cover, green herbaceous cover, and cover height were greatest. It is difficult to believe that forage quantity was diminished in my study area during the population decline, but the extremely large populations of voles may have been able to deplete forage resources that were available to them. That would be difficult to demonstrate without a manipulative experiment. I suggest that microtine caused, decreased forage quality may be the critical factor causing population declines. Further evidence for this hypothesis comes from the fact that the 2 grids with the largest populations (grids 2 and 4) had no individuals for 4 and 3 months, respectively, after the population decline in 1990. At the same time, grids 3 and 5 had no individuals for only 1 month. Population recovery took 3 - 4 times longer on grids 2 and 4, suggesting that larger populations depleted a critical resource (or resources) more severely than smaller populations, and that resource recovery was a function of severity of depletion. 48 I did not study the effects of predation or secondary compounds on vole populations. My findings are not entirely inconsistent with either of those hypotheses. Research designed to specifically address those hypotheses will be necessary to adequately understand their influence(s) on vole populations. Reithrodontomys megalotis Largest numbers of harvest mice within each grid were associated with lower cover height of non woody vegetation, a finding in contrast to those of other researchers (Abramsky et al. 1979, O'Farrell and Clark 1986, Ribble and Samson 1987, Kaufman et al. 1988). This was likely due to the fact that, unlike the above studies, my grids were all located on relatively undisturbed areas and/or that those studies were conducted in arid or semi- arid environments, whose productivity was much less than on my study area. Similar to my findings, Kaufman et al. (1988) found R. megalotis populations increased during periods of peak seed abundance. Some researchers have suggested that R. megalotis is a socially subordinate species in small mammal communities (Blaustein 1980, Spevak 1983). Most authors examining potential competitive interactions between R. megalotis and other larger rodents in mammalian communities have reported negative correlations between abundance or reproductive activity of R. megalotis and abundance of potential 49 competitors (e.g., Blaustein 1980, Heske et al. 1984, Johnson and Gaines 1988, Swihart and Slade 1990). My findings of only positive correlations between abundance of voles and harvest mice are contrary to the above studies, although harvest mouse populations (with the exception of April and May 1990) were always much smaller than those of voles. The 2 months that R. megalotis populations were larger (coinciding with the vole annual population low) could be interpreted as comptetive release. I suggest that the much smaller harvest mice needed fewer food resources than voles and that when vole populations declined, sufficient food existed (for a short time) for harvest mice. Although R. megalotis habitat requirements rarely have been studied, some authors have quantified vegetation species composition in areas inhabited by R. megalotis (e.g., Ribble and Samson 1987). Other researchers have compared R. megalotis abundance among habitats such as sage, greasewood, and marsh-meadow (O'Farrell and Clark 1986) or forested and meadow habitats (Ribble and Samson 1987). My ability to correctly predict R. megalotis abundance (88% correct classification) based on vegetation variables within a relatively homogeneous habitat supports the contention that it is a microhabitat specialist as suggested by Ribble and Samson (1987). These findings are not conclusive, and will require further study to refute or corroborate, especially if they are to be applied to other areas. TERRITORY SIZE REGULATION IN BLACK-SHOULDERED KITES INTRODUCTION The relationship between raptors and their prey has been studied extensively (e.g., Craighead and Craighead 1956; Village 1982, 1987; Korpimaki 1984; 1985a,b, 1988; Newton et al. 1986). Most raptor-prey studies have either focused on raptor abundance (Baker and Brooks 1981, Village 1982, Cully 1991) or raptor reproductive success (Hammerstrom 1979; Smith et al. 1981; Korpimaki 1986, 1988; Ridpath and Brooker 1986; Ward 1990) in relation to prey abundance. Theoretical predictions of territory size include that territory size is inversely related to food abundance and biomass (Schoener 1968). Furthermore, Myers et al. (1979) presented two hypotheses regarding territory size regulation. First, individuals establish territories of a size that contain adequate resources to meet their energetic needs. Second, an individual will defend an area as large as it can, constrained by competition with other individuals. These hypotheses are not mutually exclusive, though they have generally been viewed as such (Myers et al. 1979).

Researchers of insectivorous birds often confront the difficult problem of accurately estimating arthropod abundance within an area (Hutto 1990, Recher 1990).

50 51

Although raptor biologists seemingly have a lesser task at hand when estimating prey abundance (usually 1-3 species make up the bulk of the diet), most raptor species have extremely large home ranges or territories, which present major logistical constraints in both estimating territory or home range size and prey abundance within those areas. The result is that prey abundance is often sampled sporadically or extrapolated from a few trap lines or grids to areas of several km2 (Phelan and Robertson 1978, Smith et al. 1981, Ridpath and Brooker 1986, Simmons et al. 1986, and others). In general, prey sampling effort often receives inadequate attention in studies that are primarily focused on describing or predicting raptor-prey relations. Understanding an animal's space use under dynamic environmental conditions is central to effective species/population conservation, management, and recovery. I studied the relationship between the size of hunting territories of the black-shouldered kite (Elanus caeruleus) and prey abundance and competitor abundance. Kites are good subjects for such a study because: 1) they have distinctive hunting habits (hovering), which allows unambiguous description(s) of areas/points used for foraging; 2) they use open to semi-open habitats, which facilitates long continuous observation periods; 3) they have relatively small home ranges or territories (Henry 1983); and 4) they rely exclusively on rodents as prey (Waian and Stendell 52

1973). They prey primarily on California voles (Microtus californicus) in California (Stendell 1972, Bammann 1975), so estimating prey abundance is an easier task than it would be with a more catholic predator. METHODS Kites were captured with a selective pole trap (Dunk 1991), then banded with uniquely colored leg-bands with vinyl tabs. Kite hunting territories were estimated opportunistically from June 1989 - November 1990. Each territory was delineated by observing a kite and documenting all locations where the bird hovered, perched, or interacted with another kite or raptor. Locations were estimated by measuring distance with an optical range finder and direction with a compass from fixed observation spots. The accuracy of the range finder was +/- 1% at 100 m and +/- 10% at 1000 m. For each location, distance, direction, and time (to the nearest 0.05 min) were recorded on a micro-cassette recorder. Each kite was observed for a minimum of 2 hrs or until the bird left the area to roost. Total observation time per territory ranged from 2 - 15 hours over a 1 - 7 day period. Mean number of locations per territory was 64.5 (SD = 24.2). Adequacy of sampling was determined graphically by plotting hunting territory size (y axis) and number of locations (x axis) until reaching an asymptote. Hunting territory size was estimated using the 95% minimum convex polygon estimator in program HOME RANGE (Ackerman et al. 1989). I attempted to delineate as many hunting territories as possible. I did not randomly choose individuals to

53 54 observe, but observed individuals as follows: 1) color banded but territory not previously measured; 2) previously unmeasured territory with unbanded individual(s) that were repeatedly associated with a specific geographic area prior to the sampling time; and 3) color banded individuals whose territories had been estimated in a prior month. Unbanded birds were identified using one or more of the following criteria: 1) unique molting patterns; 2) perches used; 3) association with (i.e., mates of) banded kites; and 4) whether adult or juvenile. Within 2 weeks of delineating each hunting territory, I sampled the vegetation structure and composition within it. After plotting territory boundaries on a 1:2400 aerial photograph, I established 100 vegetation plots in a random systematic design in each territory, in order to have complete coverage of the territory. In each plot I measured percentage of the plot covered by: 1) green grass (Graminaea); 2) brown grass; 3) green herbaceous vegetation; 4) brown herbaceous vegetation; 5) seed heads on grasses (not exclusive of grass cover); 6) green rush (Cyperaceae); 7) brown rush. These were visually estimated and placed into 1 of 6 cover classes: 0%, 1 - 20%, 21 - 40%, 41 - 60%, 61 - 80%, or 81 - 100%. Parameters 6 and 7 were estimated beginning in December 1990. Cover height, an estimate of the height at which there was 100% cover horizontally, was also measured in the center of each plot 55 by pushing a meter stick through the vegetation, flush to the ground, and recording the height (cm) of the intersection between plot center and the meter stick. Plots falling on flooded areas were classified as flooded with no vegetation characteristics estimated within them. Potential competitors of kites were counted each month within 14 days of estimating each territory. From a central location, I counted the numbers of each raptor species observed by scanning the entire study area (approximately 95% of the area could be viewed). Counts were terminated when 3 consecutive counts yielded identical results. All counts were made within 1 hr of sunset because previous observations showed that raptors in the area were most active at this time. Barn Owls (Tyto alba) were counted by looking in the 2 barns on the study area. M. californicus were trapped within 4 trapping grids each month. All grids consisted of 40 traps spaced by 5 m and arranged in an 8 x 5 pattern. Traps were locked open for 24 - 36 hrs prior to being set each month. All traps were then baited with bird seed and polyester pillow stuffing, and set. Traps were checked twice daily for 4 consecutive days. At each capture location individuals were identified to species, uniquely toe-clipped, weighed (g), sexed, and released (see methods of small mammal - vegetation relationships for a more detailed description of small mammal trapping). Within one week of trapping, 56 vegetation was sampled within each trapping grid each month. The same parameters were estimated as in kite hunting territories. To estimate vole abundance within each kite territory I categorized each vegetation plot within a territory as being of high, medium, or low quality habitat for voles. This was done after modelling vole-vegetation associations using stepwise discriminant analysis each month (see methods of small mammal - vegetation relationships). This procedure was effective at predicting vole abundance. High, medium, and low quality vole habitats were defined relative to the largest number of individuals caught within a month (e.g., high quality habitat in June may have been 100 ind/ha whereas in September it could have been 900 ind/ha). The percentage of plots in each habitat quality category was multiplied by territory area, then by naive density (# individuals caught/grid area) of voles for that category. These were added together to estimate vole abundance for each territory. Three methods of relating kite territory size to vole abundance were used: 1) regressing territory size on a monthly index of vole abundance expressed as number of individuals caught per 640 trap nights for all grids combined; 2) regressing territory size on the estimate of the number of vole within each territory; 3) regressing territory size on the estimate of vole density within each 57 territory. I also examined the effects of potential competitor abundance on kite territory size by regressing territory size on the estimate of competitor abundance for the month closest to the time the territory was delineated. I used multiple regression analysis and partial correlation analysis to simultaneously examine the effects of food abundance and competitor abundance on territory size. Partial correlation analysis examines the effects of each independent variable while statistically controlling for the effects of variables already entered into the multiple regression model (Neter et al. 1989). RESULTS

In all, 26 kite hunting territories were estimated; three individuals' territories were estimated more than once. Territory size ranged from 1.6 to 21.5 ha. Mean number of voles per territory was 1483 (SE = 163, n = 25). Vole density within kite territories ranged from 0/ha to 602/ha (mean = 277.8, SE = 33.2). Potential competitor abundance ranged from 7 - 45 individuals (mean = 28.9, SE = 5.7, n = 26, Table 9), or 4.8 to 31.0 individuals per km2. The correlation between kite territory size and the estimate of prey abundance within each territory was not significant (r = 0.23, P = 0.261, n = 25). Kite territory size was negatively correlated with both total raptor abundance and the index of prey abundance (r = -0.78 and -0.75, respectively, P < 0.001 for both). Territory size also was negatively correlated with the estimated density of voles within each territory (r = -0.71, P < 0.001, n = 25). A significant negative correlation was found between territory size and abundance of conspecifics (r = -0.64, P < 0.001, n = 26). I found a highly significant correlation between total raptor abundance and the index of vole abundance each month (all grids combined) (r = 0.90, P < 0.001, n = 19). Three stepwise regression models were built. First, 58 Table 9. Monthly raptor species abundances at the Fay Slough Wildlife Area, Eureka, California. June 1989 - December 1990.

Month Species Ju Ju Au Se Oc No De Ja Fe Ma Ap Ma Ju Ju Au Se Oc No De black-shouldered kite 8 11 15 17 20 17 22 19 14 14 5 4 8 7 10 15 15 12 12 (Elanus caeruleus) red-tailed hawk 2 2 3 3 3 2 3 3 3 3 1 2 2 3 2 2 2 3 2 (Buteo jamaicensis) red-shouldered hawk 1 1 2 2 3 3 2 3 1 1 0 2 1 2 2 2 2 3 2 (B. linneatus) rough-legged hawk 0 0 0 0 2 1 1 2 0 0 0 0 0 0 0 0 0 0 0 (B. laqopus) northern harrier 2 3 4 5 7 7 7 7 8 5 1 1 2 3 3 3 5 6 4 (Circus cvaneus) American kestrel 0 0 2 3 3 2 4 4 4 3 0 0 0 0 0 1 2 3 3 (Falco sparvarius) short-eared owl 0 0 0 0 0 3 5 6 6 4 0 0 0 0 0 0 0 6 8 (Asio flammeus) barn owl 0 0 1 1 1 1 1 1 1 1 0 0 0 0 0 1 1 1 2 (Tyto alba)

Total 13 17 27 31 39 36 45 45 37 31 7 9 13 15 17 24 27 34 33 59 60 I regressed the estimate of vole density within kite territories and competitor abundance (total raptors) against kite territory size. Only competitor abundance was entered into this model (r = -0.76, P < 0.001, n = 25). Partial correlation revealed that once competitor abundance was controlled statistically, prey density did not explain a significant amount (r = -0.25, P > 0.20, n = 25) of the remaining variation in territory size (Table 10). When I forced both variables to enter the model, the partial correlation of competitor abundance remained significant (r = -0.45, P < 0.05, n = 25) when prey density was statistically controlled. Second, I regressed the estimate of prey density and kite abundance (conspecific competitors) against kite territory size. Prey density was the only variable entered into the model this time (r = -0.71, P < 0.001, n = 25). When I forced both variables to enter the model (kite abundance was entered first), partial correlation revealed prey density to continue to be significantly correlated with kite territory size (r = -0.47, P < 0.05, n = 25, Table 10). Lastly, I regressed the index of vole abundance and total raptor abundance against kite territory size. Raptor abundance was the only variable entered into this model (r = -0.780, P < 0.001, n = 26). When both variables were forced into the model, partial correlation revealed that raptor abundance continued to be significantly correlated with kite Table 10. Partial correlation analysis results using black-shouldered kite territory size as the dependent variable and competitor abundance and prey abundance as independent variables.

Independent Partial Adjusted variables r correlation P variable r2 Model P n

Raptor abundance -0.780 -0.446 <0.02 0.613 <0.001 26 Index # Microtus -0.746 -0.305 0.10 < 0.20 Microtus density -0.711 -0.468 <0.02 0.380 <0.001 25 Kite abundance -0.637 -0.246 0.20 < 0.50 Raptor abundance -0.762 -0.448 <0.05 0.569 <0.001 25 Microtus density -0.711 -0.246 0.20 < 0.50 61

62 territory size (r = -0.446, P < 0.02, n = 26, Table 10). DISCUSSION At least three other investigators examined the relationships among food abundance, competitor abundance, and territory size of birds using approaches similar to mine (partial correlation analysis). Myers et al. (1979) found that once the interaction between prey density and intruder (competitor) density was controlled statistically, prey density had no significant effect on territory size of sanderlings (Caladris alba). Those findings are identical to mine when all raptors are considered as competitors. Conversely, McFarland (1986) and Temeles (1987) found that food supply, not intruder pressure, determined territory size of New Holland honeyeaters (Phvlidonvris novaehollandiae) and northern harriers (Circus cvaneus), respectively. I propose that kite territory size was proximately regulated by competitor abundance and ultimately regulated by prey abundance because raptor abundance was regulated by vole abundance. Myers et al. (1979) reached the same conclusion about sanderling territory size. Within the beaches that sanderlings inhabited, food abundance changed rapidly. Myers et al. (1979) suggested that sanderlings should exhibit one of two behavioral responses: 1) adjust territory size to the present prey abundance or 2) establish a large enough territory to encompass enough prey to support

63 64 them through periods of low prey abundance. They argued that sanderlings exhibited the latter. Kites exhibited aggressive behavior towards many of the raptor species listed in Table 9. Although I did not examine numbers of interactions between territory owners and competitors, competitor abundance may have been related positively to number of interactions between owners and competitors. Similar relationships between territory size and (total) competitor abundance were found in my study and that of Myers et al. (1979) (r = -0.78 for kites and -0.72 for sanderlings). Kites apparently defend as large a territory as possible, constrained by competition with other raptors. My findings show that kites can and do adjust territory boundaries with respect to current local conditions. My results seem somewhat contradictory depending on how competitors are defined. When all raptors were treated as competitors, competitor abundance but not prey abundance influenced territory size of kites. Conversely, when only kites were considered as competitors, prey abundance but not competitor abundance influenced territory size. During 19 months of small mammal trapping only 4 species of small mammals were captured: M. californicus, western harvest mice (Reithrodontomys megalotis), house mice (Mus musculus), and vagrant shrews (Sorex vagrans). M. californicus were consistently the most abundant small mammal and constituted even more of the small mammal biomass. This along with my 65 findings that M. californicus abundance explained 81% of the variation in total raptor abundance strongly suggests that they were the major food source for almost all of the raptors on my study area. Thus, I considered all species that were using the same food sources to be competitors, as opposed to only conspecifics. One of my most striking findings was the temporal stability of the estimate of vole abundance within territories. Mean number of voles per territory was 1483 suggesting that kites need about 1500 voles within their territories. Kites should abandon territories that have much less than this number of voles. It appeared that prey abundance and prey availability were correlated in this instance. Microtus apparently need a threshold level of cover in order for large populations to survive (Birney et al. 1976). Thus, within areas of dense and tall vegetation kites "needed" about 1500 voles (abundance) so that some were available to foraging kites. Larger vole populations were found in areas with shorter vegetation (see results of small mammal - vegetation relationships), areas that probably simultaneously maximized their abundance and availability to raptors. I suggest that kites defend as large an area as is energetically feasible at any one time, but that increased competitor abundance reduces this feasibility. Several pieces of circumstantial evidence support this contention. 66 Raptors in the Arcata-Eureka area undergo regular annual fluctuations in abundance (Table 9), generally increasing in September - October then remaining relatively constant until March - April when numbers decrease dramatically (Table 9). The largest number of voles estimated to be in a kite territory was 3340 during mid August 1990 (territory size was 15.6 ha). That was a time when vole abundance was relatively large, yet many of the "winter resident" raptors had not arrived on the study area. Also, on one occasion I was able to document the precise day that a territory was abandoned by its owner; on that same day the entire territory was incorporated into the territory of a neighboring kite. Despite the importance of competitors in limiting territory size, the best overall predictor of raptor abundance and space use within an area may be food abundance. This is supported by an overwhelming number of raptor-prey studies that have reported positive correlations between raptor and prey abundances (Baker and Brooks 1981; Village 1982, 1987; Korpimaki 1985b; Cully 1990; and others). Other factors that influence raptor territory size include body mass and hunting style. Generally, larger species maintain larger territories (Newton 1979). Raptors generally adopt one or more of the following hunting strategies: 1) perch and pounce (e.g., red-tailed hawks in 67 many areas); 2) active pursuit (e.g., peregrine falcons [Falco peregrinus)); or 3) hovering (e.g., black-shouldered kites), which may affect territory size. For example, red- tailed hawk territories were positively correlated with abundance of perch sites as opposed to prey abundance (Janes 1984). Mendelsohn (1981) reported much larger territory sizes for black-shouldered kites (though prey abundances were not reported) in South Africa where they primarily perch hunt (Mendelsohn 1981). Because kites primarily hover hunt in California, I believe their relatively small territory size on my study area was a function of both prey abundance and the fact that they can hunt 100% of their territories. Previous researchers of black-shouldered kites have reported them to be both territorial and not territorial (Henry 1983). The birds I studied were territorial, although I only observed territorial birds. Hence, I am unable to address the role or existence of non-territorial kites. Mendelsohn (1981) suggested that kite territoriality would break down at some upper threshold of prey density, but found that South African kites abandoned territories most often when prey populations were low. It may be that prey populations on my study area never reached this lower threshold. Instead, my findings suggest that territories simply become smaller as a result of increasing competitor abundance (the proximate factor regulating territory size) 68 which is strongly correlated with food abundance (the ultimate factor controlling kite territory size). LITERATURE CITED

Abramsky, Z. 1978. Small mammal community ecology. Oecologia 34:113-123. ----, M. L. Dyer, and P. D. Harrison. 1979. Competition among small mammals in experimentally perturbed areas of the short grass prairie. Ecology 60:530-536. Ackerman, B. B., F. A. Leban, E. 0. Garton, and M. D. Samuel. 1989. User's manual for program Home Range. Second edition (draft copy). Tech. Rept. 15, Forestry, Wildlife and Range Experiment Station, Univ. Idaho, Moscow, ID. 81pp. Andersson, M. and S. Jonasson. 1986. Rodent cycles in relation to food on an alpine heath. Oikos 46:93-106. Baker, J. A. and R. J. Brooks. 1981. Raptor and vole populations at an airport. J. Wildl. Manage. 45:390-396. Bammann, A. R. 1975. Ecology of predation and social interactions of wintering white-tailed kites. M. S. Thesis, Humboldt State University, Arcata, CA. 81pp. Batzli, G. O. 1986. Nutritional ecology of the California vole: effects of food quality on reproduction. Ecology 67:406-412. ----. 1988. The role of nutrition and carrying capacity in models of small mammal population dynamics. Oikos 52:224-226. ----, and F. A. Pitelka. 1970. Influence of meadow mouse populations on California grassland. Ecology 51:1027- 1039. Bergeron, J. M. and L. Jodoin. 1987. Defining "high quality" food resources of herbivores: the case for meadow voles (Microtus pennsylvanicus). Oecologia 71:510-517. Birney, E. C., W. E. Grant, and D. D. Baird. 1976. Importance of vegetative cover to cycles of Microtus populations. Ecology 57:1043-1051.

69 Blaustein, A. R. 1980. Behavioral aspects of competition in a three-species rodent guild of coastal Southern California. Behay. Ecol. Sociobiol. 6:247-255. ----. 1981. Population fluctuations and extinctions of small rodents in coastal Southern California. Oecologia 48:71-78. Brownie, C., J. E. Hines, and J. D. Nichols. 1986. Constant parameter capture-recapture models. Biometrics 42:561-574. Boonstra, R. 1977. Predation on Microtus townsendii populations: impact and vulnerability. Can. J. Zool. 55:1631-1643. Craighead, J. J. and F. C. Craighead. 1956. Hawks, owls and wildlife. Stackpole Co. Pennsylvania. 443pp. Cully, J. F. Jr. 1991. Response of raptors to reduction of a Gunnison's prairie dog population by plague. Am. Midl. Nat. 125:140-149. Desy, E. A., G. 0. Batzli, and J. Liu. 1990. Effects of food and predation on behaviour of prairie voles: a field experiment. Oikos 58:159-168. Dixon, W. J. 1985. BMDP Statistical Software. University of California Press. 734pp. Dunk, J. R. 1991. A selective pole trap for raptors. Wildl. Soc. Bull. 19:208-210. Eadie, W. R. 1953. Response of Microtus to vegetative cover. J. Mammal. 34:263-264. Ford, R. G. and F. A. Pitelka. 1984. Resource limitation in populations of the California vole. Ecology 65:122- 136. Ford, S. D. 1977. Range, distribution and habitat of the Western harvest mouse, Reithrodontomys megalotis, in Indiana. Am. Midl. Nat. 98:422-432. Getz, L. L. 1985. Habitats. pp 286-309 in R. H. Tamarin ed. Biology of New World Microtus. Spec. Publ. Am. Soc. Mammal. No. 8. Hammerstrom, F. 1979. Effect of prey on predator: voles and Harriers. Auk 96:370-374.

70 71 Henry, M. E. 1983. Home range and territoriality in breeding white-tailed kites. M. S. Thesis. San Diego State Univ., San Diego, California. 132pp. Heske, E. J., R. S. Ostfeld, and W. Z. Lidicker, Jr. 1984. Competitive interactions between Microtus californicus and Reithrodontomys megalotis during two peaks of Microtus abundance. J. Mammal. 65:271-280. ----, and ----. 1988. Does social behavior drive vole cycles? An evaluation of competing models as they pertain to California voles. Can. J. Zool. 66:1153-1159. Hestbeck, J. B. 1986. Multiple regulation states in populations of the California vole, Microtus, californicus. Ecol. Mono. 56:161-181. Hornfeldt, B., 0. Lofgren, and B. -G. Carlsson. 1986. Cycles in voles and small game in relation to variations in plant production indices in Northern Sweden. Oecologia 68:496-502. Hutto, R. J. 1990. Measuring the availability of food resources. pp 20-28 in Morrison, M. L., C. J. Ralph, J. Verner, and J. R. Jehl Jr. eds. Avian foraging: theory, methodology, and applications. Stud. Avian Biol. No. 13. Janes, S. W. 1984. Influences of territory composition and interspecific competition on red-tailed hawk reproductive success. Ecology 65:862-870. Johnson, M. L. and M. S. Gaines. 1988. Demography of the western harvest mouse, Reithrodontomys megalotis, in eastern Kansas. 75:405-411. Jolly, G. M. 1965. Explicit estimates from capture- recapture data with both death and immigration-stochastic model. Biometrika 52:225-247. Jones, E. N. 1990. Effects of forage availability on home range and population density of Microtus pennsylvanicus. J. Mammal. 71:382-389. Kaufman, G. A., D. W. Kaufman, and E. J. Finck. 1988. Influence of fire and topography on habitat selection by Peromvscus maniculatus and Reithrodontomys megalotis in ungrazed tallgrass prairie. J. Mammal. 69:342-352. Keller, B. L. 1985. Reproductive patterns. pp 725-778 in R. H. Tamarin ed. Biology of new world Microtus. Spec. pub. Am. Soc. Mammal. No. 8. 72 Klatt, B. J. and L. L. Getz. 1987. Vegetation characteristics of Microtus ochrogaster and M. pennsylvanicus habitats in East-Central Illinois. J. Mammal. 68:569-577. Koplin, J. R. and R. S. Hoffman. 1968. Habitat overlap and competitive exclusion in voles (Microtus). Amer. Midl. Nat. 80:494-507. Korpimaki, E. 1984. Population dynamics of birds of prey in relation to fluctuations in small mammal populations in western Finland. Ann. Zool. Fenn. 21:287-293. ----• 1985a. Prey choice strategies of the kestrel Falco tinnunculus in relation to available small mammals and other Finnish birds of prey. Ann. Zool. Fenn. 22:91- 104. • 1985b. Rapid tracking of microtine populations by their avian predators: possible evidence for stabilizing predation. Oikos 45:281-284. ----, 1986. Diet variation, hunting habitat and reproductive output of the Kestrel (Falco tinnunculuc) in the light of the optimal diet theory. Ornis. Fenn. 63:84-90. ----. 1988. Effects of territory quality on occupancy, breeding performance and breeding dispersal in Tengmalm's owl. J. Anim. Ecol. 57:97-108. Krebs, C. J. 1966. Demographic changes in fluctuating populations of Microtus californicus. Ecol. Mono. 36:239-273. ----, M. S. Gaines, B. L. Keller, J. H. Myers, and R. H. Tamarin. 1973. Population cycles in small rodents. Science 179:35-41. Krohne, D. T. 1982. The demography of low-litter-size populations of Microtus californicus. Can. J. Zool. 60:368-374. Lack, D. 1954. The natural regulation of animal numbers. Oxford Univ. Press, Oxford, 343pp. Lidicker, W. Z. Jr. 1973. Regulation of numbers in an island population of the California vole, A problem in community dynamics. Ecol. Mono. 43:271-302. 73 McFarland, D. C. 1986. Determinants of feeding territory size in the New Holland honeyeater Phylidonyris novaehollandiae. Emu 86:180-185. Mendelsohn, J. M. 1981. A study of the blackshouldered kite Elanus caeruleus. Ph. D. Thesis. Univ. of Natal, Pietermaritzburg, South Africa. 285pp. Mossman, A. S. 1955. Light penetration in relation to small mammal abundance. J. Mammal. 36:564-566. Myers, J. P., P. G. Connors, and F. A. Pitelka. 1979. Territory size in wintering sanderlings: the effects of prey abundance and intruder density. Auk 96:551-561. Neter, J., W. Wasserman, and M. H. Kutner. 1989. Applied linear regression models. Irwin Inc. Homewood, IL. 667pp. Newton, I. 1979. Population ecology of raptors. Buteo Books, Vermillion, South Dakota. 399pp. ----, I. Wyllie, and R. Mearns. 1986. Spacing of sparrowhawks in relation to food supply. J. Anim. Ecol. 55:361-370. O'Farrell, M. J. and W. A. Clark. 1986. Small mammal community structure in Northeastern Nevada. Southwest. Nat. 31:23-32. Ostfeld, R. S. 1985. Experimental analysis of aggression and spacing behavior in California voles. Can. J. Zool. 63:2277-2282. ----. 1986. Territoriality and mating system of California voles. J. Anim. Ecol. 55:691-706. ----, and L. L. Klosterman. 1986. Demographic substructure in a California vole population inhabiting a patch environment. J. Mammal. 67:693-704. ----, and R. H. Tamarin. 1986. The role of seasonality in vole cycles. Can. J. Zool. 64:2871-2872. ----, W. Z. Lidicker, Jr., and E. J. Heske. 1985. The relationship between habitat heterogeneity, space use, and demography in a population of California voles. Oikos 45:433-442. Pearson, 0. P. 1966. The prey of carnivores during one cycle of mouse abundance. J. Anim. Ecol. 35:217-233. 74 1971. Additional measurements of the impacts of carnivores on California voles (Microtus californicus). J. Mammal. 52:41-49. Phelan, F. J. S. and R. J. Robertson. 1978. Predatory responses of a raptor guild to changes in prey density. Can. J. Zool. 56:2565-2572. Pitelka, F. A. 1964. The nutrient-recovery hypothesis for arctic microtine cycles. I. Introduction, Pp. 55-66 in Grazing in terrestrial and marine environments. D. J. Crisp, ed. Blackwell Sci. Publ., Oxford, 322pp. Pitcher, E. J. and B. L. Keller. 1979. Aspects of interspecific interaction between Reithrodontomvs megalotis and Microtus montanus in one acre enclosures. Res. Popul. Ecol. 21:81-104. Recher, H. F. 1990. Specialist or generalist: avian response to spatial and temporal changes in resources. pp 333-336 in Morrison, M. L., C. J. Ralph, J. Verner, and J. R. Jehl Jr. eds. Avian foraging: theory, methodology, and applications. Stud. Avian Biol. No. 13. Ribble, D. 0. and F. B. Samson. 1987. Microhabitat associations of small mammals in Southeastern Colorado, with special emphasis on Peromvscus (Rodentia). Southwest. Nat. 32:291-302. Ridpath, M. G. and M. G. Brooker. 1986. The breeding of the wedge-tailed Eagle Aquila audax in relation to its food supply in arid Western Australia. Ibis 128:177- 194. Schoener, T. W. 1968. Sizes of feeding territories among birds. Ecology 49:123-141. Simmons, R., P. Barnard, B. MacWhirter, and G. L. Hansen. 1986. The influence of microtines on polygyny, productivity, age, and provisioning of breeding northern harriers: a 5-year study. Can J. Zool. 64:2447-2456. Smith, D. G., J. R. Murphy, and N. D. Woffinden. 1981. Relationships between Jackrabbit abundance and ferruginous hawk reproduction. Condor 83:52-56. Spevak, T. A. 1983. Population changes in a mediterranean scrub rodent assembly during drought. Southwest. Nat. 28:47-52. 75 Stendell, R. C. 1972. The occurrence, food habits, and nesting strategy of white-tailed kites in relation to a fluctuating vole population. Ph. D. Thesis, University of California, Berkeley, Berkeley, California. 224pp. Swihart, R. K. and N. A. Slade. 1990. Long-term dynamics of an early successional small mammal community. Am. Midl. Nat. 123:372-382. Taitt, M. J. and C. J. Krebs. 1981. The effect of extra food on small rodent populations:II. Voles (Microtus townsendii). J. Anim. Ecol. 50:125-137. ---- and ----. 1983. Predation, cover, and food manipulations during a spring decline of Microtus townsendii. J. Anim. Ecol. 52:837-848. ---- and ----. 1985. Population dynamics and cycles. pp 567-620 in R. H. Tamarin ed. Biology of new world Microtus. Spec. Publ. Am. Soc. Mammal. No. 8. Tamarin, R. H. 1985. Ed. Biology of new world Microtus. Spec. pub. no. 8. Amer. Soc. Mammal. 893pp. Temeles, E. J. 1987. The relative importance of prey availability and intruder pressure in feeding territory size regulation by harriers, Circus cvaneus. Oecologia 74:286-297. Village, A. 1982. The home range and density of kestrels in relation to vole abundance. J. Anim. Ecol. 51:413- 428. ----. 1987. Numbers, territory-size and turnover of short-eared owls (Asio flammeus) in relation to vole abundance. Ornis. Scand. 18:198-204. Waian, L. B. and R. C. Stendell. 1970. The white-tailed kite in California with observations of the Santa Barbara Population. Calif. Fish and Game 56:188- 198. Ward, J. P. Jr. 1990. Spotted owl reproduction, diet and prey abundance in Northwest California. M. S. Thesis, Humboldt State University, Arcata, California. 70pp.