Citation
Population status of mountain plover in South Park, Colorado

Material Information

Title:
Population status of mountain plover in South Park, Colorado
Creator:
Pierce, Allison Katherine ( author )
Publication Date:
Language:
English
Physical Description:
1 electronic file (59 leaves) : ;

Subjects

Subjects / Keywords:
Plovers -- Colorado ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Review:
Mountain Plovers (Charadrius montanus) are migratory shorebirds of conservation concern that breed on grasslands and xeric tablelands along the western edge of the Great Plains. Here, I provide an updated status for a population of plovers breeding on high-elevation intermountain rangelands in South Park, Park County, Colorado. To estimate demographic changes over time, I estimated plover breeding density using distance sampling data collected from historic surveys conducted from 2000 to 2006 and from a survey I conducted in 2016. Mean density was highest in 2001 (12.8 plovers km-2, 95% CI = 6.57–21.7, 35% CV), over 4.5 times higher than the estimated mean density from 2016 (2.8 plovers km-2, 95% CI = 1.54–4.57, 31% CV). Estimates across years provide weak evidence for a negative trend in the density of Mountain Plovers in South Park. I also estimated the effects of time and weather on daily nest survival probability. Mean expected nest survival probability for 2015 and 2016 was 0.22. Daily minimum temperature best predicted variation in daily nest survival in South Park; survival odds declined with decreasing temperature. This finding, coupled with reported negative effects of maximum temperatures on survival of nests at lower elevations in Colorado suggests that extreme temperatures may be a limiting factor for plover nest survival.
Thesis:
Thesis (M.S..)--University of Colorado Denver, 2017.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: Adobe Reader.
Statement of Responsibility:
by Allison Katherine Pierce.

Record Information

Source Institution:
University of Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
on10059 ( NOTIS )
1005933393 ( OCLC )
on1005933393

Downloads

This item has the following downloads:


Full Text
POPULATION STATUS OF MOUNTAIN PLOVER IN SOUTH PARK, COLORADO
by
ALLISON KATHERINE PIERCE
B.S., University of Colorado Denver, 2013
A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Master of Science Biology Program
2017


This thesis for the Master of Science degree by Allison Katherine Pierce has been approved for the Biology Program by
Michael B. Wunder, Chair Michael Greene
Stephen J. Dinsmore


Pierce, Allison Katherine (M.S., Biology Program)
Population Status of Mountain Plover Breeding in South Park, Colorado Thesis directed by Associate Professor Michael B. Wunder
ABSTRACT
Mountain Plovers (Charadrius montanus) are migratory shorebirds of conservation concern that breed on grasslands and xeric tablelands along the western edge of the Great Plains. Here, I provide an updated status for a population of plovers breeding on high-elevation intermountain rangelands in South Park, Park County, Colorado. To estimate demographic changes over time, I estimated plover breeding density using distance sampling data collected from historic surveys conducted from 2000 to 2006 and from a survey I conducted in 2016. Mean density was highest in 2001 (12.8 plovers km'2, 95% Cl = 6.57-21.7, 35% CV), over 4.5 times higher than the estimated mean density from 2016 (2.8 plovers km'2, 95% Cl = 1.54-4.57,31% CV). Estimates across years provide weak evidence for a negative trend in the density of Mountain Plovers in South Park. I also estimated the effects of time and weather on daily nest survival probability. Mean expected nest survival probability for 2015 and 2016 was 0.22. Daily minimum temperature best predicted variation in daily nest survival in South Park; survival odds declined with decreasing temperature. This finding, coupled with reported negative effects of maximum temperatures on survival of nests at lower elevations in Colorado suggests that extreme temperatures may be a limiting factor for plover nest survival.
The form and content of this abstract are approved. I recommend its publication.
Approved: Michael B. Wunder
m


DEDICATION
To my mother and grandmother, women of admirable strength; and to my partner life, Jesse Calkin, for his unconditional love and encouragement.


ACKNOWLEDGEMENTS
I am thankful for support and funding that was provided for this work by the University of Colorado Denver, Bird Conservancy of the Rockies, and the Denver Field Ornithologists. I would also like the thank the wonderful folks with Colorado Parks and Wildlife for providing logistical support, especially Mark Lamb and Karl Copeman. I am grateful to Angela Dwyer for her assistance with field logistics and funding. I thank Emery Young for his field assistance and company. I am also grateful to the late Fritz Knopf for enthusiastic encouragement he gave me in our correspondence during my first field season. His excitement for plovers instilled in me a great fondness for the species.
In addition, I thank my committee members for their advice, guidance, and encouragement with the development and execution of this work. I thank Mike Greene for his infectious enthusiasm and assistance in navigating logistics of proposed laboratory work. I thank Steve Dinsmore for his expertise not only in plover ecology but also in modeling techniques used in this project. And of course, I am profoundly grateful to my advisor and mentor Mike Wunder for continually challenging me to grow beyond the limits Ive set for myself and having faith in my success. I look forward to my next academic adventure working with you.
I am grateful to Libby Pansing and all my other wonderful student colleagues too numerous to name who helped form this work either through thoughtful feedback or being my sounding board when I needed it.
All fieldwork and animal handling protocols for this work were approved by the University Institutional Animal Care and Use Committee under protocol number: 92014(05)1C.
v


TABLE OF CONTENTS
CHAPTER
I. DENSITY TRENDS IN THE HIGH-ELEVATION POPULATION OF MOUNTAIN
PLOVERS IN COLORADO................................................... 1
Introduction...........................................................1
Methods................................................................3
Results................................................................7
Discussion.............................................................8
II. MOUNTAIN PLOVER NEST SURVIVAL IN A HIGH-ELEVATION
POPULATION............................................................16
Introduction..........................................................16
Methods...............................................................18
Results...............................................................23
Discussion............................................................25
REFERENCES..................................................................35
APPENDIX....................................................................38
A. Distance sampling model selection results for all years...........38
B. R output summaries for most parsimonious distance sampling models.40
C. Parameter estimates for all nest survival models..................44
D. Distance sampling analyses R code.................................46
E. Nest survival analyses R code.....................................50
vi


CHAPTER I
DENSITY TRENDS IN THE HIGH-ELEVATION POPULATION OF MOUNTAIN
PLOVERS IN COLORADO Introduction
The Mountain Plover is an uncommon upland shorebird that is considered a species of conservation concern in many of the states where it breeds and overwinters, due in part to apparent population declines (Colorado Parks and Wildlife 2015; Andres and Stone 2010). This plover breeds along the western edge of the Great Plains and winters in areas along the southern border of the United States and northern Mexico (Knopf and Wunder 2006). The global population size is estimated to range from 11000 to 14000, but may number up to 18000 (Plumb et al. 2005; Tipton et al. 2009; Andres and Stone 2010). Colorado is estimated to support more than half of the global population of breeding plovers (Tipton et al. 2009), thus, conservation of plovers in Colorado is of particular interest.
Anecdotal records suggest Mountain Plover population declines dating back to the early 1900s (Cooke 1914; Abbott 1939). More contemporary quantitative studies such as the Breeding Bird Survey estimate that continental populations have declined 3.1% annually from 1996 to 2012 (Sauer et al. 2012). These steep declines have corresponded with observed declines in the population on the Pawnee National Grassland (PNG)once considered the stronghold of the Colorado breeding population (Graul and Webster 1976; Knopf 1994; Knopf and Wunder 2006). The PNG decline is mainly attributed to altered grazing regimes of large herbivores and reduced populations of prairie dogs, both of which contribute to maintaining the bare ground and low vegetation habitats preferred by plovers.
1


Despite local PNG declines, comparatively large breeding populations in other areas of eastern Colorado and South Park, Park County remain. Much of the Colorado population breeds on the eastern plains (>8600 individuals). However, the breeding population of-2300 in South Park occurs at the highest density across the range, 7.90 0.90 (SE) plovers/ km2 (Wunder et al. 2003; Tipton et al. 2009), likely attributed to differences between breeding habitats. The high elevation intermountain grasslands in South Park are geographically separated and constrained by mountainous and forested habitat, whereas breeding habitats in eastern Colorado comprise a patchwork of short-grass prairie, black-tailed prairie dog colonies (Cynomys ludovicianus), and fallow agricultural fields spread out over -81,200 km2 (Tipton et al. 2009). Furthermore, South Park has a relatively milder climate and has experienced minor historic change compared to habitats on the eastern plains (Wunder et al. 2003).
Because of these differences between habitats, continental trends may not reflect local population dynamics and trajectories. Moreover, large scale roadside surveys often produce unreliable estimates for cryptic species with low detectability such as the Mountain Plover. Developing statistically robust surveys that cover the entirety of the range of a sparsely and variably distributed bird to accurately estimate the direction and magnitude of population change is financially and logistically unrealistic. However, long term monitoring of key local populations across space and time may be manageable, yet still informative about population dynamics of the species. Although previous research suggested the South Park population size was stable (Wunder et al. 2003; Dinsmore et al. 2010), the population size had not been estimated since 2006. The goal of this study was to provide an updated status of the population of Mountain Plover breeding in South Park, Colorado. Here I provide estimates
2


for density and abundance of plovers for part of the South Park population and directly compare them with previous estimates for the same area.
Methods
Study area
I studied Mountain Plovers breeding on public lands in South Park, Park County, Colorado (39 9' 19.296"N, 105 50' 51.72"W). Breeding habitat in South Park is comprised of high-elevation intermountain grassland characterized by short stature bunch grasses and small shrubs (see Wunder et al. 2003 for more detail). The primary area of interest of my study is located within a corridor of Mountain Plover habitat adjacent to a forested ridge in the South Park Basin (Fig. 1). This area comprises the western portion of the James Mark Jones State Wildlife Area (JMJSWA) as well as grazing pastures managed by the Bureau of Land Management (BLM). This area was selected for re-establishing a long-term study of Mountain Plover in South Park because it represents publically accessible habitat identified as supporting a high density of breeding plovers and was of high conservation value (Grunau and Wunder 2001).
Historic surveys
To evaluate the current population status of Mountain Plover in South Park, I used historic survey data collected from 2000 to 2006 to compare with estimates generated in 2016. Historically, Mountain Plover density and abundance were estimated using distance sampling along eight randomly generated transects in potentially suitable habitat on public and private lands throughout South Park (Wunder et al. 2003). This method models detection probability as a function of perpendicular distance of an object of interest from the transect line to adjust for imperfect detection during surveys (Buckland et al. 2001). Assumptions for
3


these models include (1) all birds on the line are detected with certainty, (2) birds are detected at their initial location, and (3) distances are measured accurately. Plovers are wary of surveyors on foot but are tolerant of slow moving vehicles. Therefore, to ensure satisfaction of the first two assumptions, surveys were conducted from a vehicle driven at <15 km hr'1. To maximize detectability, sampling was conducted early in the morning and during the brood rearing stage in July when plovers are conspicuous. When broods or small flocks were encountered, chicks and juveniles were easily identified and not included as detections in surveys to ensure estimates reflected the density of potentially breeding plovers. Perpendicular distances were derived from distances measured from the observer using a laser rangefinder and a standard compass to measure sighting angle (Wunder et al. 2003). Contemporary surveys
The study in 2016 was designed for consistency with historic surveys but with changes to increase estimate precision and spatial coverage. Buckland et al. (2001) recommend a minimum of 10-20 transects to reliably estimate encounter variance, more if species are patchily distributed. Mountain Plovers are loosely colonial and thus are unlikely to be randomly distributed (Graul 1975). I therefore systematically placed 38 short east-west parallel transects across the study area. Transects ranged in length from 0.6 km to 3.6 km and were spaced 500 m apart. We intended to survey all transects two times but were only able complete the second survey for 16 transects due to time constraints. Survey effort totaled 82 km and was conducted by a single observer (AKP).
Transects were placed in an east-west orientation so that they ran perpendicular to the forested ridge that comprises the western edge of the study area. Plovers are less likely to occur near forested habitat, so if an abundance gradient in plover distribution existed, the
4


survey lines would run perpendicular to that gradient. Transects extended across the width of the study area and terminated at fence borders on the east, and at the edge of forested habitat on the west.
Surveys were conducted in 2016 between 5 July and 15 July to maximize the probability of meeting the closure assumption. Similar to historic surveys, sampling was conducted from a vehicle driven at <15 km hr'1 and constrained to ~4 hours following sunrise or preceding sunset during low wind conditions with <10% cloud cover to maximize the detectability of plovers. I surveyed transects in order from North to South to maximize survey efficiency and alternated transect direction to adjust for potential bias from sun glare. To increase number of detections within the small study area, I surveyed transects in the morning and in the evening. The number of surveys completed in one day varied due to transect length and conditions during morning and evening survey periods. Perpendicular distance of a plover to the line was derived from sighting distance and angle using a laser ranger finder (Bushnell Corporation, Overland Park, Kansas; rated to 500 m) and a standard compass. In all surveys, detection distance was not fixed and all detections regardless of distance from the line were recorded.
Statistical analysis
To estimate historic density and abundance I used distance sampling data collected in 2000 to 2006. Because my study area represents a small portion of Mountain Plover habitat in South Park, I only included historic data from transects that intersected my smaller study area, reducing the number of transects from eight to three. Logistic and financial restraints in 2006 prevented a complete re-survey and data were available from only two transects for that year.
5


I used the R package Distance (R version 3.3.2; Miller 2016) to estimate Mountain Plover density in the study area for 2000-2006 and 2016.1 modeled detection probability as a function of exact distances using key functions suggested by Buckland et al. to be robust models of detection: 1) Half-normal with cosine adjustments, 2) Half-normal with hermite adjustments, 3) Hazard-rate with polynomial adjustments, and 4) Uniform with cosine adjustments (2001:155). Buckland et al. (2001:103) also suggest that in cases where detection width is not fixed, data may be right truncated to exclude 5-10% of furthest detections or where detection probability is estimated at 0.15 to improve the fit of the detection function. I truncated one dataset (2006) to 300 m, which removed two detections. I did not truncate other datasets as there were no obvious outliers and when I compared models with truncated data vs untruncated data, there was no improvement in detection model fit but some loss in precision with density estimates due to reduction in detection sample size. Because of differences in observers across years, I estimated the detection function separately for each year. The ds() function in package:Distance fits key functions using a forward stepping procedure for inclusion of the specified adjustment terms and selects the most parsimonious model based on Akaike Information Criterion (AIC) scores. This process resulted in some candidate model sets containing key function models with no adjustments (Appendix A). I used AIC to rank the models in the set for each year (Burnham and Anderson 2002), and evaluated model fit to the data using a Cramer-von Mises test (P-value range: 0.53-0.95; see Appendices A and B). Most candidate models were within 2 AAIC of each other so I averaged density and abundance predictions using Akaike weights. I constructed unconditional log-normal 95% confidence intervals for each averaged estimate to
6


account for increased variance from model selection uncertainty (Buckland et al. 2001: 77; Burnham and Anderson 2002: 164; Buckland et al. 2015: 106).
I estimated plover abundance in the study area from density estimates extrapolated to area of potentially suitable habitat within the study site. This area was defined earlier (see Wunder et al. 2003), and was marginally larger than the area covered by the 2016 transect design (30.8 km2 and 38.8 km2 respectively).
Results
I recorded 51 detections of Mountain Plover along 82 km of transects in 2016. The average number of detections from the sub-sampled historic data set was 80 (SD = 25.4), and ranged from a low of 26 (in 2006) to a high of 101 (in 2002). The uniform key function with cosine adjustments was the top ranked model for every year except 2003 and 2004, however, almost all key functions modeled the data well (Fig. 2; Buckland et al. 2015: 61).
Estimated detection probability ranged from 0.39 (2001) to 0.64 (2004); 95% lognormal confidence intervals around these estimates overlapped (Fig. 3). The mean estimated distance that all plovers within are detected (effective strip width) was 138 m (SD = 2.3). Histograms of years 2002, 2004, 2005, and 2016 suggest some avoidance of the line by plovers evidenced by less detections in the first bin (Fig. 2). However, estimation bias from this behavior was likely minimal because avoidance did not appear to move plovers beyond the detection range and fitted detection functions were monotonically constrained (Buckland etal 2015:204).
The highest estimated density within the study area was in 2001 (D = 12.8 plovers km'2, 95% Cl = 6.57-21.7, 35% CV), more than 4.5 times greater than the estimated density from 2016 (2.8 plovers km'2, 95% Cl = 1.54-4.57, 31% CV). Overall, there was a weak
7


negative trend in density within the study area over time (Fig. 4). Historic point estimates of density constrained to the 2016 study area were marginally greater than those estimated from distance sampling data from all eight transects distributed across South Park but all 95% confidence intervals overlapped (Fig. 4). By extrapolation, I estimated 110 plovers (95% Cl 60-177) occupied the 38.8 km2 of potentially suitable habitat within my study area in 2016.
Discussion
My results suggest that Mountain Plover density has declined in the study area during the past 17 years. Point estimates for density from 2000 to 2006 for all of South Park suggest a similar trend (Fig. 4). The density estimate of 2.8 plovers km'2 (95% Cl = 1.54-4.57) in 2016 is lower than the park wide density of 7.89 plovers km'2 (95% Cl = 4.9-9.9) for 2000-2003 as reported by Wunder et al. (2003). Historic estimates of density from the study area were similar to those estimated from data collected across the South Park region (Fig. 4), suggesting that monitoring at this study site may provide a reasonable index of population trends in the overall population of South Park. Reducing survey scope to a smaller patch of habitat from sites representative to the entire South Park area has allowed us to employ a more robust distance sampling design, increased power to detect temporal variation in plover density, and decreased logistical demands on surveys. A limitation of this strategy is that habitat in my study area may not always be representative of the entirety of South Park in the future because I will not detect any changes in distribution or density of plovers in areas of South Park outside of my study site.
Although current South Park density estimates are lower than historic estimates they still compare with those reported elsewhere. Plover density was estimated at 2.0 1.61 (SE) plovers km'2 for grazed rangelands on the PNG for the period from 1990 to 1994 (Knopf and
8


Wunder 2006). More recently, Tipton et al. (2009) estimated plover densities in three different habitat types across eastern Colorado: 2.26 plovers km'2 (95% Cl 2.15-5.13) on prairie dog colonies, 0.45 plovers km'2 (95% Cl 0.44-0.53) on agricultural fields, and 0.23 plovers km'2 (95% Cl 0.17-1.76) on grassland habitats, with an overall density of 0.93 plovers km'2 across all habitat types. Plumb et al. (2005) estimated 4.47 0.55 plovers km'2 on grassland habitats in Wyoming. McConnell et al. (2009) estimated 0.46 0.13 plovers km'2 in agricultural habitats in Oklahoma. Augustine (2011) estimated densities of 6.8 plovers km'2 (95% Cl 4.3-10.6), 2.0 plovers km'2 (95% Cl 0.8-5.0), and 5.6 plovers km'2 (95% Cl 3.5-9.1) for active prairie dog colonies, inactive prairie dog colonies, and prescribed burns, respectively, on the PNG.
Those estimates of Mountain Plover density or abundance are results of single studies conducted in a single or a few years. In Montana, where long-term study of Mountain Plover has been established, multiple estimates of density exist. Plover density estimates on prairie dog colonies have ranged from a high of 7.20 0.42 (SE) plovers km'2 in 2008 to a low of 1.28 0.06 (SE) plovers km'2 in 1995 (Childers and Dinsmore 2008). Due to strong associations with prairie dog habitat in Montana, changes in breeding distribution among colonies is expected because of sylvatic plague dynamics and colony size fluctuations (Dinsmore et al. 2005; Dinsmore and Smith 2010). Across five years on the PNG, plover densities declined by as much as 70% in the following year on colonies affected by plague and on prescribed burns, supporting the idea that breeding density tracks fluctuations in disturbance patterns (Augustine and Skagen 2014).
Despite high local fidelity to breeding areas, local variation in distribution and availability of suitable nesting areas shifts breeding densities within breeding areas.
9


Characteristics such as vegetation height and bare ground cover are preferentially used by plovers for nesting on both microsite and landscape scales (Goguen 2012). In comparison to previous years, precipitation was unusually high in 2015 resulting in excessive vegetative growth in 2016 (MBW, AKP, pers. obs.). Because my study is restricted to small area of publicly owned and managed rangelands it is possible that breeding density in my area shifted to private lands where grazing pressures are heavier, thus creating potentially more favorable nesting areas (Grunau and Wunder 2001). Alternatively, changes in vegetation structure may have influenced detection probability. However, I cannot definitively make any conclusions about either scenario because I did not explicitly measure any vegetative characteristics.
I estimated 110 breeding adults occupied my study area in 2016. Although my abundance estimate was extrapolated from the year with the lowest estimated breeding density, my study area only represents a small portion of comparably suitable plover breeding habitat, suggesting that South Park still supports a comparatively large portion of the breeding population. For comparison, McConnell et al. (2009) estimated the entire breeding plover population of Oklahoma consists of only 68-91 breeding individuals spread out across a fragmented cultivated landscape. Monitoring population fluctuations in the breeding population of South Park has obvious implications for Mountain Plover conservation and continued efforts over time can provide insight into factors that influence breeding population variation, particularly when combined with concurrent monitoring efforts across the range.
Mounting evidence suggests that Mountain Plover populations are declining, much like other grassland bird species, but to what extent remains unclear. This places difficulty on
10


planning and enacting conservation actions that are efficient and effective. This is further complicated by their patchy distribution in differential habitats across their range creating the potential for large demographic variation between local populations. Establishment of comparative long-term studies range wide is warranted to identify mechanisms driving these demographic differences. Re-establishment of plover monitoring in South Park is a small piece in a larger puzzle and future efforts in monitoring plovers in other breeding areas as well as in habitats used during stopover and winter are key steps that should be taken to characterize and mitigate further population declines.
11


Longitude
Fig. 1. Study area boundaries in South Park, Park County, Colorado. Park County shown in dark grey on inset map of Colorado counties (lower left) and study area location is marked with an asterisk. Enlarged map depicts study area boundaries within the western portion of the James Mark Jones State Wildlife Area (JMJSWA) and adjacent BLM pastures.
12


Distance (km)
Fig. 2. Histograms of detections with most parsimonious fitted detection functions for historic surveys (2000-2006, A-G) and the more recent survey (2016, H). Individual detection data points are depicted with open circles. See Appendix A for full list of detection model selection results.
13


1.00-
H 0.75-ro
o
L_
Q.
8 o-5-
o
CD
CD
Q 0.25-
2000 2004 2008 2012 2016
Year
Fig. 3. Detection probability estimates and 95% log-normal confidence intervals from historic (2000-2006) distance sampling data and data from the survey conducted in 2016.
Year
Fig. 4. Mountain Plover density estimates and 95% log-normal confidence intervals from transects within the 2016 study (black circles) and from transects ranging across South Park (grey triangles). Horizontal dashed line is drawn at upper limit of density estimate for 2016 for comparison to previous years.
14


Table 1. Estimates and coefficients of variation (CV) of density, detection probability, encounter rate (ER), and effective strip width (ESW) from historic (2000-2006) distance sampling data sub-sampled from the study site and data from the contemporary survey conducted in 2016.
Year Density (km-2) Detection probability ER ESW (m)
2000 8.71 1.46 0.54 0.06 2.71 0.06 156 2
17% CV 11% CV 12% CV
2001 12.8 4.5 0.39 0.03 2.98 1.01 116 1
35% CV 8% CV 34% CV
2002 11.9 2.4 0.51 0.04 3.42 0.62 144 1
20% CV 8% CV 18% CV
2003 7.85 3.99 0.43 0.05 1.79 0.88 115 1
50% CV 12% CV 49% CV
2004 7.19 1.95 0.64 0.07 2.51 0.62 175 2
27% CV 11% CV 25% CV
2005 9.01 1.92 0.57 0.09 2.44 0.33 136 2
21% CV 15% CV 14% CV
2006 5.02 1.18 0.5 0.07 1.5 0.26 150 2
23% CV 16% CV 17% CV
2016 2.82 0.89 0.59 0.07 0.62 0.18 110 1
31% CV 13% CV 28% CV
15


CHAPTER II
MOUNTAIN PLOVER NEST SURVIVAL IN A HIGH-ELEVATION POPULATION
Introduction
Broad-scale climate gradients and localized weather phenomena are important ecological drivers of grassland ecosystems in the Great Plains of North America that influence habitat structure and animal community composition (Samson and Knopf 2004). Considerable variation in both climate and weather is characteristic of this region; creating breeding conditions for ground-nesting grassland birds that can be unpredictable in both time and space; affecting nesting success (Skagen and Yackel Adams 2012). Thus, understanding factors that limit nest survival in a grassland species requires consideration of the relative intensity and variability of climate and other sources of disturbance across their breeding range, particularly if the species is widely but patchily distributed.
Mountain Plovers (Charadrius montanus) are migratory shorebirds that breed on grasslands and xeric tablelands scattered along the western edge of the Great Plains (Knopf and Wunder 2006). Mountain Plovers use areas frequently disturbed by stochastic events such as weather, grazing, tillage and burning; processes that affect the short vegetation and bare-ground habitats preferred for nesting. Populations of this species appear to have markedly declined since the 1960s and as a result Mountain Plovers have been identified as a species of conservation concern in many of the states where they breed and overwinter (Knopf 1994; Knopf and Wunder 2006; Andres and Stone 2010).
In Colorado, comparatively large populations of plovers can be found breeding on the eastern plains (Tipton et al. 2009) and also on intermountain grasslands in South Park, Park County (Wunder et al. 2003). Breeding habitat in South Park is comprised of high elevation
16


(-2700 m) rangelands that have experienced little historic change in land use compared to the rangelands on the eastern plains of Colorado. At lower elevations in eastern Colorado (-1470 m), plovers breed on a patchwork of habitat including grazed shortgrass prairie, burned rangelands, black-tailed prairie dog colonies (Cynomys ludovicianus), and fallow dryland crop fields; all of which vary annually in extent and availability. In addition to differences in habitat availability, differences in elevation create different climates and alter breeding schedules.
On lower elevation sites across the breeding range, Mountain Plovers breed earlier and nest survival is negatively influenced by increased precipitation and warmer temperatures, but the relative strength of weather effects and temporal variation in nest survival differs between locales (Dinsmore et al. 2002; Dreitz et al. 2012). In north-central Montana, at the northern periphery of the plovers range, Dinsmore et al. (2002) found relatively strong negative effects of increased precipitation on nest survival. In addition to precipitation, Dinsmore et al. (2002) also modeled nest survival as a function of daily maximum temperature, year, time in season (linear and quadratic), nest age, and sex of the incubating adult (adult plovers of both sexes independently incubate nests). Overall, daily nest survival in Montana was higher for male-tended nests, decreased over the season, increased over the incubation period, and decreased with precipitation. In eastern Colorado, increased precipitation was also associated with lower nest survival but survival also declined with increased maximum temperatures (Dreitz et al. 2012). However, in this study, variation in survival due to precipitation was best explained by 10-day periods of drought, not daily precipitation totals. Dreitz et al. (2012) also considered annual and seasonal time trends and
17


similar to the study in Montana, they found that nest survival decreased over the season but did not vary by year.
Results of these studies suggest that weather and time in season are important factors influencing Mountain Plover nest survival. Specifically, early and mid-season nests in cooler and drier conditions have more favorable odds of survival at lower elevations. However, no studies have examined the relationship between weather and temporal variation in plover nest survival on high elevation habitats which have distinctively different climate, habitat stability, and nesting phenology. I monitored nests of a population of Mountain Plover breeding at high elevation in South Park, Colorado to study influences on daily nest survival as compared to those from lower elevation breeding sites. I modelled daily variation in nest survival as a function of temperature, precipitation, and annual and seasonal time trends. I hypothesized that at high elevation cold would limit survival rather than heat, therefore I considered minimum temperatures in my nest survival models in place of maximum temperatures. To test how well models from Dinsmore et al. (2002) and Drietz et al (2012) described nest survival variation in my high elevation population, I also evaluated models constructed from my variable set that closely resembled the most parsimonious model from each study.
Methods
Study area
I studied Mountain Plovers breeding at high elevation (2786 19 m SD) on publicly owned and managed rangelands in Park County, Colorado (39 9' 19.296"N, 105 50' 51.72"W). My study area encompassed ~55 km2 of lands within the western portion of the James Mark Jones State Wildlife Area and adjacent pastures managed by the Bureau of Land
18


Management (BLM). In this area, plover breeding habitat is comprised of high-elevation intermountain grasslands grazed in winter by native herbivores such as elk (Cervus canadensis) and pronghorn (Antilocapra americana), and in summer by domestic cattle (Bos taurus). Although plovers are commonly associated with black-tailed prairie dogs in other areas of their range (Knowles et al.1982), there are no black-tailed prairie dogs in South Park (Wunder et al. 2003). Dominant grasses on the study area included species of muhly (Muhlenbergia filiculmis and Muhlenbergia montana), blue grama (Bouteloua gracilis), and Arizona fescue (Festuca arizonica). The most common forbs and shrubs included fringed sage (Artemisiafrigida) and yellow rabbit brush (Chrysothamnus viscidiflorus), respectively. Nest monitoring
I followed methods used by Dinsmore et al. (2002) to locate nests. During June and July of 2015 and 2016,1 conducted semi-systematic ground surveys of pastures on all-terrain vehicles to flush incubating plovers. Once a plover was sighted, it was observed until it returned to the nest or its behavior indicated no nest was present. Upon discovery, nest location was recorded using a GPS unit (Garmin Montana 680t, Olathe, KS, USA) and eggs were floated to estimate nest age (Dinsmore et al. 2002). To aid in relocation, a picture of the nest was taken and natural markers such as a small stack of dried cattle dung or a small rock pile were placed 10 m north and south of the nest. Nests were checked every 3-5 days until the eggs hatched (at least one chick produced), or the nest failed. Mountain Plover chicks are precocial and leave the nest within 24 hours of hatching. Therefore, in cases where chicks were not found in the nest cup or with the incubating adult nearby, nesting outcome was determined based on the presence or absence of small pip chips in the nesting cup (Mabee 1997). Nests were recorded as depredated if large shell fragments were scattered near the
19


nest site or if eggs were missing before the expected hatch date and no pip chips were found in the nest material. Some nests failed due to abandonment, which was confirmed by the persistent absence of the incubating adult, eggs that were cool to the touch, and/or lack of progression of development evidenced by egg floatation. In such cases, failure was recorded as the first day the nest was suspected to be abandoned.
As part of a concurrent studies, incubating adults were trapped using a walk-in wire mesh cage placed over the nest. Each bird was banded with a USGS aluminum leg band and unique color band combination and morphometric measurements were recorded. Some plovers were also fitted with a GPS datalogger (GPS PinPoint 50; Lotek, Newmarket, Ontario, Canada) attached using a leg-loop harness (Rappole and Tipton 1991). Tagging efforts were focused on plovers with late stage nests to mitigate potential risks of nests abandonment due to tag placement.
Weather data
I obtained temperature and precipitation data from a weather station administered by the National Oceanic and Atmospheric Administration (NOAA National Centers for Environmental Information; http://www.ncdc.noaa.gov/) located 18.4 km from the study area, in similar habitat and elevation (2718 m; Antero Reservoir station ID# USC00050263; 38 59' 35.8794"N, 105 53' 30.84"W). Minimum temperature values for 3 days were missing for this station in 2015, therefore, I substituted these values with measures from another station nearby at an altitude of 3086 m (Fairplay CO station ID# USC00052816; 39 13' 18.8394"N, 105 59' 36.96"W). Daily measures were recorded at 07:30 hours and reflected minimum temperature and total precipitation from the previous 24-hr period. In cases where I was interested in same day weather effects, I adjusted dates backward one day
20


so measures reflected the 24-hr period beginning at 07:30 hours on that date.
Model development and survival analysis
I estimated daily survival rate (DSR) of nests using the nest survival model available in Program MARK (White and Burnham 1999) using the R package RMark (R version 3.3.2; Laake 2013). I omitted 2 nests from analysis; one that was discovered shortly after hatching and another that was accidentally destroyed on discovery. Because I was more interested in evaluating a specific set of hypotheses rather than determining the importance of covariates, I limited my analyses to a set of a priori biological hypotheses constructed from the following covariates previously found to influence nest survival in Mountain Plovers (Table 1).
1) Year. I included models to allow nest survival to vary by year to account for annual variation that may not be captured in my covariates such as habitat structure, habitat quality, and predator abundance. Annual variation in nest survival has been previously documented in Mountain Plover (Augustine and Skagen 2014).
2) Seasonal time trends. Nest survival may also vary within a season. In eastern Colorado, daily survival rates of Mountain Plover nests declined over the season (Dreitz and Knopf 2007; Dreitz et al. 2012). In Montana, daily survival rates were lowest late season but also depressed early in early season (Dinsmore et al. 2002). Based on this, I considered models incorporating either a linear time trend or a quadratic time trend.
3) Nest age. DSR may vary over the incubation period, increasing as nests approach hatch. Vulnerable nests may be lost early in incubation, or parental behaviors may
21


influence predation rates as investment increases (Klett and Johnson 1982). DSR has also been shown to increase over incubation in plovers (Dinsmore et al. 2002).
4) Minimum temperature and precipitation. Dreitz et al. (2012) reported increased DSR during 10 day periods of drought, and reduced daily survival with high maximum temperatures in Mountain Plover nests in eastern Colorado. In Montana, Dinsmore et al. (2002) found reduced nest survival with increased daily precipitation. I hypothesized that nest survival would also be influenced by daily precipitation and temperature in South Park, but because of differences in climate at higher elevation, survival would be limited by minimum temperatures instead of maximum temperatures. I also considered minimum temperature and precipitation from the previous day to incorporate potential one day lag effects on survival. As uniparental incubators, plovers may be restricted to nests during hot, cold, or wet conditions and thus may increase duration or frequency of off-bouts the following day to forage and recoup energetic reserves (Skrade and Dinsmore 2012), leaving nests vulnerable to predators (Smith et al. 2012).
5) GPS tag placement. As part of concurrent studies some incubating adults were fitted with a GPS tag which may have affected their behavior and reduced nest survival. Thus, I also considered a model that allowed DSR to vary between nests with tagged and untagged incubators.
I limited the complexity of these models because of comparatively small sample size, and did not want to consider additional models at random that included all other combinations of temporal and weather variables. To provide a basis for comparison to previous Mountain Plover nest survival work with more complex models, I constructed models from my set of
22


covariates that approximated the most parsimonious model from Dinsmore et al. (sex of incubating adult, nest age, quadratic time trend and precipitation; 2002) and Dreitz et al. (linear time trend, 10-day drought, and maximum temperature; 2012). My model equivalent to Dinsmore et al. (2002) did not include sex of the incubating adult as I did not have these data for nests in 2016.1 substituted previous day precipitation for 10-day drought in my model equivalent to Dreitz et al. (2012) because I did not observe any 10-day periods without precipitation in the two years of my study. I also replaced maximum temperature with minimum temperature because extremely hot temperatures do not occur in South Park, but temperatures can regularly fall below freezing even in summer months.
Results
Data summary and nesting phenology
I monitored 89 Mountain Plover nests on actively grazed rangelands during June and July in 2015 (n = 52) and in 2016 (n = 37). Of these 89 nests, at least one egg hatched in 35 (39.3%), 46 were lost to predation (51.7%), 7 were abandoned (7.9%), and one was accidentally destroyed upon discovery (1.1%). The naive point estimate for nest success in 2015 was higher than in 2016 (48.1% vs. 35.1%, respectively).
Nests were monitored over a 53 day exposure period (01 June-24 July). Mean ( 1 SE) nest initiation dates were similar between years (2015, 03 June 1.3 days; 2016, 29 May 1.6 days). Assuming all discovered nests survived the 29-day incubation period from their estimated initiation date the expected mean hatch date was 01 July 1.0 days, which was similar to the observed mean hatch date of successful nests (03 July 1.7 days).
23


Nest survival
The model including only daily minimum temperature best predicted variation in DSR, followed by additive effects of minimum temperature and precipitation. Models incorporating partial and combined effects of temperature and precipitation from the previous day were slightly more parsimonious than model counterparts without lagged weather effects (Table 2). Although AICc values were similar between all models in my set, a minimum temperature effect (either same day or lagged) was present in all models within 2 AAICc units of the most parsimonious model (Table 2). Furthermore, same day minimum temperature and previous day minimum temperature were the only parameter estimates with a 95% confidence intervals that did not overlap zero (See Appendix C). GPS tag placement did not have a measureable effect on DSR; estimated 95% Cl for coefficients of tag presence overlapped zero (Appendix C). Of the adults incubating 87 nests, 30 adults were fitted with a GPS tag. Of those 30 nests, 14 (47%) hatched at least one chick; this was higher than the overall naive estimate of hatching success for the population.
The unadjusted mean daily nest survival probability for nests in my study was 0.949 (95% Cl, 0.934-0.961) and total probability of survival across a typical incubation period of 29 days based on constant DSR was 0.218 (95% Cl 0.139-0.317). My most parsimonious model predicted that for every 1C increase in previous day minimum temperature adjusted daily odds of survival increased by a factor of 1.17 (95%CI 1.04-1.31; Table 3). An equally competitive model predicted a similar adjustment factor in odds for every 1C increase in previous day minimum temperature (1.19, 95% Cl, 1.05-1.34) as well as an additive adjustment in odds of survival by a factor of 0.97 (95% Cl 0.92-1.01) for every 1 mm in precipitation (Table 3). Holding prevTMIN and prevPRCP at their average observed values
24


(4.5C and 1.7 mm), the top two weather models (prevTMIN and prevTMIN + prevPRCP; Table 2) predicted a mean DSR of 0.950 and 0.952, respectively. Total probability of survival across a typical incubation period of 29 days as 0.231 and 0.242 for the prevTMIN model and the prevTMIN + prevPRCP model, respectively (Table 2). Effects of prevTMIN on DSR were strongest at temperatures below freezing (Fig. 1) and with heavy precipitation (Fig. 2).
Discussion
Extreme high temperatures have negative effects on Mountain Plover nest survival at lower elevations in Colorado (Dreitz et al. 2012), and extreme low temperatures have negative effects on nest survival at high elevation (this study) sites. In my study, extreme cold decreased odds of survival for nests, and that effect increased with cold intensity (Fig. 1; Table 2). At lower elevations in eastern Colorado, Drietz et al. (2012) reported a negative effect of extreme heat on plover nest survival. However, in the northern edge of the breeding range in north-central Montana, Dinsmore et al. (2002) did not find a meaningful effect of maximum temperature on nest survival. Drietz et al. (2012) suggests that this may be attributable to the lower intensity of extreme heat that occurs at a higher latitude. Dinsmore et al. (2002) did not consider minimum temperature effects, so it is uncertain if cold temperatures also affect DSR at this latitude.
I also found limited support that precipitation affected daily nest survival, but only in conjunction with minimum temperature (Fig. 2). Moreover, although the estimated effect was negative, it was not statistically different from zero (Table 2). It is possible that I did not have a sufficient sample size to estimate the potentially more subtle effect of precipitation. Dinsmore et al. (2002) and Drietz et al. (2012) both found negative effects of precipitation
25


but also had much larger sample sizes, 641 nests and 936 nests respectively, as compared with 87 in this study.
Models analogous to the most parsimonious models from previous Mountain Plover nest survival studies by Dinsmore et al. (2002) and Drietz et al. (2012) which incorporated temporal and weather covariates were outperformed by simpler temperature and precipitation only models (Table 2). This is unsurprising given differences between habitats. However, negative effects of extreme temperature and precipitation may be generalities that exist between locales but are potentially mediated through different mechanisms at varying temporal scales.
Evidence of annual or seasonal temporal trends in nest survival was not supported by my data. Many factors that I did not explicitly measure could be attributed to annual variation in nest survival, such as changes in habitat structure, climate, predator abundance, or plover abundance. Augustine and Skagen (2014) found that Mountain Plover nests in one year had lower DSR than those in the subsequent year and that breeding density of plovers were similar between years. They did not include weather variables in their model set, and as they suggested, annual variation in nest survival may have been attributable to climatic differences between years, as nest survival was higher in the cooler dryer year. The notion that annual variation in DSR may be better explained by weather is supported by the lack of annual trends in aforementioned studies where weather variables are modeled.
DSR may vary as a function of time in season due to many factors such as vegetation changes over the season, predator and prey abundance, re-nesting activity, or breeding experience. Dinsmore et al. (2002) found that DSR of plover nests in Montana was highest later in the season and lowest mid-season. Conversely, Drietz et al. (2012) and, Augustine
26


and Skagen (2014) found that DSR of plover nests decreased over the season in eastern Colorado. The lack of seasonal trends in my study may be attributable to the comparatively short season length at high elevation. Nest monitoring exposure periods in previous plover nest survival studies spanned 77 days (Dinsmore et al. 2002), 66 days (Augustine and Skagen 2014), and up to 110 days (Drietz et al. 2012), as contrasted with 53 days in this study. Between years, my observed mean nest initiation date was 02 June 1.0 days (SE), similar to reported mean initiation dates for male-tended nests at higher latitudes in Montana (02 June 3.2 days; Dinsmore et al 2002). The constrained breeding period in South Park likely limits the ability of plovers to adjust their nesting schedule and constrains potential for re-nesting attempts. Therefore, with my sample size, it is not surprising that I did not uncover any evidence of temporal variation in nest survival that was not better explained by seasonal changes in temperature and precipitation. Models that incorporated temporal and weather covariates in the most parsimonious models of nest survival from previous work were not competitive in my analysis (Table 2). This supports my conclusion that annual and seasonal variation in nest survival at my location is best described by minimum temperature and precipitation, and any additional factors that I considered may have effects too small to be detected with my sample.
My estimate of overall mean nest survival assuming a constant DSR was 0.218 (95%CI, 0.139-0.317), almost half of the observed proportion of hatched nests (naive estimate) included in analysis (0.39). This naive estimate is biased high because it does not account for nests lost prior to discovery. My estimate of nest survival is low compared to nest survival and hatch rates reported in Mountain Plover literature which range from 0.27 to 0.70 and is much lower than previously reported naive estimates of nest success rates of 0.60 in
27


South Park (Graul 1975; Dinsmore et al. 2002; Mettenbrink et al. 2006; Dreitz and Knopf 2007; Dinsmore et al. 2010; Dreitz et al. 2012). I cannot rule out that this low rate may be partially attributed to differences in climatic conditions between years of this study (2015, 2016) and when plover nests were last monitored in South Park (2000-2006). Minimum temperatures were not unusual during the period of my study, however, total precipitation was usually high in 2015. In May and June of 2015, precipitation totaled 113.4 mm and 73.4 mm, respectively, which was much higher the mean totals for May (18.3 mm) and June (22.1 mm) between 2000 and 2006. Lagging effects of precipitation on prey abundance or habitat structure may have had influence on nest survival that was not captured in the changes in daily measures that I measured.
Overall, my results suggest that cold temperatures limit Mountain Plover nest survival in South Park. The mechanism behind this relationship is less clear. Nest predation is the primary cause of nest failure in ground-nesting grassland birds (Martin 1993), which was also true for nests in my study. As such, cold temperatures likely affect nest survival in South Park by increasing predation risk. In my study area, the predator community consists largely of avian predators such as Common Ravens (Corvus corax) and small mammals such as thirteen-lined ground squirrels (Ictidomys tridecemlineatus). Due to the high elevation and low ambient temperatures, South Park is notably devoid of snakesa common nest predator on low elevation breeding sites. Effects of weather and climate on predation risk are contingent on the composition of the nest predator community because different predators respond differently to environmental conditions. For example, Dinsmore et al. (2002) hypothesized that precipitation would enhance olfactory cues that snakes use to locate nests increasing predation risk but this would not be the case for avian predators which rely on
28


visual cues to find nests. These nest predator community composition differences may partly explain why precipitation did not have a strong significant effect on nest survival in South Park. As for temperature, minimums usually occur at night so it is unlikely that these temperature drops increase predation risk by influencing behavior of visual predators like ravens. Increased mammalian predation during periods of low temperature also seems unlikely as plovers actively defend nests from ground squirrels (MBW, AKP pers. Obs.) or perform distraction displays (Graul 1975) to lure them away from the nest.
It is more likely that temperature effects on predation are mediated by effects on incubation behavior. In some uniparental arctic shorebirds, maintaining stable incubation temperatures during cold periods may increase energy expenditure inducing an increase in duration of subsequent foraging bouts to balance energetic condition (Tulp and Schekkerman 2006). These long or frequent off-bouts from the nest decrease nest vigilance which may in turn increase predation rates (Smith et al. 2012). For example, within species comparisons in Sanderlings (Calidris alba) demonstrated that uniparental incubators expend more energy in maintaining egg temperatures than biparental counterparts (Reneerkens et al. 2011). Additionally, Reneerkens et al. (2011) found that arthropod abundance increased with warming temperatures and only uniparental Sanderlings increased their recess frequency as temperatures warmed. Overall, Reneerkens et al. (2011) observed decreased nest attendance, increased recess frequency, and longer recess duration in uniparental Sanderlings compared to biparental. As uniparental incubators, Mountain Plovers may also employ similar compensatory behaviors to balance incubation and personal energetic maintenance. In Montana, plovers regularly leave the nest to forage in the middle of the night and at crepuscular times (Skrade and Dinsmore 2012). However, arthropod activity may be
29


depressed during cold evenings in South Park, limiting productive foraging to daylight hours. In turn, after long or intense cold periods nests may be left unattended longer or more frequently during daylight when most nest predators are active. Energy expenditure may also be increased to maintain nest and individual body temperatures in below freezing conditions. However, maintaining egg temperatures may not be as costly in Mountain Plovers due to unusually thick shells which may act as insulation during colder temperatures (Dinsmore et al. 2002). Precipitation during these cold periods could potentiate effects on both energy expenditure and prey availability. As such, effects on incubation behavior are likely contingent on the duration and timing of extreme cold and precipitation events. This information was not available for the weather data I collected in my study, which might explain why I saw little differentiation between same day and one day lag effects of both minimum temperature and precipitation. From my study, I cannot definitively conclude why minimum temperature and to a limited degree, precipitation, are associated with lower nest survival, but alterations in incubation behavior induced by changes in prey availability or thermal maintenance seem to be plausible mechanisms that warrant future study on how cold temperatures negatively influence Mountain Plover nest survival in South Park.
My study demonstrates the importance of considering differences between weather and phenology among breeding locations. To better understand limitations in adaptability to disturbance, it is important to sample across spatial gradients where these disturbances occur. For example, to understand impacts of weather on Mountain Plover nest survival, studies should be conducted in other nesting areas across their geographic range, particularly in New Mexico and Mexico. Within the central breeding range on the Great Plains, it is likely that climatic extremes will become more intense and frequent in the future (Skagen and Yackel
30


Adams 2012). Therefore, understanding the differential effects on plover productivity will aid in conservation planning.
31


Table 1. Candidate models and notation.
Candidate models Model notation
Constant D SR Intercept only
Annual trend Year
Linear time trend Time
Quadratic time trend Time2
Annual and linear time trend Year + Time
Annual and quadratic time trend Year + Time2
Nest age NestAge
Current day min. temp (C) TMIN
Previous day min. temp (C) prevTMIN
Current day precip total (mm) PRCP
Previous day precip total (mm) prevPRCP
Current day min temp (C) and precip total (mm) TMIN + PRCP
Previous day min temp (C) and precip total (mm) prevTMIN + prevPRCP
Linear time, previous precip total (mm), and min temp Time + prevPRCP + TMIN
Nest age, quadratic time, and current day precip total (mm) NestAge + Time2 + PRCP
GPS tag on incubating adult Tag
Table 2. Summary of nest survival model selection results for Mountain Plovers in South Park, Colorado (2015 to 2016). The TMIN and PRCP variables refer to daily minimum temperature and precipitation respectively. Variables preceded by prev refer to one day lag effects.
Model Parameters AlCe AAICe Wi Deviance
prevTMIN 2 298.413 0 0.23 294.401
TMIN 2 298.683 0.27 0.201 294.67
prevTMIN + prevPRCP 3 299.062 0.65 0.166 293.038
TMIN + PRCP 3 300.359 1.946 0.087 294.335
Year 2 301.06 2.647 0.061 297.048
Time + prevPRCP + TMIN 4 301.37 2.957 0.052 293.329
Year + Time 3 301.988 3.575 0.038 295.963
Tag 2 302.565 4.152 0.029 298.553
Year + Time2 3 302.658 4.245 0.028 296.634
Constant (Intercept only) 1 302.74 4.327 0.026 300.736
Time 2 303.257 4.844 0.02 299.244
NestAge 2 303.41 4.997 0.019 299.398
Time2 2 304.109 5.696 0.013 300.096
prevPRCP 2 304.503 6.09 0.011 300.491
PRCP 2 304.726 6.313 0.01 300.714
NestAge + Time2 + PRCP 4 305.238 6.825 0.008 297.197
32


Table 3. Parameter estimates, SE, and 95% confidence intervals for lag effect weather models on the logit scale used to calculate nest survival probability over a 29-day incubation period.
Model Parameter Estimate SE 95% Cl
prevTMIN Intercept 2.2553 0.2667 (1.7325, 2.7781)
prevTMIN 0.1546 0.0597 (0.0377, 0.2716)
prevTMIN + prevPRCP Intercept 2.2627 0.2658 (1.7417, 2.7836)
prevTMIN 0.1738 0.0620 (0.0523, 0.2952)
prevPRCP -0.0358 0.0237 (-0.0823, 0.0107)
Fig. 1. Estimated daily survival rate of Mountain Plover nests across varying daily minimum temperatures lagged by one day in South Park, Colorado.
33


Daily Survival Rate (DSR)
Fig. 2. Estimated daily survival rate of Mountain Plover nests across varying daily minimum temperatures and precipitation totals lagged by one day in South Park, Colorado.
34


REFERENCES
Abbott, C.G. 1939. Mountain Plover at San Diego, California. Condor 42:125.
Andres, B.A., and K.L. Stone. 2010. Conservation Plan for the Mountain Plover (iCharadrius montanus). Version 1.1. Manomet Center for Conservation Sciences, Manomet, MA.
Augustine, D.J. 2011. Habitat selection by mountain plovers in shortgrass steppe. Journal of Wildlife Management. 75:297-304.
Augustine, D.J., Skagen, S.K. 2014. Mountain plover nest survival in relation to prairie dog and fire dynamics in shortgrass steppe. Journal of Wildlife Management. 78:595-602.
Buckland, S.T., Anderson, D.R., Burnham, K.P., Laake, J.L., Borchers, D.L., and L. Thomas. 2001. Introduction to distance sampling estimating abundance of biological populations. Oxford University Press, Oxford, UK.
Buckland, S.T., Rexstad, E.A., Marques, T.A., snd C.S. Oedekoven. 2015. Distance Sampling: Methods and Applications. Springer, New York, NY, USA.
Burnham, K.P., and D.R. Anderson. 2002. Model selection and multimodel inference: a practical information-theoretic approach. 2nd ed. Springer, New York, NY,US A.
Childers, T.M., and S.J. Dinsmore. 2008. Density and abundance of Mountain Plovers in northeastern Montana. Wilson Journal of Ornithology. 120:700-707.
Colorado Parks and Wildlife. 2015. State Wildlife Action Plan: A Strategy for conserving wildlife in Colorado. Denver, Colorado, USA.
Cooke, W.W. 1914. Our Shorebirds and their Future. Yearbook of the Department of Agriculture 1914:275-294.
Dinsmore, S.J., G. C. White, and F. L. Knopf, 2005. Mountain Plover population responses to black-tailed prairie dogs in Montana. The Journal of Wildlife Management 69:1546- 1553.
Dinsmore, S.J., and M.D. Smith. 2010. Mountain Plover responses to plague in Montana. Vector-Borne Zoonotic Disease. 10:37-45.
Dinsmore, S.J., White, G.C., and F.L. Knopf. 2002. Advanced techniques for modeling avian nest survival. Ecology 83:3476-3488.
Dinsmore, S.J., Wunder, M.B., Dreitz, V.J., and F.L. Knopf. 2010. An assessment of factors affecting population growth of the Mountain Plover. Avian Conservation and Ecology. 5:13.
35


Dreitz, V.J., Conrey, R.Y., and S.K. Skagen. 2012. Drought and cooler temperatures are associated with higher nest survival in Mountain Plovers. Avian Conservation and Ecology. 7(1):6.
Dreitz, V.J., and F.L. Knopf. 2007. Mountain Plovers and the politics of research on private lands. Bioscience 57:681-687.
Graul, W.D. 1975. Breeding biology of the Mountain Plover. Wilson Bulletin. 87:6-31.
Graul, W.D., and L.E. Webster. 1976. Breeding status of the Mountain Plover. Wilson Bulletin. 78:265-267.
Grunau, L., and M.B.Wunder. 2001. Conservation assessment for Mountain Plover
{Charadrius montanus) in South Park, Colorado. Colorado Natural Heritage Progarm.
Klett, A.T. and D.H. Johnson. 1982. Variability in nest survival rates and implications to nesting studies. The Auk. 99:77-87.
Knopf, F.L. 1994. Avian assemblages on altered grasslands. Studies in Avian Biology. 15:247-257.
Knopf, F.L., andB.J. Miller. 1994. Charadrius montanus: montane, grassland, or bare-ground plover? The Auk. 111:504-506.
Knopf, F.L., and M.B. Wunder. 2006. Mountain Plover (Charadrius montanus). In: The Birds of North America Online (A. Poole, ed.) Cornell Lab of Ornithology, Ithaca, NY.
Knowles, C.J., Stoner, C.J. and S.P. Gieb. 1982. Selective use of black-tailed prairie dog towns by Mountain Plovers. Condor. 84(1)71-74.
Laake, J.L. 2013. RMark: An R interface for analysis of capture-recapture data with MARK. AFSC Processed Report. 2013-01, 25p. Alaska Fish. Sci. Cent., NOAA, Natl. Mar. Fish. Serv., 7600 Sand Point Way NE, Seattle WA 98115.
Mabee, T.J., 1997. Using eggshell evidence to determine nest fate of shorebirds. The Wilson Bulletin. 109(2): 3 07-313.
Martin, T. E. 1993. Nest predation among vegetation layers and habitat types: revising the dogmas. The American Naturalist. 141 (6):897913.
McConnell, S., OConnell, T.J., Leslie, David M.J., and J.S. Shackford. 2009. Mountain plovers in Oklahoma: distribution, abundance, and habitat use. Journal of Field Ornithology 80:27-34.
36


Mettenbrink, C.W., Dreitz, V.J., Knopf, F.L., and T. Brush. 2006. Nest success of Mountain Plovers relative to anthropogenic edges in Eastern Colorado. The Southwest Naturalist. 51:191-196.
Miller, D.L. 2016. Distance: Distance Sampling Detection Function and Abundance
Estimation. R package version 0.9.6. https://CRAN.R-project.org/package=Distance
Plumb, R.E., Knopf F.L., and S.H. Anderson. 2005. Minimum population size of Mountain Plovers breeding in Wyoming. The Wilson Bulletin. 117(1): 15-22.
Rappole, J.H., and A.R. Tipton. 1991. New harness design for attachment of radio
transmitters to small passerines (Nuevo Diseno de Arnes para Atar Transmisores a Passeriformes Pequenos). Journal of Field Ornithology. 335-337.
Reneerkens, J., Grond, K., Schekkerman, FL, Tulp, I. and T. Piersma. 2011. Do uniparental sanderlings Calidris alba increase egg heat input to compensate for low nest attentiveness?. PLOS one, 6(2):p.el6834.
Skagen, S. K., and A. A. Yackel Adams. 2012. Weather effects on avian breeding performance and implications of climate change." Ecological Applications 22(4): 1131-1145.
Skrade, P.D.B., and S.J. Dinsmore. 2012. Incubation patterns of a shorebird with rapid
multiple clutches, the Mountain Plover (iCharadrius montanus). Canadian Journal of Zoology. 90:257-266.
Smith, P.A., Tulp, I., Schekkerman, H., Gilchrist, H.G., and M.R. Forbes. 2012. Shorebird
incubation behaviour and its influence on the risk of nest predation. Animal Behavior 84(4): 83 5-842.
Tipton, H.C., Doherty Jr P.F., and V.J. Dreitz.2009. Abundance and density of mountain plover (Charadrius montanus) and burrowing owl (Athene cunicularia) in eastern Colorado. The Auk 126:493-499.
Tulp, I., and H. Schekkerman. 2006. Time allocation between feeding and incubation in uniparental arctic-breeding shorebirds: Energy reserves provide leeway in a tight schedule. Journal of Avian Biology. 37:207-218.
White, G.C., and K.P. Burnham. 1999. Program MARK: survival estimation from populations of marked animals. Bird study 46:S120-S139.
Wunder M.B., Knopf, F.L., and C.A. Pague. 2003. The high-elevation population of Mountain Plovers in Colorado. Condor 105:654.
37


APPENDIX A
Distance sampling model selection results for all years. P-values listed are from a Cramer-von Mises goodness of fit test. Estimated detection probabilities with SE are also listed.
2000
Key function C-vM P-value Detection SE JAIC
Uniform with cosine adjustment term of order 1 0.72 0.52 0.02 0.00
Hazard-rate 0.95 0.59 0.04 0.52
Half-normal with cosine adjustment term of order 2 0.79 0.54 0.09 1.47
Half-normal 0.43 0.48 0.05 1.50
2001
Key function C-vM P-value Detection SE JAIC
Uniform with cosine adjustment terms of order 1,2 0.64 0.39 0.03 0.00
Half-normal 0.69 0.39 0.03 0.59
Hazard-rate 0.54 0.44 0.04 3.24
2002
Key function C-vM P-value Detection SE JAIC
Uniform with cosine adjustment term of order 1 0.96 0.52 0.02 0.00
Half-normal 0.70 0.48 0.04 1.12
Hazard-rate 0.97 0.58 0.04 1.98
2003
Key function C-vM P-value Detection SE JAIC
Hazard-rate 0.62 0.45 0.05 0.00
Uniform with cosine adjustment terms of order 1,2 0.47 0.40 0.04 1.75
Half-normal 0.49 0.41 0.04 2.21
38


2004
Key function C-vM P-value Detection SE JAIC
Half-normal 0.84 0.65 0.07 0.00
Uniform with cosine adjustment term of order 1 0.79 0.63 0.06 0.39
Hazard-rate 0.87 0.67 0.09 1.95
2005
Key function C-vM P-value Detection SE JAIC
Uniform with cosine adjustment term of order 1 0.45 0.61 0.04 0.00
Hazard-rate 0.92 0.53 0.10 0.17
Half-normal with cosine adjustment term of order 2 0.92 0.51 0.07 0.37
Half-normal 0.42 0.62 0.05 0.85
2006
Key function C-vM P-value Detection SE JAIC
Uniform with cosine adjustment term of order 1 0.85 0.50 0.06 0.00
Half-normal 0.90 0.49 0.08 0.44
Hazard-rate 0.87 0.53 0.11 3.28
2016
Key function C-vM P-value Detection SE JAIC
Uniform with cosine adjustment term of order 1 0.42 0.59 0.05 0.00
Hazard-rate 0.50 0.58 0.10 0.56
Half-normal 0.41 0.59 0.06 0.88
39


APPENDIX B
R output summaries for most parsimonious distance sampling models.
2000
Summary for distance analysis
Number of observations : 80
Distance range : 0 0.291488
Model : Uniform key function with cosine adjustment term of order 1
Strict monotonicity constraints were enforced.
AIC : -238.9163
Detection function parameters Scale coefficient (s) :
NULL
Adjustment term coefficient(s):
estimate se
cos, order 1 0.9377728 0.05963779
Estimate SE CV
Average p 0.5160564 0.01588239 0.03077646
N in covered region 155.0218229 12.96680032 0.08364500
2001
Summary for distance analysis
Number of observations : 88
Distance range : 0 - 0.2961807
Model : Uniform key function with cosine adjustment terms of order 1,2
Strict monotonicity constraints were enforced.
AIC : -291.5907
Detection function parameters Scale coefficient (s) :
NULL
Adjustment term coefficient(s):
estimate se
cos, order 1 1.2716211 0.09524336 cos, order 2 0.3222461 0.09614585
Estimate SE Average p 0.3855248 0.02748264 N in covered region 228.2603065 25.07168197
CV
0 07128631 0.10983812
40


2002
Summary for distance analysis
Number of observations : 101
Distance range : 0 - 0.278
Model : Uniform key function with cosine adjustment term of order 1
Strict monotonicity constraints were enforced.
AIC : -311.5709
Detection function parameters Scale coefficient (s) :
NULL
Adjustment term coefficient(s): estimate se
cos, order 1 0.937892 0.05672406
Estimate SE CV
Average p 0.5160246 0.01510456 0.02927101
N in covered region 195.7270912 14.71032996 0.07515735
2003
Summary for distance analysis Number of observations : 53 Distance range : 0 0.266
Model : Hazard-rate key function AIC : -183.3245
Detection function parameters Scale coefficient(s):
estimate se
(Intercept) -2.300607 0.1541386
Shape coefficient (s) :
estimate se
(Intercept) 1.402911 0.2549164
Estimate SE CV
Average p 0.4534208 0.05302137 0.1169363
N in covered region 116.8892006 18.10346193 0.1548771
41


2004
Summary for distance analysis
Number of observations : 74
Distance range : 0 - 0.2708349
Model : Half-normal key function AIC : -204.4621
Detection function parameters Scale coefficient(s):
estimate se
(Intercept) -1.891013 0.1521052
Estimate SE CV
Average p 0.6476029 0.06810742 0.1051685
N in covered region 114.2675534 14.37344785 0.1257877
2005
Summary for distance analysis
Number of observations : 87
Distance range : 0 - 0.24096
Model : Uniform key function with cosine adjustment term of order 1
Strict monotonicity constraints were enforced.
AIC : -265.9016
Detection function parameters Scale coefficient (s) :
NULL
Adjustment term coefficient(s): estimate se
cos, order 1 0.6328666 0.1195913
Estimate SE CV
Average p 0.6124199 0.04485368 0.07324008
N in covered region 142.0593913 14.07682630 0.09909113
42


2006
Summary for distance analysis
Number of observations : 24
Distance range : 0 - 0.3
Model : Uniform key function with cosine adjustment term of order 1
Strict monotonicity constraints were enforced.
AIC : -68.46056
Detection function parameters Scale coefficient (s) :
NULL
Adjustment term coefficient(s): estimate se
cos, order 1 1 0.2570877
Estimate SE CV
Average p 0.5 0.06427194 0.1285439
N in covered region 48.0 9.27740302 0.1932792
2016
Summary for distance analysis Number of observations : 51
Distance range : 0 - 0.188
Model : Uniform key function with cosine adjustment term of order 1
Strict monotonicity constraints were enforced.
AIC : -183.1659
Detection function parameters Scale coefficient (s) :
NULL
Adjustment term coefficient(s): estimate se
cos, order 1 0.6989065 0.144215
Estimate SE CV
Average p 0.5886139 0.04996563 0.08488693
N in covered region 86.6442314 10.70755315 0.12358068
43


APPENDIX C
Parameter estimates for all nest survival models. Estimates are reported on the logit scale with SE and upper (UCL) and lower (LCL) 95% confidence limits. Effects with confidence intervals that do not overlap zero are highlighted in bold text.
Model Parameter Estimate SE LCL UCL
prevTMIN Intercept 2.2553 0.2667 1.7325 2.7781
prevTMIN 0.1546 0.0597 0.0377 0.2716
TMIN Intercept 2.2160 0.2823 1.6627 2.7692
TMIN 0.1603 0.0629 0.0371 0.2836
prevTMIN + prevPRCP Intercept 2.2627 0.2658 1.7417 2.7836
prevTMIN 0.1738 0.0620 0.0523 0.2952
prevPRCP -0.0358 0.0237 -0.0823 0.0107
TMIN + PRCP Intercept 2.2137 0.2806 1.6637 2.7637
TMIN 0.1708 0.0652 0.0431 0.2985
PRCP -0.0191 0.0290 -0.0759 0.0376
Year Intercept 3.1325 0.1898 2.7604 3.5046
Year2016 -0.5538 0.2848 -1.1120 0.0043
Time + prevPRCP + TMIN Intercept 2.2684 0.3128 1.6553 2.8815
Time -0.0051 0.0154 -0.0353 0.0251
prevPRCP -0.0444 0.0287 -0.1007 0.0118
TMIN 0.1977 0.0818 0.0373 0.3580
Year + Time Intercept 2.8394 0.3340 2.1847 3.4940
Year2016 -0.5246 0.2863 -1.0858 0.0365
Time 0.0138 0.0134 -0.0125 0.0402
Tag Intercept 2.7636 0.1698 2.4309 3.0963
Tag 0.4423 0.3065 -0.1585 1.0431
Year + Time2 Intercept 3.0276 0.2485 2.5404 3.5147
Year2016 -0.5385 0.2859 -1.0988 0.0219
Time2 0.0002 0.0003 -0.0004 0.0007
Constant D SR Intercept 2.9189 0.1412 2.6422 3.1956
Time Intercept 2.5895 0.2974 2.0066 3.1723
Time 0.0162 0.0134 -0.0101 0.0425
44


Parameter estimates for all nest survival models contd.
Model Parameter Estimate SE LCL UCL
NestAge Intercept 3.3376 0.4011 2.5514 4.1238
NestAge -0.0218 0.0190 -0.0589 0.0154
IT Intercept 2.7949 0.2070 2.3892 3.2006
Time2 0.0002 0.0003 -0.0003 0.0008
prevPRCP Intercept 2.9607 0.1629 2.6413 3.2800
prevPRCP -0.0182 0.0311 -0.0792 0.0427
PRCP Intercept 2.9065 0.1642 2.5847 3.2284
PRCP 0.0062 0.0438 -0.0797 0.0921
NestAge + Time2 + PRCP Intercept 3.3397 0.4149 2.5265 4.1529
NestAge -0.0357 0.0213 -0.0775 0.0060
Time2 0.0005 0.0003 -0.0002 0.0011
PRCP 0.0049 0.0419 -0.0772 0.0870
45


APPENDIX D
Distance sampling analyses R code.
#################################################################################
# #
# Mountain Plover distance sampling analysis 2017 #
# author: Allison Pierce #
# #
#################################################################################
library(tidyverse) library(Distance) library(AICcmodavg)
#read in data and format years to factors, mopl <- read.csv(file = 'JMJDSAllYears.csv') mopl$year <- factor(mopl$year)
# area of suitable MOPL habitat in study site
mopl$Area <- 38.77328
# fit the following detection models for each year:
# Half-normal with cosine adjustments
# Half-normal with hermite adjustments
# Hazard-rate with polynomial adjustments
# Uniform with cosine adjustments
#
# Note: The ds() function will try different orders of adjustments (including key-only)
# and select the one with the best AIC.
#
# Truncation set at 3% for 2006 due to large outlier, trucation did not improve fit
# for any other years
#
moplOO hn <- ds(subset(mopl,year =="2000"), key = "hn", adjustment = "cos", truncation =
0%)
moplOO hnh <- ds(subset(mopl,year =="2000"), key = "hn", adjustment = "herm", truncation =
0%)
moplOO hr <- ds(subset(mopl,year =="2000"), key = "hr", adjustment = "poly", truncation =
0%)
moplOO unif <- ds(subset(mopl,year =="2000"), key = "unif", adjustment = "cos", truncation
0%)
moplOl hn <- ds(subset(mopl,year =="2001"), key = "hn", adjustment = "cos", truncation =
0%)
moplOl hnh <- ds(subset(mopl,year =="2001"), key = "hn", adjustment = "herm", truncation =
0%)
moplOl hr <- ds(subset(mopl,year =="2001"), key = "hr", adjustment = "poly", truncation =
0%)
moplOl unif <- ds(subset(mopl,year =="2001"), key = "unif", adjustment = "cos", truncation
0%)
mopl02 hn <- ds(subset(mopl,year =="2002"), key = "hn", adjustment = "cos", truncation =
0%)
mopl02 hnh <- ds(subset(mopl,year =="2002"), key = "hn", adjustment = "herm", truncation =
0%)
mopl02 hr <- ds(subset(mopl,year =="2002"), key = "hr", adjustment = "poly", truncation =
0%)
mopl02 unif <- ds(subset(mopl,year =="2002"), key = "unif", adjustment = "cos", truncation
0%)
mopl03 hn <- ds(subset(mopl,year =="2003"), key = "hn", adjustment = "cos", truncation =
0%)
mopl03 hnh <- ds(subset(mopl,year =="2003"), key = "hn", adjustment = "herm", truncation =
0%)


mopl03 hr <- 0%) mopl03 unif 0%) ds(subset(mopl,year == "2003"), key = ' 'hr", adjustment = " poly", truncation =
<- ds(subset(mopl,year : ==2003; ) key = = "unif", adjustment = "cos ", truncation
mopl04 hn <- 0%) mopl04 hnh < 0%) mopl04 hr <- 0%) mopl04 unif 0%) ds(subset(mopl,year == 2004), key = ' 'hn", adjustment = " cos", truncation =
- ds ( subset (mopl, year =: =2004), , key = "hn", adjustment = "herm", truncation =
ds(subset(mopl,year == 2004), key = ' 'hr", adjustment = " poly", truncation =
<- ds(subset(mopl,year : ==2004; ) key = = "unif", adjustment = "cos ", truncation
mopl05 hn <- 0%) mopl05 hnh < 0%) mopl05 hr <- 0%) mopl05 unif 0%) ds(subset(mopl,year == 2005), key = ' 'hn", adjustment = " cos", truncation =
- ds ( subset (mopl, year =: =2005), , key = "hn", adjustment = "herm", truncation =
ds(subset(mopl,year == 2005), key = ' 'hr", adjustment = " poly", truncation =
<- ds(subset(mopl,year : ==2005; ) key = = "unif", adjustment = "cos ", truncation
mopl06 hn <- ds(subset(mopl,year == 2006), key = ' 'hn", adjustment = " cos", truncation = 0
mopl06 hnh < - ds ( subset (mopl, year =: =2006), , key = "hn", adjustment = "herm", truncation =
0.3)
mopl06 hr <- ds(subset(mopl,year == 2006), key = ' 'hr", adjustment = " poly", truncation =
0.3)
mopl06 unif <- ds(subset(mopl,year : ==2006; ) key = = "unif", adjustment = "cos ", truncation
0.3)
mopl2016 hn <- ds(subset(mopl,year : ==2016; ) key = = "hn", adjustment = "cos", truncation =
list(left = 0%, right = 0%))
mopl2016 hnh <- ds(subset(mopl,year ==2016 ' ) key = "hn", adjustment = "herm ", truncation
list(left = 0%, right = 0%))
mop!2016 hr <- ds(subset(mopl,year : ==2016 ; ) key = = "hr", adjustment = "poly" , truncation :
list(left = "0%", right = "0%"))
mopl2016 unit <- ds(subset(mopl,year =="2016"), key = "unit", adjustment = "cos", truncation = list(left = 0%, right = 0%))
# create list of models for each year to average estimates
mlistOO <- list(mopl00 hn, moplOO hnh, moplOO hr, moplOO unif)
mlistOl <- list(mopl01 hn, moplOl hr, moplOl unif)
mlist02 <- list(mopl02_hn, mopl02_hr, mopl02_unif)
mlist03 <- list(mopl03 hn, mopl03 hr, mopl03 unif)
mlist04 <- list(mopl04 hn, mopl04 hr, mopl04 unif)
mlist05 <- list(mopl05 hn, mopl05 hnh, mopl05 hr, mopl05 unif)
mlist06 <- list(mopl06 hn, mopl06 hr, mopl06 unif)
mlistl6 <- list(mopl2016_hn, mopl2016_hr, mopl2016_unif)
#function that pulls estimates from ds model objects getEstimates <- function(model){
summary <- summary(model) p <- summary[[1]]$average p p.se <- summary[[1]]$average.p.se
D <- summary[[2]][[1]]$D$Estimate D.lower <- summary[[2]][[1]]$D$lcl D.upper <- summary[[2]][[1]]$D$ucl
N <- summary[[2]][[1]]$N$Estimate N.lower <- summary[[2]][[1]]$N$lcl N.upper <- summary[[2]][[1]]$N$ucl
LL <- logLik(model)
47


data <- data.frame( estimate = c("p", "D","N","LL"),value = c(p,D,N,LL), lowerCI = c (NA, D. lower,N. lower,NA) upperCI = c(NA,D.upper,N.upper,NA))
# function that calculates model average of estmates and their log-normal CIs from list of models
# this uses the AICcmodavg package avgModels <- function(mlist){
estimates <- lapply(mlist, FUN = getEstimates)
LLs <- sapply(seg along(mlist), function(i) mlist[[i]]$ddf$lnl)
modNames <- sapply(seg along(mlist), function(i) mlist[[i]]$ddf$name.message)
estimates.Pa <- sapply(seg along(mlist), function(i) mlist[[i]]$dht$individuals$average.p) estimates.D <- sapply(seg along(mlist), function(i) mlist[[i]]$dht$individuals$D$Estimate) estimates.N <- sapply(seg along(mlist), function(i) mlist[[i]]$dht$individuals$N$Estimate) estimates.esw <- estimates.Pa sapply(seg along(mlist), function(i) mlist [ [ i] ]$ddf$ds$aux$int.range[2] )
se.Pa <- sapply(seg along(mlist), function(i) summary(mlist[[i]])[[1]]$average.p.se)
se.D <- sapply(seg along(mlist), function(i) mlist[[i]]$dht$individuals$D$se)
se.N <- sapply(seg along(mlist), function(i) mlist[[i]]$dht$individuals$N$se)
se.esw <- se.Pa sapply(seg along(mlist), function(i) mlist[[i]]$ddf$ds$aux$int.range[2])
K <- sapply(seg along(mlist), function(i) length(mlist[[i]]$ddf$par)) nobs <- sapply(seg along(mlist), function(i) nrow(mlist[[i]]$ddf$data))
logLimits <- function(Est,SE){
logVar <- loglp ( SE/',2/Est/',2 )
C <- exp(1.96 *sqrt(logVar))
return(c(logLL = Est/C, LogUL = Est C))
detectionProb <- modavgCustom(logL = LLs,K = K,modnames = modNames,estimate = estimates.Pa,se = se.Pa, nobs = nobs,second.ord = F)
density <- modavgCustom(logL = LLs,K = K, modnames = modNames, estimate = estimates.D, se = se.D, nobs = nobs, second.ord = F)
nhat <- modavgCustom(logL = LLs, K = K, modnames = modNames, estimate = estimates.N, se = se.N, nobs = nobs, second.ord = F)
esw <- modavgCustom(logL = LLs, K = K, modnames = modNames, estimate = estimates.esw, se = se.esw, nobs = nobs, second.ord = F)
print(detectionProb) print(density) print(nhat) print(esw)
avgEstimates <- data.frame(estimate = c("Pa","D","N", "ESW", "ER"),
modavg = c(detectionProb$Mod.avg.est, density$Mod.avg.est, nhat$Mod.avg.est, esw$Mod.avg.est, mlist[[1]]$dht$individuals$summary$ER),
SE = c(detectionProb$Uncond.SE, density$Uncond.SE, nhat$Uncond.SE, esw$Uncond.SE, mlist[[1]]$dht$individuals$summary$se.ER),
LL = c(detectionProb$Lower.CL, density$Lower.CL, nhat$Lower.CL,
esw$Lower.CL, NA),
esw$Upper.CL, NA)
UL = c(detectionProb$Upper.CL, density$Upper.CL, nhat$Upper.CL,
logLL =
c(logLimits(detectionProb$Mod.avg.est,detectionProb$Uncond.SE)[1],
logLimits(density$Mod.avg.est,density$Uncond.SE)[1], logLimits(nhat$Mod.avg.est,nhat$Uncond.SE)[1], logLimits(esw$Mod.avg.est,esw$Uncond.SE)[1], NA),
logUL =
c(logLimits(detectionProb$Mod.avg.est,detectionProb$Uncond.SE)[2],
logLimits(density$Mod.avg.est,density$Uncond.SE)[2],
logLimits(nhat$Mod.avg.est,nhat$Uncond.SE)[2], logLimits(esw$Mod.avg.est,esw$Uncond.SE)[2],NA))
48


avgEstimates$cv <- avgEstimates$SE/avgEstimates$modavg return(avgEstimates)
}
#get model averaged estimates for each year ModavgOO <- avgModels(mlistOO)
ModavgOl <- avgModels(mlistOl)
Modavg02 <- avgModels(mlist02)
Modavg03 <- avgModels(mlist03)
Modavg04 <- avgModels(mlist04)
Modavg05 <- avgModels(mlist05)
Modavg06 <- avgModels(mlist06)
Modavgl6 <- avgModels(mlistl6)
#combine it all into one dataframe
ModavgAll <- bind rows("2000" = ModavgOO, "2001" = ModavgOl, "2002" = Modavg02
2003 = Modavg03, 2004 = Modavg04, 2005 = Modavg05
2006 = Modavg06, 2016 = Modavgl6, .id = "year")
49


APPENDIX E
Nest survival analyses R code.
#################################################################################
# #
# Mountain Plover nest survival analysis 2017 #
# author: Allison Pierce #
# #
#################################################################################
library(RMark) library(tidyverse) library(forcats) library(readxl) library(lubridate)
#read in nest monitoring data load("mopl.rda")
#read in weather data
SPWeather <- read.csv("SPWeather.csv", na.strings = c("9999","-9999")) SPWeather$date <- as Date ( as character ( SPWeather$DATE ), format = ,,%Y%m%d")
#select relevant columns and subset by station
weather <- dplyr::select(SPWeather,STATION_NAME,date,PRCP,TMIN) antero <- filter(weather,STATION_NAME == "ANTERO RESERVOIR CO US) fairplay <- filter(weather,STATION_NAME != "ANTERO RESERVOIR CO US)
#24 period ends at observation time (7:30 AM)
#Prev night/day (reported on day of record)
prevdayweatherl5 <- subset(antero, date > as.Date("2015-06-01") & date < as.Date("2015-07-
24") )
prevdayweatherl6 <- subset(antero, date > as.Date("2016-06-01") & date < as.Date("2016-07-
24") )
#current day reported on next day of record
weatherl5 <- subset(antero, date > as.Date("2015-06-02") & date weatherl6 <- subset(antero, date > as.Date("2016-06-02") & date
as.Date("2015-07-25"))
as.Date("2016-07-25"))
#add previous day measures to data frame for each year
weatherl5$prevTMIN <- prevdayweatherl5$TMIN weatherl6$prevTMIN <- prevdayweatherl6$TMIN weatherl5$prevPRCP <- prevdayweatherl5$PRCP weatherl6$prevPRCP <- prevdayweatherl6$PROP
## Get weather for FAIRPLAY to fill in NA's in 2015
fpprevdayweatherl5 <- subset(fairplay, date > as.Date("2015-06-01") & date < as.Date("2015-07-24"))
fpweatherl5 <- subset(fairplay, date > as.Date("2015-06-02") & date < as.Date("2015-07-25")) fpweatherl5$prevTMIN <- fpprevdayweatherl5$TMIN
#Fill in NA's for 2015 with data from Fairplay missing <- which(is.na(weatherl5), arr.ind = T) weatherl5[missing] <- as.numeric(fpweatherl5[missing])
#bind weather in repeating order for adding the design data weatherbind <- rbind(weather15,weather16)
#prepare nest data to add in time varying covariates for RMark
mopl dp <- process.data(mopl,nocc=max(mopl$LastCheeked),mode1="Nest", groups=c("Year","Tag")) mopl ddl <- make.design.data(mopl dp)
#create data frame of time varying covariates
occasion year <- data.frame(group = mopl ddl$S$group, time = mopl ddl$S$time, PRCP =
weatherbind$ PRCP,
prevPRCP = weatherbind$prevPRCP, prevTMIN = weatherbind$prevTMIN, TMIN = weatherbind$TMIN, date = weatherbind$date)
50


#merge time varying covariates to design matrix
mopl ddl$S <- merge design.covariates(mopl ddl$S,occasion year, bytime = T, bygroup = T)
#create guadratic time covariate
mopl_ddl$S$TT<-as.numeric(mopl_ddl$S$Time)/"2
#define nest survival models for RMark mopl models <- function(){
#DSR is constant
dot <- mark(mopl dp,mopl ddl,nocc=max(mopl$LastChecked),model = "Nest" profile.int = T) ###### Temporal models (year and seasonal)
#DSR is constant over season but varies by year
Year <- mark(mopl dp,mopl ddl,nocc=max(mopl$LastCheeked),mode1="Nest",
model. parameters = list ( S=list ( formula=~Year ) ) groups=,,Year" profile .int = T )
#DSR varies linearly over time and Year
Year.T <- mark(mopl dp,mopl ddl,nocc=max(mopl$LastCheeked),mode1="Nest",
model. parameters = list ( S=list ( formula=~Year + Time ) ) groups = ,,Year" ,
profile.int = T)
#DSR varies guadratically over time and Year
Year.TT <- mark(mopl dp,mopl ddl,nocc=max(mopl$LastCheeked),mode1="Nest",
model. parameters = list ( S=list ( formula=~Year + TT ) ) groups = ,,Year" ,
profile.int = T)
#DSR varies linearly over time
Time <- mark(mopl dp,mopl ddl,nocc=max(mopl$LastCheeked),mode1="Nest",
model. parameters = list ( S=list ( formula=~Time ) ) groups=,,Year" profile.int = T)
#DSR varies guadratically over time
TT <- mark(mopl dp,mopl ddl,nocc=max(mopl$LastCheeked),mode1="Nest",
model. parameters = list ( S=list ( formula=~TT ) ) groups=,,Year" profile.int = T)
#DSR varies by age of nest
NestAge <- mark(mopl dp,mopl ddl,nocc=max(mopl$LastCheeked),mode1="Nest",
model.parameters=list(S=list(formula=~NestAge)), profile.int = T)
###### Weather models
#DSR varies with day of precipitation
PRCP <- mark(mopl dp,mopl ddl,nocc=max(mopl$LastCheeked),mode1="Nest",
model.parameters=list(S=list(formula=~PRCP)), profile.int = T)
#DSR varies with prev day precipitation
prevPRCP <- mark(mopl dp,mopl ddl,nocc=max(mopl$LastCheeked),mode1="Nest",
model.parameters=list(S=list(formula=~prevPRCP)), profile.int = T)
#DSR varies with day of min temperature
TMIN <- mark(mopl dp,mopl ddl,nocc=max(mopl$LastCheeked),mode1="Nest",
model.parameters=list(S=list(formula=~ TMIN)), profile.int = T)
#DSR varies with prev day min temperature
prevTMIN <- mark(mopl dp,mopl ddl,nocc=max(mopl$LastCheeked),mode1="Nest",
model.parameters=list(S=list(formula=~ prevTMIN)), profile.int = T)
#DSR varies with day of min temp and precip (additive)
TMIN.PRCP <- mark(mopl dp,mopl ddl,nocc=max(mopl$LastCheeked),model="Nest",
model.parameters=list(S=list(formula=- TMIN + PRCP)), profile.int = T)
#DSR varies with prev day min temp and precip (additive)
prevTMIN.prevPRCP <- mark(mopl dp,mopl ddl,nocc=max(mopl$LastCheeked),mode1="Nest",
model.parameters=list(S=list(formula=~ prevTMIN + prevPRCP)),
profile.int = T)
### GPS Tag
#DSR varies with application of GPS tag
51


Tag <- mark(mopl dp,mopl ddl,nocc=max(mopl$LastCheeked),mode1="Nest",
model.parameters=list(S=list(formula=~ Tag)), groups = "Tag", profile.int = T)
### models analogous to other papers
drietz
= T)
mark(mopl dp,mopl ddl,nocc=max(mopl$LastCheeked),mode1="Nest",
model.parameters=list(S=list(formula=~ Time + prevPRCP + TMIN)), profile.int
dinsmore = T)
mark(mopl dp,mopl ddl,nocc=max(mopl$LastCheeked),mode1="Nest",
model.parameters=list(S=list(formula=~ NestAge + TT + PRCP)), profile.int
return(collect.models(table = T))
#run models
mopl results <- mopl models()
#collect beta estimates from all models getBetas <- function(models){ betatable <- data.frame(NULL) for(i in 1:length(models)){
betas <- models[[i]]$results$beta
betas$parameter <- dimnames(models[[i]]$results$beta)[[1]] betas$model <- models[[i]]$model.name betas$AICc <- models[[i]]$results$AICc betatable <- rbind(betatable, betas)
}
return(betatable)
betadata <- getBetas(mopl results) %>% arrange(AICc)
betadata <- betadata[c(6,5,1,2,3,4,7)] lereate DSR plots #### DSR prev tmin
s.temp <- coef(mopl results$prevTMIN)
tminrange <- seg(min(weatherbind$prevTMIN),max(weatherbind$prevTMIN),0.5) logit.values <- s.temp[l,l] + tminrange s.temp[2,l] deriv <- matrix(1,ncol=2,nrow=length(tminrange)) deriv[,2] <- tminrange
std.errors <- sqrt(diag(deriv%*%mopl_results$prevTMIN$results$beta.vcv[l:2,l:2]%*%t(deriv) ) ) lcl.logit <- logit.values-1.9 6*std.errors ucl.logit <- logit.values + 1.9 6*std.errors
tminmodel <- data.frame(Surv = plogis(logit.values),
tminrange = tminrange,
LCL = plogis(lcl.logit),
UCL = plogis(ucl.logit))
ggplot(data = tminmodel, aes(y = Surv, x = tminrange, ymin = 0.7, ymax = 1)) + geom ribbon(aes(ymin = LCL, ymax = UCL), alpha = 0.5) + geom line(size = 0.8) + theme classic() +
scale x continuous(limits = c(min(weatherbind$prevTMIN),max(weatherbind$prevTMIN)), breaks =
seq(ceiling(min(weatherbind$prevTMIN)),floor(max(weatherbind$prevTMIN)),2)) + scale y continuous(breaks = seq(0.75,1,0.05)) +
labs(x = expression("Daily Minimum Temperature ("*degree*C*")"), y = "Daily Survival Rate (DSR)") +
theme(legend.position = c(0.7,0.3),
text = element text(size = 10))
#### DSR as a function of prev tmin by prev precip (none, light, med, heavy) s.temp <- coef(mopl results$prevTMIN.prevPRCP)
tminrange <- seq(min(weatherbind$prevTMIN),max(weatherbind$prevTMIN),0.5) logit.values.N <- s.temp[l,l] + tminrange s.temp[2,l] + 0 s.temp[3,l]
52


logit.values.L <- s.temp[l,l] + tminrange * s.temp[2,l] + 1.25 s.temp[3,l]
logit.values.M <- s.temp[l,l] + tminrange * s.temp[2,l] + 5 s.temp[3,l]
logit.values.H <- s.temp[l,l] + tminrange * s.temp[2,l] + 25 s.temp[3,l]
deriv <- matrix(1,ncol=2,nrow=length(tminrange)) deriv[,2] <- tminrange std.errors <-
sqrt(diag(deriv%* %mopl_resuits$prevTMIN.prevPRCP$results$beta.vcv[1:2,1:2] lcl.logit.N <- logit.values.N-l.9 6*std.errors ucl.logit.N <- logit.values.N+l.9 6*std.errors
It(deriv)))
lcl
ucl
lcl
ucl
logi
logi
logi
logi
t.L <- logi
t.L <- logi
t.M <- logi t.M <- logi
t.values.L-l.9 6* std.errors t.values.L+l.9 6* std.errors
t.values.M-l.9 6* std.errors t.values.M+l.9 6* std.errors
lcl.logit.H <- logit.values.H-l.9 6*std.errors ucl.logit.H <- logit.values.H+l.9 6*std.errors
preciplevel <- factor(rep(c("no","light","mod","heavy") each = length(logit.values.N))) preciplevel <- fct relevel(preciplevel,"heavy","mod","light","no") tminrange <- rep(tminrange,4)
tminprcpmodel <- data.frame(Surv =
plogis(c(logit.values.N,logit.values.L,logit.values.M,logit.values. H ) ) ,
tminrange = tminrange,
LCL = plogis(c(lcl.logit.N,lcl.logit.L,lcl.logit.M,lcl.logit.H)), UCL = plogis(c(ucl.logit.N,ucl.logit.L,ucl.logit.M,ucl.logit.H)) : preciplevel)
ggplot(data = tminprcpmodel, aes(y = Surv, x = tminrange, color = Precipitation, ymin = 0.70, ymax = 1)) +
geom line(size = 0.8) + theme classic() +
scale x continuous(limits = c(min(weatherbind$prevTMIN),max(weatherbind$prevTMIN)), breaks =
seq(ceiling(min(weatherbind$prevTMIN)),floor(max(weatherbind$prevTMIN)),2)) + scale y continuous(breaks = seq(0.75,1,0.05)) + scale color grey(name = "Total Precipitation",
labels = c("Heavy (25mm)", "Moderate (5 mm)", "Light (1.25 mm)","None (0
mm)"),
start = 0, end = 0.8) +
labs(x = expression("Daily Minimum Temperature ("*degree*C*")"), y = "Daily Survival Rate
(DSR)") +
theme(legend.position = c(0.7,0.3),
text = element text(size = 10))
53


Full Text

PAGE 1

POPULATION STATUS OF MOUNTAIN PLOVER IN SOUTH PARK, COLORADO by ALLISON KATHERINE PIERCE B.S., University of Colorado Denver, 2013 A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of t he requirements for the degree of Master of Science Biology Program 2017

PAGE 2

ii This thesis for the Master of Science degree by Allison Katherine Pierce h as been approved for the Biology Program b y Michael B. Wunder, Chair Michael Greene Stephen J. Dinsm ore Date: May 13, 2017

PAGE 3

iii Pierce, Allison Katherine (M.S., Biology Program ) Population S tatus of Mountain Plover B reeding in South Park, Colorado Thesis directed by Associate Professor Michael B. Wunder ABSTRACT Mountain Plovers ( Charadrius montanus ) are mi gratory shorebirds of conservation concern that breed on grasslands and xeric tablelands along the western edge of the Great Plains Here, I provide an updated status for a population of plovers breeding on high elevation intermountain rangelands in South Park, Park County, Colorado To estimate demographic changes over time, I estimated plover breeding density using distance sampling data collected from historic surveys conducted from 2000 to 2006 and from a survey I conducted in 2016. Mean d ensity was hig hest in 2001 (12.8 plovers km 2 95% CI = 6.57 21.7, 35% CV), over 4.5 times higher than the estimated mean density from 2016 (2.8 plovers km 2 95% CI = 1.54 4.57 31 % CV ). Estimates across years provide weak evidence for a ne gative trend in the densi ty of Mountain P lovers in South Park. I also estimated the effects of time and weather on daily nest survival probability. Mean expected n est survival probability for 2015 and 2016 was 0. 22 D aily minimum temperature best predicted variation in daily nest survival in South Park ; survival odds declined with decreasing temperature. This finding coupled with reported negative effects of maximum temperatures on survival of nests at lower elevations in Colorado suggests that extreme temperatures may be a limiti ng factor for plover nest survival The form and content of this abstract are approved. I recommend its publication. Approved: Michael B. Wunder

PAGE 4

iv DEDICATION To my mother and grandmother, women of admirable strength ; and to my par tner in life, Jesse Calki n, for his unconditional love and encouragement

PAGE 5

v ACKNOWLEDGEMENTS I am thankful for s upport and funding that was provided for this work by the University of Colorado Denver, Bird Conservancy of the Rockies, and the Denver Field Ornithologists. I would also like the thank the wonderful folks with Colorado Parks and Wildlife for providing logistical support, especially Mark Lamb and Karl Copeman I am grateful to Angela Dwyer for her assistance with field logistics and funding. I thank Emery Young for his field assistance and company. I am also grateful to the late Fritz Knopf for enthusiastic encouragement he gave me in our correspondence during my first field season. His excitement for plovers instilled in me a great fondness for the species In additio n, I thank my committee members for their advice, guidance, and encourage ment with the development and execution of this work. I thank Mike Greene for his infectious enthusiasm and assistance in navigating logistics of proposed laboratory work I thank Ste ve Dinsmore for his expertise not only in plover ecology but also in modeling techniques used in this project. And of course, I am profoundly grateful to my advisor and mentor Mike Wunder for continually challenging me to grow beyond the limits I've set fo r myself and having faith in my success I look forward to my next academic adventure working with you. I am grateful to Libby Pansing and all my other wonderful student colleagues too numerous to name who helped form this work either through thoughtful feedback or being my sounding board when I needed it. All fieldwork and animal handling protocols for this work were approved by the University Institutional Animal Care and Use Committee under p rotocol number: 92014(05)1C.

PAGE 6

vi TABLE OF CONTENTS CHAPTER I. DENSITY TRENDS IN TH E HIGH ELEVATION POPULATION OF MOUNTAIN PLOVERS IN COLORADO ................................ ................................ ........................ 1 Introduction ................................ ................................ ................................ ................... 1 Methods ................................ ................................ ................................ ......................... 3 Results ................................ ................................ ................................ ........................... 7 Discussion ................................ ................................ ................................ ..................... 8 II. MOUNTAIN PLOVER NEST SURVIVAL IN A HIGH ELEVATION POPULATION ................................ ................................ ................................ ........... 16 Introduction ................................ ................................ ................................ ................. 16 Methods ................................ ................................ ................................ ....................... 18 Results ................................ ................................ ................................ ......................... 23 Discussion ................................ ................................ ................................ ................... 25 REFERENCES ................................ ................................ ................................ ....................... 35 APPENDIX ................................ ................................ ................................ ............................. 38 A. Distance sampling model selection results for all years ................................ ....... 38 B. R outpu t summaries for most parsimonious distance sampling models ............... 40 C. Parameter estimat es for all nest survival models ................................ .................. 44 D. Di stance sampling analyses R code ................................ ................................ ...... 46 E. Nest survival a nalyses R code ................................ ................................ ............. 50

PAGE 7

1 CHAPTER I DENSITY TRENDS IN TH E HIGH ELEVATION POPULATION OF MOUNTAIN PLOVERS IN COLORADO Introduction The Mountain Plover is an uncommon upland shorebird that is considered a species of conservation conc ern in many of the states where it breeds and overwinters, due in part to apparent population declines (Colorado Parks and Wildlife 2015 ; Andres and Stone 2010) This plover breeds along the western edge of the Great Plains and winters in areas along the southern border of the United States an d northern Mexico (Knopf and Wunder 2006) The global population size is estimated to range from 11000 to 14000, but may number up to 18000 (Plumb et al. 2005; Tipton et al. 2009; Andres and Stone 2010) Colorado is estimated to support more than half of the global population of breeding plovers (Tipton et al. 2009) thus, conservation of plove rs in Colorado is of particular interest. Anecdotal records suggest Mountain Plover population declines dating back to the early 1900s (Cooke 19 14; Abbott 1939) More contemporary quantitative studies such as the Breeding Bird Survey estimate that continental populations have declined 3.1% annually from 1996 to 2012 (Sauer et al. 2012). These steep declines have corresponded with observed declines in the population on the Pawnee National Grassland (PNG) once considered the stronghold of the Colorado breeding population (Graul and Webster 1976; Knopf 1994; Knopf and Wunder 2006) The PNG decline is mainly attributed to altered grazing regimes of large herbivores and reduced populations of prairie dogs, both of whic h contribute to maintaining the bare ground and low vegetation habitats preferred by plovers.

PAGE 8

2 Despite local PNG declines, comparatively large breeding populations in other areas of eastern Colorado and South Park, Park County remain. Much of the Colorado population breeds on the eastern plains (!8600 individuals) H owever, the breeding population of ~2300 in South Park occurs at the highest density across the range, 7.90 0.90 (SE) plovers/ km 2 (Wunder et al. 2003; Tipton et al. 2009) likely attributed to differences between breeding habitats. The high elevation intermountain grasslands in South Park are geographically separated and constrained by mount ainous and forested habitat, whereas breeding habitats in eastern Colorado comprise a patchwork of short grass prairie, black tailed prairie dog colonies ( Cynomys ludovicianus ), and fallow agricultural fields spread out over ~81,200 km 2 (Tipton et al. 2009) Furthermore, South Park has a relatively milder climate and has experienced minor historic change compared to habitats on the eastern plains (Wunder et al. 2003) Because of these differences between habitats, continental trends may not reflect local population dynamics and trajectories. Moreover, large scale roadside surveys often produce unreliable estimates for cryp tic species with low detectability such as the Mountain Plover. Developing statistically robust surveys that cover the entirety of the range of a sparsely and variably distributed bird to accurately estimate the direction and magnitude of population change is financially and logistically unrealistic. However, long term monitoring of key local populations across space and time may be manageable, yet still informative about population dynamics of the species. Although previous research suggested the South Par k population size was stable (Wunder et al. 2003; Dinsmore et al. 2 010) the population size had not been estimated since 2006. The goal of this study was to provide an update d status of the population of Mountain Plover breeding in South Park, Colorado. Here I provide estimates

PAGE 9

3 for density and abundance of plovers for p art of the South Park population and directly compare them with previous estimates for the same area. Methods Study a rea I studied Mountain Plovers breeding on public lands in South Park, Park County, Colorado ( 39¡ 9' 19.296" N, 105¡ 50' 51.72" W) Breedi ng habitat in South Park is comprised of high elevation intermountain grassland characterized by short stature bunch grasses and small shrubs (see Wunder et al. 2003 for more detail ). The primary area of interest of m y study is located within a corridor of M ountain P lover habitat adjacent to a forested ridge in the South Park Basin (Fig 1 ). This area comprises the western portion of the James Mark Jones State Wildlife Area (JMJSWA) as well as grazing pastures managed by the Bureau of Land Management (BLM). This area was selected for re establishing a long term study of Mountain Plover in South Park because it represents publically accessible habitat identified as supporting a high density of breeding plovers and was of high conservation value (Grunau and Wunder 2001) Histo ric s urveys To evaluate the current population status of Mountain Plover in South Park I used historic survey data collected from 2000 to 2006 to compare with estimates generated in 2016. Historically, Mountain Plover density and abundance were estimated using distance sampling along eight randomly generated transects in potentially suitable habitat on public and private lands throughout South Park (Wunder et al. 2003) This method models detection probability as a function of perpendicular distance of an object of interest from the transect line to adjust for imperfect detection during surveys (Buckland et al. 2001) A ssumption s for

PAGE 10

4 these models include (1) all birds on the line are detected with certainty, (2) birds are detected at their initial location, and (3) distances are measured accurately. Plovers are wary of surveyors on foot but are tolerant of slow moving vehicles T herefore, to ensure satisfaction of the first two assumptions, surveys were conducted from a vehicle driven at <15 km hr 1 To maximize detectability, sampling was conducted early in the morning and during the brood rearing stage in July when plovers are conspicuous. When broods or small flocks were encountered, chicks and juveniles were easily identified and not included as detections in surveys to ensure estimates reflected the density of poten tially breeding plovers. Perpendicular distances were derived from distances measured from the observer using a laser rangefinder and a standard compass to measure sighting angle (Wunder et al. 2003) Contemporary s urveys The study in 2016 was designed for consistency with historic surveys but with changes to increase estimate precision and spatial coverage Buckland et al. (2001) recommend a minimum of 10 20 transects to reliably estimate encounter variance, more if species are patchily distributed. Mountain Plovers are loosely colonial and thus are unlikely to be randomly distributed (Graul 1975) I therefore systematically placed 38 s hort east west parallel transects across the study area. Transects ranged in length from 0.6 km to 3.6 km and were spaced 500 m apart. We intended to survey all transects two times but were only a ble complete the second survey for 16 transects due to time constraints Survey effort totaled 82 km and was conducted by a single observer (AKP). Transects were placed in an east west orientation so that they ran perpendicular to the forested ridge that c omprises the western edge of the study area. Plovers are less likely to occur near forested habitat, so if an abundance gradient in plover distribution existed, the

PAGE 11

5 survey lines would run perpendicular to that gradient. Transects extended across the width of the study area and terminated at fence borders on the east, and at the edge of forested habitat on the west. Surveys were conducted in 2016 between 5 July and 15 July to maximize the probability of meeting the closure assumption. Similar to historic s urveys, s ampling was conducted from a vehicle driven at <15 km hr 1 and constrained to ~4 hours following sunrise or preceding sunset during low wind conditions with <10% cloud cover to maximize the detectability of plovers. I surveyed transects in order f rom North to South to maximize survey efficiency and alternated transect direction to adjust for potential bias from sun glare. To increase number of detections within the small study area, I surveyed transects in the morning and in the evening The number of surveys completed in one day varied due to transect length and conditions during morning and evening survey periods. Perpendicular distance of a plover to the line was derived from sighting distance and angle using a laser ranger finder ( Bushnell Cor po ration, Overland Park, Kansas; rated to 500 m ) and a standard compass. In all surveys, detection distance was not fixed and all detections regardless of distance from the line were recorded. Statistical a nalysis To estimate historic density and abundance I used distance sampling data collected in 2000 to 2006. Because m y study area represents a small portion of Mountain Plover habitat in South Park, I only included historic data from transects that intersected my smaller study area, reducing the number of t ransects from eight to three. Logistic and financial restraints in 2006 prevented a complete re survey and data were available from only two transects for that year.

PAGE 12

6 I used the R package Distance' ( R version 3.3.2; Miller 2016) to estimate Mountain Plover den sity in the study area for 2000 2006 and 2016. I modeled detection probability as a function of exact distances using key functions sugges ted by Buckland et al. to be robust models of detection : 1) Half normal with cosine adjustments 2) Half normal with hermite adjustments 3) Hazard rate with polynomial adjustments and 4) Uniform with cosine adjustments (2001:155) Buckland et al. (2001:103) also suggest that in cases where detection width is not fixed, data may be right truncated to exclude 5 10% of furthest detect ions or where detection probability is estimated at 0.15 to improve the fit of the detection function. I truncated one dataset (2006) to 300 m, which removed two detections. I did not truncate other datasets as there were no obvious outliers and when I com pared models with truncated data vs untruncated data, there was no improvement in detection model fit but some loss in precision with density estimates due to reduction in detection sample size. Because of differences in observers across years, I estimated the detection function separately for each year. The ds() function in package:Distance fits key functions using a forward stepping procedure for inclusion of the specified adjustment terms and selects the most parsimonious model based on Akaike Informatio n Criterion (AIC) scores. This process resulted in some candidate model sets containing key function models with no adjustments (Appendix A) I used AIC to rank the models in the set for each year (Burnham and Anderson 2002) and evaluated model fit to the data using a CramÂŽr von Mises test (P value range: 0.53 0.95 ; see Appendices A and B ). Most candidate models were within 2 "AIC of each other so I averaged den sity and abundance predictions using Akaike weights. I constructed unconditional log normal 95% confidence intervals for each averaged estimate to

PAGE 13

7 account for increased variance from model selection uncertainty (Buckland et al. 2001: 77; Burnham and Anderson 2002: 164; Buckland et al. 20 15: 106) I estimate d plover a bundance in the study area from density estimates extrapolated to area of potentially suitable habitat within the study site. This area was defined earlier (see Wunder et al. 2003), and was marginally larger than the area cov ered by the 2016 transect design (30.8 km 2 and 38.8 km 2 respectively). Results I recorded 51 detections of Mountain Plover along 82 km of transects in 2016. The average number of detections from the sub sampled historic data set was 80 (SD = 25.4) and ra nged from a low of 26 (in 2006) to a high of 101 (in 2002). The uniform key function with cosine adjustments was the top ranked model for every year except 2003 and 2004, however, almost all key functions modeled the data well (Fig. 2; Buckland et al. 2015 : 61). Estimated detection probability ranged from 0.39 (2001) to 0.64 (2004) ; 95% log normal confidence intervals around these estimates overlapped (Fig 3 ). The mean estimated distance that all plovers within are detected (effective strip width) was 138 m (SD = 2.3). Histograms of years 2002, 2004, 2005, and 2016 suggest some avoidance of the li ne by plovers evidenced by less detections in the first bin (Fig. 2). However, estimation bias from this behavior was likely minimal because avoidance did not app ear to move plovers beyond the detection range and fitted detection functions were monotonically constrained (Buckland et al 2015:204). The highest estimated density within the study are a was in 2001 (D = 12.8 plovers km 2 95% CI = 6.57 21.7, 35% CV) m ore than 4.5 times greater than the estimated density from 2016 (2.8 plovers km 2 95% CI = 1.54 4.57 31 % CV ). Overall, there was a weak

PAGE 14

8 negative trend in density within the study area over time (Fig. 4). Historic point estimates of density constrained to the 2016 study area were marginally greater than those estimated from distance sampling data from all eight transects distributed across South Park but all 95% confidence intervals overlapped (Fig. 4 ). By extrapolation, I est imated 110 plovers (95% CI 6 0 177) occupied the 38.8 km 2 of potentially suitable habitat within m y study area in 2016. Discussion My results suggest that Mountain Plover density has declined in the study area during the past 17 years. Point estimates for density from 2000 to 2006 for all of South Park suggest a similar trend (Fig. 4). The density estimate of 2.8 plovers km 2 ( 95% CI = 1.54 4.57) in 2016 is lower than the pa rk wide density of 7.89 plovers km 2 ( 95% CI = 4.9 9.9) for 2000 2003 as reported by Wunder et al. (2003 ). Historic estimates of density from the study area were similar to those estimated from data collected across the South Park region (Fig. 4), suggesting that monitoring at this study site may provide a reasonable index of population trends in the overall population of S outh P ark. Reducing survey scope to a smaller patch of habitat from sites representative to the entire South Park area has allowed us to employ a more robust distance sampling design, increased power to detect temporal variation in plover d ensity, and decreased logistical demands on surveys. A limitation of this strategy is that habitat in my study area may not always be representative of the entirety of South Park in the future because I will not detect any changes in distribution or densit y of plovers in areas of South Park outside of my study site. Although c urrent South P ark density estimates are lower than historic estimates they still compare with those reported elsewhere. Plover density was estimated at 2.0 1.61 (SE) plovers km 2 for grazed rangelands on t he PNG for the period from 1990 to 1994 (Knopf and

PAGE 15

9 Wunder 2006) More recently, Tipton et al. (2009) estimated plover densities in three different habitat types across eastern Colorado: 2.26 plovers km 2 (95% CI 2.15 5.13) on prairie dog colonies, 0.45 plovers km 2 (95% CI 0.44 0.53) on agricultural fields, and 0.23 plovers km 2 (95% CI 0.17 1.76) on grass land habitats, with an overall density of 0.93 plovers km 2 across all habitat types. Plumb et al. (2005) estimated 4.47 0.55 plovers km 2 o n grassland habitats in Wyoming. McConnell et al. (2009) estimated 0.46 0.13 plovers km 2 in ag ricultural habitats in Oklahoma. Augustine (20 11) est imated densities of 6.8 plovers km 2 (95% CI 4.3 10.6), 2.0 plovers km 2 (95% CI 0.8 5.0), and 5.6 plovers km 2 (95% CI 3.5 9.1) for active prairie dog colonies, inactive prairie dog colonies, and prescribed burns, respectively, on the PNG. Those estimates of Mountain Plover density or abundance are results of single studies conducted in a single or a few years. In Montana, where long term study of Mountain Plover has been established, multiple estimates of density exist. Plover density estimates o n prairie dog colonies have ranged from a high of 7.20 0.42 (SE) plovers km 2 in 2008 to a low of 1.28 0.06 (SE) plovers km 2 in 1995 (Childers and Dinsmore 2008) Due to strong associations with prairie dog habitat in Montana, changes in breeding distribution among colonies is expected because of sylvatic plague dynamics and colony size fluctuations ( Dinsmore et al. 2005; Dinsmore and Smith 2010) Across five years on the PNG, plover densities declined by as much as 70% in the following year on colonies affected by plague and on prescribed burns suppo rting the idea that breeding density tracks fluctuations in disturbance patterns (Augustine and Skagen 2014). Despite high local fidelity to breeding areas, local v ariation in distribution and availability of suitable nesting areas shifts breeding densiti es within breeding areas.

PAGE 16

10 Characteristics such as vegetation height and bare ground cover are preferentially used by plovers for nesting on both microsite and landscape scales (Goguen 2012). In comparison to previous years, precipitation was unus ually high in 2015 resulting in excessive vegetative growth in 2016 (MBW, AKP, pers. obs.). Because m y study is restricted to small area of publicly owned and managed rangelands it is possible that breeding density in m y area shifted to private lands where grazing p ressures are heavier, thus creating potentially more favorable nesting areas (Gr unau and Wunder 2001 ) Alternatively, changes in vegetation structure may have influenced detection probability However I cannot definitively make any conclusions about eithe r scenario because I did not explicitly measure any vegetative characteristics. I estimated 110 breeding adults occupied m y study area in 2016. Although m y abundance estimate was extrapolated from the year with the lo west estimated breeding density, m y st udy area only represents a small portion of comparably suitable plover breeding habitat suggesting that South Park still supports a comparative ly large portion of the breeding population. For comparison, McConnell et al. (2009) estimated the entire breedi ng plover population of Oklahoma consists of only 68 91 breeding individuals spread out across a fragmented cultivated landscape. Monitoring population fluctuations in the breeding population of South Park has obvious implications for Mountain Plover conse rvation and continued efforts over time can provide insight into factors that influence breeding population variation, particularly when combined with concurrent monitoring efforts across the range. Mounting evidence suggests that Mountain Plover populati ons are declining, much like other grassland bird species, but to what extent remains unclear. This places difficulty on

PAGE 17

11 planning and enacting conservation actions that are efficient and effective. This is further complicated by the ir patchy distribution i n differential habitats across their range creating the potential for large demographic variation between local populations. Establishment of comparative long term studies range wide is warranted to identify mechanisms driving these demographic differences Re establishment of plover monitoring in South Park is a small piece in a larger puzzle and future efforts in monitoring plovers in other breeding areas as well as in habitats used during stopover and winter are key steps that should be taken to characte rize and mitigate further population declines.

PAGE 18

12 Fig. 1. Study area boundaries in South Park, Park County, Colorado. Park County shown in dark grey on inset map of Colorado counties (lower left) and study area location is marked with an asterisk. Enlarge d map depicts study area boundaries within the western portion of the James Mark Jones State Wildlife Area (JMJSWA) and adjacent BLM pastures. JMJSWA BLM 39.08 39.12 39.16 39.20 105.900 105.875 105.850 105.825 105.800 Longitude Latitude

PAGE 19

13 Fig. 2. Histograms of detections with most parsimonious fitted detection functions for historic surveys (2000 2 006, A G) and the more recent survey (2016, H). Individual detection data points are depicted with open circles. See Appendix A for full list of detection model selection results. Detection probability 0.00 0.10 0.20 0.30 0.0 0.2 0.4 0.6 0.8 1.0 ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! A. 2000 Detection probability 0.00 0.10 0.20 0.30 0.0 0.2 0.4 0.6 0.8 1.0 ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! B. 2001 Detection probability 0.00 0.10 0.20 0.0 0.2 0.4 0.6 0.8 1.0 ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! C. 2002 Detection probability 0.00 0.10 0.20 0.0 0.2 0.4 0.6 0.8 1.0 ! ! ! ! ! ! ! ! ! ! ! ! ! D. 2003 Detection probability 0.00 0.10 0.20 0.0 0.4 0.8 ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! E. 2004 Detection probability 0.00 0.05 0.10 0.15 0.20 0.25 0.0 0.4 0.8 ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! ! F. 2005 Detection probability 0.00 0.10 0.20 0.30 0.0 0.2 0.4 0.6 0.8 1.0 ! ! ! ! ! ! G. 2006 Detection probability 0.00 0.05 0.10 0.15 0.0 0.2 0.4 0.6 0.8 1.0 ! ! ! ! ! ! ! ! ! ! ! ! H. 2016 Distance (km) Detection Probability

PAGE 20

14 Fig. 3. Detection probability estimates and 95% log normal confidence intervals from historic (2000 2006) distance sampling data and data from the survey conducted in 2016. Fig. 4. Mountain Plover density estimates and 95% log normal confidence intervals from transects within the 2016 study (black circles) and from trans ects ranging across South Park (grey triangles). Horizontal dashed line is drawn at upper limit of density estimate for 2016 for comparison to previous years.

PAGE 21

15 Table 1. Estimates and coefficients of variation ( CV ) of density, detection probability, encounter rate (ER), and effective s trip width (ESW) from historic (2000 2006) distance sampling data sub sampled from the study site and data from the contemporary survey conducted in 2016. Year Density ( km 2 ) Detection probability ER ESW (m) 2000 8.71 1.46 17% CV 0.54 0.06 11% CV 2. 71 0.06 12% CV 156 2 2001 12.8 4.5 35% CV 0.39 0.03 8% CV 2.98 1.01 34% CV 116 1 2002 11.9 2.4 20% CV 0.51 0.04 8% CV 3.42 0.62 18% CV 144 1 2003 7.85 3.99 50% CV 0.43 0.05 12% CV 1.79 0.88 49% CV 115 1 2004 7.19 1.95 2 7% CV 0.64 0.07 11% CV 2.51 0.62 25% CV 175 2 2005 9.01 1.92 21% CV 0.57 0.09 15% CV 2.44 0.33 14% CV 136 2 2006 5.02 1.18 23% CV 0.5 0.07 16% CV 1.5 0.26 17% CV 150 2 2016 2.82 0.89 31% CV 0.59 0.07 13% CV 0.62 0.18 28% C V 110 1

PAGE 22

16 CHAPTER II MOUNTAIN PLOVER NEST SURVIVAL IN A HIGH ELEVATION POPULATION Introduction Broad scale climate gradients and localized weather phenomena are important e cological drivers of grassland ecosystems in the Great Plains of North America t hat influence habitat structure and animal community composition (Samson and Knopf 2004) Considerable variat ion in both climate and weather i s characteristic of this region; creating breeding conditions for ground nesting grassland birds that can be unpre dictable in both time and space ; affecting nesting success (Skagen and Yackel Adams 2012 ) Thus, understanding factors that limit nest survival in a grassland species requires consideration of the relative intensity and variability of climate and other sou rces of disturbance across their breeding range particularly if the species is widely but patchily distributed Mountain Plover s ( Charadrius montanus ) are migratory shorebirds that breed on grasslands and xeric tablelands scattered along the western edge of the Great Plains (Knopf and Wunder 2006) Mountain P lovers use area s frequently disturbed by stochastic events such as weather grazing, tillage and burning ; p rocesses that affect the short vegetation and bare gro und habitats preferred for nesting Populations of this species appear to have markedly declined since the 1960s and as a result Mountain Plovers ha ve been identified as a species of conservation concern in many of the states where they breed and overwinte r (Knopf 1994; Knopf and Wunde r 2006; Andres and Stone 2010) In Colorado, comparatively large populations of plovers can be found breeding on the eastern plains (Tipton et al. 2009) and also on intermountain grasslands in South Park, Park County (Wunder et al. 2003) Breeding habitat in South Park is comprised of high elevation

PAGE 23

17 (~2700 m) rangelands that have experienced little historic change in land use compared to the rangelands o n the eastern plains of Colorado At lower elevations in eastern Colorado (~1470 m), plovers breed on a patchwork of habitat including grazed shortgrass prairie burn ed rangelands black tailed prairie dog colonies ( Cynomys ludovicianus ), and fallow dryla nd crop fields; all of which vary annually in extent and availability. In addition to differences in habitat availability, differences in elevation create different climates and alte r breeding schedules. On lower elevation sites across the breeding range, Mountain Plover s breed earlier and nest survival is negatively influenced by increased precipitation and warmer temperatures but the relative strength of weather effects and temporal variation in nest survival differs between locales (Dinsmore et al. 2002; Dreitz et al. 2012) In north central Montana, at t he northern periphery of the plover's range, Dinsmore et al. (2002) found relatively strong negative effects of increased precipitation on nest survival In addition to precipitation, Dinsmore et al. (2002) also modeled nest survival as a function of daily maximum temperature, year, time in season (linear and quadratic), nest age, and sex of the incubating adult (adult plovers of both sexe s independently incubate nests). Overall, daily nest survival in Montana was hig her for male tended nests, decreased ove r the season, increased over the incubation period, and decreased with precipitation. In eastern Colorado, increased precipitation was also associated with lower nest survival but survival also declined with increased maximum temperatures (Dreitz et al. 20 12). However, i n this study variation in survival due to precipitation was best explained by 10 day periods of drought not daily precipitation totals Dreitz et al. (2012) also considered annual and season al time tre nd s and

PAGE 24

18 s imilar to the study in Montan a, they found that nest survival decreased over the season but did not vary by year Results of these studies suggest that weather and time in season are important factor s in fluencing Mountain Plover nest survival Specifically, early and mid season nests in cooler and drier conditions have more favorable odds of survival at lower elevations However, no studies have examined t he relationship between weather and temporal variation in plover nest survival on high elevation habitats which have distinctively different climate, habitat stability, and nesting phenology. I m onitored nests of a population of Mountain Plover breeding at high elevation in South Park, Colorado to study influences on daily nest survival as compare d to those from lower elevation breedi ng sites. I modelled daily variation in nest survival as a function of temperature, precipitation a nd annual and seasonal time trends. I hypothesized that at high elevation cold would limit survival rather than heat, therefore I considered minimum tempera tures in m y nest survival models in place of maximum temperatures. To test how well models from Dinsmore et al. (2002) and Drietz et al (2012) described nest survival variation in m y high elevation population, I also evaluated models constructed from m y va riable set that closely resembled the most parsimonious model from each study. Methods Study a rea I studied Mountain Plovers breeding at high elevation (2 786 19 m SD ) on public ly owned and managed rangelands in Par k County, Colorado ( 39¡ 9' 19.296" N, 10 5¡ 50' 51.72" W ) My study area encompassed ~55 km 2 of lands within the western portion of the James Mark Jones State Wildlife Area and adjacent pastures managed by the Bureau of Land

PAGE 25

19 Management (BLM) In this area, plover breeding habitat is comprised of h igh elevation intermountain grasslands grazed in winter by native herbivores such as elk ( Cervus canadensis ) and pronghorn ( Antilocapra americana ) and in summer by domestic cattle ( Bos taurus ) Although plovers are commonly associated with black tailed pr airie dogs in other areas of their range ( Knowles et al.1982 ), there are no black tailed prairie dogs in South Park (W under et al. 2003) Dominant grasses on the study area included species of muhly ( Muhlenbergia filiculmis and Muhlenbergia montana ), blue grama ( Bouteloua gracilis ) and Arizona fescue ( Festuca arizonica ) The most common f orbs and shrubs included fringed sage ( Artemisia frigida ) and yellow rabbit brush ( Chrysothamnus viscidiflorus ) respectively Nest m onitoring I followed methods used by Dinsmore et al. (2002) to locate nests. During June and July of 2015 and 2016, I conducted semi systematic ground surveys of pastures on all terrain vehicles to flush incubating plovers. Once a plove r was sighted, it was observed until it returned to the nest or its behavior indicated no nest was present. Upon discovery, nest location was recorded using a GPS unit (Garmin Montana 680t, Olathe, KS, USA ) and eggs were floated to estimate nest age (Dinsmore et al. 2002) To aid in relocation, a picture of the nest was taken and natural markers such as a small stack of dri ed cattle dung or a small rock pile were placed 10 m north and south of the nest. Nests were checked every 3 5 days until the eggs hatched (at least one chick produced) or the nest failed Mountain Plover chicks are precocial and leave the nest within 24 hours of hatching. T herefore, in cases where chicks were not found in the nest cup or with the incubating adult nearby, nesting outcome was determined b ased on the presence or absence of small pip chips in the nesting cup ( Mabee 1997 ). Nests were recorded as depredated if large shell fragments were scattered near the

PAGE 26

20 nest site or if eggs were missing before the expected hatch date and no pip chips were found in the nest material. Some nests failed due to abandonment, which was confirmed by the persistent ab sence of the incubating adult, eggs that were cool to the touch and /or lack of progression of development evidenced by egg floatation. In such case s, failure was recorded as the first day the nest was suspected to be abandoned. As part of a concurrent stu d ies incubating adults were trapped using a walk in wire mesh cage placed over the nest. Each bird was banded with a USGS aluminum leg band and unique color band combination and morphometric measurements were recorded. Some plovers were also fitted with a GPS datalogger (GPS PinPoint 50; Lotek, Newmarket, Ontario, Canada) attached using a leg loop harness (Rappole and Tipton 1991 ) Tagging efforts were focused on plovers with late stage nests to mitigate potential risks of nests abandonment due to tag placement. Weather d ata I obtained temperature and prec ipitation data fr om a weather station administered by the National Oceanic and Atmospheric Administration ( NOAA National Centers for Environme ntal Information; http://www.ncdc.noaa.gov/) located 18.4 km from the study area in similar habitat an d elevation ( 2718 m ; Antero Reservoir station ID# USC00050263 ; 38¡ 59' 35.8794" N, 105¡ 53' 30.84" W ) Minimum temperature values for 3 days were missing for this station in 2015, therefore, I substituted these values with measures from another station nearby at an alt itude of 3086 m (Fairplay CO station ID# USC00052816; 39¡ 13' 18.8394" N, 105¡ 59' 36.96" W). D aily measures were recorded at 07:30 hours and reflected minimum temperature and total precipitation from the previous 24 h r period In cases where I was intereste d in same day weather effects, I adjusted dates backward one day

PAGE 27

21 so measures reflected the 24 hr period beginning at 07:30 hours on that date Model development and s urvival a nalysis I estimated daily survival rate (DSR) of nests using the nest survival model available in Program MARK (White and Burnham 1999) using the R package RMark' (R version 3.3.2; Laake 201 3 ) I omitted 2 nests from analysis; one that was discovered shortly after hatching and another that was accidentally destroyed on discovery. Because I was more interested in evaluating a specific set of hypotheses rather than determining the importance o f co varia tes I limited m y analyses to a set of a priori biological hypotheses constructed from the following covariates previously found to influence n est survival in Mountain Plovers (Table 1) 1) Year. I included models to allow nest survival to vary by y ear to account for annual variation that may not be captured in m y covariates such as habitat structure, habitat quality, and predator abundance. Annual variation in nest survival has been previously documented in Mountain Plover (Augustine and Skagen 2014 ). 2) Seasonal time trends. Nest survival may also vary within a season. In eastern Colorado, daily survival rates of Mountain Plover nests declined over the season (Dreitz and Knopf 2007; Dreitz et al. 2012) In Montana, daily survival rates were lowest late season but also depressed early in early season (Dinsmore et al. 2002). Based on this, I considered models incorporating either a linear time trend or a quadratic time trend. 3) Nest age. DSR may vary over the incubation period, increasing as nests approach hatch. Vulnerable nests may be lost early in inc ubation, or parental behaviors may

PAGE 28

22 influence predation rates as investment increases (Klett and Johnson 1982). DSR has also been shown to increase over incubation in plovers (Dinsmore et al. 2002). 4) Minimum temperature and precipitation. Dreitz et al. (201 2) reported increased DSR during 10 day periods of drought, and reduced daily survival with high maximum temperatures in Mountain Plover nests in eastern Colorado. In Montana, Dinsmore et al. (2002) found reduced nest survival with increased daily precipit ation. I hypothesized that nest survival would also be influenced by daily precipitation and temperature in South Park, but because of differences in climate at higher elevation survival would be limited by minimum temperatures instead of max imum temperat ures. I also considered minimum temperature and precipitatio n from the previous day to incorporate potential one day lag effects on survival. As uniparental incubators, plovers may be restricted to nests during hot, cold, or wet conditions and thus may inc rease duration or frequency of off bouts the following day to forage and rec oup energetic reserves (Skrade and Dinsmore 2012) leaving nests vulnerable to predators (Smith et al. 2012) 5) GPS tag placement. As part of concurrent studies s ome incubating adults were fitted with a GPS tag which may have affected their behavior a nd re duced nest survival. Thus, I also considered a model that allowed DSR to vary between nests with tagged and untagged incubators I limited the complexity of these models because of comparatively small sample size and did not want to consider additional mo dels at random that included all other combinations of temporal and weather variables To provide a basis for comparison to previous Mountain Plover nest survival work with more complex models, I constructed models from m y set of

PAGE 29

23 covariates that approxima t ed the most parsimonious model from Dinsmore et al ( sex of incubating adult nest age, quadratic time trend and precipitation ; 2002) and Dreitz et al. ( linear time trend, 10 day drought, and maximum temperature; 2012) My mod el equivalent to Dinsmore et a l. (2002) did not include sex of the incubating adult as I did not have these data for nests in 2016. I substituted previous day precipitation for 10 day drought in m y model equivalent to Dreitz et al. (2012) because I did not observe any 10 day periods wi thout precipitation in the two years of m y study I also replaced maximum temperature with minimum temperature because extremely hot temperatures do not occur in South Park, but temperatures can regularly fall below freezing even in summer months Results Data summary and nesting p henology I monitored 89 Mountain Plover nests on actively grazed rangelands during June and July in 2015 (n = 52) and in 2016 (n = 37 ). Of these 89 nests, at least one egg hatched in 35 (39.3%) 4 6 were lost to predation ( 51.7 %), 7 were abandoned (7.9%), and one was accidentally destroyed upon discovery ( 1.1%) The na•ve point estimate for nest success in 2015 was higher than in 2016 (48.1% vs. 35.1%, respectively). Nests were monitored over a 53 day exposure period (01 June 24 July). Me an ( 1 SE ) nest initiation dates were similar between years (2015, 03 Jun e 1.3 d ays ; 2016, 29 May 1.6 d ays ). Assuming all discovered nests survived the 29 day incubation period from their estimated initiation date the expected mean hatch date was 01 July 1.0 days which was similar to the observed mean hatch date of successful nests (03 July 1.7 days).

PAGE 30

24 Nest survival The model including only d aily minimum temperature best predicted variation in DSR, followed by additive effects of minimum temperature and precipitation. Models incorporating partial and combined effects of temperature and precipitation from the previous day were slightly more parsimonious than model c ounterparts without lagged weather effects (Table 2). Although AICc values were similar between all models in m y set, a minimum temperature effect (either same day or lagged ) was present in all models within 2 AICc units of the most parsimonious model (Table 2 ) Furthermore, same day minimum temperature and previous day minimum temperature w ere the only parameter estimate s with a 95% confid ence interval s that did not overlap zero (See Appendix C ). GPS tag place ment did not have a measureable e ffect on DSR ; estimated 95% CI for coefficients of tag presence overlapped zero (Appendix C) Of the adults incubating 87 nests, 30 adults were fitted with a GPS tag Of those 30 nests, 14 ( 47% ) ha tched at least one chick; this was higher than the overall na•ve estimate of hatching success for the population. The unadjusted mean daily nest survival probability for nests in m y study was 0.949 (95% CI, 0.934 0.961 ) and total probability of survival across a typical incubation period of 29 days based on constant DSR was 0.218 (95% CI 0.139 0.317). My most parsimonious model predicted that for every 1 ¡ C increase in previous day minimum temperature adjusted daily odds of survival increased by a factor of 1.17 (95%CI 1 .0 4 1 .31 ; Tab le 3 ) An equally competitive model predicted a similar adjustment factor in odds for every 1 ¡ C increase in previous day minimum temperature (1.19, 95% CI 1 .0 5 1.34) as well as an additive adjustment in odds of survival by a factor of 0.97 (95% CI 0.92 1. 01) for every 1 mm in precipitation (Table 3) Holding prevTMIN and prevPRCP at their average observed values

PAGE 31

25 ( 4.5 ¡ C and 1.7 mm), the top two weather models (prevTMIN and prevTMIN + prevPRCP; Table 2) predicted a mean DSR of 0.95 0 and 0.952, respectively. T otal probability of survival across a typical incubation period of 29 days as 0.231 and 0.242 for the prevTMIN model and the prevTMIN + prevPRCP model respectively (Table 2 ). Effects of prevTMIN on DSR were strongest at temperatures below freezing (Fig. 1) and with heavy precipitation (Fig. 2). Discussion Extreme high temperatures have negative effects on Mountain Plover nest survival at lower elevation s in Colorado (Dreitz et al. 2012) and extreme low temperatur es have negative effects on nest survival at high elevation (this study) sites In m y study, extreme cold decreased odds of survival for nests, and that effect increased with cold intensity (Fig 1 ; Table 2 ). At lower elevations i n eastern Colorado, Drietz et al. (2012) reported a negative effect of extreme heat on plover nest survival. However, in the northern edge of the breeding range in north central Montana, Dinsmore et al. (2002) did not find a meaningful effect of maximum t empe rature on nest survival. Drietz et al. (2012) suggests that this may be attributable to the lower intensity of extreme heat that occurs at a higher latitude. Dinsmore et al. (2002) did not consider minimum temperature effects so it is uncertain if cold te mperatures also affect DSR at this latitude. I also found limited support that precipitation affected daily nest survival, but only in conjunction with minimum temperature (Fig. 2) Moreover, a lthough the estimated effect was negative, it was not statistic ally different from zero (Table 2) It is possible that I did not have a sufficient sample size to estimate the potentially more subtle effect of precipitation. Dinsmore et al. (2002) and Drietz et al. (2012) both found negative effects of precipitation

PAGE 32

26 bu t also had much larger sample sizes, 641 ne sts and 936 nests respectively as compared with 87 in this study Models analogous to the most parsimonious models from previous Mountain Plover nest survival studies by Dinsmore et al. (2002) and Drietz et al. (2012) which incorporated temporal and weather covariates were outperformed by simpler temperature and precipitation only models (Table 2). This is unsurprising given differences between habitats. However, negative effects of extreme temperature and precip itation may be generalities that exist between locales but are potentially mediated through different mechanisms at varying temporal scales. Evidence of annual or seasonal temporal trends in nest survival was not supported by m y data. Many factors that I d id not explicitly measure could be attributed to annual variation in nest survival such as changes in habitat structure, climate, predator ab undance, or plover abundance. Augustine and Skagen (2014) found that Mountain Plover nests in one year had lower DSR than those in the subsequent year and that breeding density of plovers were similar between years They did not includ e weather variables in their model set, and as they suggested, annual variation in nest survival may have been attributable to climatic differences between years, as nest survival was higher in the cooler dryer year The notion that annual variation in DSR may be better explained by weather is supported by the lack of annual trends in aforementioned studies where weather variables are modeled. DSR may vary as a function of time in season due to many factors such as vegetation changes over the season preda tor and prey abundance, re nesting activity or breeding experience. Dinsmore et al. (2002) found that DSR of plover nests in Montana was highest later in the season and lowest mid season. Conversely, Drietz et al. ( 2012 ) and Augustine

PAGE 33

27 and Skagen ( 2014 ) f ound that DSR of plover nests decreased over the season in eastern Colorado. The lack of seasonal trends in m y study may be attributable to the comparatively short season length at high elevation. Nest monitoring exposure periods in previous plover nest su rvival studies spanned 77 days (Dinsmore et al. 2002), 66 days (Augustine and Skag en 2014) and up to 110 days (Drietz et al. 2012), as co ntrasted with 53 days in this study. Between years, m y observed mean nest initiation date was 02 June 1.0 days (SE) similar to reported mean initiation dates for male tended nests at higher latitu des in Montana (02 June 3.2 d ays; Dinsmore et al 2002). The constrained breeding period in South Park likely limits the ability of plovers to adjust their nesting schedule and c o nstrains potential for re nest ing attempts. Therefore with m y sample size, it is not surprising that I did not uncover any evidence of temporal variation in nest survival that was not better explained by seasonal changes in temperature and precipitation. Models that incorporated temporal and weather covariates in the most parsimo nious models of nest survival from previous work were not competitive in m y analysis (Table 2 ). This supports m y conclusion that annual and seasonal variation in nest survival at m y location is best described by minimum temperature and precipitation, and a ny additional factors that I considered may have effects too small to be detected with m y sample. My estimate of overall mean nest survival assuming a constant DSR was 0.218 (95%CI, 0.139 0.317), almost half of the observed proportion of hatched nests (na• ve estimate) included in analysis (0.39). This na•ve estimate is biased high because it does not account for nests lost prior to discovery. My estimate of nest survival is low compared to nest survival and hatch rates reported in Mountain Plover literature which range from 0.27 to 0.70 and is much lower than previously reported na•ve estimates of nest success rates of 0.60 in

PAGE 34

28 South Park (Graul 1975; Dinsmore et al. 2002; Mettenbrink et al. 2006; Dreitz and Knopf 2007; Dinsmore et al. 2010; Dreitz et al. 2012) I cannot rule out that this low rate may be partially attributed to differences in climatic conditions between years of this s tudy (2015, 2016) and when plover nests were last monitored in South Park (2000 2006). Minimum temperatures were not unusual during the period of my study, however, total precipitation was usually high in 2015. In May and June of 2015, precipitation totale d 113.4 mm and 73.4 mm, respectively, which was much higher the mean totals for May (18.3 mm) and June (22.1 mm) between 2000 and 2006. Lagging effects of precipitation on prey abundance or habitat structure may have had influence on nest survival that was not captured in the changes in daily measures that I measured. Overall, my results suggest that cold temperatures limit Mountain Plover nest survival in South Park. The mechanism behind this relationship is less clear. Nest predation is the primary cause of nest failure in ground nesting grassland birds (Martin 1993), which was also true for nests in my study. As such, cold temperatures likely affect nest survival in South Park by increasing predation risk. In my study area, the predator community consists largely of avian predators such as Common Ravens ( Corvus corax ) and small mammals such as thirteen lined ground squirrels ( Ictidomys tridecemlineatus ). Due to the high elevation and low ambient temperatures, South Park is notably devoid of snakes a common nest predator on low elevation breeding sites. Effects of weather and climate on predation risk are contingent on the composition of the nest predator community because different predators respond differently to environmental conditions. For example, Dins more et al. (2002) hypothesized that precipitation would enhance olfactory cues that snakes use to locate nests increasing predation risk but this would not be the case for avian predators which rely on

PAGE 35

29 visual cues to find nests These nest predator commun ity composition differences may partly explain why precipitation did not have a strong significant effect on nest survival in South Park. As for temperature, m inimums usually occur at night so it is unlikely that these temperature drops increase predation risk by influencing behavior of visual predators like ravens. Increased mammalian predation during periods of low temperature also seems unlikely as plovers actively defend nests from ground squirrels (MBW, AKP pers. Obs.) or perform distraction displays (Graul 1975) to lure them away from the nest. It is more likely that temperature effects on predation are mediated by effects on incubation behavior. In some uniparental arctic shorebirds, maintaining stable incubation temperatures during cold perio ds may increase energy expenditure inducing an increase in duration of subsequent foraging bouts to balance energetic condition (Tulp and Schekkerman 2006) These long or frequent off bouts from the nest decrease nest vigilance which may in turn increase predation rates (Smith et al. 2012) For example, within sp ecies comparisons in Sanderlings ( Calidris alba ) demonstrated that uniparental incubators expend more energy in maintaining egg temperatures than biparental counterparts (Reneerkens et al. 2011). Additionally, Reneerkens et al. (2011) found that arthropod abundance increased with warming temperatures and only uniparental Sanderlings increased their recess frequency as temperatures warmed. Overall, Reneerkens et al. (2011) observed decreased nest attendance, increased recess frequency, and longer recess dura tion in uniparental Sanderlings compared to biparental. As uniparental incubators, Mountain Plovers may also employ similar compensatory behaviors to balance incubation and personal energetic maintenance. In Montana, plovers regularly leave the nest to for age in the middle of the night and at crepuscular times (Skrade and Dinsmore 2012) Ho wever, arthropod activity may be

PAGE 36

30 depressed during cold evenings in South Park, limiting productive foraging to daylight hours. In turn, after long or intense cold periods nests may be left unattended longer or more frequently during daylight when most nest predators are active. Energy expenditure may also be increased to maintain nest and individual body temperatures in below freezing conditions. However, maintaining egg temperatures may not be as costly in Mountain Plovers due to unusually thick shells whi ch may act as insulation during colder temperatures (Dinsmore et al. 2002). Precipitation during these cold periods could potentiate effects on both energy expenditure and prey availability. As such, effects on incubation behavior are likely contingent on the duration and timing of extreme cold and precipitation events. This information was not available for the weather data I collected in my study, which might explain why I saw little differentiation between same day and one day lag effects of both minimum temperature and precipitation. From my study I cannot definitively conclude why minimum temperature and to a limited degree, precipitation, are associated with lower nest survival, but alterations in incubation behavior induced by changes in prey availab ility or thermal maintenance seem to be plausible mechanisms that warrant future study on how cold temperatures negatively influence Mountain Plover nest survival in South Park. My study demonstrates the importance of considering differences between weath er and phenology among breeding locations To better understand limitations in adaptability to disturbance, it is important to sample across spatial gradients where these disturbances occur For example to understand impacts of weather on Mountain Plover nest survival studies should be conducted in other nesting areas across their geographic range, particularly in New Mexico and Mexico. Within the central breeding range on the Great Plains it is likely that climatic extremes will become more intense and frequent in the future ( Skagen and Yackel

PAGE 37

31 Adams 2012) T herefore, understanding the differential effects on plover productivity will aid in conservation planning.

PAGE 38

32 Table 1. C andidate models and notation. Candidate m odel s Model notation Constant DSR Inte rcept only Annual trend Year Linear time trend Time Quadratic time trend Time 2 Annual and linear time trend Year + Time Annual and quadratic time trend Year + Time 2 Nest age NestAge Current day min. temp (C¡) TMIN Previous day min. temp (C¡) prevTM IN Current day precip total (mm) PRCP Previous day precip total (mm) prevPRCP Current day min temp (C¡) and precip total (mm) TMIN + PRCP Previous day min temp (C¡) and precip total (mm) prevTMIN + prevPRCP Linear time, previous precip total (mm), and min temp Time + prevPRCP + TMIN Nest age, quadratic time, and current day precip total (mm) NestAge + Time 2 + PRCP GPS tag on incubating adult Tag Table 2 Summary of nest survival model selection results for Mountain Plovers in South Park, Colorado (2015 to 2016). The TMIN and PRCP variables refer to daily minimum temperature and precipitation respectively. Variables preceded by prev' refer to one day lag effects. Model Parameters AIC c AIC c w i Deviance prevTMIN 2 298.413 0 0.23 294.401 TMIN 2 298.683 0.27 0.201 294.67 pre v TMIN + prevPRCP 3 299.062 0.65 0.166 293.038 TMIN + PRCP 3 300.359 1.946 0.087 294.335 Year 2 301.06 2.647 0.061 297.048 Time + prevPRCP + TMIN 4 301.37 2.957 0 .052 293.329 Year + Time 3 301.988 3.575 0.038 295.963 Tag 2 302.565 4.152 0.029 298.553 Year + Time 2 3 302.658 4.245 0.028 296.634 Constant (Intercept only) 1 302.74 4.327 0.026 300.736 Time 2 303.257 4.844 0.02 299.244 NestAge 2 303.41 4.997 0.019 299.398 Time 2 2 304.109 5.696 0.013 300.096 prevPRCP 2 304.503 6.09 0.011 300.491 PRCP 2 304.726 6.313 0.01 300.714 NestAge + Time 2 + PRCP 4 305.238 6.825 0.008 297.197

PAGE 39

33 Table 3 Parameter estimates, SE, and 95% confidence intervals for lag effect wea ther models on the logit scale used to calculate nest survival probability over a 29 day incubation period. Model Parameter Estimate SE 95% CI prevTMIN Intercept 2.2553 0.2667 ( 1.7325 2.7781 ) prevTMIN 0.1546 0.0597 ( 0.0377 0.2716 ) prevTMIN + pr evPRCP Intercept 2.2627 0.2658 ( 1.7417 2.7836 ) prevTMIN 0.1738 0.0620 ( 0.0523 0.2952 ) prevPRCP 0.0358 0.0237 ( 0.0823 0.0107 ) Fig. 1. Estimated daily survival rate of Mountain Plover nests across varying daily minimum temperatures lagged by o ne day in South Park, Colorado.

PAGE 40

34 Fig. 2. Estimated daily survival rate of Mountain Plover nests across varying daily minimum temperatures and precipitation totals lagged by one day in South Park, Colorado.

PAGE 41

35 REFERENCES Abbott C G. 1939. Mountain Plover at San Diego, California. Condor 42:125. Andres B A and K.L. Stone. 2010. Conservation Plan for the Mountain Plover ( Charadrius montanus ). Version 1.1. Manomet Center for Conservation Sciences, Manomet, MA. Augustine D J. 2011. Habitat selection by mountain plovers in shortgrass steppe. Journal of Wildlife Manage ment 75:297 304. Augustine D J Skagen S K. 2014. Mountain plover nest survival in relation to prairie dog and fire dynamics in shortgrass steppe. Journal of Wildlife Management. 78:595 602. Buckland S T Anderson D R Burnham K P Laake J L Borchers D L and L. Thomas 2001. Introduction to distance sampling estimating abundance of biological populations. Oxford University Press Oxford, UK. Buckland S T Rexstad E A Marques T A snd C.S. Oedekoven. 2015. Distance Sampling: Methods and Applications. Springer, New York, NY, USA. Burnham K P ., and D.R. Anderson 2002. Model selection and multimodel inference: a practical information theoretic approach. 2nd ed. Springer, New York, NY,USA Childers T M and S.J. Dinsmore 2008. Density and abundance of Mountain Plovers in northeaste rn Montana. Wilson Journal of Ornithology. 120:700 707. Colorado Parks and Wildlife. 2015. State Wildli fe Action Plan: A Strategy for c onserving w ildlife in Colorado. Denver, Colorado, USA. Cooke W W. 1914. Our Shorebirds and their Future. Yearbook of the Department of Agriculture 1914:275 294. Dinsmore, S.J., G. C. White, and F. L. Knopf, 2005. Mountain Plover population responses to black tailed prairie dogs in Montana. The Journal of Wildlife Management 69:1546 1553. Dinsmore S J and M.D. Smith 2010. Mountain Plover r esponses to p lague in Montana. Vector Borne Zoonotic Dis ease 10:3 7 45. Dinsmore S J White G C and F.L. Knopf 2002. Advanced techniques for modeling avian nest survival. Ecology 83:3476 3488. Dinsmore S J Wunder M B Dreitz V J and F.L. Knopf 2010. An a ssessment of f actors a ffecting p opulation g rowth of the Mountain Plover. Avian Conservation and Ecol ogy 5:13.

PAGE 42

36 Dreitz V J Conrey R Y and S.K. Skagen 2012. Drought and c ooler t emperatures a re a ssociated with h igher n est s urvival in Mountain Plovers. Avian Conserv ation and Ecology. 7(1): 6. Dreitz V J and F.L. Knopf 2007. Mountain Plovers and the p olitics of r esearch on p rivate l ands. Bioscience 57:681 687 Graul W D. 1975. Breeding b iology of the Mountain Plover. Wilson Bull etin 87:6 31. Graul W D and L.E. Webster 1976. Breeding s tatus of the Mountain Plover. Wilson Bull etin 78:265 267. Grunau L and M.B.Wunder 2001. Conservation a ssessment for Mountain Plover ( Charadrius montanus ) in South Park, Colorado. Colorado Natural Heritage Progarm. Klett, A.T. and D.H. Johnson. 1982. Variability in nest survival rates and implications to nesting studies. The Auk 99:77 87. Knopf F L 1994. Avian a ssembla ges on a ltered g rasslands. Studies in Avian Biol ogy 15:247 257. Knopf F L and B.J. Miller 1994. Charadrius montanus : montane, grassland, or b are ground plover? The Auk 111:504 506. Knopf F L and M.B. Wunder 2006. Mountain Plover ( Charadrius montanus ). In: The Birds of North America Online (A. Poole, ed.) Cornell Lab of Ornithology, Ithaca, NY. Knowles, C.J., Stoner, C.J. and S.P. Gieb. 1982. Selective use of black tailed prairie dog towns by Mountain Plovers. Condor 84(1)71 74. Laake J L. 2013. RMark: An R i nterface for a nalysis of c apture r ecapture d ata with MARK. AFSC Processed Report. 2013 01, 25p. Alaska Fish. Sci. Cent., NOAA, Na tl. Mar. Fish. Serv., 7600 Sand Point Way NE, Seattle WA 98115. Mabee, T.J., 1997. Using eggshell evidence to determine nest fate of shorebirds. The Wilson Bulletin 109(2): 307 313. Martin, T. E. 1993. Nest predation among vegetation layers and habitat t ypes: revising the dogmas. The American Naturalist 141 (6):897 913. McConnell S O'Connell T J Leslie, David M J and J.S. Shackford. 2009. Mountain plovers in Oklahoma: distribution, abundance, and habitat use. Journal of Field Ornithology 80:27 3 4.

PAGE 43

37 Mettenbrink C W Dreitz V J Knopf F L and T. Brush 2006. Nest success of Mountain P lovers relative to anthropogenic edges in Eastern Colorado. The Southwest Nat uralist 51:191 196. Miller D L. 2016. Distance: Distance Sampling Detection Func tion and Abundance Estimation. R package version 0.9.6. https://CRAN.R project.org/package=Distance P lumb, R E Knopf F L and S.H. Anderson 2005. Minimum population size of Mountain Plovers breeding in Wyoming. The Wilson Bull etin 117 (1) :15 22. Rapp ole J H and A.R. Tipton. 1991. New harness design for attachment of radio transmitters to small passerines (Nuevo Dise–o de ArnŽs para Atar Transmisores a Passeriformes Peque–os) Journal of Field O rnithology. 335 337. Reneerkens, J., Grond, K., Schekk erman, H., Tulp, I. and T. Piersma. 2011. Do uniparental sanderlings Calidris alba increase egg heat input to compensate for low nest attentiveness?. PLOS one 6 (2): p.e16 834. Skagen, S. K., and A A. Yackel Adams. 2012. Weather effects on avian breeding p erformance and implications of climate change." Ecological Applications 22(4):1131 1145. Skrade P D B and S.J. Dinsmore 2012. Incubation patterns of a shorebird with rapid multiple clutches, the Mountain Plover ( Charadrius montanus ). Canadian Journal of Zoology. 90:257 266. Smith P A Tulp I Schekkerman H Gilchrist H G and M.R. Forbes 2012. Shorebird incubation behavi our and its influence on t he risk of nest predation. Ani ma l Behavior 84 (4) :835 842. Tipton H C Doherty Jr P.F., and V.J Dreitz 2009. Abundance and density of mountain plover ( Charadrius montanus ) and burrowing owl ( Athene cunicularia ) in eastern Colorado. The Auk 126:493 499. Tulp I and H. Schekkerman 2006. Time allocation between feeding and incubation in uniparenta l arctic breeding shorebirds: Energy reserves provid e leeway in a tight schedule. Journal of Avian Biol ogy 37:207 218. White G C and K.P. Burnham 1999. Program MARK: survival estimation from populations of marked animals. Bird study 46:S120 S139. W under M B Knopf F L and C.A. Pague 2003. The high elevation population of Mountain Plovers in Colorado. Condor 105:654.

PAGE 44

38 APPENDIX A Distance sampling model selection results for all years. P values listed are from a CramÂŽr von Mises goodness of fi t test. Estimated detection probabilities with SE are also listed. 2000 Key function C vM P value Detection SE AIC Uniform with cosine adjustment term of order 1 0.72 0.52 0.02 0.00 Hazard rate 0.95 0.59 0.04 0.52 Half normal with cosine adjustment term of order 2 0.79 0.54 0.09 1.47 Half normal 0.43 0.48 0.05 1.50 2001 Key function C vM P value Detection SE AIC Uniform with cosine adjustment terms of order 1,2 0.64 0.39 0.03 0.00 Half normal 0.69 0.39 0.03 0.59 Hazard rate 0.54 0.44 0.04 3 .24 2002 Key function C vM P value Detection SE AIC Uniform with cosine adjustment term of order 1 0.96 0.52 0.02 0.00 Half normal 0.70 0.48 0.04 1.12 Hazard rate 0.97 0.58 0.04 1.98 2003 Key function C vM P value Detection SE AIC Hazard r ate 0.62 0.45 0.05 0.00 Uniform with cosine adjustment terms of order 1,2 0.47 0.40 0.04 1.75 Half normal 0.49 0.41 0.04 2.21

PAGE 45

39 2004 Key function C vM P value Detection SE AIC Half normal 0.84 0.65 0.07 0.00 Uniform with cosine adjustment term of o rder 1 0.79 0.63 0.06 0.39 Hazard rate 0.87 0.67 0.09 1.95 2005 Key function C vM P value Detection SE AIC Uniform with cosine adjustment term of order 1 0.45 0.61 0.04 0.00 Hazard rate 0.92 0.53 0.10 0.17 Half normal with cosine adjustment term o f order 2 0.92 0.51 0.07 0.37 Half normal 0.42 0.62 0.05 0.85 2006 Key function C vM P value Detection SE AIC Uniform with cosine adjustment term of order 1 0.85 0.50 0.06 0.00 Half normal 0.90 0.49 0.08 0.44 Hazard rate 0.87 0.53 0.11 3.28 201 6 Key function C vM P value Detection SE AIC Uniform with cosine adjustment term of order 1 0.42 0.59 0.05 0.00 Hazard rate 0.50 0.58 0.10 0.56 Half normal 0.41 0.59 0.06 0.88

PAGE 46

40 APPENDIX B R output summaries for most parsimonious distance samplin g models 2000 Summary for distance analysis Number of observations : 80 Distance range : 0 0.291488 Model : Uniform key function with cosine adjustment term of order 1 Strict monotonicity constraints were enforced. AIC : 238.9163 Detection function parameters Scale coefficient(s): NULL Adjustment term coefficient(s): estimate se cos, order 1 0.9377728 0.05963779 Estimate SE CV Average p 0.5160564 0.0 1588239 0.03077646 N in covered region 155.0218229 12.96680032 0.08364500 2001 Summary for distance analysis Number of observations : 88 Distance range : 0 0.2961807 Model : Uniform key function with cosine adjustment terms of order 1, 2 Strict monotonicity constraints were enforced. AIC : 291.5907 Detection function parameters Scale coefficient(s): NULL Adjustment term coefficient(s): estimate se cos, order 1 1.2716211 0.09524336 cos, order 2 0.3222461 0.09614585 Estimate SE CV Average p 0.3855248 0.02748264 0.07128631 N in covered region 228.2603065 25.07168197 0.10983812

PAGE 47

41 2002 Summary for distance analysis Number of observations : 101 Distance r ange : 0 0.278 Model : Uniform key function with cosine adjustment term of order 1 Strict monotonicity constraints were enforced. AIC : 311.5709 Detection function parameters Scale coefficient(s): NULL Adjustment term coefficient(s ): estimate se cos, order 1 0.937892 0.05672406 Estimate SE CV Average p 0.5160246 0.01510456 0.02927101 N in covered region 195.7270912 14.71032996 0.07515735 2003 Summary for distance analysis Number of observations : 53 Distance range : 0 0.266 Model : Hazard rate key function AIC : 183.3245 Detection function parameters Scale coefficient(s): estimate se (Intercept) 2.300607 0.15 41386 Shape coefficient(s): estimate se (Intercept) 1.402911 0.2549164 Estimate SE CV Average p 0.4534208 0.05302137 0.1169363 N in covered region 116.8892006 18.10346193 0.1548771

PAGE 48

42 2004 Summary for distance analysis Number of observations : 74 Distance range : 0 0.2708349 Model : Half normal key function AIC : 204.4621 Detection function parameters Scale coefficient(s): estimate se ( Intercept) 1.891013 0.1521052 Estimate SE CV Average p 0.6476029 0.06810742 0.1051685 N in covered region 114.2675534 14.37344785 0.1257877 2005 Summary for distance analysis Number of observations : 87 Distance range : 0 0.24096 Model : Uniform key function with cosine adjustment term of order 1 Strict monotonicity constraints were enforced. AIC : 265.9016 Detection function parameters Scale coefficient(s): NULL Adjustment term coefficient(s): estimate se cos, order 1 0.6328666 0.1195913 Estimate SE CV Average p 0.6124199 0.04485368 0.07324008 N in covered region 142.0593913 14.07682630 0.09909113

PAGE 49

43 2006 Summary for distance analysis Number of observations : 24 Distance range : 0 0.3 Model : Uniform key function with cosine adjustment term of order 1 Strict monotonicity constraints were enforced. AIC : 68.46056 Detection f unction parameters Scale coefficient(s): NULL Adjustment term coefficient(s): estimate se cos, order 1 1 0.2570877 Estimate SE CV Average p 0.5 0.06427194 0.1285439 N in cov ered region 48.0 9.27740302 0.1932792 2016 Summary for distance analysis Number of observations : 51 Distance range : 0 0.188 Model : Uniform key function with cosine adjustment term of order 1 Strict monotonicity constraints wer e enforced. AIC : 183.1659 Detection function parameters Scale coefficient(s): NULL Adjustment term coefficient(s): estimate se cos, order 1 0.6989065 0.144215 Estimate SE CV Average p 0.5886139 0.04996563 0.08488693 N in covered region 86.6442314 10.70755315 0.12358068

PAGE 50

44 APPENDIX C Parameter estimates for all nest survival models. Estimates are reported on the logit scale with SE and upper (UCL) and lower (LCL) 95% confide nce limits. Effects with confidence intervals that do no t overlap zero are highlighted in bold text. Model Parameter Estimate SE LCL UCL prevTMIN Intercept 2.2553 0.2667 1.7325 2.7781 prevTMIN 0.1546 0.0597 0.0377 0.2716 TMIN Intercept 2.2160 0 .2823 1.6627 2.7692 TMIN 0.1603 0.0629 0.0371 0.2836 prevTMIN + prevPRCP Intercept 2.2627 0.2658 1.7417 2.7836 prevTMIN 0.1738 0.0620 0.0523 0.2952 prevPRCP 0.0358 0.0237 0.0823 0.0107 TMIN + PRCP Intercept 2.2137 0.2806 1.6637 2.76 37 TMIN 0.1708 0.0652 0.0431 0.2985 PRCP 0.0191 0.0290 0.0759 0.0376 Year Intercept 3.1325 0.1898 2.7604 3.5046 Year2016 0.5538 0.2848 1.1120 0.0043 Time + prevPRCP + TMIN Intercept 2.2684 0.3128 1.6553 2.8815 Time 0.0051 0.015 4 0.0353 0.0251 prevPRCP 0.0444 0.0287 0.1007 0.0118 TMIN 0.1977 0.0818 0.0373 0.3580 Year + Time Intercept 2.8394 0.3340 2.1847 3.4940 Year2016 0.5246 0.2863 1.0858 0.0365 Time 0.0138 0.0134 0.0125 0.0402 Tag Intercept 2.7636 0.1698 2.4309 3.0963 Tag 0.4423 0.3065 0.1585 1.0431 Year + Time 2 Intercept 3.0276 0.2485 2.5404 3.5147 Year2016 0.5385 0.2859 1.0988 0.0219 T ime 2 0.0002 0.0003 0.0004 0.0007 Constant DSR Intercept 2.9189 0.1412 2.6422 3.1956 Time Intercept 2.5895 0.2974 2.0066 3.1723 Time 0.0162 0.0134 0.0101 0.0425

PAGE 51

45 Parameter estimates for all nest survival models cont'd. Model Parameter Estimate SE LCL UCL NestAge Intercept 3.3376 0.4011 2.5514 4.1238 NestAge 0.0218 0.0190 0.05 89 0.0154 TT Intercept 2.7949 0.2070 2.3892 3.2006 T ime 2 0.0002 0.0003 0.0003 0.0008 prevPRCP Intercept 2.9607 0.1629 2.6413 3.2800 prevPRCP 0.0182 0.0311 0.0792 0.0427 PRCP Intercept 2.9065 0.1642 2.5847 3.2284 PRCP 0.0062 0.0438 0.0797 0.0921 NestAge + Time 2 + PRCP Intercept 3.3397 0.4149 2.5265 4.1529 NestAge 0.0357 0.0213 0.0775 0.0060 T ime 2 0.0005 0.0003 0.0002 0.0011 PRCP 0.0049 0.0419 0.0772 0.0870

PAGE 52

46 APPENDIX D Distance sampling analyses R code # ################################################################################ # # # Mountain Plover distance sampling analysis 2017 # # author: Allison Pierce # # # ################################################################################# library(tidyverse) l ibrary(Distance) library(AICcmodavg) #read in data and format years to factors, mopl < read.csv(file = 'JMJDSAllYears.csv') mopl$year < factor(mopl$year) # area of suitable MOPL habitat in study site mopl$Area < 38.77328 # fit the following detectio n models for each year: # Half normal with cosine adjustments # Half normal with hermite adjustments # Hazard rate with polynomial adjustments # Uniform with cosine adjustments # # Note: The ds() function will try different orders of adjustments (including key only) # and select the one with the best AIC. # # Truncation set at 3% for 2006 due to large outlier, trucation did not improve fit # for any other years # mopl00_hn < ds(subset(mopl,year =="2000"), key = "hn", adjustment = "cos", truncation = "0%") mopl00_hnh < ds(subset(mopl,year =="2000"), key = "hn", adjustment = "herm", truncation = "0%") mopl00_hr < ds(subset(mopl,year =="2000"), key = "hr", adjustment = "poly", truncation = "0%") mopl00_unif < ds(subset(mopl,year =="2000") key = "unif", adjustment = "cos", truncation = "0%") mopl01_hn < ds(subset(mopl,year =="2001"), key = "hn", adjustment = "cos", truncation = "0%") mopl01_hnh < ds(subset(mopl,year =="2001"), key = "hn", adjustment = "herm", truncation = "0%") mopl01_ hr < ds(subset(mopl,year =="2001"), key = "hr", adjustment = "poly", truncation = "0%") mopl01_unif < ds(subset(mopl,year =="2001"), key = "unif", adjustment = "cos", truncation = "0%") mopl02_hn < ds(subset(mopl,year =="2002"), key = "hn", adjustment = "cos", truncation = "0%") mopl02_hnh < ds(subset(mopl,year =="2002"), key = "hn", adjustment = "herm", truncation = "0%") mopl02_hr < ds(subset(mopl,year =="2002"), key = "hr", adjustment = "poly", truncation = "0%") mopl02_unif < ds(subset(mopl,year =="2002"), key = "unif", adjustment = "cos", truncation = "0%") mopl03_hn < ds(subset(mopl,year =="2003"), key = "hn", adjustment = "cos", truncation = "0%") mopl03_hnh < ds(subset(mopl,year =="2003"), key = "hn", adjustment = "herm", truncation = "0% ")

PAGE 53

47 mopl03_hr < ds(subset(mopl,year =="2003"), key = "hr", adjustment = "poly", truncation = "0%") mopl03_unif < ds(subset(mopl,year =="2003"), key = "unif", adjustment = "cos", truncation = "0%") mopl04_hn < ds(subset(mopl,year =="2004"), key = "hn", adjustment = "cos", truncation = "0%") mopl04_hnh < ds(subset(mopl,year =="2004"), key = "hn", adjustment = "herm", truncation = "0%") mopl04_hr < ds(subset(mopl,year =="2004"), key = "hr", adjustment = "poly", truncation = "0%") mopl04_unif < ds(subset (mopl,year =="2004"), key = "unif", adjustment = "cos", truncation = "0%") mopl05_hn < ds(subset(mopl,year =="2005"), key = "hn", adjustment = "cos", truncation = "0%") mopl05_hnh < ds(subset(mopl,year =="2005"), key = "hn", adjustment = "herm", trunca tion = "0%") mopl05_hr < ds(subset(mopl,year =="2005"), key = "hr", adjustment = "poly", truncation = "0%") mopl05_unif < ds(subset(mopl,year =="2005"), key = "unif", adjustment = "cos", truncation = "0%") mopl06_hn < ds(subset(mopl,year =="2006"), ke y = "hn", adjustment = "cos", truncation = 0.3) mopl06_hnh < ds(subset(mopl,year =="2006"), key = "hn", adjustment = "herm", truncation = 0.3) mopl06_hr < ds(subset(mopl,year =="2006"), key = "hr", adjustment = "poly", truncation = 0.3) mopl06_unif < ds (subset(mopl,year =="2006"), key = "unif", adjustment = "cos", truncation = 0.3) mopl2016_hn < ds(subset(mopl,year =="2016"), key = "hn", adjustment = "cos", truncation = list(left = "0%", right = "0%")) mopl2016_hnh < ds(subset(mopl,year =="2016"), ke y = "hn", adjustment = "herm", truncation = list(left = "0%", right = "0%")) mopl2016_hr < ds(subset(mopl,year =="2016"), key = "hr", adjustment = "poly", truncation = list(left = "0%", right = "0%")) mopl2016_unif < ds(subset(mopl,year =="2016"), key = "unif", adjustment = "cos", truncation = list(left = "0%", right = "0%")) # create list of models for each year to average estimates mlist00 < list(mopl00_hn, mopl00_hnh, mopl00_hr, mopl00_unif) mlist01 < list(mopl01_hn, mopl01_hr, mopl01_unif) mlist02 < list(mopl02_hn, mopl02_hr, mopl02_unif) mlist03 < list(mopl03_hn, mopl03_hr, mopl03_unif) mlist04 < list(mopl04_hn, mopl04_hr, mopl04_unif) mlist05 < list(mopl05_hn, mopl05_hnh, mopl05_hr, mopl05_unif) mlist06 < list(mopl06_hn, mopl06_hr, mopl06_un if) mlist16 < list(mopl2016_hn, mopl2016_hr, mopl2016_unif) #function that pulls estimates from ds model objects getEstimates < function(model){ summary < summary(model) p < summary[[1]]$average.p p.se < summary[[1]]$average.p.se D < summary[[2]][[1]]$D$Estimate D.lower < summary[[2]][[1]]$D$lcl D.upper < summary[[2]][[1]]$D$ucl N < summary[[2]][[1]]$N$Estimate N.lower < summary[[2]][[1]]$N$lcl N.upper < summary[[2]][[1]]$N$ucl LL < logLik(model)

PAGE 54

48 data < d ata.frame( estimate = c("p","D","N","LL"),value = c(p,D,N,LL), lowerCI = c(NA,D.lower,N.lower,NA), upperCI = c(NA,D.upper,N.upper,NA)) } # function that calculates model average of estmates and their log normal CIs from list of models # this uses the AICc modavg package avgModels < function(mlist){ estimates < lapply(mlist, FUN = getEstimates) LLs < sapply(seq_along(mlist), function(i) mlist[[i]]$ddf$lnl) modNames < sapply(seq_along(mlist), function(i) mlist[[i]]$ddf$name.message) estimat es.Pa < sapply(seq_along(mlist), function(i) mlist[[i]]$dht$individuals$average.p) estimates.D < sapply(seq_along(mlist), function(i) mlist[[i]]$dht$individuals$D$Estimate) estimates.N < sapply(seq_along(mlist), function(i) mlist[[i]]$dht$individual s$N$Estimate) estimates.esw < estimates.Pa sapply(seq_along(mlist), function(i) mlist[[i]]$ddf$ds$aux$int.range[2]) se.Pa < sapply(seq_along(mlist), function(i) summary(mlist[[i]])[[1]]$average.p.se) se.D < sapply(seq_along(mlist), function( i) mlist[[i]]$dht$individuals$D$se) se.N < sapply(seq_along(mlist), function(i) mlist[[i]]$dht$individuals$N$se) se.esw < se.Pa sapply(seq_along(mlist), function(i) mlist[[i]]$ddf$ds$aux$int.range[2]) K < sapply(seq_along(mlist), function(i) length(mlist[[i]]$ddf$par)) nobs < sapply(seq_along(mlist), function(i) nrow(mlist[[i]]$ddf$data)) logLimits < function(Est,SE){ logVar < log1p(SE^2/Est^2) C < exp(1.96 *sqrt(logVar)) return(c(logLL = Est/C, LogUL = Est C)) } detectionProb < modavgCustom(logL = LLs,K = K,modnames = modNames,estimate = estimates.Pa,se = se.Pa, nobs = nobs,second.ord = F) density < modavgCustom(logL = LLs,K = K, modnames = modNames, estimate = estimates.D, se = se.D, nobs = nobs, second .ord = F) nhat < modavgCustom(logL = LLs, K = K, modnames = modNames, estimate = estimates.N, se = se.N, nobs = nobs, second.ord = F) esw < modavgCustom(logL = LLs, K = K, modnames = modNames, estimate = estimates.esw, se = se.esw, nobs = nobs, second.ord = F) print(detectionProb) print(density) print(nhat) print(esw) avgEstimates < data.frame(estimate = c("Pa","D","N", "ESW", "ER"), modavg = c(detectionProb$Mod.avg.est, density$Mod.avg.est, nhat$Mod .avg.est, esw$Mod.avg.est, mlist[[1]]$dht$individuals$summary$ER), SE = c(detectionProb$Uncond.SE, density$Uncond.SE, nhat$Uncond.SE, esw$Uncond.SE, mlist[[1]]$dht$individuals$summary$se.ER), LL = c (detectionProb$Lower.CL, density$Lower.CL, nhat$Lower.CL, esw$Lower.CL, NA), UL = c(detectionProb$Upper.CL, density$Upper.CL, nhat$Upper.CL, esw$Upper.CL, NA), logLL = c(logLimits(detectionProb$Mod. avg.est,detectionProb$Uncond.SE)[1], logLimits(density$Mod.avg.est,density$Uncond.SE)[1], logLimits(nhat$Mod.avg.est,nhat$Uncond.SE)[1], lo gLimits(esw$Mod.avg.est,esw$Uncond.SE)[1], NA), logUL = c(logLimits(detectionProb$Mod.avg.est,detectionProb$Uncond.SE)[2], logLimits(density$Mod.avg.est,density$Uncond.SE)[2], logLimits(nhat$Mod.avg.est,nhat$Uncond.SE)[2], logLimits(esw$Mod.avg.est,esw$Uncond.SE)[2],NA))

PAGE 55

49 avgEstimates$cv < avgEstimates$SE/avgEstimates$modavg return(avgEstimates) } #get mod el averaged estimates for each year Modavg00 < avgModels(mlist00) Modavg01 < avgModels(mlist01) Modavg02 < avgModels(mlist02) Modavg03 < avgModels(mlist03) Modavg04 < avgModels(mlist04) Modavg05 < avgModels(mlist05) Modavg06 < avgModels(mlist06) Mod avg16 < avgModels(mlist16) #combine it all into one dataframe ModavgAll < bind_rows("2000" = Modavg00,"2001" = Modavg01,"2002" = Modavg02, "2003" = Modavg03,"2004" = Modavg04,"2005" = Modavg05, "2006" = Moda vg06,"2016" = Modavg16, .id = "year")

PAGE 56

50 APPENDIX E Nest survival analyses R code ################################################################################# # # # Mountain Plover nest survival analysis 2017 # # author: Allison Pierce # # # #################### ############################################################# library(RMark) library(tidyverse) library(forcats) library(readxl) library(lubridate) #read in nest monitoring data load("mopl.rda") #read in weather data SPWeather < read.csv("SPWeather.csv ", na.strings = c("9999"," 9999")) SPWeather$date < as.Date(as.character(SPWeather$DATE),format = "%Y%m%d") #select relevant columns and subset by station weather < dplyr::select(SPWeather,STATION_NAME,date,PRCP,TMIN) antero < filter(weather,STATION_N AME == "ANTERO RESERVOIR CO US") fairplay < filter(weather,STATION_NAME != "ANTERO RESERVOIR CO US") #24 period ends at observation time (7:30 AM) #Prev night/day (reported on day of record) prevdayweather15 < subset(antero, date > as.Date("2015 06 0 1") & date < as.Date("2015 07 24")) prevdayweather16 < subset(antero, date > as.Date("2016 06 01") & date < as.Date("2016 07 24")) #current day reported on next day of record weather15 < subset(antero, date > as.Date("2015 06 02") & date < as.Date("20 15 07 25")) weather16 < subset(antero, date > as.Date("2016 06 02") & date < as.Date("2016 07 25")) #add previous day measures to data frame for each year weather15$prevTMIN < prevdayweather15$TMIN weather16$prevTMIN < prevdayweather16$TMIN weather15$p revPRCP < prevdayweather15$PRCP weather16$prevPRCP < prevdayweather16$PRCP ## Get weather for FAIRPLAY to fill in NA's in 2015 fpprevdayweather15 < subset(fairplay, date > as.Date("2015 06 01") & date < as.Date("2015 07 24")) fpweather15 < subset(fair play, date > as.Date("2015 06 02") & date < as.Date("2015 07 25")) fpweather15$prevTMIN < fpprevdayweather15$TMIN #Fill in NA's for 2015 with data from Fairplay missing < which(is.na(weather15), arr.ind = T) weather15[missing] < as.numeric(fpweather15[ missing]) #bind weather in repeating order for adding the design data weatherbind < rbind(weather15,weather16) #prepare nest data to add in time varying covariates for RMark mopl_dp < process.data(mopl,nocc=max(mopl$LastChecked),model="Nest", groups=c( "Year","Tag")) mopl_ddl < make.design.data(mopl_dp) #create data frame of ti me varying covariates occasion_year < data.frame(group = mopl_ddl$S$group, time = mopl_ddl$S$time, PRCP = weatherbind$PRCP, prevPRCP = weatherbind$p revPRCP, prevTMIN = weatherbind$prevTMIN, TMIN = weatherbind$TMIN, date = weatherbind$date)

PAGE 57

51 #merge time varying covariates to design matrix mopl_ddl$S < merge_design.covariates(mopl_ddl$S,occasion_year, bytime = T, bygroup = T) #create quadratic time covariate mopl_ddl$S$TT< as.numeric(mopl_ddl$S$Time)^2 #define nest survival models for RMark mopl_models < function(){ #DSR is constant dot < mark(mopl_dp,mopl_ddl,nocc=max(mopl$LastChecked),model="Nest", profile.int = T) ###### Temporal models (year and seasonal) #DSR is constant over season but varies by year Year < mark(mopl_dp,mopl_ddl,nocc=max(mopl$LastChecked),model="Nest", model.parameters=list(S=list(formula=~Year)), groups="Year", p rofile.int = T) #DSR varies linearly over time and Year Year.T < mark(mopl_dp,mopl_ddl,nocc=max(mopl$LastChecked),model="Nest", model.parameters=list(S=list(formula=~Year + Time)), groups="Year", profile.int = T) #DSR varies quadratically over time and Year Year.TT < mark(mopl_dp,mopl_ddl,nocc=max(mopl$LastChecked),model="Nest", model.parameters=list(S=list(formula=~Year + TT)), groups="Year", profile.int = T) #DSR varies linearly over time Time < mark(mopl_dp,mopl_ddl,nocc=max(mopl$LastChecked),model="Nest", model.parameters=list(S=list(formula=~Time)), groups="Year", profile.int = T) #DSR varies quadratically over time TT < mark(mopl_dp,mopl_ddl,nocc=max(mopl$LastChecked), model="Nest", model.parameters=list(S=list(formula=~TT)), groups="Year", profile.int = T) #DSR varies by age of nest NestAge < mark(mopl_dp,mopl_ddl,nocc=max(mopl$LastChecked),model="Nest", model.parameters=list(S=l ist(formula=~NestAge)), profile.int = T) ###### Weather models #DSR varies with day of precipitation PRCP < mark(mopl_dp,mopl_ddl,nocc=max(mopl$LastChecked),model="Nest", model.parameters=list(S=list(formula=~PRCP)), profile.in t = T) #DSR varies with prev day precipitation prevPRCP < mark(mopl_dp,mopl_ddl,nocc=max(mopl$LastChecked),model="Nest", model.parameters=list(S=list(formula=~prevPRCP)), profile.int = T) #DSR varies with day of min tempera ture TMIN < mark(mopl_dp,mopl_ddl,nocc=max(mopl$LastChecked),model="Nest", model.parameters=list(S=list(formula=~ TMIN)), profile.int = T) #DSR varies with prev day min temperature prevTMIN < mark(mopl_dp,mopl_ddl,nocc=max(mopl$L astChecked),model="Nest", model.parameters=list(S=list(formula=~ prevTMIN)), profile.int = T) #DSR varies with day of min temp and precip (additive) TMIN.PRCP < mark(mopl_dp,mopl_ddl,nocc=max(mopl$LastChecked),model="Nest", model.parameters=list(S=list(formula=~ TMIN + PRCP)), profile.int = T) #DSR varies with prev day min temp and precip (additive) prevTMIN.prevPRCP < mark(mopl_dp,mopl_ddl,nocc=max(mopl$LastChecked),model="Nest", model.parameters=list(S=list(formula=~ prevTMIN + prevPRCP)), profile.int = T) ### GPS Tag #DSR varies with application of GPS tag

PAGE 58

52 Tag < mark(mopl_dp,mopl_ddl,nocc=max(mopl$LastChecked),model="Nest", model.parameters=li st(S=list(formula=~ Tag)), groups = "Tag", profile.int = T) ### models analogous to other papers drietz < mark(mopl_dp,mopl_ddl,nocc=max(mopl$LastChecked),model="Nest", model.parameters=list(S=list(formula=~ Time + prevPRCP + T MIN)), profile.int = T) dinsmore < mark(mopl_dp,mopl_ddl,nocc=max(mopl$LastChecked),model="Nest", model.parameters=list(S=list(formula=~ NestAge + TT + PRCP)), profile.int = T) return(collect.models(table = T)) } #run model s mopl_results < mopl_models() #collect beta estimates from all models getBetas < function(models){ betatable < data.frame(NULL) for(i in 1:length(models)){ betas < models[[i]]$results$beta betas$parameter < dimnames(models[[i]]$results$ beta)[[1]] betas$model < models[[i]]$model.name betas$AICc < models[[i]]$results$AICc betatable < rbind(betatable, betas) } return(betatable) } betadata < getBetas(mopl_results) %>% arrange(AICc) betadata < betadata[c(6,5,1,2,3,4,7) ] #create DSR plots #### DSR prev tmin s.temp < coef(mopl_results$prevTMIN) tminrange < seq(min(weatherbind$prevTMIN),max(weatherbind$prevTMIN),0.5) logit.values < s.temp[1,1] + tminrange s.temp[2,1] deriv < matrix(1,ncol=2,nrow=length(tminrange) ) deriv[,2] < tminrange std.errors < sqrt(diag(deriv%*%mopl_results$prevTMIN$results$beta.vcv[1:2,1:2]%*%t(deriv))) lcl.logit < logit.values 1.96*std.errors ucl.logit < logit.values+1.96*std.errors tminmodel < data.frame(Surv = plogis(logit.values), tminrange = tminrange, LCL = plogis(lcl.logit), UCL = plogis(ucl.logit)) ggplot(data = tminmodel, aes(y = Surv, x = tminrange, ymin = 0.7, ymax = 1)) + geom_ribbon(aes(ymin = LCL, ymax = UCL), alpha = 0.5) + geom_line(size = 0.8) + theme_classic() + scale_x_continuous(limits = c(min(weatherbind$prevTMIN),max(weatherbind$prevTMIN)), breaks = seq(ceiling(min(weatherbind$prevTMI N)),floor(max(weatherbind$prevTMIN)),2)) + scale_y_continuous(breaks = seq(0.75,1,0.05)) + labs(x = expression("Daily Minimum Temperature ("*degree*C*")"), y = "Daily Survival Rate (DSR)") + theme(legend.position = c(0.7,0.3), text = element_ text(size = 10)) #### DSR as a function of prev tmin by prev precip (none, light, med, heavy) s.temp < coef(mopl_results$prevTMIN.prevPRCP) tminrange < seq(min(weatherbind$prevTMIN),max(weatherbind$prevTMIN),0.5) logit.values.N < s.temp[1,1] + tminran ge s.temp[2,1] + 0 s.temp[3,1]

PAGE 59

53 logit.values.L < s.temp[1,1] + tminrange s.temp[2,1] + 1.25 s.temp[3,1] logit.values.M < s.temp[1,1] + tminrange s.temp[2,1] + 5 s.temp[3,1] logit.values.H < s.temp[1,1] + tminrange s.temp[2,1] + 25 s.temp[ 3,1] deriv < matrix(1,ncol=2,nrow=length(tminrange)) deriv[,2] < tminrange std.errors < sqrt(diag(deriv%*%mopl_results$prevTMIN.prevPRCP$results$beta.vcv[1:2,1:2]%*%t(deriv))) lcl.logit.N < logit.values.N 1.96*std.errors ucl.logit.N < logit.values.N+ 1.96*std.errors lcl.logit.L < logit.values.L 1.96*std.errors ucl.logit.L < logit.values.L+1.96*std.errors lcl.logit.M < logit.values.M 1.96*std.errors ucl.logit.M < logit.values.M+1.96*std.errors lcl.logit.H < logit.values.H 1.96*std.errors ucl.log it.H < logit.values.H+1.96*std.errors preciplevel < factor(rep(c("no","light","mod","heavy"), each = length(logit.values.N))) preciplevel < fct_relevel(preciplevel,"heavy","mod","light","no") tminrange < rep(tminrange,4) tminprcpmodel < data.frame(S urv = plogis(c(logit.values.N,logit.values.L,logit.values.M,logit.values.H)), tminrange = tminrange, LCL = plogis(c(lcl.logit.N,lcl.logit.L,lcl.logit.M,lcl.logit.H)), UCL = plogis(c(ucl.logit.N,ucl.logit.L,ucl.logit.M,ucl.logit.H)), Precipitation = preciplevel) ggplot(data = tminprcpmodel, aes(y = Surv, x = tminrange, color = Precipitation, ymin = 0.70, ymax = 1)) + geom_line(size = 0.8) + t heme_classic() + scale_x_continuous(limits = c(min(weatherbind$prevTMIN),max(weatherbind$prevTMIN)), breaks = seq(ceiling(min(weatherbind$prevTMIN)),floor(max(weatherbind$prevTMIN)),2)) + scale_y_continuous(breaks = seq(0.75,1,0.05 )) + scale_color_grey(name = "Total Precipitation", labels = c("Heavy (25mm)", "Moderate (5 mm)", "Light (1.25 mm)","None (0 mm)"), start = 0, end = 0.8) + labs(x = expression("Daily Minimum Temperature ("*degree *C*")"), y = "Daily Survival Rate (DSR)") + theme(legend.position = c(0.7,0.3), text = element_text(size = 10))