Looking at the Bigger Picture: How Abundance of Nesting and Brooding Habitat Influences Lek-Site Selection by Lesser Prairie-Chickens

ABSTRACT.—Lesser Prairie-Chicken (Tympanuchus pallidicinctus) populations have declined throughout most of their distribution since the mid-1980s. These declines are largely attributed to loss of habitat through the conversion and expansion of cropland, construction of oil wells and other anthropogenic features on the landscape, and grazing intensification. Changes inhabit availability and quality are seemingly having a disproportionate effect on the reproductive habitat of Lesser Prairie-Chickens, as some populations continue to decline.Nest and brood survival are crucial to population growth of Lesser Prairie-Chickens, with adequate reproductive habitat vital to population persistence. To better understand the influence of reproductive habitat availability on populations, we quantified the composition of reproductive habitat in lek landscapes across the northern extent of the Lesser Prairie-Chicken range. We measured vegetation at six study sites in Kansas and Colorado from 2013–2016. We sought to quantify available nest and brooding habitat adjacent to leks, investigate the relationship between reproductive habitat availability and lek attendance by males at several spatial scales, and examine vegetation characteristics that influence lek attendance. Within 5 km of a lek, 25% (2546/10,320 points) and 26% (2682/10,320 points) of random locations provided nesting and brooding habitat, respectively. Changes to reproductive habitat at both scales affected male attendance at leks. Visual obstruction of vegetation was the main predictor of male lek attendance at both spatial scales and limited the amount of reproductive habitat in lek landscapes. Accordingly, management should increase visual obstruction throughout theLesser Prairie-Chicken range to increase reproductive success and improve populations to facilitate achieving the conservation goal set by the Western Association of Fish and WildlifeAgencies of a 10 y average Lesser Prairie-Chicken population of 67,000 birds.

INTRODUCTION

Many grouse species (Tympanuchus cupido, T. phasianellus, Centrocercus urophasianus),including the Lesser Prairie-Chicken (T. pallidicinctus), use a lek mating system in which male grouse gather at a centralized location, or lek, to display and breed with females(Johnsgard, 1973). In the case of Lesser Prairie-Chickens, mated females subsequently nest and raise their brood within 5 km of the lek (Giesen, 1994; Pitman et al., 2006; Bell et al.,2010; Winder et al., 2015; Sullins et al., 2018a, Lautenbach et al., 2019); henceforth, the area immediately surrounding a lek (,5 km radius) will be referred to as the lek landscape. With all reproductive activity occurring in such a concentrated area, the lek landscape must contain vegetation composition and structure that meets the criteria of each reproductive stage for Lesser Prairie-Chickens to be successful in their reproductive efforts. Knowledge of the importance of leks and use of adjacent areas is well established in the literature (Taylor and Guthery, 1980; Bradbury, 1981; Giesen, 1994; Pitman et al., 2006); however, there is limited information on why certain leks and lek landscapes are more successful compared to others. From a conservation perspective, this lack of information can be costly as the preservation of grassland areas surrounding leks has not always been effective (Boal andHaukos, 2016). For conservation to be successful, knowledge of the vegetative characteristics within lek landscapes and the landcover types that preserve them is necessary.

Lesser Prairie-Chickens have differing habitat requirements for each breeding stage: lekking(mating stage), nesting (egg laying and incubation stage), and brooding (chick-rearing stage).Leks require unobstructed open areas with short vegetation at relatively high elevation or in alkali flats (Giesen, 1998). In contrast females select nest sites that have relatively tall and dense vegetation and greater litter depth, while avoiding anthropogenic structures and trees on the landscape (Giesen, 1994; Johnson et al., 2004; Lautenbach et al., 2017; Plumb et al., 2019).Vegetation cover at nests is typically comprised of mid-height or tall grasses, particularly bunchgrasses (e.g., little bluestem [Schizachyrium scoparium], sideoats grama [Bouteloua curtipendula])and shrubs (sand sagebrush [Artemisia filifolia], sand shinnery oak [Quercus havardii]), given these species provide adequate overhead and lateral cover that allow the female to detect predators and escape when necessary for her survival (Haukos and Smith, 1989; Riley et al.,1992; Lautenbach et al., 2019). Females with broods require forb-rich patches (7-37% forbcover) that can be of moderate shrub and subshrub cover to allow for easy movement and access to arthropods, yet provide sufficient overhead cover (Hagen et al., 2013; Lautenbach,2015; Sullins et al., 2018b). Broods also require tall grasses to provide appropriate amounts of thermal cover in relatively cooler microhabitats on the landscape (Lautenbach, 2017).

Before nesting and brooding, female Lesser Prairie-Chickens choose a lek or lek complex to attend. There are several hypotheses as to how leks are formed and what scales are important in driving lek attendance, including the hotshot and hotspot hypotheses(Bradbury and Gibson, 1983; Dastagir et al., 1997). The hotshot hypothesis states attractiveness of males on a lek brings females to a lek (Dastagir et al., 1997). For a female to evaluate attractiveness of a male, characteristics of the lek and its immediate surroundings may be important, as an elevated area with short vegetation that can project sound can be ideal (Nooker and Sandercock, 2008). In contrast the hotspot hypothesis states males congregate in areas where females are known to occur (Bradbury and Gibson,1983). Females gather in areas representing quality nesting and brooding habitat, typically found within a 5 km radius of a lek (Winder et al., 2015; Sullins et al., 2018a). The hotspot hypothesis emphasizes the importance of the broad landscape surrounding a lek, whereas the hotshot hypothesis may emphasize the importance of only the lekking grounds themselves. Based on hierarchy theory, both broad-scale and emergent fine-scale factors can constrain the occupancy and abundance of Lesser Prairie-Chickens on leks, but it is unclear how factors pertinent to both hypotheses operating at multiple scales influence LesserPrairie-Chicken lek attendance (Bradbury and Gibson, 1983; Dastagir et al., 1997).

In order to understand habitat factors important to Lesser Prairie-Chicken reproductive success, we first identified factors that predict lek attendance. We tested landcover type,habitat type, and vegetation characteristics at the lek landscape scale as well as vegetation characteristics of the lek at the micro-habitat (100 m radius) scale. At the 5 km radius, we predicted that greater percent grassland landcover would influence lek attendance, given previous studies indicating the benefits of grasslands for overall resilience in prairie grouse populations (Niemuth, 2000; Ross et al., 2016; Sullins et al., 2018a). Based on the hotspot hypothesis and other grouse studies (White, 1993; Gibson, 1996; Doherty et al., 2010), we hypothesized greater availability and quality of nesting and brooding cover would drive lek attendance. Subsequently, we expected vegetation height and density (visual obstruction readings) to drive lek attendance, given the relationship between nesting habitat and visual obstruction (Lautenbach et al., 2019). Visual obstruction is a common method used to evaluate grassland structure that accounts for both vegetation height and density simultaneously and is a well-regarded descriptor of prairie-chicken habitat (Robel et al.,1970). To describe nesting and brooding habitat abundance, we estimated percent vegetation composition among study sites and examined which, if any, vegetation communities may be less conducive to prairie-chicken reproductive success. This study provides an initial description of the relative composition of available nesting and brooding habitat associated with active leks in the currently occupied northern Lesser Prairie-Chicken range. Our findings allow us to assess the quality of grassland with respect to nesting and brooding habitat of theLesser Prairie-Chicken to gain a better understanding of how to sustain their populations.

METHODS

STUDY AREA

From 2013 to 2016, we sampled vegetation on and surrounding 62 leks at six different study sites throughout the northern extent of the Lesser Prairie-Chicken range in westernKansas and eastern Colorado (244,453 ha in total; Fig. 1; Table A1). The study sites represented three of the four Eco-regions occupied by Lesser Prairie-Chickens as defined byMcDonald et al. (2014), including Mixed-Grass Prairie, Sand Sagebrush Prairie, and Short-Grass Prairie/Conservation Reserve Program (CRP) Mosaic eco-regions. From the western- to the easternmost study site, average annual precipitation ranged from 37–69 cm (Table A1).

The Red Hills study site (49,111 ha) represented the eastern border of the Lesser Prairie-Chicken range and was located in Comanche and Kiowa counties, Kansas within the Mixed-Grass Prairie Ecoregion. The Clark study site (47,466 ha), primarily located in western ClarkCounty, Kansas, was comprised of two privately owned ranches, one within the CimarronRiver floodplain (32,656 ha) and the other 20 km north surrounded by rolling hills (14,810ha). This site was within the Mixed-Grass Prairie Ecoregion, but the Cimarron Riverfloodplain had a substantial sand sagebrush (Artemisia filifolia) component within the Mixed-Grass Prairie Ecoregion. The Northwest Kansas study area (129,762 ha) encompassed bothGove County and Logan County study sites and was located in the Short-Grass Prairie/CRPMosaic Ecoregion. One Colorado study site was located on the border of Prowers and Bacacounties (1146 ha). The site was located within the Sand Sagebrush Prairie Ecoregion;however, Lesser Prairie-Chickens at this site predominantly relied on CRP grasslands fornesting. The Colorado study site in Cheyenne County (16,968 ha) was comprised of large expanses of lightly and heavily grazed sand sagebrush prairie and for which 30yr precipitation averages were the lowest of all the study sites at 37 cm (Grisham et al., 2016).

DATA COLLECTION

Leks were located by the presence of displaying males during February–May, 2013–2016. We defined a lek as an area occupied by two or more displaying males and recorded locations of all leks within our study areas detected through intensive searches each spring. Counts of individual Lesser Prairie-Chickens attending each lek were frequently recorded during March–May (averaging 21 unique observations at each of the 62 lek sites). Lek locations (center of area of displaying males) were recorded using a Global Positioning System (GPS; Garmin Vista, Kansas City, MO, U.S.A.). Lek area was estimated based on the distribution of males on a lek. Observers marked 10 m intervals in each cardinal direction from the lek center until the perimeter of displaying males was reached. Average lek area was estimated by averaging the area of each lek based on the number of 10 m intervals that included displaying males throughout lekking period.

Vegetation sampling within 100 m of leks.—We sampled vegetation within 100 m of leks at 35leks in Kansas during spring of 2014 and 2015. Sampling occurred at 10 m intervals along a100 m transect starting at the center of each lek and spanning out in all four cardinaldirections (Johnson et al., 2004; Hunt and Best, 2010). We chose 10 m intervals to ensuremeasurement independence, as each visual obstruction reading was taken 4 m from the Robelpole (Robel et al., 1970). The lek center was determined visually from observations amongmultiple days as the midpoint of northernmost, southernmost, easternmost, and westernmostmales. At lek center and each 10 m interval, we measured visual obstruction (VOR) at 100%obstruction to the nearest decimeter (dm; meaning the entire dm was fully obstructed from view)using a Robel pole (Robel et al., 1970). In addition we randomly placed a 60 360 cm modifiedDaubenmire (1959) frame alternately on the left and right side of the transect. Within the frame we measured grass height (cm), litter depth (cm), and estimated percent ground cover of grass, forb, bare ground, litter, and shrub. Litter was identified as dead or dormant herbaceous vegetation lying parallel with the ground. All measurements taken were classified as on or off the lek based on whether Lesser Prairie-Chickens were observed displaying as far out as the measurement was taken. We then averaged all measurements by lek.

Vegetation sampling within 5 km of leks.—At the 5 km scale, we sampled vegetation at points on the lek landscape that were available to Lesser Prairie-Chickens. These points were randomly generated throughout the study sites at a rate of 1 per 4 ha with a maximum of 10points per individual patch. Patches were defined as areas of homogenous grassland vegetation .2 ha in area, categorized based on vegetation type (e.g., native working grassland, cropland, and CRP) via aerial imagery in ArcGIS 10.5 (product of: ESRI, i-cubed,USDA FSA, USGS, AEX, GeoEye, Getmapping, Aerogrid, and IGP), and confirmed uponground truthing (Robinson et al., 2016). Vegetation was measured at random points during the spring/early summer, with the earliest points being measured on 26 March, and the latest points measured on 4 June. Vegetation was also measured within a stratified random sample of 20% of patches later in the summer with the earliest point measured on 2 June and continuing until 12 October. Measurements were also taken during the fall/winter, with the earliest points measured on 7 October and the latest on 18 March. We limited random points to only include those within 5 km of any established lek as these points were within the estimated nesting and brooding range of Lesser Prairie-Chickens (Bell et al., 2010;Sullins et al., 2018a). We measured the distance between each random location and nearest established lek center using the ‘Near’ tool in ArcGIS 10.5.

Vegetation characteristics.—At each random location within the lek landscape, we estimated percent cover of forbs, bare ground, grass, and shrub at the point center and 4 m from the point center at each cardinal direction within a 60 360 cm modified Daubenmire frame( five estimates/point). We recorded visual obstruction and vertical vegetation density 4m from point center at each cardinal direction by estimating the dm at which 50% and 75%obstruction (based on Lautenbach et al., 2019) occurred using a Robel pole (four estimates/point; Robel et al., 1970). Litter depth (cm) was measured at 0.5 m increments stretching 4m north, east, south, and west of point center (32 estimates/point; Davis et al., 1979). Maximum grass height (cm) was measured at each random point location.

Landcover type classification and percent abundance.—We estimated percent landcover type within each lek landscape using landcover data that differentiated among cropland, CRP, native working grassland, and ‘‘other’’ , over each study site lek landscape in ArcGIS 10.5 using the 2011 National Landcover dataset and a unique shape file identifying the distribution of CRP grasslands in 2014 (Homer et al., 2015). Native working grassland was used to describe grassland that was not CRP and typically used as pastureland. We added another category labeled grassland, which combined CRP and native working grassland values in our subsequent analyses.

Nesting and brooding habitat abundance estimation.—Based on criteria in previous literature(Lautenbach, 2015; Lautenbach et al., 2019), we quantified the proportion of random points within the lek landscape as potential nesting habitat, brooding habitat, or reproductive habitat (meet the criteria for either nesting or brooding). We classified points that had both10% bare ground and an average 75% visual obstruction between 1.5–3.5 dm as nesting habitat. Similarly, we classified random points that had 50% visual obstruction between 2–5dm and for cover between 7% –37% as brooding habitat. We estimated the proportion of reproductive habitat type by dividing the number of random points within each 5 km lekbuffer that met each breeding habitat criteria (nesting, brooding, and reproductive) overthe total number of random points within each 5 km lek buffer. We sampled between 49 and2812 random points within each lek landscape to obtain proportion of reproductive habitat types representative of the lek landscape. We assessed if samples could be biased by variation in the number of random samples at each lek landscape by randomly selecting 47 samples from each lek landscape (to represent the smallest sample size), we then compared means and standard deviation of 75% visual obstruction to the lek landscape with the largest sample size (2812). We chose the mean and SD of VOR at 75% obstruction as representative of the spread of other vegetation measurements. Because mean and SD did not differ basedon sample sizes, we used all random locations in our analyses. We grouped all points thatmet each reproductive habitat type within each lek landscape into six groups based on thestudy areas in which the leks were found.

STATISTICAL ANALYSES

Predictors of lek attendance within 100 m of leks.—We evaluated lek attendance (averagenumber of birds on a lek) as a function of vegetation characteristics measured within 100 mof each lek. We used measurements from vegetation covariates (as described in Vegetationsampling within 100 m of leks) that were considered off the lek, as vegetation immediatelysurrounding the lek may be responsible for attracting Lesser Prairie-Chickens onto a lek. We used single variable and quadratic Poisson regressions to evaluate the relative influence of measured covariates on lek attendance and included a null intercept-only model. We tested the influence of each vegetation covariate on lek attendance in a single variable and quadratic model for a total of 16 models, including the constant null model. We ranked allmodels using Akaike’s Information Criterion corrected for small sample sizes (AICc;Burnham and Anderson, 2002). If the top ranked model was null, we declared vegetationwithin 100 m surrounding the lek center was not related to lek attendance. We assessed models for significance based on the co-efficients and concluded there was no relationship between the covariate and lek attendance if the slope co- efficient overlapped with zero at the 85% confidence interval (Arnold, 2010). If models were within 2 AICcof each other, we considered them, in the absence of spurious variables, to be competing models and both potential predictors of lek attendance.

Predictors of lek attendance within 5 km of leks.—We tested the response of lek attendance as a function of vegetation characteristics measured, percent landcover composition (nativeworking grassland, CRP, and grassland), and percent abundance of nesting, brooding, and reproductive habitat within a 5 km radius of a lek using Poisson regression in three separate suites of models. We also used a null model in each suite. If the null model ranked highest,we declared lek attendance was not explained by the covariates we examined. To assess predictors of lek attendance at the 5 km scale, we used the same statistical approach as at the100 m scale. We ranked all models using AICcand assessed model slope co-efficients for overlap with zero at the 85% confidence interval (Burnham and Anderson, 2002; Arnold,2010). If competing models were within 2 AICc, in the absence of spurious variables, we considered both to be potential predictors of lek attendance.

Vegetation characteristics.—Our first suite of models included seven covariates of vegetation characteristics (as described in Vegetation sampling within 5 km of leks). We averaged all repeated measurements at each random point and used the average as covariates in Poisson regression models to identify relationships with lek counts of males. Each vegetation characteristic covariate was then developed into a single variable and quadratic effect candidate model for a total of 15 models. Based on niche theory, we expected that many variables would be quadratic and optimized at some value. For example there is extensive evidence that grass visual obstruction can be too short or too tall (Lautenbach et al., 2019).

Landscape characteristics.—In the second suite of models, we tested the effects of land cover composition on lek attendance by males using seven single variable and quadratic effect candidate models. Previous studies have documented an increase in Lesser Prairie-Chickenhabitat after an increase in overall grassland abundance following the implementation ofCRP (Hagen et al., 2004; Spencer et al., 2017; Sullins et al., 2018a). We assessed the influence of native working grassland, CRP, and grassland (native working grassland þCRP)abundance in a landscape on Lesser Prairie-Chicken lek attendance.

Habitat type.—We estimated the amount and variance of nesting and brooding habitat surrounding each lek and at each site to provide one of the first estimates of available habitat composition in areas occupied by Lesser Prairie-Chickens based on structural and composition characteristics known to be selected by Lesser Prairie-Chickens. We then used the proportional estimates of available habitat within each lek landscape to evaluate the relationship with lek attendance. We calculated proportions of nesting, brooding, and reproductive habitat within each lek landscape and site based on the vegetative characteristics at random points and criteria of Lautenbach (2015) and Lautenbach et al., (2019). We used adjusted proportions of each habitat type to account for the fact that all random points sampled were in grasslands, given not all lek landscapes were centered only within grasslands and many included cropland or other landcover types. To assess the relationship of nest and brooding habitat abundance on lek attendance, we developed a suite of seven candidate models that included covariates of nesting,brooding, and reproductive habitat abundance as single variable models. To provide some inference on the configuration of nesting and brooding habitat within lek landscapes, we examined how the abundance of nesting and brooding habitat varied in relation to distance from lek by modeling points that met nesting and brooding criteria (as defined in Nesting and brooding habitat abundance estimation) against how these points varied in distance from a lek. We illustrated these relationships using linear models.

CONTINUED…..

FOR FULL RESULTS CLICK ON READ FULL ARTICLE BELOW.