Inadequate thermal refuge constrains landscape habitability for a grassland bird species

Study area

The rolling plains of Texas, USA (Gould, 1969) is a semi-arid physiographic region in northwestern Texas that historically supported large bobwhite subpopulations that are now declining in abundance (Peterson, Wu & Rho, 2002). Topography is characterized by flat to gently rolling plains intersected by streams that flow in an east to southeasterly direction; elevation ranges from 215 to 950 m (Gould, 1969). Soils vary from coarse sands to tightly packed clays, with substantial fine scale variability in soil type. Annual precipitation increases in an easterly direction, from roughly 550 mm to nearly 760 mm; May and September are the wettest months, with a summer dry period. Plant communities were dominated historically by mid or tall bunch grass species, including little bluestem (Schizachyrium scoparium), big bluestem (Andropogon gerardii), sand bluestem (Andropogon hallii), Indian grass (Sorghastrum nutans), and sideoats grama (Bouteloua curtipendula). Riparian areas supported woody plant communities that included oaks (Quercus spp.), eastern cottonwood (Populus delitoides), elms (Ulmus spp.), and junipers (Juniperus spp.). Extensive woody plant encroachment due to overgrazing and/or fire suppression has resulted in reduction of grass-dominated range sites. Honey mesquite (Prosopis glandulosa) is a common woody species throughout the rolling plains, with shinnery oak (Quercus havardii) and sand sage (Artemisia filifolia) primarily on sandy soils. As a result, rangeland comprised of mixed native grass-herbaceous vegetation with interspersed shrubs characterizes much of the land cover. Although livestock grazing operations dominate land use across this region as a whole, many areas have been converted to cropland (USDA-NASS, 2007).

We conducted our study on six ranches dispersed across the rolling plains, which range in size from roughly 1,600 to 78,200 ha. These ranches capture much of the landscape variability (i.e., geology, soil type, vegetation, and topography) experienced by bobwhites in the rolling plains ecoregion. All six ranches were managed for a combination of livestock grazing and fee hunting. All ranches experienced drier conditions during our study than normal due to an ongoing drought. Three sites experienced wildfires during 2011 that removed all vegetation on ≥60% of those sites.

Data collection

We collected temperature and relative humidity data for two years (January 2012–2014) using a spatially nested (multi-scalar) sampling design along sampling transects. On each ranch, we established sampling points with an array at each point, and a minimum separation of 1 km between points. We established nine and 10 sample arrays on two ranches because of transect length constraints, and 11 sample arrays on the other four ranches (n = 63). Thus, transects were nine, 10, and 11 km in length, respectively. We used model DS1923 Hygrochron iButton temperature/humidity sensors (hereafter sensors) (Dallas Semiconductor, 2011) mounted inside the open cavity on the underside of screw-on type electric fence post insulators (Fig. 1) on a 1 m piece of steel rebar. The fence post insulators protected sensors against direct insolation, weather (e.g., hail, flood), and animal damage. Mounted sensors were aligned toward the west. Four sensors at each sample array collected hourly temperature and relative humidity data at two heights (10 and 60 cm) in two locations (random point and apparently suitable cover). The 10 cm height corresponded with approximate the height of a bobwhite nest, and a height of 60 cm was used as an estimate of ambient temperature. For the purposes of this study, we defined apparently suitable nesting cover as the center of the clump of native, warm-season bunch grass suitable for bobwhite nesting nearest to the matching random point (Rader et al., 2007). Although we recognize the ability of other structures (e.g., bunch grass–brush complexes) to serve as bobwhite nesting cover, we used a structure that is ubiquitous for bobwhite nesting across the species distribution. We moved nesting sensors before each nesting season if senescence and degradation of the previous year’s nesting cover had occurred. Random points were placed 25 m from the sampling transect every 1 km of transect travelled, alternating left or right. Points were checked to ensure they were no closer together than every 1 km. These transects representatively traversed every vegetation cover type on each ranch in proportion to their abundance. Although assigned randomly, these random points were generally located on bare soils; only 3% occurred in vegetation (i.e., woody plants, grasses) that may have provided some effect on in situ temperature and relative humidity. We collected data on relative habitat quality surrounding sensor locations as it related to ability of bobwhites to use a landscape. We used ocular estimation to assess percent cover of bare ground, herbaceous cover, woody vegetation, and available nesting structures. We ranked sites from 1 to 5, indicating habitat quality that ranged from poor (1) to excellent (5) based on accepted standards attested for the rolling plains of Texas and Oklahoma in the literature (Guthery et al., 2001; Jackson, 1969; Webb & Guthery, 1982). Sensors collected temperature and relative humidity hourly for two years (January 2012–2014). We downloaded data from sensors every four months using the Thermodata Viewer Software (Thermodata Corporation, Whitewater, WI, USA). During download, we removed sensors from mounts, inspected mounts for damage, replaced damaged mounts, and replaced foam inserts and cable ties. Raw data were later reformatted, and imported into a relational database.

Analyses

We compared temperature and relative humidity between cover types and sensor heights using repeated-measures, split-block analyses of variance (ANOVA) with one within-factor (time as month) and two between-factors (sensor height and sensor location). This method accounts for lack of independence when repeated observations are obtained from the same experimental units (Tzilkowski & Storm, 1993; von Ende, 2001; Zar, 2010). Environmental heterogeneity among ranches was controlled by blocking (Gotelli & Ellison, 2012; Zar, 2010). We examined differences in temperature and relative humidity during daily heat maxima (i.e., 1300–1500 h) in bobwhite nesting season (1 May–31 September) to compare potential for harmful heat exposure: (1) between random and nesting locations, (2) between ambient and nesting heights, and (3) across months. Critical assumptions were assessed prior to hypothesis testing (see next paragraph). All analyses were conducted in JMP Pro 11.0.0 (SAS Institute, Cary, NC, USA), and we considered p < 0.05 significant.

Plots of normality and residual variance indicated that temperature and relative humidity data were approximately normally distributed and homoscedastic. Mauchley’s test for sphericity (circularity; Mauchly, 1940) revealed a lack of sphericity in the temperature variance–covariance matrix (W = 0.38; Χ² = 27,697.11, p < 0.001) and in the relative humidity variance–covariance matrix (W = 0.70; Χ² = 10142.52, p < 0.001). We therefore used the more conservative Greenhouse–Geisser corrected F-test (Greenhouse & Geisser, 1959; Huynh & Feldt, 1970) to assess treatment effects for temperature (ε = 0.38) and relative humidity (ε = 0.70) data. We visually evaluated trends between treatment factors and habitat condition using both plots of mean temperature and relative humidity during daily heat maxima to assess the ability of nesting cover to reduce in situ heat below harmful levels (i.e., temperatures ≥40 °C).

Due to obvious differences in habitat condition within and across ranches due to differences in land use, precipitation, and other factors, we conducted an analysis to investigate the influence of habitat condition on thermal suitability of nesting substrates. We ranked relative habitat conditions for bobwhites at each sampling point on a scale from 1 to 5 based on ocular estimates of grass and forb availability, total percent ground cover, and diversity of vegetation from photo points taken each May and September. We conducted a one-way ANOVA to test for differences in temperature and relative humidity by habitat condition. A post-hoc test using a pairwise Student’s t-test comparison between all ranked habitat classes to assess whether differences in temperature in situ at all sensors were significant between different assessed habitat conditions.

Temperature

When examining the effect of between-subjects factors (location, height) on temperature by month, we found a significant main effect between both locations and heights (Table S1). There also was a significant interaction between location and height, which indicates that monthly temperatures at both nesting and random locations were strongly influenced by height of sensor, rather than placement in suitable nesting cover or not. When considering the within-subjects factor (time) on in situ temperatures, we found a significant main effect of month (Table S1). Nevertheless, the interactions of month and location, month and height, and month, location, and height were significant, indicating that time, alone, was not the driving influence on temperature.

Mean temperature was higher at nest height in random locations ($\overline{x}=36.81$, SE = 0.0683) than at nest height in nesting cover locations ($\overline{x}=35.99$, SE = 0.07). Mean temperatures at ambient height in nesting and random locations were essentially identical ($\overline{x}=32.78$, SE = 0.06, and $\overline{x}=32.99$, SE = 0.06, respectively). During the course of our study, 54.2% of temperature readings occurred between 0 and 40 °C ($\overline{x}=34.65$, SE = 0.03), a temperature range suitable for bobwhites.

Relative humidity

For between-subjects factors influencing mean relative humidity by month, we found a significant main effect between heights, but not locations (Table S2). The presence of a significant interaction between location and height indicated that differences between locations depended primarily on height. Although we found a significant main effort for the within-subjects factor (month) on relative humidity (Table S2), interactions among month and location, month and height, and month, height, and location were significant, indicating that time was not the driving force in mean relative humidity. Thus, while the influence of time was statistically significant, in situ relative humidity is largely dependent on height, and to a lesser extent, location.

Overall, mean relative humidity was higher at nest height in nesting cover locations ($\overline{x}=34.53$, SE = 0.11) than at nest height in random locations ($\overline{x}=33.35$, SE = 0.12). Similarly, ambient height in nesting locations exhibit higher mean relative humidity ($\overline{x}=36.22$, SE = 0.10) than ambient height in random locations ($\overline{x}=35.75$, SE = 0.10). During the course of our study, 80.0% of relative humidity readings occurred between 19% and 100% relative humidity ($\overline{x}=34.96$, SE = 0.05).

Monthly and daily trends

Monthly means plots indicate divergent mean temperature and relative humidity across months during the nesting season based primarily on height (Figs. 2 and 3). Trends for both temperature and relative humidity were similar, regardless of sensor height or location, although the magnitude of the differences was based primarily on height. Ambient heights exhibited similar trends, and experienced coolest mean temperatures and lowest mean relative humidity throughout the day. Differences in temperatures and relative humidity between height/location combinations were smallest at the beginning and end of the nesting season.

Nest height in nesting locations experienced cooler temperatures and higher relative humidity than nest height in random locations during daily heat maxima. Nevertheless, both nesting and random locations at nest height appeared to maintain harmful and/or lethal temperatures throughout much of the nesting season during daily heat maxima whether viewed by month (Figs. 2 and 3). Cumulatively, 257 days experienced temperatures ≥40 °C at one of our sensors during the 304 sampling days.

Habitat influences

Trends in temperature and relative humidity at nest height in nesting and random locations when viewed by habitat condition (Fig. 4) revealed that temperatures were below harmful levels (∼40 °C) longer in nesting cover than at sites with relatively better habitat conditions (Figs. 5 and 6). Student’s t-test revealed that temperature at all habitat classifications were significantly different from one another (p < 0.0001). Similarly, a Student’s t-test revealed that relative humidity was significantly different from one another at all habitat classifications at a p < 0.0001 level.