Skip to main content
Dryad logo

Fine-scale habitat selection limits trade-offs between foraging and temperature in a grassland bird


Londe, David et al. (2021), Fine-scale habitat selection limits trade-offs between foraging and temperature in a grassland bird, Dryad, Dataset,


Many species are frequently faced with the decision about how to balance the use of thermal refuge against access to food resources. We evaluated the habitat use of female greater prairie-chickens (Tympanuchus cupido) to assess the potential for trade-offs between thermal conditions and food resources during the habitat selection process. Our objectives were to 1) compare near-ground temperatures, invertebrate availability, and vegetation characteristics at sites used by greater prairie-chickens to conditions at random landscape locations in various time since fire patches and 2) assess changes in conditions at used sites throughout the day to determine if selection for resources changes relative to ambient conditions resulting in trade-offs between foraging sites and thermal refuge. We found that greater prairie-chickens primarily used patches 0-12 months post fire that had relatively high abundances and biomasses of invertebrates compared to the landscape. Greater prairie-chickens further modified their selection at relatively fine spatial scales within these food-rich patches to select for areas with cooler temperatures during the hottest part of the day. The use of thermal refuge did not appear to influence access to food resources as invertebrate abundance and biomass at used sites were consistent throughout the day. Our results show that food resources and thermal cover influences habitat selection for greater prairie-chickens, but there was little evidence for trade-offs during the habitat selection process. Consideration of spatial and temporal scales is critical for evaluating trade-offs in habitat selection for animals and this research provides insights into the decision-making process by prairie-chickens.


Thermal, Invertebrate and Vegetation Sampling

Prairie-chicken locations- On every day with stable weather conditions (sunny and no rain), we randomly selected either a brooding or non-brooding prairie-chicken to evaluate habitat use. For the selected prairie-chicken, we selected four GPS locations from a recent day with similar weather conditions (Figure 1). Locations were selected such that two GPS locations occurred during the morning active period (6:30 – 10:30) and two GPS locations were recorded during the afternoon refuge period (12:30 – 16:30). To account for the fact that changing weather conditions throughout the day could influence invertebrate activity and thermal conditions at a site, we divided our thermal and invertebrate data collection efforts into two sampling periods that matched the two activity periods (Figure 1). We then randomly assigned one telemetry location from the active period and one telemetry location from the refuge period to be sampled during the morning sample period (6:30-10:30). The remaining two points from that day (one from the active period and one from the refuge period) were assigned to be sampled during the afternoon sample period (12:30-16:30; Figure 1). The actual sampling of vegetation, thermal environment, and food resources (insects) occurred within 3-4 days of the day the telemetry locations were recorded (mean =3.74 days). This sampling methodology allowed us to assess thermal conditions and invertebrate resources under ambient weather conditions similar to when the individual was actually at the location as well as the period of the day that the prairie-chicken was not present at a location.  

To account for error associated with the transmitters (up to 18 meters), we established sampling arrays consisting of nine sample points around each of the used GPS locations where vegetation and thermal measurements would occur. The use of multiple sampling points per site allowed us to characterize the average vegetation and thermal conditions available at a given location. We delineated the area around each telemetry location into four quadrants using two perpendicular 36-meter transects (2x the error of the transmitters) that intersected at the telemetry location and were aligned with the cardinal directions. In addition to a sample point at the telemetry location (center of the plot), we established two sample points in each of the four quadrants that were located at a random distance (1-6 meters) and cardinal direction from the center of the quadrant for a total of 9 sample points per telemetry point. All subsequent vegetation and temperature measurements occurred at these nine sample points for each plot. 

We estimated thermal conditions at prairie-chicken locations using black-bulb temperature probes. Black-bulb temperature incorporates information about ambient temperature, solar radiation, wind, and convective heating into a single temperature measurement providing a more accurate approximation of what an organism experiences compared to ambient conditions alone (Bakken 1992). The black-bulb temperature probes consisted of a temperature sensor placed in the center of a 101.6mm diameter, 20-gauge steel sphere painted matte black that was connected to a HOBO data logger (U12-008, Onset Corporation, Bourn, Massachusetts, USA; Guthery et al. 2005). We deployed black-bulb sensors at all nine sample locations at both telemetry locations (one active, one refuge location) assigned to a sampling period and programed the data loggers to record a temperature reading every minute for the duration of the four-hour sample period. Black-bulb sensors were placed on the ground or the surface of the litter layer at the base of any vegetation structure directly above the sample point. If no plants were above the sample point the black-bulb was placed in the open.

While the thermal sensors were recording at each prairie-chicken location, we used sweep-net transects to sample the invertebrate community along the two perpendicular transects used to delineate the plots. While sweep-net samples may underestimate certain orders of insects, we choose this method as we felt it provided an adequate index of the overall availability of invertebrates for prairie-chickens, particularly insect orders that are known to be important for prairie-chickens (Orthoptera and Lepidoptera). To account for changing invertebrate activity throughout the day, we timed our sweep net samples based on the timestamps from the used prairie-chicken telemetry locations. For example, at locations where the sampling period matched the telemetry location time (ex. an active location sampled during the morning sampling period), we conducted sweep net samples within 30 minutes of the approximate time the prairie-chicken had previously been at the location. At locations where the prairie-chicken activity period did not match the sampling period (ex. an afternoon refuge location sampled in the morning sampling period), we conducted the sweep-net samples two hours before or after the other sweep-net sample for a given sampling period. The two-hour delay between samples ensured sweep net transects corresponded approximately to the time the prairie-chicken was at the location and to ensure the samples were evenly spaced throughout the four-hour sampling period. To perform sweep-net samples, we walked each transect at a moderate pace, taking one sweep of the sweep-net with each step (Doxon et al. 2011). We transferred sweep-net samples to plastic bags and stored them in a freezer until they could be sorted. We sorted invertebrate specimens to the level of order and size class (eg., 0-5 mm, 6-10 mm, etc.) for each sample and we recorded the total number of insects and mass (grams) to estimate abundance and biomass for each order and size class combination at a location.

After the thermal sensors had been removed from a site, we collected vegetation data at each of the nine sample points. We centered a 0.5 m2 vegetation sampling frame over each point where the black-bulb temperature sensor was located, and in each frame, we estimated the percent cover of grass, sericea lespedeza (an introduced invasive forb), forbs (excluding sericea), shrub, litter, and bare ground using standard Daubenmire cover classes (Daubenmire 1959). Additionally, we recorded the height of the tallest vegetation in the frame (cm), litter depth (cm, taken 10 cm west of the black-bulb location) and visual obstruction using a Nudd’s profile board (Nudds 1977).

Random Landscape Locations

To assess vegetation, thermal conditions and invertebrate availability across the landscape, we collected the same suite of data collected at prairie-chicken locations at random sampling locations that were stratified over the three time since fire categories (0-12 months post fire, 13-24 months post fire, and >24 month post fire). Each random site was composed of a cluster of four locations. We generated clusters by generating a single starting location followed by three additional points that were a random direction and distance from the previous location, with the restriction that random points must be >36 meters apart and within the same time since fire patch as the first random point. Distances between sample locations were based on distributions of observed distances between sequential prairie-chicken telemetry locations. Within each cluster of 4 random locations, we randomly assigned two of the locations to be sampled during the morning sampling period and two locations were assigned to the afternoon sampling period. Sampling at each of the random locations followed the same sampling designs as that used at prairie-chicken locations.

Usage Notes

The data is organized into three csv files with a file for the vegetation data, thermal, data, and invertebrate data.

All three files have several columns in common that allow users to relate vegetation, thermal and invert data across files.

All three files have a location.type column indicating if the location corresponds to a brooding hen (Brood), non-brooding hen (Adult), or random location (random). The column Location.type2 divides these point types further into prairie-chicken locations recorded during the morning active period (Brood.AM, Adult.AM), prairie-chicken locations recorded in the afternoon refuge period (Brood.PM, Adult.PM) or a random location in each of the three time since fire classes (0=random location in patches 0-12 months post fire, 1=random location in patches 13-24 months post-fire, 2=random locations >24 months post fire). 

Sample.period indicates if the data was recorded in the morning sampling period (AM) or afternoon sample period (PM). 

The columns band.random and indicate different sampling clusters (sets of four telemetry locations or four random locations). Each band.random/ combination will have four values in the chicken.time column that indicates which sample location from the sample cluster the data belongs to. For prairie-chicken locations these values in the chicken.time column also indicate the approximate time the GPS location was recorded, but for random locations these values are randomly assigned.  

For the vegetation data unique combinations for quadrant and port indicates unique sample points at a sample location. For each location there should be 5 values in quadrant (center, NW, NE, SW, SE), that have a corresponding value for port that indicates one of two points in a quadrant. Vegetation data is found in columns K:T.  Grasslike=% cover of grass, Forb=%cover of forbs, litter=% cover of Litter, bare=% cover of bare ground, and shrub=% cover of shrubs. All cover classes are shown as the midpoint of Daubenmire cover classes (0-5% cover, 6-25% cover, 26-50% cover, 51-75% cover, 76-95% cover, 96-100% cover). Litter.Depth is measured in centimeters 5 cm west of the sample location. Veg.Height.Frame is the height of the tallest vegetation it the point. Nudds.average is the average of the four Nudds board readings at a site, and is measured in decimeters.

In the thermal file each band.random/ combination will have an hourly black bulb reading (hourly mean;degrees celcius). The hourly mean is the average of all nine blackbulb sensors at a location for each hour. Thermal data was only recorded for four hours during either the morning (AM) or afternoon (PM) sample period. A corresponding air temperature (degrees Celcius) and solar radiation (watts/meter-squared) value were recorded by an onsite mesonet weather station. 

The invertebrate datasheet shows the invertebrates collected at each location. Similar to other files band.random/ combinations indicate individual sample locations. Within each sample location there are were two transects (Transect =E-W or N-S). For each transect, we identified invertebrates by order and size class, and recorded count (Count) and biomass (Invert Weight [grams]) of each order/size class combination. Size.class is broken into 5mm size categories (0-5mm, 6-10 mm, 11-15 mm, 16-20 mm, 21-25mm >25 mm). Categories in size class is prefixed with sc. to prevent excel from converting to date format, this can be removed when working in R or other programs. Bug.Sample.Time indicates the actual time of day the samples were collected.