Richardson’s ground squirrel (Urocitellus richardsonii, Sabine 1822) is a widespread burrowing mammal on the Northern Great Plains. This species is a key prey item for a variety of predators, and its burrows provide important habitat for other wildlife; however, Richardson’s ground squirrel is also considered an economically damaging agricultural pest. Despite the ecological importance of Richardson’s ground squirrels and their status as pests, there are gaps in our knowledge about large-scale habitat associations for this species. In 2011 and 2012, we conducted 1840 road-side surveys in a 130,000-km² area of southern Saskatchewan, Canada to understand which habitat features are associated with Richardson’s Ground Squirrel occurrence. Ground squirrels were observed during only 8% of the surveys. Probability of ground squirrel occurrence was highest in areas with moderate amounts of grassland (approximately 30%), areas that were developed by humans (>30%), and had a high proportion of clay loam soils, presumably for burrowing. Our study highlights the importance of heterogeneous landscapes and that areas disturbed by humans may provide suitable vegetation structure for ground squirrels. This information can help to identify important habitat for species that rely on Richardson’s ground squirrels and identify areas where they could come into conflict with agriculture.

We undertook surveys in an area covering approximately 130,000 km² (Fig. 1). Richardson’s ground squirrels regularly occur north of our study area (Michener and Koeppl 1985); however, we could not conduct surveys further north given the limited time available for surveys and so our inferences are limited to the southern part of Richardson’s ground squirrel range in Saskatchewan. We divided the area into 25 x 25-km quadrants and placed a 16-km transect along a road near the centre of every second quadrant. Transect orientation alternated between running north-south and east-west among the grid rows. Surveys were conducted at 1.6-km intervals (10 survey points per transect) along transects. This scale was selected to match the legal land division system such that counts took place in the middle of every second quarter section along the transect. Ground squirrel counts were conducted in a 200 x 200-m plot directly adjacent to the road with the observer in the middle of the outside edge.

Surveys

In 2011, 49 transects were surveyed in the eastern half of the study area between 4 June - 20 June, and 43 transects were surveyed in the western half of the study area between 26 July - 10 August (940 points). In 2012, all north-south running transects from the 2011 sites (n=48) were re-surveyed twice: once between 3 - 29 June, and again between 7 - 24 August (960 points). Dates were chosen to avoid periods of population fluctuation such as emergence of juveniles (Michener 1992). However, these dates exclude most adult females and many adult males who would have entered hibernation at this time. In total, 1840 points surveyed, distributed equally between 2011 (N=920) and 2012 (N=920; Fig. 1) were used in our analysis.

The number of ground squirrels was determined at each survey plot using a call-playback method (Downey et al. 1994). The area was scanned with binoculars for a minimum of 30 seconds and continued until all visible squirrels were counted or a full scan of the area was complete (Fagerstone 1983). A wildlife caller (FoxPro Spitfire, sound pressure level 102-106 dB) programmed with a Richardson’s ground squirrel alarm call (repeated whistles; provided by Dr. James Hare, Department of Biological Sciences, University of Manitoba) was played for the duration of the count, causing squirrels to stand up/call back (Hare and Atkins 2001). Surveys were performed when there was no precipitation, wind was < 40 km/h, and temperature was < 30°C (Michener and Koeppl 1985; Downey et al. 1994). In addition, all surveys were conducted at least 75 mins after sunrise or before sunset to coincide with peak periods of activity. Study methods were approved by the University of Regina’s, President’s Committee on Animal Care in accordance with the guidelines of the Canadian Council on Animal Care.

Habitat characterization

                We chose three scales, 100-, 200-, and 300-m radii around survey plot centres, to assess the habitat within and around the survey plots. While the 100-m scale likely represents habitat exactly where ground squirrels were observed, the 300-m scale represents the larger context within which a colony resides. We used currently available Geographic Information System layers to characterize the habitat within, and landscape surrounding, each survey site. We used Agriculture and Agri-Food Canada’s (AAFC) Land Cover for Agricultural Regions of Canada, circa 2000 (http://open.canada.ca/data/en/dataset/16d2f828-96bb-468d-9b7d-1307c81e17b8, Accessed January 2016). The AAFC layer covers the entire agricultural region in Saskatchewan at 30-m resolution and is currently the best available and most current land cover layer available. At all three scales, we calculated the proportion of cropland, combining categories of cultivated agricultural land, annual cropland, and perennial cropland and pasture. We also calculated the proportions of grassland (which includes both native and tame grasslands) and trees and shrubs; the latter combining categories of shrubland, forest/trees, coniferous forest, deciduous forest, and mixed forest. The AAFC layer did not accurately depict other non-cropland areas that are disturbed by humans such as towns, cities, farmyards, or well pads. Instead we used Saskatchewan’s Southern Digital Landcover (SDLC) raster-based mapping product. Saskatchewan’s SDLC also occurs at a 30-m resolution, but was constructed in 1994. While some areas around urban centres have likely expanded since 1994, the SDLC represents the best available information to our knowledge on footprints representing areas that are not cropland or roads but are still impacted by humans. At all three scales, we calculated the proportion of area that was developed by humans using this SDLC layer using a GIS.  

To assess soil properties at survey points, we used the Soil Landscapes of Canada (version 3.2; http://sis.agr.gc.ca/cansis/nsdb/slc/v3.2/index.html, Accessed January 2016) and Saskatchewan’s Detailed Soil Survey layers (both layers were converted to a 30-m resolution raster). We calculated the percentage of soil parent orders: vertisolic, solonetzic, regosolic, gleysolic, and chernozemic from the Soil Landscapes of Canada database. We used the Detailed Soil Survey of Saskatchewan to assess the percentage of undifferentiated, sandy loam, silty loam, silty clay loam, silty clay, sandy, loamy sand, loam, heavy clay, gravelly sandy loam, gravelly loam sand, gravelly loam, fine sand, fine loam, clay loam, and clay. Using the same Detailed Soil Survey of Saskatchewan, we also calculated the proportion of soil that had: no stones (class 0), was slightly stony (class 1), moderately stony (class 2), very stony (class 3), and exceedingly stony (class 4).  We did not assess soil properties at multiple scales, because in all cases the soil property at the centre of the survey point was the same at all three scales.

Yearly variation in habitat characteristics

                None of the explanatory variables above varied by year because the GIS datasets we used were static. However, 2011 was an extremely wet year in parts of Saskatchewan with significant local flooding that could have impacted the presence or absence of ground squirrels along our survey routes. Therefore, we used the yearly AAFC cropland inventory layers from 2011 and 2012 (30-m resolution) and calculated the proportion of the 300-m radius around the survey point that was classified as “Too Wet to be Seeded” (hereafter, wetness). Wetness quantifies the extent of areas that have received above-average spring moisture and result in the inability of agricultural producers to seed an area for crop production (Agriculture and Agri-Food Canada 2015). In 2011, 2,108,640.5 ha of the prairie ecoregion was considered too wet to be seeded, whereas in 2012 only 141,064 ha were considered too wet to be seeded.