Habitat Suitability

We modeled habitat suitability for each of the three WSRRI targets in different ways, based on expert input from each target co-production working group.

Greater Sage-Grouse

To map the spatial priorities of the Greater Sage-grouse, we trained a habitat suitability using the MAXENT algorithm (Phillips et al. 2010) empirically in a use-availability study design using habitat predictors and 93,474 observations of grouse in this landscape going back to the 1980s. The habitat predictors included climate variables (mean annual temperature, mean annual precipitation, climatic moisture deficit, growing degree days, etc.), fractional vegetation cover data (perennial grass, sagebrush cover, annual grass cover), landcover data (xeric/mesic grass/shrub, developed, agricultural classes, etc.), topography data (slope, topographic wetness index, heat load index, etc.) and the human footprint (powerlines, roads, railroads, wind turbines, urbanization, etc.). The model exhibited a strong relationship to sage grouse occurrence (area under the receiver-operator curve = 0.90), with high accuracy (0.83), sensitivity (0.83), and specificity (0.82).

Dry (Xeric) Shrubsteppe Ecosystem

To map habitat for the Xeric Ecosystem, we first computed an ecological integrity score largely following the Sagebrush Conservation Design (Doherty et al, 2022). Specifically, we fit curves to fractional cover datasets produced by TerrAdapt on invasive annual grass cover, perennial grass cover as well as the human footprint to calculate a q score, which is a measure of habitat quality (Figure 17). All data was computed on an annual basis for each 100m grid cell across the study area that were classified as either shrubland or grassland by TerrAdapt’s Landcover model.

Wet (Mesic) Shrubsteppe Ecosystem

To map Wet-Mesic habitat, we first created a layer of habitat quality and a layer of ‘wetland potential’. Habitat quality was driven from the human footprint and constrained to an area defined by either 1) the wetlands landcover (emergent or woody wetland) classes defined in TerrAdapt’s landcover model OR 2) low lying areas (defined by a height above nearest drainage less than 15m) that also were classified as mesic vegetation (mesic grass/shrub or forest). Wetland potential was defined as a linear function of the Normalized Difference Wetness Index (NDWI) calculated from Landsat imagery. NDWI is a measure of vegetation moisture, with values ranging from –1 to 1. Values above 0 are extremely moist environments likely to be inundated by water. Values in the range of –0.3 to 0 represent vegetation with ample moisture, indicating high water availability at or near the surface. We created a Wetland Potential Index ranging linearly from 0 (for NDWI values < -0.3) to 1 (for NDWI values >= 0). Only pixels with a landcover class that could become a wetland if restored (irrigated or non-irrigated row crops, pasture, fallow, forest, or mesic grass/shrub) that were also in low-lying places (height above nearest drainage < 15m) were allowed to have wetland potential > 0. In this way, the Wetland Potential Index reflects low lying areas of potentially restorable landcover that has access to surface moisture.