Species Distribution Modeling:
Predicting Snowy Plover Habitat, Great Salt Lake
Introduction:
Snowy plover (Charadrius alexandrinus) are small wading birds found throughout the Nearctic and Neotropics. In the western United States, plovers use the Great Salt Lake, UT as important breeding grounds. The shallow playas and wetlands surrounding the lake offer important habitat for protecting and raising offspring. Unfortunately, the IUCN recognizes snowy plovers as near threatened. Fluctuating lake levels and human development along the Wasatch Front have indirectly threatened plover habitat. Water is frequently diverted before reaching the GSL shrinking wetland and shore habitat. Further, many plover predators (such as skunk and foxes) are well adapted for urban environments, which have increasingly infringed upon plover habitat. Scientists have had difficulty detecting snowy plovers around the Bear River Refuge of the Great Salt Lake (GSL) without targeted monitoring efforts. Due to these low detection rates, monitoring may be improved by better predicting plover habitat use surrounding the GSL. Therefore, species distribution models were developed to examine snowy plover habitat preferences, to hopefully improve future monitoring efforts.
Methods:
Several species distribution models (SDMs) were developed using the 'sdm' package within R v3.3.3. More information about the R 'sdm' package is available at biogeoinformatics.org. Using snowy plover presence data, provided by John Cavitt at Weber State University, and several predictor variables, SDMs were developed using generalized linear regression (glm), random forest (rf), and maxent. Predictor variables were derived from slope, vegetation type, wetland presence, soil type, and land cover. Predictor variables were obtained from Utah's Automated Geographic Reference Center. All predictor variables were converted to raster data and resampled to 100 x 100 m grid cells. For each different modeling technique, 101 presence data points were combined with 500 background generated pseudo-absences. Several replications were used within each model, drawing 25% random subsamples each.
+ Plover observations
Predictor variables
Results:
Predicted Habitat
An ensemble SDM was developed by combining the glm, rf and maxent models using a weighted average of the TSS statistic. By combining the outputs of several techniques, predictive power draws from all three models.
Discussion:
Here an ensemble SDM was developed for Snowy plover, based on glm, rf, and maxent. The predictions may offer new information relevant to field biologists and conservation planners. Given the uncertainty of the future associated with climate change, coupled with continued development across the Wasatch Front, information provided by SDMs, like the example here, are important considerations for biodiversity conservation, and sustainable development.