This dataset depicts the predicted probability of Sahara Mustard (Brassica tournefortii) occurrence in the Sonoran Desert Ecoregion, based on a MaxEnt model. This model was trained using occurrence points derived from multiple sources, and using prediction surfaces derived from a digital elevation model, distance to highways, precipitation, temperature, existing vegetation (LANDFIRE EVT v1.1), surficial geology, and soils (SSURGO and STATSGO).
Model AUC is 0.857 (indicates fair rate of prediction accuracy relative to the training data).
Due to potentially wide variability in the quality of training plots used for this prediction, and the coarse resolution of several predictors, caution should be exercised when interpreting this dataset. In particular, the training plots are unevenly sampled within the ecoregion, and often have spatial bias that may or may not be indicative of site fidelity of Sahara mustard. For example, highway roadsides are sampled heavily; Sahara mustard is known to occur in some of these locations, but other areas where Sahara mustard are known to occur are not adequately represented within the training data (Jim Weigand, BLM, personal communication, 2011).
This dataset has not been masked to remove areas where Sahara mustard is predicted to occur but is highly unlikely to occur (e.g., impervious surfaces).
The full results from MaxEnt modeling are available in an associated folder. This includes details about the relative contribution of the predictive surfaces.