Habitat Suitability Modeling for the Houston Toad

Authors: Brandon C. Bowers, Danielle K. Walkup, Toby J. Hibbitts, Paul S. Crump, A. Michelle Lawing, Wade A. Ryberg

Our objectives were to update the Buzo (2008) habitat suitability model for the Houston Toad (Anaxyrus houstonensis) with current spatial data and habitat variables, quantify suitable habitat amount, develop a Species Distribution Model (SDM) to better understand environmental variable importance, and conduct connectivity modeling to prioritize areas for conservation and recovery actions. For the updated Buzo habitat suitability model, we constructed scripts to generate models that incorporate tree canopy cover, soils, and geology at varying weights and initially evaluated it with models for Bastrop County. We used four different model setups designed to estimate occurrence, in order of most restrictive to least restrictive: (1) Occurrence-informed Presence Prediction Models, (2) Evenly Weighted Presence Prediction Models, (3) Integrated Substrate Presence Prediction Models, and (4) Restoration and Reintroduction Potential Prediction Models. We then ran the four models for each of the remaining 12 counties and mapped the resulting outputs and quantified the suitable habitat amounts (Appendix 1).

We also constructed five ensemble species distribution models using different subsets of the Houston Toad occurrence data. We found that all five ensemble models generally identified consistent core areas for the Houston Toad, with variation among models occurring outside of those core areas. We found the proportion of deep sand (60-100 cm) in soil samples to be the most important variable in every model, generally having at least 6x the variable importance of the next most important variable. While the core areas contain generally acknowledged Houston Toad presence areas, there were some additional sites that stood out as having a higher predicted probability of presence (> 0.5) despite having no known occurrences in these areas. These areas include southern Bastrop County (south of the Colorado River) and areas north of known occurrences in Robertson and Leon counties. One important conclusion from our species distribution modeling results is that the variables included in the Buzo model (i.e., soil and forest cover) were also the most important variables in the species distribution models (in addition total March precipitation), despite the inclusion of other climate, vegetation, and elevation variables. This suggests the Buzo model is truly capturing the key elements of the Houston Toad’s habitat.

Finally, connectivity analysis showed little habitat connectivity among several or two core areas for threshold probability of presence values greater than 0.5 or 0.75, respectively. The isolation of these core areas suggests that population connectivity among them will have to be achieved through the captive breeding process for this species.

County Model Rasters

• Robertson
• Milam
• Waller
• Leon
• Colorado
• Harris
• Burleson
• Austin
• Brazos
• Lee
• Freestone
• Lavaca
• Bastrop

 Ensemble Models

• ESDM5 Uncertainty
• ESDM5 Probability
• ESDM4 Uncertainty
• ESDM4 Probability
• ESDM3 Uncertainty
• ESDM3 Probability
• ESDM2 Uncertainty
• ESDM2 Probability
• ESDM1 Uncertainty
• ESDM1 Probability

Suggested Citation