For effective conservation, managers first must understand where species occur. A useful tool for understanding species ranges are species distribution models (SDMs), which make predictions by relating occurrences with environmental factors. SDMs assume that occurrence data provide an accurate representation of a species range; however, SDMs commonly use museum specimens or self-reported data, which may contain biases and not meet this assumption. Therefore, we conducted a study to assess the influence of data collection methods on SDM performance and determined whether SDM performance differs between taxa with different life history and physical traits in coastal habitats. We used the Dwarf Seahorse (Hippocampus zosterae) as a standard to compare because knowledge of its distribution is poorly understood, thus conservation of the species is stymied. Our research utilized varying sources of occurrence data along the Texas Coast, systematic and random sampling provided by the Coastal Fisheries Division of Texas Parks and Wildlife (TPWD) and opportunistic sampling provided by the Fishes of Texas Database (FoTx). We then compared seahorse SDMs to a suite of other species representing different traits from the seahorse that could influence SDM performance, including large-bodied, long-lived, vagile, and widely distributed species.
Results/Conclusions
SDMs of dwarf seahorses demonstrated that data source (TPWD vs. FoTx) affects model predictions. The AUC (indicators of model strength) values for TPWD = 0.763 ± 0.06 and (FoTx = 0.9273 ± 0.03. Although the FoTx data had a high AUC value, indicating strong predictive performance, these data were more haphazardly collected than TPWD, likely causing an overfit model rather than depicting the full seahorse distribution. Predictions based on TPWD data suggested a wider distribution and range of the Dwarf Seahorse compared to the FoTx output, which indicated a concern for the future of the species. Predictions differed between the Dwarf Seahorse SDMs and SDMs for other taxa with traits that different from the seahorse, suggesting the sensitivity of species traits along with accuracy of occurrence data. Overall, our research demonstrated an impact of data collection method and species traits on SDM performance. Thus, these factors should be considered when using SDMs. Improving effectiveness of SDM predictions improves conservation, management, and research.