P-318 Zooming in on Environmental Predictors of an Overfished Species in the Gulf of Mexico Using GIS and Classification Trees
Our multivariate model retained 18 environmental predictors. The most influential were bathymetric followed by geographic and biological predictors; physical and anthropogenic categories ranked lowest. Notably, complexity in predictors was related to scale and location. For instance, the “Western Coastal” cluster required the most partitions and was the only cluster where both anthropogenic variables, oil/gas platform density and shrimping effort, were useful predictors. 21% of the observations, were classified by the first split as absence of RS attributed to habitat that was either too shallow (depth strata < 13m) or too deep (> 73m). The “Western Offshore” cluster had the highest probability of predicting presence of RS. In this group, a few shallower and deeper strata, corresponding to ranges of 15-18m and 27-55m, respectively, had very high probability of RS presence (0.86), accounting for about 1/3 of the presence cases. Overall, a good agreement was achieved between the classification tree and the hotspot approaches, as both identified the western region of the nGOM as the most important habitat for RS. In addition, the hotspot analysis suggested that a change in high density sites may have occurred from HSE to LSE years. We speculate that the difference in hotspot location may have resulted from the post-Katrina reduction of shrimping effort and recommend further investigation to refine the analysis and include the most recent years (2009-2010). Our findings may provide useful insights to prioritize EFH for red snapper in the nGOM.