NCCOS Assessment: Predicted habitat suitability for Leiopathes glaberrima in the US Gulf of Mexico, 2016-01-17 to 2017-09-30 (NCEI Accession 0276866)

Submission Package ID: BR43JR

Methods: Habitat suitability for Leiopathes glaberrima in the U.S. Gulf of Mexico was modeled using a Maximum Entropy (MaxEnt) modeling framework. Models used the values of spatially explicit environmental predictor variables describing conditions that may influence the distribution of Leiopathes glaberrima at locations where Leiopathes glaberrima was observed to estimate the relationships between these environmental conditions and the presence of Leiopathes glaberrima. These relationships were then used to predict and map habitat suitability for Leiopathes glaberrima across the U.S. Gulf of Mexico. Records of Leiopathes glaberrima occurrence in the U.S. Gulf of Mexico were obtained from the NOAA National Database for Deep-Sea Corals and Sponges (https://deepseacoraldata.noaa.gov/). Data records with identifiable positional errors were removed following quality control. Environmental predictor variables included measures of seafloor topography derived from depth data, seafloor substrate derived from surficial sediment survey data, and physical and biological oceanography derived from in situ data and remotely sensed data. A stepwise model selection procedure was used to select a best model that balanced predictive performance with model complexity. At each step a model was fit and predictive performance and complexity were calculated. The area under the receiver operating characteristic curve (AUC) was calculated using data withheld from model fitting in order to assess model performance. Model complexity was assessed using Akaike’s information criterion corrected for small sample size (AICc). The least important environmental predictor variable was identified and removed prior to the next step of model fitting using a jackknife test that determined which predictor variable resulted in the smallest reduction in AUC when omitted from the model. Following the stepwise procedure, the models were ranked in terms of AUC (highest AUC = best performing model) and AICc (lowest AICc = least complex model) and the best model was selected based on these rankings. To allow direct comparisons to predictions of habitat suitability for other deep-sea coral taxa, the relative habitat suitability (i.e., the logistic output from MaxEnt) was reclassified into a series of habitat suitability classes using breakpoints calculated using specific ratios of the cost of false positive errors versus the cost of false negative errors. Essentially, the higher the habitat suitability class the greater the penalty for overpredicting the area considered to be suitable habitat. An additional robust very high habitat suitability class was assigned to model grid cells predicted to be in the highest class of habitat suitability for each of the replicate model runs of the best model. For additional details, see Etnoyer et al. (2018).

File Information:Total File Size: 34.2 MB total, 15 files in 1 folder (unzipped), 18.3 MB (zipped)
Data File Format(s): Geotiff .TIF (and ancillary files .TFW, .CPG, .DBF)
Data File Compression: zip
Data File Resolution: 370.65 x 370.65 meters #GIS Projection: WGS 84 UTM Zone 15N
Data Files: Leiopathes_Classified_Predicted_Habitat_Suitability Leiopathes_Classified_Predicted_Habitat_Suitability_Variability Leiopathes_Predicted_Habitat_Suitability Leiopathes_Robust_Very_High_Predicted_Habitat_Suitability Documentation Files: Leiopathes_model_output.JPG DataDocumentation.PDF