To meet its statutory requirements under the Endangered Species Act and Marine Mammal Protection Act, the Navy requires analytical tools with which to predict the distribution of marine mammals at spatial and temporal scales relevant to training exercises. The Navy must be able to conduct such exercises to maintain readiness, and the best way to avoid potentially adverse effects on these protected species is to conduct exercises in times and areas where the probability of encountering marine mammals is low.

The objectives of this project were to: (1) develop and test the robustness of various models of marine mammal habitat suitability, as predicted by physical conditions of the marine environment; (2) design a hierarchical framework for analyzing patterns of marine mammal habitat suitability across seasonal time frames; (3) assemble a Spatial Decision Support System (SDSS) that enables Navy users to examine and analyze model outputs and original input data across multiple time scales; and (4) test how well predictive marine mammal habitat and density models perform at spatial scales relevant to Navy training exercises. The overall intent of this study was to provide the Navy with tools that would allow quantitative predictions of the presence of marine mammals in particular regions of interest.

Spatially explicit statistical techniques were used to determine how features of the physical habitat influence the distribution of marine mammals. A subset of both traditional and novel statistical model approaches were evaluated to determine which are most robust in understanding marine mammal habitats under different oceanographic conditions and data limitations. The tests were structured by the spatial and temporal constraints of the observational and environmental data, taxonomic group of marine mammals evaluated, and geographic regions assessed. These approaches provided a flexible framework for spatio-temporal data, which is highly appropriate for modeling multi-scale and multi-temporal ecological processes.

Data on the distribution of marine mammals generated in dedicated surveys contained in the online OBIS-SEAMAP marine data archive were used to develop predictive habitat models for guilds of marine mammals. An online, flexible SDSS was developed to deliver model outputs. The SDSS is a browser-based, interactive mapping application that enables the user to view the model results, together with the original survey effort and marine mammal observations. During model development a suite of multivariate statistical models (CART, GLM, GAM and Bayesian approaches) was fitted to observations from at-sea surveys, together with remotely sensed environmental data (bathymetry, sea-surface temperature, chlorophyll), as well as derived variables such as slope, temperature fronts, and chlorophyll aggregations.

In total, 33 models were generated, representing 16 cetacean guilds, using environmental data from the Jet Propulsion Laboratory physical oceanographic data archive. The model results are presented as predictive maps for the likelihood of encounter with marine mammal guilds, together with estimates of the associated standard errors. A supplemental comparison of habitat models and Navy at-sea operating area density estimates (NODE) to observation data revealed insufficient data in two Navy training areas for all but two species.

In general, the NODE models performed better for data-rich species (bottlenose [genus Tursiops] and spotted [genus Stenella] dolphins) than rare species (pilot whales and Risso’s dolphins). Outside the spring and summer months, few observations were available and the comparisons in these other seasons should be treated with caution. In spring and summer, the NODE model outputs demonstrated a general positive relationship with the relative density estimates, particularly with the aerial survey data sets that contained a larger number of observations.

To access end-user products developed through this research, please visit the Marine Mammal Conservation section on the RC Tools and Training page.

Utilizing the SDSS enables users to delineate regions of interest and extract summary statistical outputs, such as histograms and model statistics, from these areas. The SDSS system also incorporates model results from a related Strategic Environmental Research and Development Program (SERDP) project: Predictive Modeling of Marine Mammal Density from Existing Survey Data and Model Validation Using Upcoming Surveys (RC-1391), as well as estimates of marine mammal density generated by Geomarine, Inc.  The SDSS enables viewing of original survey effort, marine mammal observations, oceanographic imagery, and habitat model results. Model results yield predictive maps for the likelihood of encounter as well as the associated standard error. Regions of interest can further be delineated within the SDSS for extracting summary statistical outputs, such as histograms and model statistics. The SDSS system enables users to view and query observation data and model results from this project as well as related SERDP projects and baseline geographic data. This work represents an important step towards understanding marine mammal habitat use with respect to regions utilized by the U. S. Navy.