In 2014, the Marine Geospatial Ecology Lab (MGEL) of Duke University began work with the Northeast Regional Ocean Council (NROC), the NOAA National Centers for Coastal Ocean Science (NCCOS), the NOAA Northeast Fisheries Science Center (NEFSC) and Loyola University Chicago, as part of the Marine-life Data and Analysis Team (MDAT), to characterize and map marine life in the Northeast region in support of the Regional Ocean Plan. In 2015, the Mid-Atlantic Regional Council on the Ocean (MARCO) contracted with MDAT to build upon and expand this effort into the Mid-Atlantic planning area, and in support of the Mid-Atlantic Regional Ocean Plan. These research groups collaborated to produce “base layer” predictive model products with associated uncertainty products for cetacean species or species guilds and avian species, and three geospatial products for fish species. Periodic updates to these base layer models and data are produced by the individual institutions in the MDAT team based on schedules set by the funders of each modeling effort.

MDAT member NCCOS developed a comprehensive synthesis of models and data on marine and coastal birds as part of a 5-year BOEM funded project “Integrative Statistical Modeling and Predictive Mapping of Marine Bird Distributions and Abundance on the Atlantic Outer Continental Shelf.” In 2017, NCCOS updated the source data, covariates, and modeling methodology to produce new models, including additional species.

The NOAA report and the full set of model products for this project can be found here:

https://coastalscience.noaa.gov/project/statistical-modeling-marine-bird-distributions/

MDAT compiled the NCCOS long-term average density model results, with two products characterizing model uncertainty. The individual species maps represent the results of predictive modeling applied to data from the Northwest Atlantic Seabird Catalog (US Fish and Wildlife Service) and the Eastern Canada Seabirds at Sea database (Canadian Wildlife Service, Environment and Climate Change Canada). The modeling framework enabled predictions beginning 1-2km offshore and extending to the US EEZ boundary along the entire US Atlantic coast. As a result, model predictions are not available for nearshore (0-2km) areas, embayments, or estuaries, such as Long Island Sound.

Density model results are the long-term average relative abundance of individuals per unit area. It is not possible to infer absolute density because of how the survey data were collected and compiled, and how the models were generated.

Relative density model results are predicted to the full extent of the study area, and a hatched mask delineating areas with no survey effort in the dataset is provided. Mid-points of survey transect segments (~4 km in length) were gridded at a 10 x 10 km resolution, and hatched areas indicate grid cells with no segment mid-points (i.e. minimal or no survey effort). Model results in these hatched areas should be interpreted with caution. See Winship et al. (2018) Section 2 for more details.

The 90% Confidence Interval and the Coefficient of Variation are provided as two statistical measures of model uncertainty.

In 2014, the Marine Geospatial Ecology Lab (MGEL) of Duke University began work with the Northeast Regional Ocean Council (NROC), the NOAA National Centers for Coastal Ocean Science (NCCOS), the NOAA Northeast Fisheries Science Center (NEFSC) and Loyola University Chicago, as part of the Marine-life Data and Analysis Team (MDAT), to characterize and map marine life in the Northeast region in support of the Regional Ocean Plan. In 2015, the Mid-Atlantic Regional Council on the Ocean (MARCO) contracted with MDAT to build upon and expand this effort into the Mid-Atlantic planning area, and in support of the Mid-Atlantic Regional Ocean Plan. These research groups collaborated to produce “base layer” predictive model products with associated uncertainty products for cetacean species or species guilds and avian species, and three geospatial products for fish species. Periodic updates to these base layer models and data are produced by the individual institutions in the MDAT team based on schedules set by the funders of each modeling effort.

MDAT member NCCOS developed a comprehensive synthesis of models and data on marine and coastal birds as part of a 5-year BOEM funded project “Integrative Statistical Modeling and Predictive Mapping of Marine Bird Distributions and Abundance on the Atlantic Outer Continental Shelf.” In 2017, NCCOS updated the source data, covariates, and modeling methodology to produce new models, including additional species.

The NOAA report and the full set of model products for this project can be found here:

https://coastalscience.noaa.gov/project/statistical-modeling-marine-bird-distributions/

MDAT compiled the NCCOS long-term average density model results, with two products characterizing model uncertainty. The individual species maps represent the results of predictive modeling applied to data from the Northwest Atlantic Seabird Catalog (US Fish and Wildlife Service) and the Eastern Canada Seabirds at Sea database (Canadian Wildlife Service, Environment and Climate Change Canada). The modeling framework enabled predictions beginning 1-2km offshore and extending to the US EEZ boundary along the entire US Atlantic coast. As a result, model predictions are not available for nearshore (0-2km) areas, embayments, or estuaries, such as Long Island Sound.

Density model results are the long-term average relative abundance of individuals per unit area. It is not possible to infer absolute density because of how the survey data were collected and compiled, and how the models were generated.

Relative density model results are predicted to the full extent of the study area, and a hatched mask delineating areas with no survey effort in the dataset is provided. Mid-points of survey transect segments (~4 km in length) were gridded at a 10 x 10 km resolution, and hatched areas indicate grid cells with no segment mid-points (i.e. minimal or no survey effort). Model results in these hatched areas should be interpreted with caution. See Winship et al. (2018) Section 2 for more details.

The 90% Confidence Interval and the Coefficient of Variation are provided as two statistical measures of model uncertainty.