Date of Award

Summer 8-18-2017

Level of Access Assigned by Author

Campus-Only Thesis

Degree Name

Master of Science (MS)


Marine Biology


Yong Chen

Second Committee Member

Robert Steneck

Third Committee Member

Mark Wells


Habitat plays a critical role in regulating the spatiotemporal dynamics of populations and communities in the marine environment, and a good understanding of spatiotemporal changes in habitat can inform us of changes in the distribution and abundance of populations. Modeling habitats can provide accurate estimates of the distribution and quality of habitat with fine resolution, armed with this information managers can make assessments of the changes in habitat for species likely to be impacted by climate change. This research evaluates and quantifies the impacts of novel variable selection and weighting techniques on traditional habitat modeling and the relevant management implications. Two species of economic and ecological importance area used as case studies: the American Lobster (Homarus americanus) and the Northern Shrimp (Pandalus borealis) in the Gulf of Maine. Fisheries independent survey data is applied to a Habitat Suitability Index (HSI) model. Two separate decision points in the modeling process are compared. First, two techniques are evaluated for deciding which environmental variables should be included in the model, a traditional literature-based approach, and a model-based selection method. Second, three different weighting techniques are compared; equal weighting and two model-derived weighting types, the first being a machine learning process called Boosted Regression Tree (BRT) analysis and the second being an Akaike Information Criterion (AIC)- based approach. All possible model outcomes are compared through cross validation for each species to ascertain which decision route gives the best model outcomes. Spatiotemporal variability in the distribution of suitable habitat is then evaluated and quantified using the best model outcome for each species.