Research Document 2020/004

Development of a Species Distribution Modelling Framework and its Application to Twelve Species on Canada’s Pacific Coast

By Nephin, J., Gregr, E.J., St. Germain, C., Fields, C., and Finney, J.L

Abstract

Species distribution models (SDMs) have become increasingly valuable as a tool for the management and conservation of marine resources and places. However, despite their utility and prevalence, critical aspects of SDM development and uncertainty assessment are routinely overlooked. Best practices are therefore warranted and their application is increasingly required. This framework, implemented with purpose-built scripts written in the R statistical programming language, has been prepared as both a tool and a set of guidelines and methods for the development of consistent, interpretable, and defensible SDMs to support DFO’s contribution to Canada's ocean policies. SDMs were built for twelve benthic species to illustrate the application of the framework, and guide emergency oil spill response planning as part of the Regional Response Plan for the Northern Shelf Bioregion. Three model building methods of increasing complexity were applied using a suite of the best available environmental predictors. Knowledge-based envelope models were produced for all species, and emphasized for those found to be data deficient. These envelope models provided guidance for the development of the subsequent data-driven models, and can be used to help evaluate uncertainty in model predictions. They also provide an avenue for engaging species experts in the process. Data-driven generalized linear (GLM) and boosted regression tree (BRT) models were generated, along with a corresponding ensemble model, for the eight species found to have adequate observational data. The highest quality predictions were generated for those species for which sample sizes were high, and observational data were well-distributed across the study area. All ensemble models performed moderately well (AUC > 0.7) when evaluated using a spatial block cross-validation approach. Thirteen recommendations were conceived as part of the development of the framework and its application. They provide guidance on the application of SDM methods related to data selection and preparation, model development and evaluation, and highlight ways to improve modelling outcomes by applying best-practices where possible. Future development of SDMs for additional species in Pacific Canada will be greatly facilitated by the set of common predictors, methods, and evaluation tools assembled here.

Research Document 2020/004 (PDF, 9,767 KB)

Accessibility Notice

This document is available in PDF format. If the document is not accessible to you, please contact the Secretariat to obtain another appropriate format, such as regular print, large print, Braille or audio version.