Optimal weighting in species habitat modeling: a case study from Ommastrephes bartramii in the Northwest Pacific Ocean

Abstract

The weighting of environmental variables in habitat modelling is important, especially for species with a poorly understood distribution. Traditional weighting schemes, such as arithmetic or geometric mean, often cause “gradient” habitat distribution patterns. We develop a new methodology that determines optimal variable weighting via a structured sensitivity analysis approach. This method considers the full spectrum of weighting combinations and uses multiple model selection criteria to select the best fit. We use a Northwest Pacific neon flying squid (Ommastrephes bartramii) fishery dataset (1998–2012) to compare our best performance habitat suitability index (BEST-HIS) with the traditional fixed methods, as well as to the more recent machine learning approach: boosted regression tree. Approaches were evaluated based on differences in habitat metrics, such as continuity, magnitude, and ratio of estimated unfavourable/favourable habitat. The BEST-HSI model generally outperformed the other three methods, though habitat metrics notably differed depending on weighting schemes used. The BEST-HSI approach is an efficient exploratory tool to investigate empirical relationships between organism presence and the environment, particularly for species with little known life history or migration information.