Evaluating effects of data quality and variable weighting on habitat suitability modelling

Abstract

Habitat modelling is important in the conservation and management of fishes and can be sensitive to data inputs and model configuration. Survey data used in Habitat Suitability Index (HSI) models may undergo changing sampling protocols over time, and these inconsistencies may impact results. Additionally, the various habitat variables included in HSI models are typically given equal weights, even though some variables may have greater influence over distribution than others. The Long River Survey, part of the Hudson River Biological Monitoring Program, in the Hudson River Estuary (HRE), has undergone considerable protocol changes, and was calibrated to address these issues in 2023. This survey and region are an excellent case study to compare two approaches in constructing HSI models: using calibrated versus uncalibrated abundance data and weighting all environmental variables equally or using a model-based weighting method. The results of this study suggest that using calibrated abundance data with unweighted habitat variables provide the most robust estimates for bay anchovy suitable spawning habitat in the HRE, which indicates that in cases when sampling has not been consistent over time, using calibrated abundance data in habitat suitability modelling may lead to improved models. Some model configurations were unable to identify a significant trend in suitable habitat over time and overestimated habitat quality illustrating the importance of carefully considering data inputs and model configuration when building habitat models to properly quantify suitable habitat and contribute to ecosystem-based fisheries management in the wake of climate change.