Freshwater fish and mussel projections in the Northeastern United States at the HUC12 scale under different climate and land use scenarios

These data comprise the results of an analysis to 1. develop freshwater fish and mussel models using survey occurrence data, land use data, stream flow data, and stream temperature data within the Northeast United states, 2. predict the impacts of climate change on vulnerable groups of freshwater fish and mussel species, and 3. assess the potential for different management interventions to mitigate impacts of climate change. Freshwater fish and mussels survey records were used to fit species specific models that predict the probability of occurrence. The list of covariates is largely the same as the covariates released in the cross-referenced data set from Rogers et al., (2025); here, the covariate file shows the covariates used in this analysis that were not also used in the prior analysis. The species projection file has the results of the predicted probabilities of occurrence for each species of freshwater fish and mussel, at the HUC12 scale, for each of the seven scenarios: 1. baseline, 2. climate change, 3. climate change and dam removal, 4. climate change and natural riparian restoration, 5. climate change and riparian impervious removal, 6. climate change and watershed forest, and 7. climate change and a combination of numbers 3 through 6. The full methods for the fish models are described in the cited manuscript, Rogers et al., (2025); briefly, zero-inflated beta models (one for each fish species) were used to simultaneously model the probability of species occurrence and the predicted relative abundance of each fish species at the NHD Version 2 stream reach scale. The predicted probability of occurrence (from the binomial portion of the model) of each fish species at the stream reach level was averaged over the HUC12 scale to get the probability of occurrence of each of the 53 fish species by HUC12. The full methods for the mussel models are also described in Rogers et al., (In Prep); briefly we fit 12 logistic regression models (one for each mussel species) using the glm function in the stats package and the family of models set to ‘binomial’. The mussel models were fit using presence and absence data at the HUC12 scale and covariate data that was averaged across the HUC12. All models, analyses, and data visualizations were developed using R version 4.2.2 (2022-10-31 ucrt).