Modeling invasions beyond equilibrium: Integrating dispersal and accessibility, physiological constraints, and ecological context to improve invasion-risk predictions

Aquatic invasive species (AIS) are commonly modeled using species distribution models (SDMs) built on climate and remotely sensed data, yet these approaches often fail to represent the dispersal and establishment processes that shape invasion risk. This limitation is particularly important in freshwater systems, where human-mediated connectivity, propagule pressure, and local environmental constraints strongly influence invasion outcomes. Here, we develop and evaluate a transferable framework for generating process-informed spatial predictors representing four key components of invasion risk: environmental suitability, physiological constraints, ecological context, and dispersal/accessibility. Using zebra mussel ( Dreissena polymorpha ) invasion across 2,203 lakes in Minnesota as a case study, we evaluated how the progressive addition of process-informed predictors altered model performance and interpretation. Models based solely on environmental suitability predictors achieved moderate performance (mean TSS = 0.76). Incorporating physiological constraint and ecological context predictors produced incremental improvements, whereas dispersal/accessibility predictors generated the largest gains in predictive performance (mean TSS = 0.94). The addition of process-informed predictors also substantially reorganized predictor importance. Variables that dominated environmental-only models declined in importance as explicit representations of dispersal, establishment constraints, and ecological context were introduced, indicating that commonly used environmental predictors can partially function as proxies for unmeasured invasion processes. When all predictor groups were included, boater movement, distance to source, and calcium concentration emerged among the most influential predictors. These results demonstrate that invasion risk is structured not only by environmental suitability but also by the processes governing access to suitable habitat and successful establishment following introduction. More broadly, the framework provides a reproducible approach for transforming ecological mechanisms into spatial predictors while offering a practical means of evaluating which ecological processes contribute most strongly to invasion-risk prediction. By extending correlative models with process-informed predictors, this approach improves both predictive performance and ecological interpretability and provides a general pathway toward process-explicit invasion-risk modeling.