Using explainable machine learning methods to evaluate vulnerability and restoration potential of ecosystem state transitions

Ecosystem state transitions can be ecologically devastating or be a restoration success. State transitions are common within aquatic systems worldwide, especially considering human-mediated changes to land use and water use. We created a transferable conceptual framework to enable multi-scale assessments of state resilience and early warnings of state transitions that can inform strategic restorations and avoid ecosystem collapse. The conceptual framework integrated machine learning predictions with ecosystem state concepts, such as state classification, gradients of vulnerability or recovery potential leading to state transitions, and investigation of possible environmental drivers. As an example, we generated prediction probabilities of submersed aquatic vegetation (SAV) presence at nearly 10,000 sites within the Upper Mississippi River, USA. Then, we used an interpretability method for explanation of model predictions to gain insights on possible environmental drivers and threshold or linear responses of SAV presence and absence. The results are also presented via an online, interactive dashboard where our habitat suitability model outputs and prediction explanations from many spatial scales (4 m-400 km of river reach) can inform research and restoration planning. This article is protected by copyright. All rights reserved.