A Cross-Scale Ecological Genomics Framework for Predicting Biodiversity Resilience Under Climate-Induced Environmental Disturbance

Abstract

Climate-induced environmental changes accelerate biodiversity loss, by disrupting ecological functions, and decreasing the capacity of species in various ecosystems to adapt. Current methods of biodiversity prediction use mostly ecological or climatic pieces of evidence and fail to consider the combination of genomic adaptation to environmental heterogeneity, which leads to low predictive success. This paper proposes a Cross-Scale Ecological Genomics Framework to integrate ecological, climatic and genomic indicators and to predict biodiversity resilience under various disturbance conditions. The framework makes predictions based on an Extreme Gradient Boosting (XG Boost) regression model of the Biodiversity Resilience Index (BRI) which is a continuum based on multi-source environmental and genomic data. Experimental assessment shows better predictive performance with a of 0.947, RMSE 0.081, MAE 0.063 as well as MSE 0.0066 compared to traditional regression and ensemble learning models. Genomic diversity, temperature anomaly, habitat connectivity and precipitation variability were the four most important resilience predictors in the analyses of feature importance. The proposed framework provides an effective decision support tool for maintaining biodiversity, creating climate adaptation strategies and assessing the resilience of ecosystems in the future's changing environment.