Spatial Identification and Dynamics of Groundwater-Dependent Ecosystems in the North China Plain: An Integrated Random Forest and Multi-Criteria Decision Analysis Approach

Abstract

Groundwater-dependent ecosystems (GDEs) play a crucial role in maintaining ecological stability and biodiversity, particularly in arid and semi-arid regions. However, in many areas, the location and extent of GDEs remain unidentified, and existing protective measures are insufficient. This study investigates terrestrial GDEs within the North China Plain (NCP), a region increasingly affected by groundwater depletion due to rapid urbanization and intensive agriculture. By integrating multi-source remote sensing datasets, Random Forest (RF) modeling, and GIS-based Multi-Criteria Decision Analysis (MCDA), we mapped the spatial distribution and assessed the temporal dynamics of GDEs across the region. Predictor variables included hydrological and vegetation indices, land cover, and topographic factors. The RF model was trained on georeferenced points and optimized through hyperparameter tuning. Spatiotemporal analysis revealed divergent trends in GDE distribution, with declines likely driven by groundwater stress or land degradation, while expansions likely attributed to improved groundwater recharge, increased ecological conservation efforts, and land use changes. Comparative analysis indicates that most of the GDEs identified in recent years have newly emerged, while a moderate proportion remained stable. Moderate-to high-probability GDE zones identified via MCDA were consistently classified as GDEs by the RF model, highlighting the robustness of this integrated framework. These findings offer critical insights into the evolving distribution of GDEs and provide a decision-support framework for ecological monitoring and sustainable groundwater management.

Keywords: Groundwater-dependent ecosystems; Terrestrial GDE; North China Plain; Random Forest; Multi-Criteria Decision Analysis; GDE dynamics