Decoding the interactive effect of water quality-land use on benthic macroinvertebrate biodiversity in rivers with an interpretable machine learning framework

Abstract

The interactions between land use and water quality play critical roles in shaping benthic macroinvertebrate biodiversity in rivers. However, existing studies struggle to effectively identify nonlinear interactions between land use and water quality. This study integrates Random Forest and SHapley Additive exPlanations to create a robust framework that identifies the nonlinear interactions among variables. Applied to the Fu River basin, a primary tributary of Poyang Lake, China, our framework identified season-specific drivers: Ammonia Nitrogen and Shannon's diversity landscape index dominated community dynamics in the wet season, while Hydrogen ion concentration and Forest were key in the dry season. Interactive analyses revealed that during the wet season, total phosphorus (TP) and cropland cover formed the most influential pair, with synergistic effects (i.e., combined impact > sum of individual effects). Notably, cropland coverage modulated TP’s impact on benthic diversity: low cropland cover favored positive effects of TP, which diminished as TP concentrations increased to 0.04 mg/L, whereas high cropland cover triggered negative effects that intensified with rising TP and stabilized at 0.10 mg/L. During the dry season, conductivity (Cond) and forest cover emerged as the most impactful pair, also exhibiting synergistic effects. Forest cover modulated Cond’s influence on benthic diversity: under high forest cover, low Cond exerted positive effects that weakened with increasing Cond to 60 μs/cm; under low forest cover, moderate Cond induced increasingly negative effects that plateaued at 80 μs/cm. This study provides a robust approach to decipher context-dependent environmental interactions, offering valuable insights for river ecological conservation and adaptive management. • Decoded nonlinear interactions of water quality/land use on benthic biodiversity in rivers. • NH 3 -N & SHDI/pH & forest dominate the biodiversity during wet/dry season. • TP-Cropland/Conductivity-Forest synergy amplifies impacts during wet/dry season.