Storage-release dynamics of microplastics during rainfall events in conduit-fissure coupled karst aquifers

Rainfall-driven surface microplastics (MPs) infiltrating karst aquifers threaten groundwater safety and ecosystem stability, and this process is difficult to trace. This study employed high-frequency monitoring of hydrometeorological, chemical, and MP variations in a conduit-fissure coupled aquifer in southern China to identify transport via correlation, gray relational analysis, and machine learning. During rainfall events, PET (66%) and PE (32%) dominated MPs, originating from livestock and agricultural sources. Conduit-fissure exchange produced a staged storage-release of MPs: (I) pre-deposited MPs were remobilized by pre-event conduit flow, peaking at 1200 N/m³ before the discharge peak; (II) post-event conduit flow preferentially transported newly infiltrated MPs toward the spring, colored PE and PET fibers (650 N/m³) under dilution; (III) slow fissure drainage induced a secondary release of stored PE and PET fibers, reducing abundance (average 528 N/m³) and diversifying sizes and colors; and (IV) a second rainfall increased abundance to 600 N/m³ . Across stages, abundance and trait indices were weakly correlated with each other, but showed high temporal synchrony with discharge and hydrochemistry (ORP, Ca²⁺, etc.). Machine learning models predicted storage-release dynamics of MP properties (R² = 0.78-0.95). SHAP indicated that ORP and Mg²⁺ ranked among the most influential predictors for MP abundance, size, and color, while discharge and TDS showed high importance for shape and polymer diversity, respectively. Across karst systems, physicochemical and hydrological factors reflect a dynamic process dependent on concurrent conditions and antecedent states, exhibiting time-lag and hysteresis. These findings support risk assessment and stage-specific interventions for MP pollution in karst systems.