Operational monitoring windows from cumulative rainfall to chlorophyll-a in subtropical reservoirs.

Extreme rainfall reshapes the underwater light field and nutrient regime, complicating the predictability of algal risk. We sampled 20 subtropical drinking-water reservoirs in Guangdong, China, in June and September 2024 and used a data-driven search to identify month-specific antecedent rainfall windows. The optimal antecedent windows showed strong seasonality, spanning 30 d in June but only 6 d in September, with contrasting cumulative rainfall thresholds (k = 26 d in June and k = 48 d in September). Then, we used piecewise SEM to disentangle the pathways of light attenuation coefficient (Kd) and nutrients (DIN, TP) influencing size-fractionated chlorophyll-a. Under the late-season window, rainfall consistently elevates turbidity and redistributes nutrients, yielding robust increases in pico/nano chlorophyll while effects on > 20 μm remain weak-together providing an actionable 7-8-week lead time. Building on these patterns, we outline a joint monitoring bundle-windowed rainfall, Kd, and TN/TP-and convert windows into practical rules for sampling frequency and station placement in storm-prone reservoirs. The framework reduces window-selection bias and is readily transferable to operational early warning.