Phytoplankton community composition and environmental associations in Liujiaxia Reservoir under a reverse seasonal water-level background

In regulated reservoirs, actual water-level dynamics do not always coincide with conventional hydrological periods, yet phytoplankton community variation under such conditions remains insufficiently understood. We investigated phytoplankton communities and environmental conditions at 21 fixed sites in Liujiaxia Reservoir during conventional wet, normal, and dry periods, while water level increased progressively across the three surveys. Seventy-eight phytoplankton species were identified, with Bacillariophyta dominating throughout. Total abundance and α-diversity declined toward the dry period, whereas β-diversity was mainly driven by species turnover, with the highest among-site dissimilarity in the dry period. Community assembly showed a relatively stable stochastic tendency only during the wet period, when co-occurrence network connectivity was also greatest. Variation Partitioning Analysis (VPA) explained 22.43% of community variation, leaving 77.57% unexplained, with a relatively large fraction jointly explained by spatial structure and hydrological period. Redundancy Analysis (RDA) identified water temperature (WT) and ammonia nitrogen (NH 3 −N) as significant correlates of community composition. A site-grouped cross-validated eXtreme Gradient Boosting (XGBoost) model achieved a predictive R 2 of 0.574 and identified WT as an important predictor of phytoplankton abundance; however, weak or negative period-specific performance limited seasonal interpretation. In pooled Partial Least Squares Path Modeling (PLS-PM), the Physical variable block was positively associated with phytoplankton abundance (β = 0.54, 95% CI: 0.29–0.80), whereas period-specific associations were highly uncertain. Overall, under the observed reverse seasonal water-level background, phytoplankton communities exhibited pronounced period-dependent variation in diversity, spatial turnover, assembly characteristics, and environmental associations.