A Guild-Based Ecological Assessment: Integrating Multi-Model Approaches to Reveal Key Environmental Associations for Waterbirds in Nansi Lake, China.

Wetlands are increasingly threatened by biodiversity loss and functional degradation. Waterbirds guilds, which reflect functional rather than taxonomic organization, provide a potentially informative yet underexplored framework for wetland assessment. In this study, we employed a multi-model approach in Nansi Lake, northern China. Environmental gradients were derived via principal component analysis (PCA), and relationships with waterbird guild densities were examined using Tweedie generalized linear mixed models (GLMM), with species included as a random effect to account for baseline density differences. Parallel random forest (RF) and LASSO regressions were applied to explore robust associations. GLMMs revealed that guild densities were consistently associated with the first two PCA gradients: PC1 (representing phytoplankton-zooplankton contrasts) and PC2 (high dissolved oxygen, low nitrogen). Notably, piscivorous birds showed a positive response to PC2 compared to other feeding guilds. Species identity explained a large proportion of density variation, confirming that controlling for interspecific differences was essential for detecting these environmental signals. Nutrient-related variables and sediment conditions exhibited negative associations with waterbird guild metrics across models, whereas fish diversity and periphyton-related variables exerted positive influences. The convergence of association patterns between linear mixed models and nonlinear machine learning methods increased our confidence, despite limited sample size. These findings suggest that guild-level patterns reflect the combined influence of abiotic conditions and biotic resource availability. Waterbird guild metrics offer useful, scalable indicators of wetland ecological condition, with potential applications in monitoring, assessment and conservation strategies.