Unraveling the nonlinear effects of urban blue-green spaces on residents' health in the United States (2000-2019).

Amid rapid urbanization and the degradation of natural landscapes, urban blue-green spaces (BGSs) have garnered increasing attention for their role in promoting health and providing nature-based therapeutic benefits. Although the general health benefits of BGSs are well-documented, their influence on mortality related digestive diseases across diverse demographic groups remains insufficiently understood. This study addresses this gap by analyzing four BGS metrics-population-weighted greenspace exposure (GE), greenspace exposure Gini coefficient (GJ), population-weighted water-distance exposure (WE), and water-distance Gini coefficient (WJ) -and their associations with digestive diseases mortality across 3111 U.S. counties from 2000 to 2019. We used least absolute shrinkage and selection operator and Boruta algorithms to identify influential environmental, meteorological, socioeconomic, and demographic metrics and subsequently trained a LightGBM model with Bayesian hyperparameter tuning. SHAP analysis revealed pronounced nonlinear relationships between BGS characteristics and mortality outcomes. All four BGS metrics showed nonlinear and negative associations with digestive diseases mortality, with clear threshold effects. WE was the most influential metric. GE exhibited consistent protective effects that intensified around the group mean value (GMV), while GJ produced increasing benefits once surpassing the GMV. For WJ, mortality reductions were observed across a narrow window-from roughly 0.05 below to 0.08 above the GMV-after which benefits plateaued. WE demonstrated diminishing marginal effects from approximately 0.6 to 2.0 above the GMV, followed by a rebound trend. Older adults experienced noticeably stronger protective effects across all four metrics. Overall, this study underscores the importance of BGS exposure and its distribution in shaping digestive-disease mortality patterns and population-specific health outcomes.