AI-driven prediction of soil trace metal contamination and ecological health in the Sundarbans mangrove ecosystem: Implications for nature-based solutions and the UN SDGs.

Mangrove ecosystems serve as critical biogeochemical buffers in tropical coastal zones, yet escalating trace metal contamination increasingly threatens their ecological integrity and carbon sequestration capacity. Here, we quantify the multi-decadal spatio-temporal dynamics of eight trace metals (Cd, Cu, Fe, Mn, Ni, Pb, Zn, and Cr) in the Sundarbans from 1995 to 2024, integrating Earth observation data with statistical and AI-driven modelling to resolve contamination patterns and forecast future risks. Results reveal a distinct seasonal regime, characterized by pre-monsoon enrichment, monsoonal dilution, and post-monsoon re-accumulation, reflecting strong hydrological control on metal distribution. Spatial analysis indicates intensifying heterogeneity and hotspot development, particularly for Zn and Fe, suggesting a shift toward increasingly complex contamination regimes. The hybrid LASSO-GA-BPNN model demonstrated the highest predictive accuracy (R2 > 0.97), outperforming Random Forest in capturing non-linear environmental interactions. Time-series Autoregressive Integrated Moving Average projections indicate a sustained increase in trace metal concentrations through 2100 under current environmental trajectories. Halophytic mangrove species exhibit contrasting responses to trace metal stress, with Ceriops decandra and Excoecaria agallocha demonstrating adaptive resilience, whereas Aegialitis rotundifolia shows heightened sensitivity to contamination stress. Collectively, these findings establish a predictive, scalable framework for coastal contamination assessment and highlight the urgency of implementing Nature-Based Solutions, including phytoremediation and eco-hydrological restoration. This research approach directly advances the Sustainable Development Goals (SDGs) 6, 13, 14, and 15 by linking data-driven environmental diagnostics with adaptive, ecosystem-based management in climate-vulnerable deltaic systems.