Spatiotemporal dynamics and machine learning‐based prediction of above‐ground biomass in the Indus Delta mangroves

Abstract

Abstract Mangrove forests are important blue carbon ecosystems, yet long‐term above‐ground biomass (AGB) dynamics in arid deltaic systems remain poorly understood. Here, we integrated field‐derived AGB data with multisource remote sensing and machine learning models (random forest, gradient‐boosted regression tree, support vector regression and classification and regression trees) to map historical patterns and predict future AGB dynamics in the Indus Delta mangroves for 2030, 2040 and 2050. AGB increased significantly over time, with mean values rising from 18.13 ± 9.1 Mg/ha in 2002 to 25.75 ± 8.32 Mg/ha in 2022, as indicated by Mann–Kendall trend analysis. Among models, GBRT performed best ( R 2 = 0.65, RMSE = 0.52) and projected continued increases in AGB to 30.31 ± 4.8 Mg/ha in 2030, 40.12 ± 6.4 Mg/ha in 2040 and 48.6 ± 7.9 Mg/ha in 2050. AGB was positively associated with vegetation indices and negatively related to land surface temperature and land‐use change. Synthesis . Mangrove AGB in the Indus Delta is increasing and is projected to continue rising under current conditions, highlighting substantial carbon sequestration potential in arid coastal systems. The strong performance of machine learning models demonstrates their utility for large‐scale biomass prediction, while the observed environmental controls emphasize the importance of sustaining freshwater input, sediment supply and restoration efforts for long‐term ecosystem resilience and blue carbon management. Read the free Plain Language Summary for this article on the Journal blog.