Monsoon-driven DOM dynamics revealed by self-organizing map and end-member mixing analysis: Spatial partitioning of autochthonous and anthropogenic sources in plain river networks.

Monsoon influences induce dynamic variations in the composition of dissolved organic matter (DOM) sources in river systems. Quantitative assessments of autochthonous versus allochthonous (natural versus anthropogenic) DOM contributions before and after monsoon seasons remain insufficient for complex urban-rural plain river networks. This study developed a novel approach by combining leave-one-out cross-validation information criterion (LOOIC) screened optimal dual-spectral-tracer end-member mixing analysis (EMMA) model with self-organizing map (SOM) modeling to resolve spatial and monsoon-induced variations of DOM sources in complex watersheds. Comparative evaluation of six tracer combinations demonstrated the humification index (HIX) and Y fluorescence index (YFI) pair as most suitable for the study area. The integrated SOM-EMMA analysis revealed that autochthonous sources constituted the dominant contributors both before and after the monsoon period, with their relative contribution increasing during the post-monsoon phase while anthropogenic sources showed corresponding dilution effects. Notable spatial differences emerged, with autochthonous sources dominating the northern region (48.7 %) versus anthropogenic predominance in the south (41.3 %), along with observed pollution accumulation effects near sluice gates. Agricultural activities represented the primary contributor among southern anthropogenic sources, consistent with the region's more developed agricultural conditions. These findings provide a new tool for watershed water quality management and pollution source tracking, suggesting that incorporating spectral indices into routine monitoring could reduce costs and optimize management strategies.