Scale‐dependent pathways from land use to fish diversity: eDNA analysis reveals water quality as key mediator in the Nanliu River Basin

Abstract Human‐induced land‐use change poses substantial risks to aquatic biodiversity; however, the multi‐scale pathways linking land‐use patterns to fish diversity remain insufficiently resolved. Focusing on the Nanliu River Basin (Guangxi, China), we integrated environmental DNA (eDNA) detection with partial least squares structural equation modeling (PLS‐SEM) across three spatial scales—1000 m circular buffers, 2000 m river (riparian) buffers, and sub‐watersheds—to quantify the effects on species richness, Shannon diversity, and Simpson diversity. We detected 36 fish species, dominated by Cypriniformes (45.8%). At the local scale (1000 m circular buffer), fish diversity was promoted indirectly via landscape configuration (landscape→diversity 0.343), which was strengthened by a higher water body proportion. Land use positively shaped landscape configuration (0.617) but degraded water quality (land use→water quality –0.422), with TN and TP as major negative contributors and a weak direct water quality →diversity path (–0.079). At the riparian scale (2000 m river buffer), water quality—proxied by electrical conductivity—was the primary positive driver of diversity (0.329). Land use affected diversity indirectly via water‐body proportion and a secondary pathway through landscape configuration (0.197), whereas the direct landscape→diversity effect was minimal (–0.021). At the sub‐watershed scale, TP exerted a strong negative influence on diversity (–0.474); land use acted via nested mediation through water body proportion and landscape configuration (0.416), and the direct landscape→diversity path was negative (–0.098). Fish communities showed pronounced spatial heterogeneity, with the highest diversity in estuaries (Shannon = 3.31, Simpson = 0.92, richness = 18) and the lowest in urban sections (Shannon = 1.94, Simpson = 0.68, richness = 5). Collectively, the models support a scale‐dependent mechanism—land use → landscape pattern → water quality → fish diversity—and provide a quantitative basis for spatially targeted biodiversity conservation and watershed ecological management in subtropical river basins.