Four Decades of Reach‐Scale Declines in Total Phosphorus Across the Yellow River Basin Driven by Vegetation Restoration and Dam Impoundment

Abstract

Phosphorus (P) is a finite resource fundamental to terrestrial and aquatic productivity. However, excessive P transfer from land to rivers threatens food production and degrades water quality. Systematic analyses of multidecadal total phosphorus (TP) variation in large river basins remain challenging because of incomplete and non‐uniform observational data. Here, we develop a potentially transferable and generalizable framework combining remote sensing and machine learning to reconstruct long‐term and reach‐scale TP variation in the Yellow River Basin from 1986 to 2024, based on synchronous multimodal and in situ data ( N  = 1,490). The results confirm that the framework can reliably estimate TP concentration, achieving Nash–Sutcliffe efficiency = 0.71 and root mean squared error = 27.1 μg L ⁻¹ on the test set. TP concentration shows a significant temporal decline in 55% of river reaches; spatial heterogeneity is evident, with concentrations in the middle reaches declining most rapidly at 4.1 μg L ⁻¹ per decade. However, TP flux (TPF) exhibits a U‐shaped temporal trajectory, increasing after 2015 as river discharge rises; the spatial gradient follows a hump‐shaped pattern, characterized by a peak of 2,094 t yr ⁻¹ at Tongguan. Vegetation restoration reduces land‐to‐river TP export, and dam impoundment impedes in‐stream TP transport, both of which largely account for the declines in TP concentration in the Yellow River Basin. This study proposes a potentially generalizable framework for riverine TP reconstruction at high spatial resolution over multiple decades and provides new insights into the anthropogenic impact on TP dynamics in big river systems.