Abstract
The trophic state index (TSI) serves as a fundamental indicator for evaluating the water quality of lake ecosystems. Under climate change and human activities, global lake TSI has changed significantly, yet its response mechanisms remain unclear. To address this challenge, we developed a generalized TSI estimation model by integrating semi‐analytical algorithms with machine learning techniques, based on a comprehensive dataset comprising 3756 pairs of in situ measurements and remote sensing observations. The developed model demonstrated superior predictive performance with R 2 of 0.77 and RMSE of 8.25 for the testing dataset. Applying the model, we reconstructed a 21‐year time series (2003–2023) TSI for 14,189 global lakes with surface area ≥ 10 km 2 . The global mean TSI was estimated to be 54.07 ± 0.31. Among the lakes, 4.1% were classified as oligotrophic (TSI ≤ 38), 18.9% as mesotrophic (38 61). Globally, TSI showed a significantly increasing trend at a rate of 0.19 per decade ( p