A novel method for assessing water quality status using MODIS images: A case study of large lakes and reservoirs in China.

• Remote sensing identified chemical indicator-induced water quality changes. • Combining FOD and spectral indices enhanced spectral sensitivity. • Optical water classification modeling effectively handled large-scale inversions. • A high-accurate remote sensing model for water quality status was developed. • Chinese lake water quality variations were computed from 2000 to 2022. The changing climate and economic development have exerted significant pressure on the water quality of global lakes and reservoirs (hereinafter referred to as lakes). Existing sporadic in-situ monitoring limits a comprehensive understanding of water quality changes at large spatiotemporal scales. Although remote sensing techniques offer efficient water quality observations, they have primarily focused on monitoring single optical active substances, making it difficult to assess water quality changes caused by chemical indicators such as total phosphorus and total nitrogen. However, changes in chemical indicators within a certain range can cause responses to physical indicators, which are comprehensively reflected in the water-leaving reflectance. Based on this physical mechanism, this study obtained the water quality status categories (from excellent to severe pollution) by calculating the water quality index (WQI) and developed a new remote sensing assessment model of water quality status using moderate resolution imaging spectroradiometer (MODIS) imagery. This model, combining fractional-order derivatives and various-dimensional spectral indices, significantly enhanced the sensitivity of spectral reflectance to water quality states, achieving a maximum correlation of 0.72. To address the applicability of the model at large spatiotemporal scales, this study proposed a classification modeling scheme based on the optimal optical water categories, resulting in a 20.69%–34.48% improvement in model accuracy after classification. Additionally, this study proposed an improved ensemble learning model combining the concepts of Stacking and Bagging, achieving an average accuracy of 82% and enhancing accuracy by 5%–15% compared to the original single model. Subsequently, the model was used for the first time to assess the water quality status of 180 large lakes in China from 2000 to 2022. The results revealed that 76.11% of the lakes exhibited excellent and good water quality, with a spatial distribution pattern showing a "better in the west, worse in the east" pattern. Over the 23-year period, 28.33% and 58.89% of the lakes showed improvement and stability trends in water quality status, with stability and improvement predominating in the western and eastern regions, respectively. The research results provide technical support for the rapid assessment of water quality status and sustainable resource management, highlighting the potential of remote sensing technology in water quality monitoring. [ABSTRACT FROM AUTHOR]