Abstract
Lakes function as receivers, regulators, reactors, and reservoirs in the global carbon cycle. However, long-term and national-scale monitoring data on lake carbon parameters are not available. Alternatively, long-term archived Landsat data have potential for enhancing the spatial and temporal retrievals of lake carbon parameters. Using Landsat reflectance during 1984-2023 and in-situ measurements at 5,503 stations, this study developed several two-step Random Forest algorithms, systematically incorporating findings from previous studies, to remotely retrieve concentrations and storage of dissolved organic carbon (DOC), particulate organic carbon (POC), and dissolved inorganic carbon (DIC) across 24,366 lakes in China. This integrated 40-year dataset provides the grid-based distributions (1.0°) of different carbon components, overcoming limitations of prior single-component studies. The accuracies of the developed algorithms were validated by comparing the remotely derived results with publicly available reference data. The remotely retrieved dataset provides grid-based spatial and temporal distributions of lake carbon concentrations and stocks during 1984-2023, offering a valuable resource for lake water environment management, lake carbon stock estimation, and global carbon balance assessment.