Annual river dataset in China: a new product with a 10 m spatial resolution from 2016 to 2023

Abstract. Rivers play important roles in ecological biodiversity, shipping trade and the carbon cycle. Owing to human disturbances and extreme climates in recent decades, river extents have altered frequently and dramatically. The development of sequential and fine-scale river extent datasets, which could offer strong data support for river protection, management and sustainable use, is urgently needed. A literature review revealed that annual river extent datasets with fine spatial resolutions are generally unavailable for China. To address this issue, the first Sentinel-derived annual China river extent dataset (CRED) from 2016 to 2023 was produced in our study. We first produced annual water maps by combining the dynamic world (DW), ESRI global land cover (EGLC) data and the multiple index water detection rule (MIWDR). For the DW and MIWDR water time series, the mode algorithm, which calculates the most common values, was used to generate yearly water maps. Then, an object-based hierarchical decision tree based on geometric features and auxiliary datasets was developed to extract rivers from the water data. The results indicated that the overall accuracies (OAs) of the CRED were greater than 96.0 % from 2016 to 2023. The user accuracies (UAs), producer accuracies (PAs) and F1 scores of the rivers exceeded 95.3 %, 91.3 % and 93.7 %, respectively. A further data intercomparison indicated that our CRED shared similar patterns with the wetland map of East Asia (EA_Wetlands), China land use/cover change (CNLUCC) and China water covers (CWaC) datasets, with correlation coefficients (R) greater than 0.75. Moreover, our CRED outperformed the three datasets in terms of small river mapping and misclassification reduction. The area statistics indicated that the river area in China was 44,948.78 km² in 2023, which was mostly distributed in coastal provinces of China. From 2016 to 2023, the river areas were characterized by an initial increase, followed by a decrease and then a slight increase. Spatially, the decreased rivers were located mainly in Southeast China, whereas the increased rivers were distributed mainly in Central China and Northeast China. In general, the CRED explicitly delineated river extents and dynamics in China, which could provide a good foundation for improving river ecology and management. The CRED dataset is publicly available at https://doi.org/10.5281/zenodo.13841910 (Peng et al., 2024a).