A 1 km daily high-accuracy meteorological dataset of air temperature, atmospheric pressure, relative humidity, and sunshine duration across China (1961–2021)

Abstract

The lack of high-accuracy, fine-resolution meteorological datasets in China has hindered progress in climate, hydrological, and ecological studies. In this study, we present a 1 km daily dataset spanning 1961–2021 across China, which includes six key variables – average, maximum, and minimum temperature, atmospheric pressure, relative humidity, and sunshine duration – to provide a reliable foundation for advancing related research and applications. The dataset was generated using a novel hierarchical reconstruction framework that leveraged daily observations from 2345 meteorological stations and incorporated topographic attributes. This approach effectively decodes the nonlinear relationships between the meteorological variables and their spatial covariates, ensuring the generation of gridded daily fields that are both high-resolution and spatially continuous. Validation against 146 independent stations confirmed the high accuracy of the dataset. For average, maximum, and minimum temperatures, the errors are minimal (median root mean square errors (RMSEs): 1.16, 1.19, 1.29 °C; median mean errors (MEs): −0.04, −0.10, −0.01 °C), and the consistency with in-situ data is very high (median correlation coefficients (CCs): 0.99, 0.99, 0.99). Atmospheric pressure also shows very small errors (median RMSE: 2.65 hPa; median ME: −0.06 hPa) and strong correlation (median CC: 0.97). Relative humidity exhibits relatively lower accuracy (median RMSE: 6.33 %; median ME: −0.52 %; median CC: 0.90), but it still exceeds standard benchmarks. Sunshine duration maintains high precision (median RMSE: 1.48 h; median ME: 0.05 h; median CC: 0.93), indicating the robustness and reliability of the dataset. Further comparison reveals that in high-altitude and topographically complex regions, the reconstructed product demonstrates higher actual accuracy than suggested by station-to-grid validation, as spatial mismatches between stations and grid cells lead to systematic underestimation. Free access to the dataset is available at https://doi.org/10.11888/Atmos.tpdc.301341 or https://cstr.cn/18406.11.Atmos.tpdc.301341 (last access: 25 November 2025) (Zhao et al., 2024).