Abstract. Accurate, high-resolution soil moisture data are critical for hydrological modeling, climate studies, and ecosystem management. Unfortunately, current existing global products suffer from inconsistencies, coverage gaps, and biases. In this study, we evaluated the surface layers of three widely used soil moisture products, including ERA5-Land, ESA-CCI (v09.1 Combined), and SMAP L4 with resolutions ranging from 0.1 to 0.25°, against in situ measurements across five networks, including ISMN, CMA, Cemaden, COSMOS-Europe, and SONTE-China. The in situ dataset, to our knowledge, represents the most extensive global soil moisture compilation to date, comprising approximately 3.8 million records, organized into a primary dataset for modern validation (2015–2020) and an independent historical dataset (1960–2015). It is found that during the primary validation period (2015–2020), ERA5-Land exhibits high correlation (with correlation coefficient of 0.69) between measured and predicted soil moisture but the data also shows significant bias. SMAP L4 provides the highest accuracy (with root mean square error (RMSE) value of 0.088 m3 m−3) and low bias, but is limited by its temporal coverage from 2015 to the present. To address these gaps, we developed an adjusted ERA5-Land dataset spanning 1950 to 2025 by fusing ERA5-Land and SMAP L4 using a mean-variance rescaling method optimized for long time-series alignment, which enhanced the spatiotemporal coverage and reduced bias. Validation against the primary validation period demonstrates a reduction in RMSE of 24.6 % and an improvement in normalized Nash-Sutcliffe Efficiency (NNSE) of 30.6 % compared to the original ERA5-Land products. Crucially, the reliability of the backward extension was verified against independent historical observations spanning 1960 to 2015, demonstrating sustained improvements over ERA5-Land with 19.7 % RMSE reduction and 26.6 % NNSE increase. This confirms the robustness of the adjustment parameters in the mean-variance rescaling method. The adjusted ERA5-Land dataset, which is publicly available, can be used as benchmark for future research and support drought monitoring, weather prediction, and water resource management, contributing to global climate resilience across diverse ecosystems. The dataset is provided for the surface layer with global coverage at a 0.1° spatial and daily temporal resolution, spanning from 1950 to 2025, at https://doi.org/10.57760/sciencedb.30546 (Wang et al., 2026).
Wang et al. (Tue,) studied this question.