The validation of satellite soil moisture (SM) products is challenged by the large scale difference between in situ-satellite based measurements. In order to tackle the significant spatial scale mismatch, this study conducted a multi-scale validation of three typical SM products (i.e., SMOS-IC, SMAP L3, and AMSR2 LPRM) over Qinghai-Tibet Plateau. First, a 25-km better-performing SM dataset was produced by combining three SM products with the extended triple collocation (ETC) and arithmetic mean method. Second, a 500 m SM dataset was derived from the 25-km SM dataset with a random forest-based downscaling model and the high-resolution datasets of other variables. Third, the 500 m SM was evaluated using in situ SM measurements, which was then aggregated to a coarse pixel scale for the assessment of coarse-resolution satellite SM products. Finally, potential factors influencing the accuracy of satellite SM products were investigated. The results indicated that the 25-km merged SM product integrated the characteristics of these three SM products. During the downscaling process, terrain factors, NDVI, and day and night temperature difference were identified as the key variables to derive high-resolution SM, which agreed well with in situ measurements over most monitoring networks. The multi-scale validation results indicate that SMAP L3 and AMSR2 LPRM performs best regarding the median values of the correlation and deviation from the pixel scale reference in the spatial domain, respectively, and SMOS-IC is always the worst. However, when the pixel-based evaluation results were focused, AMSR2 LPRM performs best in most cases, followed by SMAP L3 and SMOS-IC. The accuracy of satellite SM products shows more dependence on slope than elevation, land cover types, and land surface temperature.

卫星土壤水分 (SM) 产品的验证受到原位卫星测量之间的大规模差异的挑战。为了解决显着的空间尺度错配问题，本研究对青藏高原上的三种典型SM产品（即SMOS-IC、SMAP L3和AMSR2 LPRM）进行了多尺度验证。首先，通过将三个 SM 产品与扩展三重搭配 (ETC) 和算术平均方法相结合，生成了一个 25 公里的性能更好的 SM 数据集。其次，500 m SM 数据集来自 25-km SM 数据集，具有基于随机森林的降尺度模型和其他变量的高分辨率数据集。第三，使用原位评估 500 m SMSM 测量，然后将其聚合到粗像素尺度，用于评估粗分辨率卫星 SM 产品。最后，研究了影响卫星SM产品精度的潜在因素。结果表明，25公里合并的SM产品综合了这三种SM产品的特点。在降尺度过程中，将地形因素、NDVI、昼夜温差确定为关键变量，得到高分辨率SM，与现场吻合较好在大多数监测网络上进行测量。多尺度验证结果表明，SMAP L3 和 AMSR2 LPRM 分别在空间域中与像素尺度参考的相关性和偏差的中值方面表现最好，而 SMOS-IC 始终是最差的。然而，当关注基于像素的评估结果时，AMSR2 LPRM 在大多数情况下表现最好，其次是 SMAP L3 和 SMOS-IC。卫星 SM 产品的精度显示出更多地依赖于坡度而不是高程、土地覆盖类型和地表温度。