Profile soil water content (SWC) is a vital variable in the atmosphere-vegetation-soil system. Although remote sensing currently can provide reliable surface SWC data (∼5 cm depth), acquiring accurate subsurface SWC data from existing reanalysis products remain challenging. In this study, we evaluated two widely used methods in estimating subsurface SWC, namely the exponential filter (ExpF) and the principle of maximum entropy (POME). The evaluation is carried out in two distinct areas on the Tibetan Plateau using ground observations collected from two monitoring networks: the Maqu network area characterized by cold humid climate and grassland and the Shiquanhe network area characterized by cold arid climate and bare ground. The results indicate that POME generally performs better than ExpF in both areas, particularly in deeper soil layers. Specifically, the accuracy of estimated SWC using the ExpF method decreases with depth, while it increases with depth using the POME method. Additionally, both methods achieve commendable performance at a depth of 10 cm in both areas. The deficiency of ExpF is mainly reflected in underestimations for dry cases, which is amplified with increasing depth. Dry cases account for 51 % in the humid area and 68 % in the arid area throughout the study period. Consequently, the ExpF method yields higher root mean square differences (RMSD) by 30 % and 113 % in the humid area at depths of 20 and 40 cm, respectively, compared to the POME method. Similarly, it results in higher RMSD values by 220 % and 200 % in the arid area. As expected, the superior performance of POME in deeper soil layers is primarily attributed to the incorporation of additional bottom and profile mean SWC observations. However, it also potentially introduces uncertainties when integrated with satellite-based data, which inherently contains errors compared to ground observations. To assess the potential of these two methods in large-scale applications combined with satellite-based datasets, this study conducted further evaluation of both methods with required input data derived from the soil moisture active and passive mission (SMAP). The results demonstrate that the performance of both methods in estimating subsurface SWC is acceptable in both humid and arid areas, although some bias is transferred from the input data. They achieve average RMSD values of 0.034 and 0.055 m3/m−3(−|−) in the humid area for the ExpF and POME methods, respectively, and 0.021 and 0.014 m3/m−3(−|−) in the arid area.

中文翻译：

---

地下土壤含水量估算方法的评估：最大熵与指数滤波器


剖面土壤含水量 （SWC） 是大气-植被-土壤系统中的一个重要变量。尽管遥感目前可以提供可靠的地表 SWC 数据（∼5 cm 深度），但从现有的再分析产品中获取准确的地下 SWC 数据仍然具有挑战性。在这项研究中，我们评估了两种广泛使用的估计地下 SWC 的方法，即指数滤波器 （ExpF） 和最大熵原理 （POME）。评价是在青藏高原的两个不同区域进行的，利用从两个监测网络收集的地面观测数据：以寒冷湿润气候和草原为特征的玛曲网区和以寒冷干旱和裸露地面为特征的石泉河网区。结果表明，POME 在这两个领域的表现通常都优于 ExpF，尤其是在较深的土壤层中。具体来说，使用 ExpF 方法估计的 SWC 的准确性随深度的增加而降低，而使用 POME 方法则随深度的增加而增加。此外，这两种方法在两个区域的 10 cm 深度都取得了值得称赞的性能。ExpF 的缺陷主要反映在对枯病病例的低估上，这种情况随着深度的增加而被放大。在整个研究期间，干旱病例在潮湿地区占 51%，在干旱地区占 68%。因此，与 POME 方法相比，ExpF 方法在 20 厘米和 40 厘米深度的潮湿区域产生的均方根差 （RMSD） 分别高出 30 % 和 113%。同样，它导致干旱地区的 RMSD 值分别高出 220 % 和 200%。正如预期的那样，POME 在更深的土壤层中的卓越性能主要归功于额外的底部和剖面平均 SWC 观测。 然而，当与基于卫星的数据集成时，它也可能会带来不确定性，与地面观测相比，这些数据本身就包含误差。为了评估这两种方法在大规模应用中结合卫星数据集的潜力，本研究使用来自土壤水分主动和被动任务 （SMAP） 的所需输入数据对这两种方法进行了进一步评估。结果表明，两种方法在估计地下 SWC 方面的性能在潮湿和干旱地区都是可以接受的，尽管从输入数据中转移了一些偏差。ExpF 和 POME 方法在潮湿地区的平均 RMSD 值分别为 0.034 和 0.055 m3/m−3（−|−），在干旱地区达到 0.021 和 0.014 m3/m−3（−|−）。