Traditionally, a monitoring network is set up to collect soil moisture information for a large area. The construction of monitoring networks is rather expensive in terms of both time and materials. However, the monitoring result is only point representativeness, and cannot satisfy a large area soil moisture mapping demand. Compared to traditional in-situ monitoring network results, information retrieved from spaceborne or airborne instruments is area representative. Moreover, the remote sensing method is much more timely and low cost. ERS (European remote sensing satellites) series satellites and METOP (meteorological operational satellite) series satellites provide global coverage, continuous, long-term, high revisit rate (2-5 days, determined by different latitudes) datasets. SCAT (scatterometer) and ASCAT (advanced catterometer) are the main scatterometer instruments onboard them respectively. Finding a practical retrieving method tailored for SCAT and ASCAT is very urgent. Referring to the TU-WIEN presented by Wolfgang in 1999, a practical method base on multi-angle long-term series change detection was developed in this paper. TU-WIEN takes full advantages of the multi-viewing capabilities of the sensor, the availability of several years of backscatter data, and a high temporal sampling rate. Taking the roughness, inhomogeneity, and vegetation cover of the land surface into account, soil moisture is retrieved by analyzing long time mass data with statistical techniques. However, there is still some weakness in the algorithm. An improvement was proposed in this paper, in which two key model parametersσ′(θ, t) andσ″(θ,t) are generated by adaption learning functions by changing a time moving window in sequence, instead of experience functions as used in the prior version. The proposed method can perform more stably and can be transplanted to different areas more easily. Besides, abnormal observations are removed from the long-term huge amounts of data to avoid fatal damage for the final output. In the experiment in the Iberian peninsula, the new function ofσ′(θ,t) which were generated by a new adapting time moving window, represented the seasonal variation ofσ′(θ). In addition, the experiment showed that the new adaption learning function could successfully take the place of the old experience one. Furthermore, the improved method was applied in Tibet Plateau area, where soil moisture is urgent needed. To validate the proposed algorithm, the result retrieved from remote sensing method was compared with in-situ observations which were collected in Maqu monitoring network in the Tibet-Obs plan. A good consistent relationship was found between the retrieval results and in-situ observations. The RMSE (root mean square error) was 0.0155, and the related coefficientR2 was 0.8361. The applicability of the algorithm was validated preliminarily. The algorithm is worthy of being applied to more needed areas to help take the advantages of satellite monitoring into practical use.%为克服传统地面监测土壤水分方法费时费力的不足，满足大范围、长时间连续、实时动态的监测要求，本文采用ERS（european remote sensing satellites）和METOP（meteorological operational satellite）卫星搭载的微波散射计对地标土壤含水量进行观测。ERS和METOP系列卫星组成了覆盖全球的高时间分辨率（根据纬度不同，大约2～5 d即可重复观测）长期连续的对地观测网；SCAT（scattorometer）和ASCAT（advanced scattorometer）散射计分别是搭载其上的微波散射计，具备全天候监测地表土壤水分的能力。该文以1999年Wolfgang提出的经典 TU-WIEN 算法为基础，改进了其中人为定义经验函数的描述模型参数季节性变化规律的不足。在伊比利亚半岛的比较中发现，用新的移动时间窗口自动生成的σ′(θ，t)函数很好的描述了σ′(θ)的季节性变化，替代原有经验函数的作用。继而，利用改进的σ′(θ，t)函数，将经典TU-WIEN算法移植到有迫切实际需求的青藏高原地区。通过和地面实测数据的比较发现，卫星反演结果与实地测量数据有着很好的一致性和较低的误差率，总体均方根误差RMSE=0.0155，相关系数R2=0.8361，证实了新算法的可行性和应用价值。
展开▼
