A Soil Moisture Retrieval Method for Reducing Topographic Effect: A Case Study on the Qinghai–Tibetan Plateau With SMOS Data

The topography can be very important for passive microwave remote sensing of soil moisture due to its complex influence on the emitted brightness temperature observed by a satellite microwave radiometer. In this study, a methodology of using the first brightness Stokes parameter (i.e., the sum of vertical and horizontal polarization brightness temperature) observed by the soil moisture and ocean salinity (SMOS) was proposed to improve the soil moisture retrieval under complex topographic conditions. The applicability of the proposed method is validated using in-situ soil moisture measurements collected at four networks (Pali, Naqu, Maqu, and Wudaoliang) on the Qinghai–Tibetan Plateau. The results over Pali, which is a typical mountainous area, showed that soil moisture retrievals using the first brightness Stokes parameter are in better agreement with the in-situ measurements (the correlation coefficient R > 0.75 and unbiased root mean square error <0.04 m3/m3) compared with that using the single-polarization brightness temperature. At the other three networks with relatively flatter terrains, soil moisture retrievals using the first brightness Stokes parameter are found to be comparable to the single-polarization retrievals. On the contrary, the maximum bias of the retrieved soil moisture caused by topographic effects exceeds 0.1 m3/m3 when using vertical or horizontal polarization alone, which is far beyond the expected accuracy (0.04 m3/m3) of SMOS satellite. In the regions on the Qinghai–Tibetan Plateau where the vegetation effect can be ignored, soil moisture retrieved using horizontal polarization brightness temperature is generally underestimated, overestimated when using vertical polarization brightness temperature. It is reasonable due to the polarization rotation effect (depolarization) caused by the topographic effects. It is concluded that the proposed method for soil moisture retrieval using the first brightness Stokes parameter has a great potential in reducing the influence of topographic effects.

T HE mountainous terrain is undulating, and the mountainous directly affect the climate and environment around it through atmospheric and hydrological processes [1], [2]. The Qinghai-Tibetan Plateau is the highest plateau in the world and also has the most complex terrain on the earth [3]. The main landcover types of the Qinghai-Tibetan Plateau include forests, grasslands, shrubs, deserts, and ice and snow, etc. The geomorphic features of the Qinghai-Tibetan Plateau include mountains, plateaus, basins and wide valleys, etc. In the Qinghai-Tibetan Plateau, not all the mountainous areas are covered by rocks. Many areas with large topographic relief are covered by meadows, shrubs or deserts in the Qinghai-Tibetan Plateau. Soil moisture plays an important role in the process of water cycle in mountainous areas of the Qinghai-Tibetan Plateau.
Many studies have shown that in the past decades, the Qinghai-Tibetan Plateau has experienced significant hydrological and climatic changes [1], [4], [5], [6], [7]. Reliable long-term soil moisture products are essential for understanding the exchanges of water and energy between the land and the atmosphere in the Qinghai-Tibetan Plateau and its impact on the climate in Eastern and Southeast Asia [1]. However, due to the complex terrain, special geographical environment and climatic conditions in the Qinghai-Tibetan Plateau (mountainous areas), it is difficult to understand the detailed process of soil moisture change due to lack of effective in-situ measurements [1], [2]. Remote sensing has become a promising method to solve these problems. Microwave remote sensing has the special capability of all-weather and all-time monitoring of soil moisture [8], [9]. Microwave L-band radiometer is considered as the optimum tool for spatial soil moisture monitoring due to its strong penetration of vegetation layer and sensitivity to surface soil moisture [10], [11], [12], [13].
At present, the soil moisture algorithms based on satellite observation mainly include the single channel algorithm (SCA) and dual-channel algorithm (DCA) [14], the multi-angular algorithm developed for SMOS [15], SMOS_IC algorithm [16], the land parameter retrieval algorithm [17], the recently developed multichannel collaborative algorithm [18] and multitemporal and multiangular approach [19], etc. The SCA algorithm uses single horizontal polarization (H_pol) or vertical polarization (V_pol) brightness temperature (TB) to retrieve soil moisture. The DCA uses H_pol and V_pol TB to simultaneously retrieve multiple parameters (such as soil moisture and vegetation optical depth). Moreover, Jackson et al. [20] and Owe et al. [17] also have tried to use H_pol and V_pol TB to retrieve multiple parameters. There is information redundancy between H_pol and V_pol TB, which makes the retrievals not robust. Konings et al. [21] found that the redundancy information from the two polarizations reduces the ability of DCA algorithm. Zhao et al. found that the degree of information (DoI) is the function of the number of observation channels (incidence angle, polarization, and frequency) [18]. Therefore, the multiangular H_pol and V_pol TB observation provided by SMOS satellite can increase the DoI for the H_pol and V_pol TBs, which provides support for simultaneous retrieval of multiple parameters (such as soil moisture and vegetation optical depth) and obtaining reliable solutions.
However, the influence of topography on the microwave emission is significant in mountainous areas [22], [23]. The microwave radiation received by the satellite radiometer may be affected by the rotation of radiation polarization plane, the shadowing effect and the multiple scattering effect between adjacent terrains and thus has a serious impact on the remote sensing of soil moisture [24].
Pierdicca et al. found that the TB bias were about 8 to 15 K at L-and X-bands due to topographic effects [25]. Luca et al. analyzed the influence of terrain slope and aspect on the microwave emission observed by satellite and found that the orientation of the linear polarization is rotated (depolarization effect) and the rotation influence on the V_pol TB is greater than that of H_pol TB [26]. Mützler et al. [27] and Kerr et al. [28] analyzed the influence of topographic effects on TB measured by spaceborne microwave radiometer, including the shadowing effect, changes of local incidence angle and the polarization rotation, which resulted in the deviation of the characteristics of the geophysical parameter to be retrieved. Talone et al. studied the changes of local incidence angle and the shadowing effects caused by the topography, and found that the change of local incidence angle caused by topography was as high as 55°when comparing incidence angle of TB values after terrain correction with that of TB under the assumed ellipsoid level of the Earth [29]. Taking the Qinghai-Tibetan Plateau as an example, Li et al. found that, due to the topographic relief, the maximum reduction of TB at V_pol was about 16 K, and the maximum increasement of TB at H_pol was 18 K, and soil moisture could be overestimated more than 0.04 m 3 /m 3 due to such topographic influence [24].
Although many researchers indicated that the topographic effects have a significant impact on the microwave TB, most soil moisture retrieval algorithms assumed that the Earth's surface is horizontal, thus may fail to retrieve soil moisture over mountain areas like the Qinghai-Tibetan Plateau [30]. The current soil moisture retrieval algorithms for satellite observations normally mask out the topographic areas [31] and few studies have demonstrated the evidence of topographic effects on soil moisture retrieval at the satellite scale [24], [29]. Therefore, it is needed to quantitatively study the influence of topography on soil moisture retrievals based on spaceborne observations and explore possible solution to minimize the topography effects. Although the topographic relief may change the process of microwave radiative transfer and affects the retrieval accuracy of land surface parameters, these effects can be minimized when using the first brightness Stokes parameter (H_pol + V_pol TB) to retrieve land surface parameter (such as soil moisture). The first Stokes parameter is widely used in ocean salinity retrieval [25], [26], [32], [33], but they are rarely used in land surface soil moisture retrieval. In this study, a novel methodology that uses the sum of H_pol and V_pol TB (the first brightness Stokes parameter) was proposed to retrieve soil moisture based on the multitemporal and multiangular (MTMA) method developed in our previous study [19], which is expected to explore a solution to reduce the uncertainty of soil moisture retrievals caused by topographic effects.
This study is composed of the following parts: the data used in this study are presented in Section II; the methodologies including three retrieval strategies (single H_pol, single V_pol) and the first brightness Stokes parameter (H_pol + V_pol TB) are described in Section III; the retrieval results were presented and analyzed in Section IV; and the discussion and conclusions are given in Sections V and VI.

A. SMOS L1C Brightness Temperature
The SMOS L1C TB data are provided in icosahedral Snyder equal area earth fixed grid (ISEA-4H9), which is referred to the discrete global grid (DGG) with nodes 15 km apart [15]. The surface is in thermal equilibrium at dawn (6 A.M., usually before the sunrise) that the temperature of vegetation is expected to be close to the temperature of soil. In this study, it is assumed that vegetation canopy temperature and soil temperature are in equilibrium and expressed by the effective soil temperature, which can reduce the number of unknown variables in the algorithm. Therefore, the SMOS L1C TB data (smos-diss.eo.esa.int) at 6 A.M. (ascending orbits) are used. The two-step regression method proposed by Zhao et al. [34] was used to preprocess the SMOS L1C data so reduce the uncertainties of observed TB due to radio frequency interference (RFI) and aliasing issues [34]. The two-step regression method does not filter the SMOS L1C TB data by setting the threshold, but makes the SMOS L1C TB data closer to the theoretical expectation by refining the SMOS L1C TB data, which means that the TB data affected by RFI may also be used for the retrieval after refinement. The details of the two-step regression method are expounded in [34].

C. Topographic Data
The shuttle radar topography mission (SRTM) data (spatial resolution of 90 m), which were jointly measured by the National Aeronautics and Space Administration and the National Imagery and Mapping Agency, were utilized to calculate the slope on the Qinghai-Tibetan Plateau [see Fig. 1

D. MODIS NDVI Product
Normalized difference vegetation index (NDVI) products (MOD13A2) in 2015-2016 from the moderate-resolution  imaging spectroradiometer (MODIS) (see Fig. 2) are used as a reference to evaluate the vegetation richness on the Qinghai-Tibetan Plateau.

E. European Centre for Medium-Range Weather Forecasts (ECMWF) Data
The ECMWF Reanalysis v5 (ERA5) data are a reanalysis dataset, which combines model data with observations across the world into a globally complete and consistent dataset. The data used in this study include soil layer temperature (0-7 cm) and cumulative precipitation (unit: m; precision: 0.1 mm). The spatial resolution of ECMWF ERA5 data is resampled to match the nodes spacing size (15 km) of SMOS L1C DGG TB data.

F. Sentinel-1A Backscatter Data
The C-band (5.405 GHz) radar data (descending, 6 A.M.) of Sentinel-1A, including the dual-polarization (VV + VH) backscatter data, are used to calculate the radar vegetation index (RVI = (4 * VH)/(VV + VH)) [37]. The temporal resolution and spatial resolution of the Sentinel-1A data at 6 A.M. are 12 to 24 days (maximum) and 10 m, respectively. The radar data of Sentinel-1 are preprocessed for radiometric calibration, thermal noise removal and terrain correction, and resampled to 25-km spatial resolution to match the spatial resolution of SMOS soil

G. In-situ Soil Moisture Network
The in-situ soil moisture data from four networks (Pali [38], Naqu [38], [39], Maqu [39], [40] and Wudaoliang) located on the Qinghai-Tibetan Plateau in China were used as validation data. The location of each validation network and the distribution of sites in each network are shown in Fig. 1(a). The Wudaoliang soil moisture and temperature monitoring network were built in 2019 with 10 sites. The range of slope and standard deviation of elevation in the area covered by each network are calculated using SRTM data with spatial resolution of 90 m to reflect the topography (see Table II). Compared with Maqu, Naqu, and Wudaoliang networks, Pali network has a wider slope range and a larger standard deviation of elevation, which means that the terrain at Pali network are more complex than the other three networks.

III. METHODOLOGY
One of the major topographic effects is the polarization rotation (depolarization) effect, which can reduce the TB values at V_pol and increase the TB values at H_pol, leading to the uncertainties of soil moisture retrieval. One possible approach for reducing topographic effects (especially the polarization rotation effect) is to apply the first brightness Stokes parameter (i.e., H_pol TB + V_pol TB) in the soil moisture retrieval algorithm. This is because the first brightness Stokes parameter remains unchanged after the polarization rotation caused by the topographic effects. The implementation of our method is composed by: 1) the SMOS L1C multi-angular TB is refined using the two-step regression method [34]; 2) the first brightness Stokes parameter is substituted into the MTMA method to replace the TB of single polarization channel [19], and then the vegetation optical depth, effective scattering albedo, soil roughness, and soil moisture are retrieved. The details of the method is described in the following sections.

A. Conception of Minimizing Topographic Effects
When the microwave radiation (TB) is affected by the topography, the polarization seen at the satellite-Earth (global reference frame) surface would be rotated with respect to that referenced to the local surface frame. The transformation from the local to the global reference frame can be expressed by the following: where ϕ represents the polarization rotation angle; T B H and T B V are the H_pol and V_pol brightness temperature at the global reference frame with global incidence angle of θ; and T B h and T B v are brightness temperature at the local reference frame with local incidence angle of θ l . The polarization rotation angle ϕ can be calculated by the following: where α l , β l are the slope angle and the azimuth angle at the local reference frame, and β is the azimuth angle at the satellite reference frame.
Equation (1) can be rewritten as follows: Therefore, before and after the polarization rotation caused by the topographic effect, the sum of the V_pol and H_pol TB, which is referred as the first brightness Stokes parameter, is unchanged. On the other hand, many researchers have indicated that the topographic effects have the depolarization effect that increases the H_pol TB and decreases the V_pol TB [24], [27], [28], [29], and thus may bring bias to the retrieval of soil moisture [10], [24], [41] when using existing retrieval algorithms usually without considering topographic effects. In this study, we propose to use the first brightness Stokes parameter (TB H + TB V = TB H+V ) to reduce topographic effects during the soil moisture retrieval.

B. Soil Moisture Retrieval Algorithm
The soil moisture retrieval algorithm used in this study is the MTMA method [19]. The MTMA method combines microwave vegetation index (MVIs) [8], [42], [43] and multitemporal method [44], [45] to decouple soil and vegetation contributions, and uses SMOS multiangular observation TB data to systematically retrieve vegetation optical depth, effective scattering albedo, surface roughness, and soil moisture.
In this study, we further introduce the first brightness Stokes parameter into to the MTMA method to minimize the topographic effects. The retrieval process of vegetation parameters is then given by the following: and the retrieval process of soil parameters are described by the following: where m p and n p are the SMOS multiangle microwave vegetation index; V e p and V a p are the vegetation emission term and attenuation term, respectively; Γ p is the vegetation transmissivity ( Γ p = exp(− − τ c p /cos(θ))) and is the function of VOD p (τ c p ); ω eff p is the effective scattering albedo; T c (K) and T s (K) are the canopy and soil temperature respectively and are assumed to be equivalent and represented by the effective soil temperature (T eff ), which could be gained from ancillary soil temperature data (0-7 cm) in the reanalysis data of ECMWF ERA5; γ p and μ p are regression constants using simulation data from the advanced integral equation model (AIEM) and their values of the two polarizations for different pairs of incidence angles are shown in the Table III; θ 1 and θ 2 are two incidence angles, for example, θ 1 is 30°and θ 2 is 40°; t is the SMOS satellite overpass time (6 A.M.); σ(T B o p ) (K) is the standard deviation of SMOS observed brightness temperature; T B p (K) is simulated brightness temperature; T B o p (K) is SMOS observed brightness temperature; N is the number of satellite overpasses; K is the number of angles of SMOS observations; A p , B p , and C p are functions of θ and obtained by regression analysis using simulation data set from AIEM [46]; E s p is the simulated soil emissivity; E o p is soil emissivity calculated using observed brightness temperature of SMOS with retrieved V OD p and ω eff p ; and X (V OD p , ω eff p , SM p , and Z s p ) are the retrieved parameters. It should be noted that (6) and (12) are derived by introducing the first brightness Stokes parameter into the MVIs developed by Cui et al. [42] and the bare soil emissivity model developed by Zhao et al. [46], respectively.
Konings et al. [21] and Zhao et al. [18] have found that the degree of information (DoI) for the H_pol and V_pol brightness temperatures is less than 2 (1< DoI <2), therefore the multitemporal TB data observed by SMOS at the 3(angle pairs)•2(polarization) = 6 channels, i.e., the dual-polarization brightness temperatures observed at three angle pairs ((15°, 30°), (20°, 35°), and (25°, 40°)) from two temporal adjacent overpasses, are used for the retrieval process in this study. Zhao et al. found that the degree of information (DoI = 1.3322 * M+0.4142, M represents the number of observation angle pairs) is the function of the number of observation angle pairs of dual-polarized TB and increases with the increase of observation angle pairs [18]. The degree of information of dual-polarized TB from six channels is 8.4. For the parameter retrieval, there are totally four parameters to be retrieved including vegetation parameters (vegetation optical depth and effective scattering albedo) and soil parameters (soil moisture and soil surface roughness). Therefore, the degree of information is sufficient to retrieve the unknown four parameters. The retrieval is performed by minimizing the cost function with the least square method.

A. Comparison With In-Situ Measurements
To validate effectiveness of topographic effects by using the first brightness Stokes parameter for the retrievals of soil moisture, we compared the MTMA_SM H+V retrieved by applying the first brightness Stoke parameter in the MTMA algorithm with in-situ measurements at Pali, Naqu and Maqu, and Wudaoliang soil moisture network. In addition, soil moisture retrieved using single-polarization TB (MTMA_SM H for H_pol and MTMA_SM V for V_pol) were also obtained. The topographic effect at the Pali network (with elevation standard deviation: 375.54 m; slope range: 0°-71.85°) is more significant than those of the other three networks (elevation standard deviation: 105.05-187.34 m; slope range: 0°-39.89°).
It should be noted that this study assumes that the MTMA method will deliver a "very dry bare soil" (low dielectric constants) output when the soil temperature is lower than 273.13 K. The soils are considered as frozen soils when the soil temperature is less than 273.15 K. Although volume scattering may exist with increasing penetration depth for frozen soils [47], [48], we assumed that it has only surface scattering as a very dry soil (low dielectric constants). Therefore, the liquid (unfrozen) soil moisture can be retrieved by using the Mironov's model (as used in this study) or other similar dielectric model suitable for frozen soils [49], [50].
As shown in Figs We further applied a normalized method, namely the CCHZ_distance between indices of simulation and observation (DISO) proposed by Hu et al. [51], [52] and Zhou et al. [53], to quantify the degree of improvement. This method is used to comprehensively evaluate the performance of algorithms by calculating the Euclidean distance between normalized metrics. The lower the DISO value, the better the performance of retrievals. The two metrics of R and ubRMSE are input into the CCHZ_DISO method to calculate the DISO values of MTMA_SM H+V , MTMA_SM V , and MTMA_SM H products. It is found that at Pali network with obvious topographic effect, the DISO value of MTMA_SM H+V is significantly lower than that of MTMA_SM H and MTMA_SM V (see Table IV). However, at the networks of Naqu, Maqu, and Wudaoliang with The topographic effect can cause the rotation of the radiation polarization plane, which will lead to the depolarization of the observed TB, and then affect the soil moisture retrieval results [24], [54], such as the case at the Pali. The existing soil moisture retrieval algorithms from spaceborne radiometer observations usually without considering topographic effects and minimize the cost function between simulated and observed TB at each polarization, leading to potential errors over complex terrains. However, this polarization rotation effects may be avoided if the first brightness Stokes parameter is used to retrieve the soil moisture as done in our work.

B. Spatial Distribution Over the Qinghai-Tibetan Plateau
In this study, the MTMA_SM H+V , MTMA_SM H , and MTMA_SM V were obtained over the Qinghai-Tibetan Plateau in 2015-2016 with averages as shown in Fig. 4. It was noted that the spatial distribution of the three soil moisture retrievals was generally consistent and the soil moisture in the southeast was higher than that in the northwest. Moreover, the spatial distribution of soil moisture of the three products also reflected the reasonable spatial variations of soil moisture in different climatic regions of Qinghai-Tibetan Plateau.
In this study, the number of effective retrievals (0.01 m 3 /m 3 ≤ retrieval ≤ field capacity (m 3 /m 3 )) of each pixel on the Qinghai-Tibetan Plateau in 2015-2016 was counted as shown in Fig. 5. The numbers of effective soil moisture retrievals of the three products are relatively small in the marginal areas of the Qinghai-Tibetan Plateau, and the numbers of retrievals are relatively large in the central area of the Qinghai-Tibetan Plateau. It is worth noting that the effective retrievals of MTMA_SM V (about 4.31 * 10 6 ) and MTMA_SM H (about 6.19 * 10 6 ) are fewer than that of MTMA_SM H+V (about 7.12 * 10 6 ). The main reason may be that the topographic effect led to the decreasing of V_pol TB and the increasing H_pol TB, leading to the soil moisture retrieval beyond its physical range.
The topography can change the radiation process of ground TB and cause the change of surface emissivity, and finally lead to the retrieval bias of soil moisture using conventional retrieval algorithm. To evaluate the influence of topographic effect on soil moisture retrievals, soil moisture difference (ΔSM p = (MTMA_SM p )−(MTMA_SM H+V )) between soil moisture retrieved without and with the consideration of topographic effect is calculated as show in Fig. 6. According to the product of 16-day MODIS NDVI and RVI, only regions from Qinghai-Tibetan Plateau with NDVI less than 0.4 and RVI less than 0.3 were extracted respectively, where the vegetation effect on soil moisture retrievals is minimal. The aim is to compare the impact of differences in vegetation information derived from optical and microwave remote sensing data on the subsequent analysis. Although the regions extracted based on NDVI threshold or RVI threshold contains the regions with less than 10°slope range, the regions are less affected by vegetation and could be used to explore the topography effects.   It is found that ΔSM H was generally smaller than 0 and ΔSM V was generally greater than 0 in most areas although the slope in some areas was less than 10°, indicating that soil moisture retrieved based on single H_pol TB was underestimated, while soil moisture retrieved by the single V_pol TB was overestimated. Similarly, Li et al. [24] have found that the topographic effect led to the decreasing of V_pol TB and the increasing of H_pol TB in most areas of the Qinghai-Tibetan Plateau (including areas with the slope <10°), which leads to overestimation or underestimation of soil moisture.
According to the statistical results, the maximum bias of soil moisture of MTMA_SM H and MTMA_SM V even exceeds 0.1 m 3 /m 3 in some areas with large terrain fluctuation (the elevation difference exceeding 1000 m and the slope exceeding 35°), which is far beyond the expected accuracy (0.04 m 3 /m 3 ) of SMOS satellite.

A. Comparison With SMOS Products
In this study, we also compared the new soil moisture retrievals considering topographic effects with the values from SMOS_L3 and SMOS_IC soil moisture products. In order to match the spatial resolution (25 km) of SMOS_L3 and SMOS_IC soil moisture products, the retrieved soil moisture with the spatial resolution of 15 km using MTMA method were resampled to 25-km spatial resolution.
According to the Fig. 7 and statistical metrics in Table V, the retrievals of MTMA_SM H+V at four networks are in better agreement with the in-situ data than SMOS_L3 and SMOS_IC products. The overall correlation coefficient (R = 0.828) is higher than that of SMOS_IC (R = 0.626) and SMOS_L3 (R = 0.619) products. The overall ubRMSE (0.041 m 3 /m 3 ) of the MTMA_SM H+V at all validation networks is also lower than that of SMOS_IC (ubRMSE = 0.057 m 3 /m 3 ) and SMOS_L3 (ubRMSE = 0.085 m 3 /m 3 ) products, which is closer to the expected accuracy of the SMOS mission. Therefore, the quantitative validation results of MTMA_SM H+V confirmed that the performance of soil moisture retrievals of the method proposed in this study are significantly better than current SMOS_L3 and SMOS_IC soil moisture products, indicating the advantage of the proposed method to reduce the topographic effects where terrain relief is strong, like in the Qinghai-Tibetan Plateau.

B. Impact of the Vegetation Effects
When the depth of vegetation layer increases, the signal from the soil layer may not be observed by the microwave sensor onboard satellites due to low penetration through the vegetation layer. Moreover, dense vegetation might significantly weaken the angle-dependent variability of the observed TB and lead to less degree of information to be used to retrieve the effective vegetation optical depth. The MTMA method used in this study is sensitive to the variability of TB with incidence angles. Therefore, in dense vegetation areas, the uncertainty of retrieved vegetation optical depth is relatively larger, since the difference in TB between different observation angles was small because of significant vegetation effects [32].
For most areas with dense vegetation cover over large terrain relief on the Qinghai-Tibetan Plateau, the microwave radiation was affected by both topography effects and vegetation effects (not shown in this study). Li et al. [24] found that the bias of TB (H_pol or V_pol) caused by topographic effects (slope and aspect) is about 10-15 K through simulation experiments. Talone et al. found that the error in TB caused by the vegetation layer also exceeded 10 K [29]. Vegetation has the depolarization effect similar as the topographic effect and it can be more significant than that from the topographic effect. The uncertainty of soil moisture retrieval associated with vegetation effects might lead to an obscuration of topography effects. For example, the overestimation or underestimation of vegetation effects at H_pol or V_pol may lead to overestimation or underestimation of soil moisture. Therefore, for densely vegetated areas, it could be very complex to distinguish vegetation effects and topography effects, and the dense vegetation areas are out of the scope in this study.

VI. CONCLUSION
Passive microwave remote sensing can provide important tool for soil moisture monitoring in mountainous areas. In this study, we proposed a methodology to apply the first brightness Stokes parameter (the sum of H_pol and V_pol TB) in the MTMA method to reduce the influence of topographic effects on soil moisture retrieval.
According to the validation results using in-situ data at the four ground networks (Pali, Naqu, Maqu, and Wudaoliang networks), it is found that soil moisture retrievals using the first brightness Stokes parameter are in better agreement with in-situ measurements (the R > 0.75 and ubRMSE < 0.04 m 3 /m 3 ) compared to the single-polarization retrievals at Pali network where severely topographic effects exist. In the other three networks with flatter terrain, the performance of soil moisture retrievals using the first brightness Stokes parameter is still comparable with that using single-polarization TB observations. In addition, by analyzing the soil moisture retrievals on the Qinghai-Tibetan Plateau for the conditions with less vegetation effects (NDVI < 0.4 or RVI < 0.3), it was found that the soil moisture using H_pol TB is generally underestimated and that of using V_pol TB is generally overestimated, which is expected due to the depolarization effect caused by the topography. However, this method may need a further detailed review in densely vegetated areas, as the effects of vegetation may obscure the effects of topography. In general, our method applying the first brightness Stokes parameter (the sum of H_pol and V_pol TB) in the MTMA method is a simple and feasible solution to deal with the topographic effects, and the results showed a considerable improvement in mountainous areas such as the Qinghai-Tibetan Plateau.