Validation of GPS-IR Soil Moisture Retrievals: Comparison of Different Algorithms to Remove Vegetation Effects

摘要

The GPS interferometric reflectometry (GPS-IR) technique can be used to estimate near-surface soil moisture from signal-to-noise ratio (SNR) data collected with standard geodetic instrumentation. However, the effects of vegetation on GPS-IR soil moisture retrievals must be considered in some environments. In situ soil moisture observations from 11 GPS sites are used to compare the performance of three different retrieval algorithms that represent vegetation effects with different degrees of complexity. A bare-soil retrieval algorithm does not perform well, even at sites where seasonal variations in vegetation water content (VWC) are less than 1 kg m-2. The range of volumetric soil moisture (VSM) is too large due to the effects of vegetation on phase of the SNR interferogram, yielding an RMSE between in situ and GPS-IR VSM of 0.055 cm3 cm-3. Errors are reduced by an algorithm that adjusts for vegetation effects using variations in the amplitude of the SNR interferogram. RMSE is 0.038 cm3 cm-3 using this algorithm, below the typical limit required for validation of satellite data. This simple vegetation algorithm performs poorly at sites where seasonal variations in VWC are 1 kg m-2 or greater. A more complex algorithm, that uses amplitude in conjunction with frequency analysis of the SNR interferogeram to predict vegetation effects, provides acceptable performance at these sites ( RMSE = 0.039 cm3 cm-3). The additional complexity of this algorithm is only warranted at sites where the simple vegetation algorithm cannot adequately represent the effects of the vegetation fluctuations.