Soil Moisture Retrievals by Combining Passive Microwave and Optical Data

Abstract

This paper aims to retrieve the temporal dynamics of soil moisture from 2015 to 2019 over an agricultural site in Southeast Australia using the Soil Moisture Active Passive (SMAP) brightness temperature. To meet this objective, two machine learning approaches, Random Forest (RF), Support Vector Machine (SVM), as well as a statistical Ordinary Least Squares (OLS) model were established, with the auxiliary data including the 16-day composite MODIS NDVI (MOD13Q1) and Surface Temperature (ST). The entire data were divided into two parts corresponding to ascending (6:00 p.m. local time) and descending (6:00 a.m. local time) orbits of SMAP overpasses. Thus, the three models were trained using the descending data acquired during the five years (2015 to 2019), and validated using the ascending product of the same period. Consequently, three different temporal variations of the soil moisture were obtained based on the three models. To evaluate their accuracies, the retrieved soil moisture was compared against the SMAP level-2 soil moisture product, as well as to in-situ ground station data. The comparative results show that the soil moisture obtained using the OLS, RF and SVM algorithms are highly correlated to the SMAP level-2 product, with high coefficients of determination (R2OLS = 0.981, R2SVM = 0.943, R2RF = 0.983) and low RMSE (RMSEOLS = 0.016 cm3/cm3, RMSESVM = 0.047 cm3/cm3, RMSERF = 0.016 cm3/cm3). Meanwhile, the estimated soil moistures agree with in-situ station data across different years (R2OLS = 0.376~0.85, R2SVM = 0.376~0.814, R2RF = 0.39~0.854; RMSEOLS = 0.049~0.105 cm3/cm3, RMSESVM = 0.073~0.1 cm3/cm3, RMSERF = 0.047~0.102 cm3/cm3), but an overestimation issue is observed for high vegetation conditions. The RF algorithm outperformed the SVM and OLS, in terms of the agreement with the ground measurements. This study suggests an alternative soil moisture retrieval scheme, in complementary to the SMAP baseline algorithm, for a fast soil moisture retrieval.

Spatial Gap-Filling of SMAP Soil Moisture Pixels Over Tibetan Plateau via Machine Learning Versus Geostatistics

“…7(b) illustrates that the TB and NDVI play important roles in the gap-filling results. This confirms the sensitivity of SMAP TB to soil moisture [21], and also the close correlation between soil moisture and NDVI [47].…”

“…This confirmed the previous findings in [55] that soil moisture was closely related to the MODIS NDVI value over the Tibetan Plateau, and the interaction between the soil moisture and vegetation was a bidirectional process. This is in line with the use of vegetation index in optical remote sensing for soil moisture retrieval and reconstruction [21,50,56,57].…”

“…The Random Forest (RF) approach as an ensemble machine learning algorithm, has been widely used in soil moisture retrieval and reconstruction [35,36]. Previous studies compared the performances of different machine learning approaches in the retrieval of soil moisture, indicating that the RF model obtained robust results [21,37,38]. The idea of RF is to train multiple decision trees for prediction.…”

“…The ascending data were for training, and the missing pixels in descending orbit due to the failed retrieval from the physical radiative transfer model were reconstructed via the developed estimator. In comparison with [21] which evaluated the potential of the Random Forest algorithm for soil moisture retrieval over a semi-arid agricultural region with a single land cover, the current paper extends the algorithms to fill the SMAP SM missed pixels under complex land cover and high altitude conditions. Particularly, the performances of the gap-filling algorithm are compared to the traditional geostatistical approaches.…”

“…Zhang et al [20] proposed an artificial neural network algorithm to reconstruct the missed data of ESA CCI from 1982 to 2015 across China by using numerous environmental variables as training data. In our previous work, we compared the Ordinary Least Squares, Support Vector Machine, and Random Forest (RF) methods to estimate soil moisture from SMAP brightness temperature [21]. The estimated SM maintained high consistency with SMAP SM, and the RF model was found to obtain the best performance.…”