Establishing an Empirical Model for Surface Soil Moisture Retrieval at the U.S. Climate Reference Network Using Sentinel-1 Backscatter and Ancillary Data

Abstract

Progress in sensor technologies has allowed real-time monitoring of soil water. It is a challenge to model soil water content based on remote sensing data. Here, we retrieved and modeled surface soil moisture (SSM) at the U.S. Climate Reference Network (USCRN) stations using Sentinel-1 backscatter data from 2016 to 2018 and ancillary data. Empirical machine learning models were established between soil water content measured at the USCRN stations with Sentinel-1 data from 2016 to 2017, the National Land Cover Dataset, terrain parameters, and Polaris soil data, and were evaluated in 2018 at the same USCRN stations. The Cubist model performed better than the multiple linear regression (MLR) and Random Forest (RF) model (R2 = 0.68 and RMSE = 0.06 m3 m-3 for validation). The Cubist model performed best in Shrub/Scrub, followed by Herbaceous and Cultivated Crops but poorly in Hay/Pasture. The success of SSM retrieval was mostly attributed to soil properties, followed by Sentinel-1 backscatter data, terrain parameters, and land cover. The approach shows the potential for retrieving SSM using Sentinel-1 data in a combination of high-resolution ancillary data across the conterminous United States (CONUS). Future work is required to improve the model performance by including more SSM network measurements, assimilating Sentinel-1 data with other microwave, optical and thermal remote sensing products. There is also a need to improve the spatial resolution and accuracy of land surface parameter products (e.g., soil properties and terrain parameters) at the regional and global scales.

Soil Moisture Estimation Using Citizen Observatory Data, Microwave Satellite Imagery, and Environmental Covariates

“…The linear regression model provided a very realistic SM distribution over the landscape that matches with the known driving forces of the SM distribution, where the RMSE ranged from 3.77 to 5.85. The results match with the ones published by [29]. The R 2 values ranged between 0.19 and 0.35, which are quite low, but more or less match the literature results at this scale.…”

“…The R 2 values ranged between 0.19 and 0.35, which are quite low, but more or less match the literature results at this scale. This was similar to [29], who reported an R 2 of 0.24. It was mainly because the model was overestimating the low values and underestimating the high ones and therefore decreased the range of the estimated values down to approximately 60% of the measured values.…”

“…Many SAR analysis and modelling studies have relied on the coefficient of determination (i.e., R 2 ) statistics, and in some cases, considering R 2 = 0.30 as the threshold criteria for accepting a given model for reliable estimation and prediction [20]. The corresponding literature of the use of SAR data for SM estimation reported R 2 values of 0.24 [29], 0.4 to 0.58 [23], and up to 0.71 [24]. However, the study setups-climate, vegetation cover and the used covariates, the accuracy and reliability of the used sensors-are so different, and it is not easy to compare the numbers alone.…”

“…The results showed a root mean square error (RMSE) value of between 2.669 and 2.701 over five months (December to April). Chatterjee et al [29] tested two empirical models: A multiple linear regression model against the cubist model using in situ soil sensor SM measurements (used as the model response), Sentinel-1 data, and various ancillary data (USGS DEM, Sentinel 1, US National Land Cover Database, and the Probabilistic Remapping of Soil Survey Geographic (POLARIS) dataset). The statistical metric used to determine the model was R 2 .…”

Estimation of surface soil moisture by integrating environmental data and remote-sensing satellites

“…The results revealed a root mean square error (RMSE) value of between 2,669 and 2,701 (December to April) for five months. Two analytical models were tested by Chatterjee et al [29]: a multiple linear regression model against the cubist model using SM measurements in situ soil sensor, Sentinel-1 data, and separate ancillary data (USGS DEM, Sentinel 1, US National Ground Cover Database, and Soil Survey Spatial Probabilistic Remapping (POLARIS) data set). Coefficient of determination (R 2 ) was the mathematical metric used for evaluating the model.…”