Combining Machine Learning and Compact Polarimetry for Estimating Soil Moisture from C-Band SAR Data

Abstract

This research aimed at exploiting the joint use of machine learning and polarimetry for improving the retrieval of surface soil moisture content (SMC) from synthetic aperture radar (SAR) acquisitions at C-band. The study was conducted on two agricultural areas in Canada, for which a series of RADARSAT-2 (RS2) images were available along with direct measurements of SMC from in situ stations. The analysis confirmed the sensitivity of RS2 backscattering (σ°) to SMC. The comparison of SMC with the compact polarimetry (CP) parameters, computed from the RS2 acquisitions by the CP data simulator, pointed out that some CP parameters had a sensitivity to SMC equal or better than σ°, with correlation coefficients up to R ≃ 0.4. Based on these results, the potential of machine learning (ML) for SMC retrieval was exploited by implementing and testing on the available data an artificial neural network (ANN) algorithm. The algorithm was implemented using several combinations of σ° and CP parameters. Validation results of the algorithm with in situ observations confirmed the promising capabilities of the ML techniques for SMC monitoring. Furthermore, results pointed out the potential of CP in improving the SMC retrieval accuracy, especially when used in combination with linearly polarized σ°. Depending on the considered input combination, the ANN algorithm was able to estimate SMC with Root Mean Square Error (RMSE) between 3% and 7% of SMC and R between 0.7 and 0.9.

A Review of Irrigation Information Retrievals from Space and Their Utility for Users

“…The S1 satellite constellation (S1-A and S1-B) offers an unprecedented (for non-commercial satellites) data coverage with a time resolution of 6 days and a pixel size of 10 × 10 m in two polarizations, VV (vertically transmitted, vertically received) and VH (vertically transmitted, horizontally received), available free of cost. Different approaches have been proposed to map irrigated areas by considering the multi-temporal information from S1 backscatter observations to detect typical signal variations of irrigated areas [88][89][90][91][92]. Gao et al [93] proposed an approach based on the direct analysis of the multi-temporal radar signal on each agricultural plot through different metrics (mean, standard deviation, correlation length, fractal dimension).…”

“…S1 also enabled the retrieval of high-resolution soil moisture estimates, which are vital for irrigation management. Approaches are based on machine learning approaches, including neural networks [88,89], change detection techniques [90,91], and also on direct inversion approaches of physical or semi-empirical models [92]. Typically, the estimates allow an accuracy of the order of 5% in volumetric soil moisture.…”

Validation of the SMOS Level 1C Brightness Temperature and Level 2 Soil Moisture Data over the West and Southwest of Iran

“…In recent years, this issue has been overcome using microwave satellite sensors, both active and passive, onboard operational satellites, for measuring soil moisture in different environmental conditions under various vegetation covers [11,12]. L-band (1-2 GHz) microwave radiometry is the most suitable approach to retrieve surface soil moisture [1,13] because it is very sensitive to soil moisture sensing.…”

“…Synthetic aperture radar (SAR) provides medium to high resolution mapping of soil moisture in all weather conditions. Sentinel-1 mission provides SAR data at C-band that can be used to retrieve and map temporal changes of soil moisture underneath vegetation cover [12,19,20].…”

A New Reflectivity Index for the Retrieval of Surface Soil Moisture From Radar Data

“…So-called active radar missions involve the use of synthetic aperture radar (SAR) data and low-resolution scatterometers. Methods based on the use of SAR data are generally applied at the scale of agricultural fields [16][17][18][19][20][21][22][23][24][25][26] or at scales close to 1 km resolution [27][28][29]; in recent years, they have become more consistent and operational thanks to the arrival of the Sentinel-1 Copernicus constellation [28][29]. In this context, there are three main approaches to the inversion of radar signals: one is based on direct inversion of physical models [30][31][32], a second is based on statistical techniques such as neural networks [33][34][35][36], and the other is based on the use of change detection algorithms [37][38][39][40].…”