SOIL MOISTURE ESTIMATION BY USING A NEW EMPRICAL PARAMETER TO MODEL ROUGHNESS EFFECT FROM RADAR BACKSCATTER COEFFICIENT

Reliable estimation of the upper part of surface soil moisture (within 10 cm) plays a fundamental role in different applications such as hydrology, agriculture, risk prediction, meteorology and climatic studies [1,2]. Many studies have shown that aerial and satellite images have high potential for soil moisture estimation [3, 4]. The active microwave Synthetic Aperture Radar (SAR) sensors are of primary interest for assessing soil moisture because of their all-weather capability, signal penetration depth through natural media, and sensitivity to surface variables (such as water content and roughness) that are difficult to estimate using optical remote sensing sensors [5]. Different studies have shown that radar backscattering coefficient (σ^0) for bare soil is highly dependent on surface roughness and soil moisture due to high dependence of the microwave dielectric constant on soil water content [6]. Surface roughness and soil moisture are typically estimated with empirical, semi-empirical or theoretical models. These models relate the measured backscatter coefficient (σ^0) into volumetric soil moisture and surface roughness [5], [7–9]. Semi-empirical models are dependent on the site and surface type on which they were developed and tested. Moreover, they have been obtained for limited ranges of incidence angle, wavelength, and soil parameters [6, 8]. Therefore, they are not able to estimate soil moisture precisely in all deferent conditions unless we calibrate them again for our test field. Many studies reveal that theoretical models are not precise either [10, 11]. Although the theoretical models are site-independent,They have limited roughness domain, and the precision of their results highly depend on precision of roughness parameters (such as correlation length or correlation function) [12, 13]. In this paper, we propose a new empirical parameter to model the surface roughness effect on radar backscattering coefficient (σ^0 ). In many studies, the effect of the correlation length (l) on the backscattering coefficient (σ^0 ) has been ignored, and just the root mean squared (rms) height has been considered [5, 6, 8, 9].