Correlation between ground measured soil moisture and RADARSAT-1 derived backscattering coefficient over an agricultural catchment of Navarre (north of Spain)

Surface soil moisture is a variable of great importance in several agronomic, hydrological and meteorological processes. The knowledge of its magnitude and its spatial distribution is essential to adequately describe and model the processes where it is involved. Radar remote sensors measure microwave energy backscattered by natural surfaces. This scattered energy depends on the geometrical and the dielectric properties of surfaces. In the case of bare soil surfaces, the dielectric properties are directly related to the soil water content, so theoretically, radar remote sensing allows the extraction of spatially distributed soil moisture information. However, the influence of surface roughness on the scattering process limits the ability to correctly estimate volumetric soil moisture values unless detailed roughness measurements are acquired. The present paper reports the results of a study where five images from the remote radar sensor on the satellite RADARSAT-1 were processed and correlated to ground measured soil moisture values over an agricultural catchment. Linear regression models were fitted between RADARSAT-1 derived backscattering coefficient sigma0 and the soil moisture at different spatial scales: point scale, field scale and catchment scale. Three soil moisture classes were identified according to their implications for crop growth: (1) low moisture values which contributed to water stress in plants; (2) medium moisture content that allowed an optimal crop growth; and (3) high moisture values which affected crop growth by other means, such as by fungal disease. Results show a direct relation between sigma0 and the soil moisture. At the catchment scale the observed correlation is high. At detailed scales, however, variability increases, causing a decrease of correlation values. The ability of sigma0 to discriminate between the considered moisture classes seems to be adequate. The accuracy of the estimation increases from the detailed to coarser scales. In the case of vegetated fields, the vegetation cover can cause a certain attenuation of the radar pulse resulting in reduced sigma0 values. In this research, vegetation-induced attenuation was considered by applying the semi-empirical 'Water Cloud' model. The presented technique is useful for crop growth monitoring and modelling at medium to large scales, particularly in early growing seasons where the attenuation of vegetation is not too high. It is also applicable to irrigation planning or crop health studies. However, regression lines are site specific and can be affected by the surface roughness variability and vegetation cover of fields.