Modified Campbell-Shepard Model to Retrieve Surface Roughness in Paddy Fields Using Radarsat-2 Data and Field Measurements

Abstract

Remote sensing technologies provide spatial data for quantifying soil moisture and surface roughness by measure a variety of surface parameters at meso - to macro-scales. It has long been recognized that satellite imagery has a unique and important role in monitoring crops and soil conditions for farm management. Traditional methods in soil physics are based on point measurements. The most critical factor, besides the enormous time and working effort, is the extrapolation of point samples to catchment (or sub-regional) scale processes. To increase the efficiency in spatial mapping of surface roughness, we evaluate the performance of Dubois and Campbell-Shepard (CS) and modify them by fitting with field measurement termed as Modified Campbell-Shepard (MCS) models to obtain surface roughness based on Radarsat-2 quad polarimetric SAR. The surface roughness was modeled using the quad-polarized backscattering intensities and fitted to surface roughness measurements at 77 sample sites. The surface roughness at the field was measured by a pin meter with a 30 cm length profile. The measurements were synchronized with Radarsat-2 acquisition time in June, August, September, and October 2014. According to the curve fitting between the model and field data, the R2 and RMSE of the Dubois model were about 0.32 and 0.37 cm, and the CS model were about 0.47 and 0.13. Meanwhile, the MCS model produced about 0.62 and 0.11 cm, respectively. The high R2 with low RMSE of the MCS originated from ground roughness independently to the biomass roughness above the ground. Therefore, the model could be used for further analyses such as soil moisture estimation in paddy field areas even though it uses a sensor with C-band frequency.