Observations of Cutting Practices in Agricultural Grasslands Using Polarimetric SAR

Abstract

In this study, TerraSAR-X dual-polarimetric HH/VV and RADARSAT-2 fully polarimetric synthetic aperture radar (SAR) parameters are compared to results from field surveys of grasslands in order to establish a methodology for the detection of grass cutting events. The experiment over grasslands in Estonia during the vegetative season of 2013 was carried out with an extensive survey measuring grass height, wet and dry biomass, and soil moisture. An entropy/alpha decomposition was applied to the data. Additionally, the polarimetric coherences between the different channels, backscatter levels, and intensity ratios were analyzed. From the numerous polarimetric parameters studied, HH/VV polarimetric coherence and the scattering entropy seemed to provide the most reliable indication about a cutting event based on the polarimetric SAR (PolSAR) time series. The behavior was more pronounced in TerraSAR-X data than in RADARSAT-2 data, possibly due to the finer temporal sampling (11 days vs. 24 days for exactly the same imaging geometry) and shorter wavelength. After the grass was cut, TerraSAR-X HH/VV polarimetric coherence magnitude decreased by 0.27 (by 0.12 on average) and scattering entropy increased by 0.21 (by 0.07 on average). The observed behavior can be well explained by field observations and modeling of the vegetation backscattering. According to a vegetation particle model, growing grass (mainly vertically oriented dipoles) corresponds to lower entropy and higher HH/VV polarimetric coherence magnitude than cut grass (more horizontally and more randomly oriented dipoles). This indicates the potential to use HH/VV polarimetric coherence magnitude and scattering entropy for the monitoring of grassland cutting practices.