Mapping rice extent and cropping scheme in the Mekong Delta using Sentinel-1A data

Abstract

Synthetic aperture radar (SAR)-based multi-temporal backscatter analysis is a common approach that has been widely used for rice mapping. Co-polarized C-band microwave backscatter (HH or VV) and the co-polarization ratio (HH/VV) are the most commonly used data sets for rice mapping due to their high data availability while the utilization of cross-polarized backscatter (HV or VH) has received less attention. In this study, Sentinel 1A time series – acquired in the dual-polarized (VV/VH) Interferometric Wide (IW) swath mode during the spring growing season (October 2015 to March 2016) in the Mekong Delta – were used to analyse the relationship between the growing cycle of rice plants and the temporal variation of SAR backscatter at different polarizations. Results show that VH backscatter is more sensitive to rice growth than VV backscatter. Several vegetation phenological parameters including beginning date, heading date and the length of the growing season were extracted from the VH backscatter time series. A decision tree approach was applied to delineate rice-cultivated areas based on seasonal phenological parameters. The classification result was validated against a 2015 land use map. The overall classification accuracy is 87.2% (kappa coefficient – κ = 0.71). In addition, the SAR-derived rice area was compared against ground statistical data at the provincial level (coefficient of determination R² = 0.98).