Synergistic Use of Radar and Optical Image Data for Improved Land Use and Land Cover Assessment: A Case Study in the North of Entre Rios Province (Argentina)

Abstract

The synergistic use of optical and radar data is already a well-known alternative in the literature for land cover characterization. The objective of this chapter is to quantify the added value of combining radar imagery from Sentinel-1 and multispectral imagery from Sentinel-2 (both at 10 m resolution) to provide information on land use and land cover change (LULC) in 2016–2017 and 2020. The Sentinel-1 image data included the Global Backscatter Model (S1GBM) for the 2016–2017 wet period, and for the driest year 2020, which were sourced from the Google Earth Engine (GEE) platform. The Sentinel-2 Global Mosaic (S2GM) service provided surface reflectance mosaic products for the same years. Sentinel-2 data were compared to derived radiometric indices and combined with Sentinel-1 imagery. The LULC classes considered for this study are three classes of Espinal ecotone (closed gallery forest, mid to open gallery forest, open low forest, and shrubland) and four classes of agricultural land defined by soil degradation processes (slight soil water erosion, moderate saline and slight soil water erosion, slight to moderate soil water erosion, and moderate to severe soil water erosion). Results show that σ0VV and σ0VH backscatter values are 1.0 to 1.8 dB lower during the 2020 drought compared to values in 2016–2017. Both σ0VV and σ0VH polarizations and the Radar Vegetation Index combined with selected optical radiometric indices for soil, vegetation, and moisture from Random Forest analysis are suitable for representing LULC changes in years with changes in moisture availability. The results showed that a significant change in LULC patterns had occurred in the driest year, 2020, in the study area.