Crop-type identification potential of Radarsat-2 and MODIS images for the Canadian prairies

Abstract

Owing to their high-frequency revisit and weather independence with high image resolution, Moderate Resolution Imaging Spectroradiometer (MODIS) and Radarsat-2 SAR (ScanSAR (synthetic aperture radar)), respectively, provide data suitable for regional-level crop-type identification in the Canadian prairies. The challenge remains in optimally combining data from the two sources, to identify crop types in individual fields. This study investigated an approach based on image fusion and a specially designed classification to obtain a result with the high spatial detail of ScanSAR and the spectral information from MODIS. The methodology employs a wavelet-IHS (intensity, hue, and saturation) combined image fusion method to enhance the spatial resolution of the MODIS data using ScanSAR data, followed by a multiresolution segmentation process supported by a road network database to generate the final classification. The fusion-classification approach yielded a result suitable for both visual and digital analysis. The overall classification accuracy of the fused data set was about 72%, higher than accuracies achieved for ScanSAR images (transformed as principal components), the MODIS data alone, or a combination of the ScanSAR principal components and MODIS data. While further investigation is warranted, this approach appears to have the attributes required for operational crop-type identification in situations where such information is required frequently and over large areas.