Evaluation of soil erosion protective cover by crop residues using vegetation indices and spectral mixture analysis of multispectral and hyperspectral data

Abstract

Crop residues are efficient in reducing erosion and surface water runoff on agricultural soils. Evaluating the crop residue cover fraction and its spatial distribution is important to scientists involved in the modelling of soil erosion and surface runoff, and also to authorities wishing to assess soil conservation adoption by farmers. This study focuses on the evaluation of four remote sensing techniques to estimate the cover fraction of cereal crop residues (i.e., wheat and corn) from multispectral and hyperspectral measurements. These are the Soil Adjusted Corn Residue Index (SACRI), the Crop Residue Index Multiband (CRIM), the Normalized Difference Index (NDI) and the spectral mixture analysis technique (SMA). Field campaigns that were carried out by the FLOODGEN project in Sainte-Angèle-de-Monnoir, Québec, Canada and in the Pays-de-Caux located in the Normandy region of France, allowed us to gather digital photographs, spectra and other measurements to determine the actual ground cover fraction. A linear regression analysis between results derived from Landsat-5 TM simulated field spectra and the actual ground cover fractions showed best results for the CRIM on the Ste-Angèle-de-Monnoir study site ( R 2 = 0.96), and equally good results for the Pays-de-Caux study site ( R 2 = 0.94). Results were not as good when SMA was applied to the same Landsat-5 TM simulated field spectra with R 2 values of 0.70 and 0.68 for both sites, respectively. However, results improved significantly when SMA was applied to the hyperspectral data in which case the R 2 values increased to 0.92 for the Sainte-Angèle-de-Monnoir site and 0.89 for the Pays-de-Caux study site. Results obtained with the NDI and SACRI from both simulated TM and hyperspectral field spectra were not conclusive.