Selection of optimal bands of AVIRIS – NG by evaluating NDVI with Sentinel-2

The AVIRIS-NG hyperspectral data consists of continuous spectral bands with low bandwidth, Sentinel-2 multispectral image has less number of bands with higher bandwidth. Several studies are carried out to calculate the Normalized Difference Vegetative Index (NDVI) of hyperspectral data. The studies considered a single band in the red and NIR region of hyperspectral data. In this present study, NDVI analysis is carried out by taking the mean reflectance of red and NIR wavelength region bands of the AVIRIS image. To choose the bands, a methodology is devised for AVIRIS image by analyzing and evaluating the NDVI between AVIRIS and Sentinel-2 image. The AVIRIS data consists of 7 red bands and 22 NIR bands. Root Mean Square Error (RMSE) between NDVI of all the 154 combinations of AVIRIS image bands and Sentinel-2 image is calculated for each Land Use Land Cover (LULC). Three mean NDVI are evaluated such as (i) mean of all bands reflectance; (ii) mean of band reflectance higher than [Mean + Standard deviation] (iii) the mean of the band reflectance involved lower than [Mean-Standard deviation] are compared using RMSE and linear regression. The bands that contribute to lower RMSE values are chosen and separated into band sets. The distinct sets of NDVI on various classes are validated with the existing studies and the other field data to check the RMSE and correlation. The proposed statistical sets are performed better than the existing models on various classes of land use and land cover.