Assessing wheat residue cover with hyperspectral remote sensing

摘要

Hyperspectral remote sensing can aid in discriminating crop residue owing to the ability of narrow bands to capture the unique absorption feature of soil and residue. The present study was carried out to find out the suitable narrow spectral bands and hyper-spectral indices for discriminating wheat residue (stubble and burnt). Ground spectra of wheat residue and the adjoining soil were collected using the ASD fieldspecTM spectroradiometer. The best spectral range was derived using the Stepwise Discriminating Analysis (SDA). 'F'statistics from one-way ANOVA was used to find out the best index for discriminating wheat residue from soil. EOl-Hyperion data over Anand-Borsad region of Gujarat state in India was acquired free of cost from USGS earth explorer website to apply the field based result over the Hyperion scene. Spectral Angle Mapper (SAM) classification scheme was used to generate the wheat residue cover over the Hyperion scene. Among the hyperspectral indices evaluated for this study the Cellulose Absorption Index (CAI) was found to be the best and hence CAI was used to classify the Hyperion scene for discriminating crop residue in field and also the burnt wheat residue. Results indicated that the wave bands at 10 nm width in the SWIR spectral region specifically from 1500-1700nm and 1900 to 2300nm are most suitable for wheat residue discrimination. The SAM classification technique is suitable for classifying the wheat residues with an overall accuracy of around 80 % whereas classification based on CAI could be used successfully to identify both wheat stubble and the burnt residues. This study concluded that wheat residue can be mapped for a large area with an accuracy of 80 % using the space borne hyperspectral data with.