Defining Optimum Narrow Bands and Band-Widths for Agricultural Applications
- Paper number
IAC-07-B1.I.03
- Author
Dr. Shibendu S. Ray, Space Applications Centre (SAC), ISRO, India
- Coauthor
Dr. Namrata Jain, Space Applications Centre (SAC), ISRO, India
- Coauthor
Ms. Anshu Miglani, Space Applications Centre (SAC), ISRO, India
- Coauthor
Dr. J.P. Singh, India
- Coauthor
Dr. A.K. Singh, Indian Agricultural Research Institute, India
- Coauthor
Dr. Sushma Panigrahy, Space Applications Centre (SAC), ISRO, India
- Coauthor
Mr. J.S. Parihar, Space Applications Centre (SAC), ISRO, India
- Year
2007
- Abstract
Hyperspectral remote sensing data has potential in crop stress detection, biophysical parameter retrieval and soil variability assessment. As hyperspectral data comes with large number of bands, with high redundancy, it is essential to define the optimum smaller set of narrow bands and the required bandwidth for various agricultural applications. This is particularly important in view of Indian space programme envisaging launching a space-borne hyperspectral sensor in near future.
In order to capture the variability in agriculture, the study included major kharif (rainy) and rabi (winter) crops like, rice, wheat, maize, pearl millet, gram, soybean, mustard and cotton. Different agronomic treatments such as, irrigation, fertilizer, variety and date of sowing were also studied. Soils with significantly different nutrition parameters were included in the analysis. 10-15 spectral reflectance profile observations were collected for each crop/treatment/soil type, using ASD handheld spectroradiometer (325-1075nm). A stepwise discriminant analysis (SDA) was carried out to identify best bands based on statistics like F values, Wilks’ Lamda and canonical correlation. The measured reflectance was aggregated to different bandwidths (3, 5, 10, 15, 20, 25 and 30nm) for selection of optimal bandwidth. The wavelength position and reflectance values were compared by computing RMSE with the original spectra at different specific locations in the total spectral region covered.
The study could identify 20 optimum bands in VNIR (400-1050 nm) region for crop and soil assessment, by combining the bands selected from individual discriminant analysis. These included bands in ultraviolet (2), blue (2), green (3), red (3), red edge (4), NIR (4) and moisture sensitive NIR (2) region. The physiological significance of these bands has been defined in the paper. The optimum bandwidth required for crop stress discrimination differed for different wavelength regions. It is required to have narrow bandwidth (5-10 nm) in red edge and early NIR region. In 500-700 and 800-900 nm regions, bandwidth better than 15 nm was found to be optimum.
The study established the optimum hyperspectral bands and bandwidths in the 400-1050 nm region for crop separation, crop stress discrimination and soil variability analysis. However, there is a need to extend this study to analyse the band specification required for biophysical parameter retrieval.
- Abstract document
- Manuscript document
IAC-07-B1.I.03.pdf (🔒 authorized access only).
To get the manuscript, please contact IAF Secretariat.