SPATIAL DISTRIBUTION MAPPING OF NITROGEN CONTENT OF ALPINE PASTURE OF WESTERN HIMALAYA USING HYPERSPECTRAL REMOTE SENSING

摘要

Nitrogen is one of the most important elements for vegetation health, a change in the concentration of which can affect the entire ecosystem. With the changing climate, it is expected that all the natural systems will be directly or indirectly affected. Alpine pastures are predicted to be the most sensitive ecosystem affected by climate change. This study attempts to develop a regression model to relate hyperspectral data with the vegetation nitrogen mass of alpine pastures of Himalaya, which is rich in biodiversity and medicinal plant treasure. Hyperion, moderate spatial resolution hyperspectral satellite data was used in the present study. Various preprocessing techniques were used to attain the reflectance satellite data. Alpine pasture region was extracted from the study area using Support Vector Machine (SVM). It was found to have accuracy of 86% and Kappa accuracy of 0.83. In the extracted alpine pasture region, field survey was carried out and in situ data (spectral signature, vegetation samples) was collected. Vegetation samples collected from the field was analyzed for nitrogen estimation using CHNS (Carbon, Hydrogen, Nitrogen and Sulphur) analyzer. Nitrogen was related with the vegetation spectra. Continuum removal technique was used for enhancement of nitrogen absorbance wavelengths. It was found that red and red edge region of electromagnetic spectra were found to be highly correlated with nitrogen content. Using this information, various vegetation indices were used to correlate the nitrogen content estimated from field. Modified Red Edge Normalized Difference Vegetation Index (MRENDVI) was found to have highest correlation of nitrogen mass with field observation and satellite data. The wavelengths used for the index were at 752nm, 701nm and 447nm of Hyperion image. The regression model developed using MRENDVI and nitrogen mass was y = 3.18x - 0.62. Using this relation, nitrogen distribution map was prepared with nitrogen mass ranging from 2 kg/ha to 12 kg/ha. The alpine pasture areas which were adjacent to the barren land had lower nitrogen concentration, whereas the areas near temperate forests had higher nitrogen concentration. The study infers that red edge region is sensitive to nitrogen in the plant and can be used to determine the health of vegetation. It can also be suggested that high spatial resolution satellite data with red edge band can be an alternative to the hyperspectral satellite data with vegetation growth monitoring.