EVALUATION OF SPECTRAL BIOPHYSICAL PARAMETERS FOR A FEW AGRICULTURAL CROPS

Abstract

With the recent development in the spectral and spatial characteristics of sensor, there has been an increasing interest in agricultural spectral properties and mapping of land use/ land cover in agricultural areas using remote sensing data. Due to increased availability and affordable cost of hyperspectral data, one envisages its widespread use for various agricultural applications. Thus, there is a need to establish not only the potential but also the limitations of hyperspectral data for land cover mapping, biophysical and biochemical parameters of agricultural crops in natural agroecosystems. Further, the possibilities of hyperspectral imaging for the extraction of information relevant to agriculture as well as vegetation demands a detailed understanding of spectral signatures in terms of position of feature specific absorption bands, shape of the spectrum, spectral variability and similarity of various types of vegetation species. In particular, credible information on the variability and uniqueness of spectral signatures has tremendous impact on the development of methods for automatic identification and classification of various agricultural crops in an area. The main aim of this study is to evaluate the potential of hyperspectral remote sensing data for mapping, measuring and identification capabilities relevant to various information classes and biophysical parameters of agricultural crops. The utility of Hyperion for land use/land cover mapping in an agricultural setting has been evaluated based on the classification accuracy. Further, the performance of the Hyperion for land use/land cover mapping has also been compared with that from LISSIII. Various feature reduction and feature selection methods (TCT, PCA and segPCA) and classifiers (MLC and SAM) have been applied on the Hyperion image to bring out the best methodology for exploitation of the Hyperion image for land use/land cover mapping within an agricultural setting. The results suggest improvement in the classification ii accuracy if Hyperion has been classified after the application of feature reduction methods. Further, the abundant resource of spectral data offered by Hyperion has been used for the improved retrieval of LAI of agricultural crops. The comparative performance of Hyperion and LISSIII for the estimation of LAI of agricultural crops has been carried out to assess the increase in reliability and accuracy of the LAI retrievals from satellite data. Results indicate that many hyperspectral bands in the NIR and red region have the great potential in developing better vegetation indices for LAI estimation. Furthermore, the possibilities of retrieving leaftotal chlorophyll and nitrogen concentrations at field or pixel level has been evaluated by up-scaling the relationships observed between in-situ and airborne spectral reflectance and leaf total chlorophyll and nitrogen concentrations at leaf and canopy level. Intensive in-situ and laboratory biophysical and biochemical measurements were made at the canopy and leaf level with specialized instrumentation. A comprehensive rigorous analysis has been carried out to identify the spectral bands and spectral indices for accurate retrieval of Chi t and LNC for cotton crop. Regression analysis has been used to determine those wavebands and waveband ratios which are most closely associated with crop biochemical parameters. The potential of these critical spectral reflectance indices has been validated with independent samples of Chi t and LNC. Anew vegetation index named as PBI (Plant Biochemical Index) has been proposed for better prediction of plant biochemicals from space-borne hyperspectral data. An assessment has been carried out to study the variability present in spectral signatures of agricultural crops (rice, sugarcane, cotton and chillies) and to identify a meaningful discrimination amongst various crops, and also for different varieties ofa single crop type. Quantitative assessment of the variability of spectral signatures of agricultural 111 crops in terms of shape of the spectrum, spectral variability and similarity of various types of crop varieties has been carried out. In-situ hyperspectral measurements have been collected for rice, cotton, sugarcane and chillies crops. It has been observed that different crops can be discriminated at variety level based on the amplitude difference of the crop spectral reflectance. The in-situ hyperspectral measurements have been used for the development of variety based crop spectral libraries. Further, the performance of these spectral libraries has been evaluated for automatic identification of unknown crop spectra. The identification of unknown crop spectra with the spectral libraries built using hyperspectral reflectance data at canopy scale (in-situ hyperspectral measurements) and at pixel scale (Hyperion data) has shown promising results with 83.3% and 91.6% of spectral variability explained by hyperspectral reflectance data at canopy scale and at pixel scale respectively. This observation highlights the possible use of hyperspectral remote sensing data for automatic identification and discrimination of various crop varieties. Finally, an attempt has been made to evaluate the potential of hyperspectral data for elimination of saturation of broadband NDVI and selection of optimal hyperspectral bands for crops characterization. In-situ hyperspectral and biophysical measurements were carried out at various phenological stages of agricultural crops. The optimal hyperspectral bands for crops characterization has been done based on the discrimination and phenological variation of NDVI and the ability to retrieve LAI during vegetative, mature and senescence stages of the selected agricultural crops. A model program has been developed for the discrimination and automatic correlation of all possible NDVI band combinations with the LAI at various phenological stages. The results suggests 11 band combinations for monitoring of crops using NDVI as the crop conditions vary due to phenological changes and other factors such as management conditions, soil characteristics, climatic conditions and cultural practices. IV