Assessment of leaf folder, Cnaphalocrosis medinalis (Guenee) damage in rice using ground based hyperspectral remote sensing techniques
Loading...
Date
2022-07-30
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Department of Entomology, OUAT, Bhubaneswar
Abstract
Field and laboratory experiments on “Assessment of leaf folder, Cnaphalocrosis medinalis (Guenee) damage in rice using ground based hyperspectral remote sensing techniques” were undertaken during Rabi 2017-18, Kharif 2018-19 and Rabi 2018-19 at ICARNRRI, Cuttack. The spectral signature from the differentially infested canopy and leaves of rice var TN1 were assessed using ASD Field Spec hyperspectral spectroradiometer. The spectral curve obtained from the pre-processed data indicated higher reflectance of healthy vegetation than the rice plants infested by C. medinalis invisible and near infrared region. Derivative analysis was adopted to find out the particular spectral bands with maximum per cent change in first order derivatives. The continuum removed spectra were also obtained using ENVI software and the specific wavelengths with the absorption dips were identified. The central wavelengths identified from derivative analysis and continuum removal were subjected to unpaired ‘t’ test for testing of significance up to 100 nm interval. Seven spectral bands i.e., 515, 582, 673, 732, 815, 898 and 1131 nm were found most sensitive to infestation of C. medinalis from both canopy and leaf level. The healthy leaves had highest concentration of total chlorophyll (3.95 mg/g) and carotenoid (0.36 mg/g) whereas these concentrations were found lowest viz., 1.467, 0.207 mg/g in severely infested rice leaves (Grade 5). The phenol content was highest (4.79 mg/g) in severely infested rice leaves due to C. medinalis but it was lowest (3.25 mg/g) in the healthy leaves. The correlation between different biochemical components and identified sensitive bands revealed that four spectral bands viz., 515, 582, 597 and 672 nm were found sensitive due to chlorophyll absorption. Similarly, the spectral band of 1131 nm was found sensitive for phenol absorption whereas for the carotenoid absorption the spectral 515 and 582 nm bands was found sensitive. Feature selection study was carried out using MATLAB by the Random Forest Elimination tool to reduce the dimensionality of the spectral data. The spectral band combination of five bands viz., 515, 582, 673, 732 and 815 nm was found to be the best band combination with highest percentage of accuracy (86.33) which could differentiate the damage by C. medinalis. Studies on five spectral vegetation indices calculated based on spectral reflectance for detecting stress in rice infested by C. medinalis revealed that the values of Normalized Difference Vegetative Index (NDVI), Leaf Area Index (LAI) and Photochemical Reflectance Index (PRI) were highest in the healthy rice vegetation and were 0.77, 3.20 and 0.17, respectively. The values of these aforementioned indices were in a declining trend with increase in infestation of C. medinalis. A gradual increase in the value of Photochemical Senescence Reflectance Index (PSRI) was observed with subsequent rise in infestation level of C. medinalis. Severely infested rice plants recorded highest value of PSRI (0.18) whereas it was lowest (0.03) for the healthy vegetation. A gradual shifting of the red edge position towards the shorter wavelength (REPI) was observed with increase in infestation level of C. medinalis. Three new indices based on two spectral combinations {(R515/R732),(R732-R515) and (R732-R582/R732+R582)} as well as three other indices based on three spectral band combination {R732/(R515+R673), R815/(R568xR732), and R815/(R673+R582)} were derived which may be used for depicting the presence of C. medinalis using satellite imagery. Results on the spectral signature obtained from the differential infestation by C. medinalis in rice variety Naveen at farmer’s field (located at Aruha, Pipili block) and under controlled conditions at ICAR-NRRI, revealed that seven spectral bands i.e., 510, 583, 674,
762, 815, 897 and 1126 nm were confirmed as the pest sensitive spectral bands for infestation of C. medinalis. The ground truth spectral data were validated using Sentinel 2A image analysis with the satellite imagery. The analytical processing of both the ground truth data and data from satellite indicated the hotspots where infestation by C. medinalis were detected through the false colour composite map with the help of various colour codes.