Detection and attribution of drought and oil-induced plant stress using hyperspectral remote sensing

Abstract

This study investigates the potential use of hyperspectral remote sensing in detecting stress in vegetation caused by both drought stress and hydrocarbon contamination of the soil. Spectral and plant biochemical measurements were taken from an in-situ field experiment and a glasshouse experiment at the James Hutton Institute, Invergowrie, U.K. The in-situ field experiment was part of an ongoing root restriction study of 10 barley genotypes, whilst the glasshouse experiment consisted of both drought and oil contamination of plants grown in pots. Two main biophysical parameters (chlorophyll and plant leaf water content) were calculated and spectral response measured with a spectroradiometer. Statistical analysis showed variation in spectral response between genotypes that were responding differently to treatments. The 1st derivative of reflectance at wavelength 705 nm, the ratio of the wavelength of the maximum 1st derivative reflectance at the green region (GEP) and the wavelength of maximum 1st derivative reflectance (REP) all showed a strong correlation with chlorophyll concentration. The ratio between normalized difference vegetation index NDVI (800,680) and the index GEP/REP was used to discriminate between the genotypes showing different levels of drought tolerance. Subsequently, using hierarchical clustering the barley genotypes were grouped into two i.e. drought sensitive and drought tolerant. The 1st derivative reflectance ratio at wavelength 520 nm and 702 nm and the normalized difference vegetation index NDVI (800, 680) showed differences in response to treatment in the oil experiment. The result of both experiments suggests that the 1st derivative ratio at wavelengths 520 nm and 702 nm, the 1st derivative reflectance at wavelength 705 nm and the 1st derivative ratio at wavelengths 1100 nm /1200 nm should be explored further as indicators of drought stress. This research demonstrates the potential of remote sensing to detect stress in vegetation in several forms, but further work is required to investigate these methods which are reliable across time, space and sensor, and are also able to differentiate stress from an early stage.

Date of Award	2018
Original language	English
Supervisor	Mark Cutler (Supervisor) & Terence Dawson (Supervisor)