Analysis Of Spectral Characteristics Of Rice Canopy Under Water Deficiency

The collected ground truth was made periodically from 6 days after stress treatments with a portable spectroradiometer (model GER-2600, Geophysical and Environmental Research Corp., New York, USA) on near cloudless days at about solar noon. There were 10 levels of soil water deficits attained from 3 groups of containers during the experimental period, with one control (the well-watered plants) for each group (Table 1). The scanner of spectroradiometer has a 10 degree field-of-view (FOV) and was held at a nadir viewing over rice canopy, at a distance of 1 m, to acquire the reflected radiance within 330-2500 nm. Reflectance spectrum was obtained by comparing the radiance of the target canopy with the radiance of a standard (the so-called Spectrolon). Six measurements of radiance spectra were made from canopy of every treatments and each measurement was set as a mean of 3 individual full-range spectral scans. A mean reflectance spectrum was calculated from those of 6 measurements. As the reflectance above wavelengths of 2400 nm had higher noise and seemed not to affect the objectives of this study, data obtained above 2400 nm were excluded from analysis. Care was taken to prevent the influence of shadow and background.

There were 4 approaches employed to determine spectral characteristics or parameters for evaluation of water deficiency in rice plants. The first approach was to acquire wavelengths of maximum correlation coefficient (r), from the analyzed wavelength domain (350-2400 nm), by computing curve of correlation coefficients for the reflectance from different treatments. The second approach was to compare difference in reflectance relative to the controls. The percent difference in reflectance over the wavelength domain between the stressed plants and the well-watered plants were calculated, and then subjected to a correlation processing as aforementioned to compare the effect of different smoothing intensity on the correlation between the wavelengths. In the third approach, 12 wavelengths were selected from the apparent peaks and valleys appeared in the mean reflectance spectrum of each controls, and characteristics wavelengths were determined from mean value of each of 12 wavelength intervals. The reflectance of 12 characteristics wavelengths from treatments were used to examine for their correlation to levels of water deficits. The final approach was to direct correlate the red edge and the normalized difference vegetation index (NDVI) to soil water potentials. The region of red-edge is defined as wavelengths between red minimum (RED) and near-infrared maximum (NIR). The position of the red-edge is determined by the wavelength of the main inflexion point of the red-edge slope, which is the maximum value of the first order derivatives in the red-edge region. Both the position and the slope of the red-edge changed between levels of stress were tested. The NDVI was calculated by the equation: (NIR-RED)/(NIR+RED).

3. Results And Discussion
The spectral reflectance in the domain of 350 to 2400 nm from rice canopy under varied water deficits and their correlation to stress levels were graphed in Figure 1. It was shown that reflectance spectrum was sensitive to water stress, while different wavebands had diverse responses to water deficiency. A dramatic difference of correlation coefficient (r) along the spectral range was found between stress treatments. The curve gives an overall picture of the correlation feature for reflectance spectra in response to water stress. By examining the differences between wave regions, the maximum values of r and the corresponding wavelengths in different regions of reflectance spectra were identified (Table 1). The percent differences in reflectance between the stressed plants and the controls were further compared in Figure 2. It shows that the intensity of the correlation curve is similar, but in less extent, to the aforementioned feature of the processing results. The most pronounced spectral characteristics in relation to water deficits was the reflectance at wavelength 2113.5 nm (r=-0.97).



Fig. 1. The spectral reflectance in the domain of 350-2400 nm from rice canopy under varied levels of water deficit and their correlation coefficients to water stress.

spectra and ratio spectra @ from rice canopies under various levels of soil water deficits.

Item	Region	Waveband	Wavelength	R
Reflectance Spectrum	Ultraviolet to visible	350-740 nm	697.3 nm	-0.90**
	Near infrared	740-1300 nm	1176.9 nm	-0.63*
	Middle infrared	1300-1800 nm	1508.3 nm	-0.95**
	Shortwave infrared	1800-2400 nm	2113.5 nm	-0.98**
Ratio Spectrum	Ultraviolet to visible	350-740 nm	692.8 nm	-0.87**
	Near infrared	740-1300 nm	1176.9 nm	-0.51
	Middle infrared	1300-1800 nm	1508.3 nm	-0.90**
	Shortwave infrared	1800-2400 nm	2113.5 nm	-0.97**