Analysis Of Spectral Characteristics Of Rice Canopy Under Water Deficiency

Chwen-Ming Yang, Muh-Rong Su
Agronomist and Assistant, Department of Agronomy
Taiwan Agricultural Research Institute
189 Chung-Cheng Road, Wufeng, Taichung Hsien, Taiwan 41301
Tel: (886)-4-330-2301ext.135 Fax: (886)-4-330-2806
E-mail: cmyang@wufeng.tari.gov.tw

Key Words: Spectral characteristics, Rice, Water deficiency, Vegetation index, Red edge

Abstract
Information of spectral characteristics of rice canopy was extracted from ground-based remotely sensed reflectance spectra of hyperspectral resolution during water stress period in the vegetative growth phase. The analyses were focused mainly on the correlation between levels of water deficiency and spectral characteristics or parameters across wavebands of 350-2400 nm. It was shown that reflectance spectrum of rice canopy was sensitive to water stress, while different wavebands had diverse responses to of water deficiency. A dramatic difference of correlation coefficients (r) along the spectral range was found between stress treatments. The correlation between reflectance and stress level can be best represented at wavelength 2113.5 nm, where the reflectance and the reflectance ratio to control were the most significant difference. Among 12 characteristics wavelengths selected to examine their correlation to water deficits, the reflectance of 8 wavelengths had significant linear relationships with stress levels. Results further indicate that changes in red-edge position and slope did not correlate to soil water potential. The relationship between NDVI and soil water potential was significant, the increasing of soil water stress with decreasing value of NDVI.

1. Introduction
The use of canopy spectral data from remote sensing for agricultural purposes such as crop growth monitoring, yield prediction, and stress evaluation has been widely adopted nowadays. Crop growth models describing the relationship between physiological status of plants and effects of environmental factors were established and modified for various applications. Before any practical utilization in agricultural fields, however, spectral characteristics or parameters should be identified and then be incorporated into algorithms linked to the relationship in this respect. On the other hand, remote sensing may yield information to the actual conditions of a crop and in turn improving the accuracy and structure of crop growth models. Currently reflectance from visible and near-infrared regions has been used as variables for estimating physical traits of a crop in modeling. Much more research has also been aimed at combing the reflectance in different wavebands in assessment of growth parameters for minimizing undesirable interference from soil background and atmosphere. Such combinations are the so-called egetation indices (Clevers, 1988, 1989; Huete, 1988; Kauth and Thomas, 1976; ichardson and Wiegand, 1977). In any case, variables employed for these applications should be able to closely represent growth characteristics or behavior.

On developing of satellite remote sensing in the 1970s, it was broad bands rather than narrow band width were used for spectral detecting due to the limitations of hardware and software techniques. With the advancement of remote sensing technology, when the spectral resolution reached to the order of 10 nm or smaller, it becomes possible to uncover minor differences in spectra data in details (Curran, 1989). Specific features of reflectance spectrum may be extracted in comparable to the specific changes of physical traits or internal chemical compounds for a crop (Elvidge, 1990; Goetz, 1991). Responses of crops to environmental stresses such as nutrient deficiency, drought, frost, and pests may also be distinguished and classified qualitatively as well as quantitatively.

The objectives of this research are to analyze remote sensing data of rice canopy from ground measurements during water stress period in the vegetative growth phase. Information of spectral characteristics was extracted from reflectance spectra of high spectral resolution. The correlation between levels of water deficit and changes in spectral parameters is studied and the potential of applying such relationship for estimating strength and status of water deficiency is evaluated.

2. Materials And Methods
The potential of using spectral characteristics for water deficiency evaluation was examined from the ground-based remotely sensed hyperspectral data over rice canopy under varied levels of soil water deficits. Rice (Oryza sativa L. cv. Tainung 67) plants were grown on a loam soil at a density of 22 plants m -2 , with row spacing of 0.3 m and plant distance of 0.15 m, in opaque plastic containers (1.0 m W × 1.2 m L × 0.55 m H). The 20-d old seedlings were transplanted with 1 plant per hill on 5 April 2000, and the containers were placed in the experimental farm of Taiwan Agricultural Research Institute at Wufeng, Taiwan. Plants were fertilized and well taken care of to avoid the infestations of weeds and pests. When reached to leaf 8 to leaf 9 stage in the vegetative growth phase, containers were divided into three groups and subjected to water-withholding treatments. At this stage, vegetative cover was considered dense with the covering percentage more than 70%.