Abstract
Multispectral data from aircraft or satellite have been widely used in agriculture as well as in other industrial sectors. To connect such data with vegetation cover for agricultural applications, generally a detailed and through ground-based should be made to characterize radiance and reflectance spectra and link it with growth and yield traits. This study was to remotely measure and monitor the seasonal variations in reflectance spectrum, in the range of 350-1100 nm, of rice (Oryza sativa L. cv. Tainung 67) canopy at the experimental field of Taiwan Agricultural Research Institute (TARI) during the cropping seasons in 1996-1997. The objectives were to examine relationship between growth traits and NDVI and to assess their potential use in estimating growth of rice crop. Attempts were also made to compare NDVI calculated from narrow-band ground-based spectral measurement and the simulated board-band multispectral satellite data. The physiological and morphological implications of the relationships were discussed.
Introduction
Satellite or aircraft remote sensing can provide a timely spatial distribution information on crop performance and be incorporated into farm management scheme. Application such as precision farming activities, growth and production estimation, and land surface characteristics evaluation are made possible (Maas, 1998). Before these practical applications, vegetation spectral properties from remote sensing must be in connection with some physical characters of plant canopy so that plant performance can reasonably be assessed. Mathematical formulae are then established from the fitting curves of these paired variables. Leaf area index (LAI), percent vegetation cover, and unit area plant biomass are among physical characters commonly used as indicators of the status of crops (Maas, 1998; Wiegand et al., 1992; Yang and Su, 1998).
Vegetation indices are mathematical transformations intended to estimate the spectral contribution of crop vegetation to multispectral observations (Elvidge and Chen,
(1995). The transformations have the effect of normalizing measurements acquired from varied environmental conditions. The formulae are derived mostly from discrete green, red and near-infrared bands, especially the two ends of the so-called chlorophyll red-edge (Jackson et al., 1983; Tucker, 1979). By correlating vegetation indices with physical characters of plant canopy, changes of vegetation feature can be potentially assessed and predicted from values of vegetation indices during the growing season (Maas, 1998; Tucker et al., 1979; Yang and Su, 1997).
In this study, variations in reflectance spectrum were monitored and analyzed and the normalized difference vegetation index (NDVI) were calculated from the ground-based remotely sensed spectral measurements of rice canopy for the growing seasons of 1996-1997. An simplified vegetation model for estimating crop growth traits from NDVI is presented, and the simulated broad-band versions of NDVI were compared with the narrow-band counterparts.
Materials and Methods
Experiments were conducted at TARI Experimental Farm (23o30'N, 120o42' E, elevation of 85 m) on a loam soil (nonacid, hyperthermic, Fluvaquentic Dystrochrept) during First and Second crops of 1996 and 1997. Cultivation and fertilization were in consistent with the existing cultural practices for rice adopted by the local region. Each season had 3 plots as replications and were furrow-irrigated. Herbicide and hand-weeding were practiced to avoid weed interference. Plants were sampled from 2 wk after transplanting till harvest periodically. Leaf area was determined and LAI, defined as the area of green leaves per unit area of land, was computed at each sampling. Leaf dry weight (LDW) was measured after oven-dried at 80oC for 72 h. Means were used for regression analyses to examine the correlation between growth traits and NDVI.
Method of spectral measurements was reported in the previous papers (Yang and Su, 1997, 1998). Spectral irradiance of the dense vegetation cover and incident solar radiation in the range of 350 to 1100 nm was measured using a LI-1800 high spectral resolution portable spectroadiometer (LI-COR inc., Nebraska, USA). The sensor head was pointed downward in a nadir-viewing and was placed horizontally 1.0 m above rice vegetation surface to scan the upward reflected radiation. Measurements were made at each plot using sunlight as illumination source on clear or near cloudless days between 11:00 to 12:00 local standard time. Date were averaged and mean values were used.
Radiance spectra of sunlight were taken before and after each replicate
measurement and the average was used as the reflectance standard. Reflectance of individual wavelength was calculated by dividing the vegetation radiance measurements with the corresponding incident solar radiation measurement with the corresponding incident solar radiation measurements. Reflectance spectra measured during the growth were grouped for each crop and the standard errors of individual wavelengths were calculated as representative of the upper and lower boundaries. The NDVI was defined as (NIR-RED)/(NIR+RED). The reflectance of the wavelengths at RED (674 nm) and NIR (754 nm) measured from LI-1800 were selected and used for calculations. Regression analysis were performed to generate fitting-curves in order to monitoring seasonal changes of NDVI as well as growth traits.
