Modeling rice growth with hyperspectral reflectance data

Field experiments were conducted to study the seasonal changes of rice reflectance spectra and approaches to model rice (Oryza sativa L.) growth with high-resolution reflectance data. Ground-based remotely sensed canopy spectra and growth parameters of rice plants, including fresh weights of leaves and shoots, dry weights of leaves and shoots, plant height, and leaf area index, were measured regularly during the first and the second cropping seasons in 2000 to 2002. Spectral indices of R-RED/R-NIR rating R-GREEN/R-NIR ratio, R-RED/R-GREEN ratio, and normalized difference vegetation index (NDVI) were calculated, where spectral characteristic R-GREEN is reflectance at green light (490-560 nm) maximum (GREEN), R-RED is reflectance at red light (640-740 nm) minimum (RED), and R-NIR is reflectance at near-infrared (740-1300 nm) peak (NIR). Seasonal patterns of these spectral attributes and growth parameters implied the existence of developmental effects as well as environmental impacts. Generally, growth parameters reached the maximum near heading when the vegetative growth was greatest, and decreased thereafter. Modeling between spectral attributes and growth parameters can be improved by separating growing period into the preheading and postheading phases. Some of the selected spectral attributes were shown to have significant relationships with growth parameters. Models established by the multiple linear regression (MLR) analysis also indicated the feasibility for estimating rice growth. These MLR models combining spectral reflectance from more than two wavebands provided flexibility in choosing the individual narrow bands, exhibited a greater sensitivity to phenological variation, and improved the models' ability to estimate plant growth.