Using canopy reflectance and partial least squares regression to calculate within-field statistical variation in crop growth and nitrogen status of rice

For the site-specific prescription of fertilizer topdressing in rice cultivation, a non-destructive diagnosis of the rice growth and nutrition status is necessary. Three experiments were done to develop and test a model using canopy reflectance for the non-destructive diagnosis of plant growth and N status in rice. Two experiments for model development were conducted, one in 2000 and another in 2003 in Suwon, Korea, including two rice varieties and four nitrogen (N) rates in 2000 and four rice varieties and 10 N treatments in 2003. Hyperspectral canopy reflectance (300-1,100 nm) data recorded at various growth stages before heading were used to develop a partial least squares regression (PLS) model to calculate plant biomass and N nutrition status. The 342 observations were split for model calibration (75%) and validation (25%). The PLS model was then tested to calculate within-field statistical variation of four crop variables: shoot dry weight (SDW), shoot N concentration (SN), shoot N density (SND) and N nutrition index (NNI) using measured canopy reflectance data from a field of 6,500 m(2) in 2004. Results showed that PLS regression using logarithm reflectance had better performance than both the PLS and multiple stepwise linear regression (MSLR) models using original reflectance data to calculate the four plant variables in year 2000 and 2003. It produced values with an acceptable model coefficient of determination (R 2) and relative error of calculation (REC). The model R-2 and REC ranged from .83 to .89 and 13.4% to 22.8% for calibration, and .76 to .87 and 14.0% to 24.4% for validation, respectively. The PLS regression model R-2 was reduced in the test data of year 2004 but the root mean square error of calculation (RMSEC) was smaller, suggesting that the PLS regression model using canopy reflectance data could be a promising method to calculate within-field spatial variation of rice crop growth and N status.