Airborne hyperspectral remote sensing was adapted to establish a general-purpose model for quantifying nitrogen content of rice plants at the heading stage using three years of data. There was a difference in dry mass and nitrogen concentration due to the difference in the accumulated daily radiation (ADR) and effective cumulative temperature (ECT). Because of these environmental differences, there was also a significant difference in nitrogen content among the three years. In the multiple linear regression (MLR) analysis, the accuracy (coefficient of determination: R 2 , root mean square of error: RMSE and relative error: RE) of two-year models was better than that of single-year models as shown by R 2 ≥0.693, RMSE≤1.405 gm -2 and RE≤9.136%. The accuracy of the three-year model was R 2 =0.893, RMSE=1.092 gm -2 and RE=8.550% with eight variables. When each model was verified using the other data, the range of RE for two-year models was similar or increased compared with that for single-year models. In the partial least square regression (PLSR) model for the validation, the accuracy of two-year models was also better than that of single-year models as R 2 ≥0.699, RMSE≤1.611 gm -2 and RE≤13.36%. The accuracy of the three-year model was R 2 =0.837, RMSE=1.401 gm -2 and RE=11.23% with four latent variables. When each model was verified, the range of RE for two-year models was similar or decreased compared with that for single-year models.The similarities and differences of loading weights for each latent variable depending on hyperspectral reflectance might have affected the regression coefficients and the accuracy of each prediction model.The accuracy of the single-year MLR models was better than that of the single-year PLSR models.However, accuracy of the multi-year PLSR models was better than that of the multi-year MLR models. Therefore, PLSR model might be more suitable than MLR model to predict the nitrogen contents at the heading stage using the hyperspectral reflectance because PLSR models have more sensitive than MLR models for the inhomogeneous results. Although there were differences in the environmental variables (ADR and ECT), it is possible to establish a general-purpose prediction model for nitrogen content at the heading stage using airborne hyperspectral remote sensing. 2 Keywords: airborne hyperspectral remote sensing, nitrogen content, heading stage, multiple linear regression (MLR), partial least square regression (PLSR) ---The similarities and differences of loading weights for each model depending on hyperspectral reflectance might have affected the regression coefficients and the accuracy of each model.The accuracy of the single-year MLR models was better than that of the single-year PLSR models.However, the accuracy of the multi-year PLSR models was better than that of the multi-year MLR models. PLSR model might be more suitable than MLR model to predict the nitrogen contents at the heading stage using the hyperspectral reflectance because not only the robust of PLSR models but also the sensiti...

Airborne hyperspectral remote sensing was used to provide data for a general-purpose model for predicting the nitrogen content of rice at panicle initiation stage using three years of data. There were significant differences between the vegetation data which were affected by the uptake of nitrogen from the soil depending on weather conditions. Therefore, the reflectance values obtained for one year may exhibit a different trend, due to the lack of vegetation. When the partial least squares regression (PLSR) models were estimated using all combination of three-year data, except for the model incorporating the data from 2005, correlation coefficients (r) were greater than 0.758, and the root mean squared error (RMSE) of prediction of the full-cross validation was less than 0.876 g m-2. The accuracy of the 2003-2004-2005 model was determined using five latent variables (PCs), with r = 0.938 and RMSEP = 0.774 g m-2. There were two different patterns for the regression coefficients associated with the NIR or red-edge regions. When the 2003-2004 model was validated using the data from 2005, the prediction error of the PLSR model was 1.050 g m-2. This became 2.378 g m-2 for the

No abstract

In this study, protein content (PC) of brown rice before harvest was established by remote sensing (RS) and analyzed to select the key management factors that cause variation of PC using a GIS database. The possibility of finding out the key management factors using GreenNDVI was tested by combining RS and a GIS database. The study site was located at Yagi basin (Japan) and PC for seven districts (85 fields) in 2006 and nine districts (73 fields) in 2007 was investigated by a rice grain taste analyzer. There was spatial variability between districts and temporal variability within the same fields. PC was predicted by the average of GreenNDVI at sampling points (Point GreenNDVI) and in the field (Field GreenNDVI). The accuracy of the Point GreenNDVI model (r 2 [ 0.424, RMSE \ 0.256%) was better than for the Field GreenNDVI model (r 2 [ 0.250, RMSE \ 0.298%). A general-purpose model (r 2 = 0.392, RMSE = 0.255%) was established using 2 years data. In the GIS database, PC was separated into two parts to compare the difference in PC between the upper (mean ? 0.5SD) and lower (mean -0.5SD) parts. Differences in PC were significant depending on the effective cumulative temperature (ECT) from transplanting to harvest (Factor 4) in 2007 but not in 2006. Because of the difference in ECT depending on vegetation term (from transplanting to sampling), PC was separated into two groups based on the mean value of ECT as the upper (UMECT) and lower (LMECT) groups. In 2007, there were significant differences in PC at LMECT group between upper and lower parts depending on the ECT from transplanting to last topdressing (Factor 2), the amount of nitrogen fertilizer at top-dressing (Factor 3) and Factor 4. When the farmers would have changed their field management, it would have been possible to decrease protein contents. Using the combination of RS and GIS in 2006, it was possible to select the key management factor by the difference in the Field GreenNDVI.

No abstract