APPLICATION OF PRINCIPAL COMPONENT ANALYSIS TO MODIS NDVI FOR ESTIMATION OF PADDY RICE YIELDS IN GYEONGGI, SOUTH KOREA

摘要

MODIS NDVI is generally used for the estimation of various kinds of crop yields including paddy rice. Efforts have been made to estimate paddy rice yield using MODIS NDVI and meteorological data in Korea. However, the choice of a representative MODIS NDVI as an independent variable for the model to estimate paddy rice yields is ambiguous due to high correlations among multi-temporal MODIS NDVI data. This study proposed a method to produce representative values from MODIS NDVI by applying a PCA technique. The method was used to develop an estimation model for paddy rice yield in Gyeonggi, one of the South Korea provinces. Application of this PCA method increased the correlation between the representatives and rice yields from 0.411 to 0.524. The robustness was also improved, as the RMSE of the models decreased from 17.32% to 5.05% by comparing estimations and actual statistics from 2012.