Suhadi Chandra


Paddy agriculture is a primary need for most of Indonesian people. Problems such as the decreasing in rice production and productivity levels drastically can affect the availability of the national’s rice stock and would give negative impact on other sectors. Therefore, it required the fast and accurate yield of crop estimation. One of the solutions offered are using remote sensing technology. Remote sensing technology used in this research include multispectral technology, the satellite, combined with hyperspectral technology, the ground based spectrometer. The objective of this paper is to estimate crop in order to assist the government to predict rice stock inventory and help the government program on food resistance based on MODIS satellite. The focus area of this research are in Karawang and Subang Regency, Indonesia. In estimate yields of rice, used three methods i.e. Principal Component Regression, Regression Ridge Regression, and Partial Least Squares. These three methods have the ability to overcome the collinearity problem, where partial least square regression give the better result from training stage. The training evaluation from PRESS  and R2 value where the partial least square regression get 0.3434 and 99,48%. Testing evaluation using real data in Subang, which has 31.281 ha areas, principle component regression give the better result. The estimation error of principle component regression from testing evaluation is 14,11%, it is smaller compare to other two methods.


Key words

Remote Sensing, MODIS, Principle Component Regression, Ridge Regression, Partial Least Square Regression.


  • There are currently no refbacks.