Abstract
The estimation of rice yield using remote sensing (RS) technology is a critical task for sustainable agriculture. RS has been widely applied to monitor the rice yield estimations in several weeks before harvesting. However, the spectral data from the common earth observation satellites with the low spectral data results in the loss of important spectral characterization data that affecting the accuracy of the evaluation of rice yields. The main objective of this research is to examine the potential of hyperspectral data to estimate the rice yield before harvesting. In addition, to remove the atmospheric noise, EO-1 Hyperion surface reflectance data was transformed into the First Derivative Spectrum (FDS). The results of the estimations of rice yield using high spectral resolution remote sensing data and Stepwise Multiple Linear Regression (SMLR) analysis revealed that FDS dataset is the best estimator variables, compared to using the surface reflectance data. The best model for estimating rice yield of specific varieties (RD-41 variety) had a coefficient of determination of 0.884 and Root Mean Square Error (RMSE) of 286.3 kg/ha. The selected FDS bands for the estimation model were centered at red edge and shortwave Infrared (SWIR) region of electromagnetic spectrum which showed the good correlation with the rice yields. Red edge is related to the amount of chlorophyll in the rice leaves and can be used to explain the variation in rice yield. The shortwave infrared is a wavelength region that can be used to analyse the health and water in plants because SWIR is defined as water absorption feature. It was found that the FDS reflectivity could explain the variability of rice yield more than the common surface reflectivity. The usefulness of transformed spectral information for the rice yield estimation provides a solution to improve the crop yield estimation method from hyperspectral remote sensing.
