The Philippines has made significant strides in developing its rice production sustainably, which has contributed to the nation's food security and sustainable agriculture. However, the sector faces various obstacles, and ensuring its long-term viability is crucial. For this reason, building a tool that allows estimating rice yield is necessary. Synthetic Aperture Radar (SAR) remote sensing data from Sentinel-1 satellites provide no cost, extensive coverage, and high spatiotemporal resolution, which has the advantage of observation in cloudy, foggy, rainy weather and independent of solar radiation. This study aimed to delineate rice crop fields and estimate the rice yield in the Rice Granary Capital of Agusan del Sur – Bayugan City, using multi-temporal Sentinel-1 data with C-band wavelength. Predictor variables derived from the Sentinel-1 image were used to model the rice yield: the VV and VH polarization backscatter value and the GLCM of VV and VH polarizations. The results showed that VH polarization produces the highest kappa coefficient of 0.93 and overall accuracy of 96.5% in delineating rice fields using the Maximum Likelihood classifier. An exponential solid relationship has been identified between the VH polarization and rice yield, producing accurate yield estimation with the highest coefficient of determination (R2) of 0.83 and the lowest root mean square error (RMSE) value of 5.29. The generated map showed the estimated rice yield value of 100.60 sacks/ha to 128.62 sacks/ha, with an average yield of 112.04 sacks/ha. Therefore, it seems reasonable to conclude that Sentinel-1A effectively estimates rice yield with its large-scale polarization's backscatter information.
|