Mapping paddy rice yield in Zhejiang Province using MODIS spectral index


Abstract: The first 19 bands of a moderate-resolution imaging spectrometer (MODIS), covering the visible to shortwave infrared spectral wavelength, were simulated by ground-level reflectance spectra. All spectral indices similar to the normalized difference vegetation index (NDVI) and ratio vegetation index (RVI) formed by every 2 bands were calculated to obtain their determination coefficients, with theoretical and real yield. Results revealed that the combinative near-infrared (NIR) index (band 2 and band 19, symbol b2 and b19, and other similar ones) and the (b16, b19) of MODIS were strongly correlated with rice yield, especially the correlative coefficient that exceeded significant levels in the maturing stage. However, combinative visible light index was correlated with rice yield strongly in the early stage and poorly in the latter stage. The best spectral indices for predicting rice yield in whole rice growth were the combinations of (b2, b19) and (b16, b19). Based on the 2-band combination of the (b2, b19) of MODIS and the constructed estimating model, the rice yield of Zhejiang Province was estimated and its spatial distribution was mapped. Estimated results based on MODIS images were validated using 8 measured validation sites. Errors in the estimated rice yield ranged from 3.2% to 20.3%, with a mean value of 10.1%. The results indicated that the 2-band combination of the (b2, b19) MODIS index was most suitable for monitoring rice yield.

Keywords: Estimation, rice yield, MODIS, spectral index

Full Text: PDF