Prediction model of rice (Oryza sativa) yield under high temperature stress based on hyper-spectral remote sensing

Abstract

Canopy spectral reflectance on flowering, filling and ripening stages of two rice (Oryza sativa L.) cultivars during heading stage under different high temperature stress as well as yield and yield components after maturation were measured in 2007 and 2008. The results showed that relativities among theoretical yield, actual yield, yield components and hyperspectral indices during flowering and filling stages were at the level of P<0.05 by contrast with ripening stage, which thus indicated these two stages could be used as key stages for forecasting rice yield. Further, difference vegetation index called as DVI[810, A(450, 560, 680)], perpendicular vegetation index called as PVI(810, 680), slope of red edge called as Dëred and Area 670-755 could be used for forecasting theoretical and actual yield of matured rice. Moreover, DVI(810, 450) and DVI(810, 560), PVI(810, 680) and Dëred could also be used for forecasting yield components of matured rice. The hyper-spectral indices with preferred fitting correlation were selected by regression analysis, and accordingly models for forecasting rice yield during different growth stages were established in the light of spectral index PVI (810,680). Finally, after testing these derived models with independently observed data, it was found that the model of flowering stage was more reliable to predict rice yield than those of filling and ripening stages.