Abstract:

Aboveground fresh biomass is an important colony qualitative index of cotton, and so it is significant for production management and yield estimation in cotton to establish the monitoring model based on hyperspectral parameter. The wavelength variation coefficient of cotton canopy hyperspectrum on aboveground fresh biomass and the correlation between the aboveground fresh biomass and reflective spectrum, the first derivative spectrum showed that the biggest value of hyperspectral variation coefficient of cotton canopy reflectance in the visible light region was in 672 nm. The biggest value of correlation coefficient between the aboveground fresh biomass and reflectance spectrum at the visible light region was in 589–700 nm and at the near infrared red region in 865-919 nm, and the former was larger than latter. The correlation coefficient between the aboveground fresh biomass and the first derivative value in the visible light had three high-value areas including 524–528 nm, 552–588 nm and 710–755 nm. According to the analysis above, the nineteen hyperspectral characteristic parameters were used to establish the hyperspectral remote sensing estimation models of the aboveground fresh biomass in cotton. The tested result of models expressed the parameters veracity of estimating the aboveground fresh biomass in cotton, including reflectance of 682 nm and 629 nm and the combination forms, was up 80%. Among them, F629[Y = 9.7914 exp(－20.738 F629)]in single band reflectance parameters was better, its veracity reached to 83.9%, RMSE was 0.64 kg m^-2, and the correlation coefficient between the estimated value and measured value was 0.94**; The monitoring model of [629, 901] [Y = 0.0986 exp(4.3696 [629, 901])] was best, and its veracity reached to 84.0%, RMSE was 0.55 kg m^-2, and the correlation coefficient between the estimated value and measured value was 0.96**, the two models above were the best among the elected models estimating cotton aboveground fresh biomass.

Key words: Cotton, Aboveround fresh biomass, Hyperspectral remote sensing, Estimation models