Biomass estimation model of winter wheat (Triticum aestivum L.) using hyperspectral reflectances
HE Jia,LIU Bing-feng,GUO Yan,WANG Lai-gang,ZHENG Guo-qing,LI Jun
DOI: https://doi.org/10.11674/zwyy.16173
2017-01-01
Abstract:[Objectives] Hyperspectral remote sensing can rapidly and nondestructively acquire vegetation canopy information.The objectives of this study were to establish wheat biomass estimation model based on winter wheat (Triticum aestivum L.) canopy hyperspectral reflectances with different rates of nitrogen or phosphorus application,and to improve the forecast precision of the biomass estimation model at different growth stages of winter wheat in the Loess Plateau of China.[Methods] Field experiments were carried out during 2009-2014 at Northwest A&F University,Yangling,China.Winter wheat cultivars were used as tested materials,and five N application rates (0,75,150,225 and 300 kg/hm2) and four P2O5 application rates (0,60,120 and 180 kg/hm2) were set.Biomass and canopy hyperstpectral reflectances were measured at the jointing,booting,heading,grain filling and maturity stages,respectively.The biomass monitoring models were constructed using correlation and regression methods.[Results] The biomass of wheat from the jointing to maturity showed a parabolic curve,and the maximum biomass was at the seed filling stage.When nitrogen or phosphorus application was sufficient,the canopy hyperspectral reflectances of wheat were reduced by 2.0％-5.0％ in the visible wavelength (P ＜ 0.05),and increased by 3.0％-21.0％ in the near infrared wavelength (P ＜ 0.05).There were significant (P ＜ 0.01) correlations between the biomass and green normalized difference vegetation index (GNDVI),ratio vegetation index (RVI),modified soil adjusted vegetation index (MSAVI),red edge triangular vegetation index (RTVI) and modified triangular vegetation index Ⅱ (MTVI Ⅱ),the range of the correlation coefficient was from 0.923 to 0.979 at different growth stages.The monitoring models based on GNDVI,RVI,MSAVI,RTVI and MTVI Ⅱ produced better estimation for biomass at the jointing,booting,heading,grain filling and maturity,respectively,and precision values of prediction R2 were respectively 0.987,0.982,0.98 l,0.985 and 0.976 (P ＜ 0.01),and standard errors (SE) were respectively 0.157,0.153,0.163,0.133 and 0.132.Meanwhile,the relative errors (RE) of the measured values and predicted values were 8.47％,7.12％,7.56％,8.21％ and 8.65％,and the root mean square errors (RMSE) were 0.141,0.113,0.137,0.176 and 0.187 kg/m2 at the jointing,booting,heading,grain filling and maturity stages,respectively.Therefore,vegetation indices of GNDVI,RVI,MSAVI,RTVI and MTVIⅡ were the most suitable indexes for monitoring winter wheat biomass at the jointing,booting,heading,grain filling and maturity stages,respectively.[Conclusions] The five tested vegetation indices show high precision in predicting the biomass of winter wheat at different growth stages,which means they can be used for monitoring biomass of winter wheat in large areas of the Loess Plateau.