冬小麦生物量高光谱遥感监测模型研究

Biomass estimation model of winter wheat(Triticum aestivum L.) using hyperspectral reflectances

【目的】高光谱遥感能快速、实时、无损监测作物长势。研究不同氮磷水平下冬小麦不同生育时期地上部生物量高光谱遥感监测模型,可提高地上部生物量高光谱监测精度。【方法】在西北农林科技大学连续进行了5年田间定位试验,设置5个施氮水平(N,0,75,150,225和300 kg/hm^2)和4个磷施用水平(P2O5,0,60,120和180 kg/hm^2),选用不同抗旱类型冬小麦品种,测定了从拔节期至成熟期生物量与冠层光谱反射率,通过相关分析、回归分析等统计方法,建立并筛选基于不同植被指数的冬小麦不同生育时期生物量分段遥感监测模型。【结果】冬小麦生物量与光谱反射率在670 nm和930 nm附近具有较高相关性,在可见光和近红外波段处均有敏感波段;在拔节期、孕穗期、抽穗期、灌浆期、成熟期,生物量与归一化绿波段差值植被指数(GNDVI)、比值植被指数(RVI)、修正土壤调节植被指数(MSAVI)、红边三角植被指数(RTVI)和修正三角植被指数Ⅱ(MTVIⅡ)均达极显著相关性(P<0.01),相关系数(r)范围为0.923~0.979;在不同生育时期,分别基于GNDVI、RVI、MSAVI、RTVI和MTVIⅡ能建立较好的生物量分段监测模型,决定系数(R2)分别为0.987、0.982、0.981、0.985、0.976;估计标准误差SE分别为0.157、0.153、0.163、0.133、0.132;预测值与实测值间相对误差(RE)分别为8.47%、7.12%、7.56%、8.21%、8.65%;均方根误差(RMSE),分别为0.141 kg/m^2、0.113kg/m^2、0.137 kg/m^2、0.176 kg/m^2、0.187 kg/m^2。【结论】在拔节期、孕穗期、抽穗期、灌浆期、成熟期可以用GNDVI、RVI、MSAVI、RTVI和MTVIⅡ监测冬小麦生物量,具有较好的年度间重演性和品种间适用性。同时,分段监测模型较统一监测模型具有较好的监测效果及验证效果,能有效改善高光谱遥感监测模型精度。

【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 during2009–2014 at Northwest AF University, Yangling, China. Winter wheat cultivars were used as tested materials, and five N application rates（0, 75, 150, 225 and 300 kg/hm^2） and four P2O5 application rates（0, 60,120 and 180 kg/hm^2） 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 by2.0%–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 indexⅡ（MTVIⅡ）, 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 MTVIⅡ 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, 0.981, 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 0.187kg/m^2 at the jointing, booting, heading, grain filling and maturity stages, respectively. Therefore, vegetation indices of GNDVI, RVI, MSAVI, RTVI and MTVIⅡwere 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.