基于高光谱遥感的叶片总初级生产力GPP反演

为探究利用高光谱植被指数反演叶片总初级生产力(GPP)的模型,以湖北省武汉大学试验田油菜和小麦叶片高光谱反射率和光照强度(PARin)为数据源,利用7种植被指数与PARin的乘积分别反演2种植被叶片GPP,构建线性及非线性回归模型,并对模型进行验证。结果表明:1)从油菜生理特点出发,需要分生育期建模。在选择的7种植被指数中,花期SR构建的一次模型效果最优,建模和验模R2分别为0.80和0.82,RMSE不超过2.85g/(m~2·d);荚果期选择CIred edge和MTCI为优选模型,建模和验模R2为0.84和0.72,RMSE3.91g/(m~2·d);全时期基于红边波段的CIred edge、MTCI为优选模型,建模集R2达到0.80,RMSE3.67g/(m~2·d),验模R2达到0.65,RMSE3.92g/(m~2·d);2)小麦中NDVI模型效果最优,建模集R2=0.59,RMSE=2.80g/(m~2·d),验模R2=0.67,RMSE=3.39g/(m~2·d)。将油菜与小麦做对比,基于红边波段的植被指数CIred edge和MTCI对2种植被差异不敏感,R2为0.72~0.73,表明CIred edge和MTCI模型可以用于小麦和油菜叶片GPP的统一反演。

Remote estimation of leaf gross primary productivity based on hyperspectral data

Aiming to establish leaf Gross Primary Productivity inversion model based on hyperspectral Vegetation Indices,the hyperspectral reflectance and light intensity (PAR_in) of rape and wheat leaves in the experimental field of Wuhan University in Hubei Province were taken as data sources.Seven vegetation indices were respectively multiplied by PAR_in to get GPP.Linear and non-linear regression model was constructed and validated.The results showed that:1) Considering the physiological characteristics of rape,it was necessary to establish a sub-growth model.SR model was the best during the flowering stage,and the model and verification''s R² were 0.80 and 0.82,and RMSE was no more than 2.85 g/(m²·d).During pod phase,CI_{red edge} and MTCI were the preferred model,in the model and verification,R² were 0.82 and 0.72,and the RMSE was less than 3.91 g/(m²·d).As for the full stage,CI_{red edge} and MTCI were also the preferred model,R² reached to 0.80 and RMSE was less than 3.67 g/(m²·d).2) In wheat,NDVI model was the best,with R² of 0.59 and RMSE of 2.80 g/(m²·d).In the prediction set R² was 0.67 and RMSE was 3.39 g/(m²·d).The results showed that CI_{red edge} and MTCI were not sensitive to the two crops,R² was ranged from 0.72 to 0.73,indicating that CI_{red edge} and MTCI could be used for the unified GPP inversion in wheat and rape.