基于冠层高光谱遥感的杂交水稻植被指数氮素营养诊断模型

Nitrogen Diagnosis Model of Vegetation Indices Based on Canopy Hyperspectral Remote Sensing for Hybrid Rice

以杂交水稻为研究对象,进行两因素裂区试验,主区为品种,副区为施氮水平,分析了4个植被指数(VIs)分别与叶片氮素含量(LNC)、叶片氮素积累量(LNA)和地上部氮素积累量(APNA)之间的相关性,并建立了以VIs为自变量的氮素营养诊断模型。结果表明,4个VIs和LNC、LNA之间均存在决定系数大于0.7的波段区域且波段区域一致,4个VIs和APNA之间的决定系数均较低,仅在0.2左右;比值植被指数(RVI)和LNC之间的决定系数最大值为0.886,对应的波段组合为694 nm和763 nm;垂直植被指数(PVI)和LNC之间的决定系数最大值为0.869,对应的波段组合为864 nm和483 nm;差值植被指数(DVI)和LNC之间的决定系数最大值为0.883,对应的波段组合为1 292 nm和1 258 nm;归一化植被指数(NDVI)和LNC之间的决定系数最大值为0.881,对应的波段组合为1 296nm和1 220 nm。最佳的氮素营养诊断模型为叶片氮素含量诊断模型,其模型表达式为LNC=1E+03NDVI2-132.55NDVI+3.72,建模集R^(2)、RMSE和RE分别为0.879、0.357%和16.267%,测试集R^(2)、RMSE和RE分别为0.895、0.331%和15.136%。

A split plot experiment of hybrid rice with two factors was carried out.There were two cultivars in the main plot and five nitrogen application levels in the sub plot.The correlation between four vegetation indices(VIs)and leaf nitrogen content(LNC),leaf nitrogen accumulation(LNA)and aerial part nitrogen accumulation(APNA)were analyzed respectively,and nitrogen diagnosis models with VIs as the independent variable were established.The results showed that there were band regions with coefficient of determination(r2)greater than 0.7 between the four VIs and LNC and LNA,respectively,and the band regions are consistent.The r2 between the four VIs and APNA were low,only about 0.2;the maximum r2 between ratio vegetation index(RVI)and LNC was 0.886,and the corresponding band combination is 694 nm and 763 nm.The maximum r2 between perpendicular vegetation index(PVI)and LNC is 0.869 and the corresponding band combination was 864 nm and 483 nm,the maximum r2 between difference vegetation index(DVI)and LNC was 0.883 and the corresponding band combination was 1292 nm and 1258 nm,the maximum r2 of normalized vegetation index(NDVI)and LNC was 0.881 and the corresponding band combination was 1296 nm and 1220 nm.The optimal nitrogen diagnosis model was LNC diagnosis model,and the model expression was LNC=1E+03NDVI2-132.55NDVI+3.72;the correlation index(R2),RMSE and RE of training set were 0.879,0.357%and 16.267%,respectively;the R2,RMSE and RE of test set were 0.895,0.331%and 15.136%,respectively.