基于高光谱的油菜叶面积指数估计

以冬油菜为研究对象,2014-2015年度设计了不同施氮水平直播油菜小区试验,在不同生育时期测量冠层光谱、土壤背景光谱以及叶面积指数(leaf area index,LAI),通过相关分析选取了12个光谱特征参数和11个植被指数,建立6叶期至角果期LAI的5种线性和非线性定量反演模型。结果表明:二次多项式反演模型比较适合估算油菜LAI苗期时以红边参数为代表的光谱特征参数,可准确估算出LAI;6叶期时红边幅值预测模型R~2为0.81,RMSEP为0.39,RPD为1.62;8叶期时红蓝边面积比归一化预测模型R~2为0.79,RMSEP为0.60,RPD为2.30;10叶期时红边幅值预测模型R~2为0.92,RMSEP为0.47,RPD为2.36;盛花期时蓝边面积预测模型R~2为0.87,RMSEP为0.34,RPD为2.57;角果期时以RDVI为代表的植被指数也可准确估算出LAI,预测模型R~2为0.74,RMSEP为0.57,RPD为1.36。油菜全生育期采用相同光谱特征参数、植被指数建模估计LAI精度明显降低,预测R~2远小于0.75,RMSEP大于0.65,RPD值均小于1.40,表明难以采用统一参数建模准确估计油菜全生育期LAI,不同生长时期需选择合适的光谱参数、植被指数分段建模估计LAI。

Prediction of rapeseed leaf area index based on hyperspectral data

Plot experiments of the winter rapeseed (Brassica napus L.) with different nitrogenous levels under direct seeding treatment were conducted in 2014-2015.The canopy spectral reflectance,soil background,LAI of each plot were measured at different stages.Correlation analysis between the canopy spectral reflectance and LAI was used to calculate eleven vegetation indices and twelve spectral parameters based on spectral position and area for optimizing five kinds of linear and nonlinear (logarithm,parabola,power and exponential) quantitative remote sensing inversion models to estimate LAI at the different and whole growth stages.The results showed that the quadratic polynomial inversion models perfectly estimated LAI of winter rapeseed using hyperspectral techniques.The spectral red edge parameters estimated accurately LAI at seedling stage.The predicted models based on Dr,NBR,Dr produced better estimation for LAI at six-leaf stage,eight-leaf stage and ten-leaf stage,respectively.R2 was 0.81,0.79 and 0.92 (P ＜ 0.01),respectively.RMSEP (root mean square error of predicted models) was 0.39,0.60 and 0.47,respectively.RPD (residual predictive deviation) was 1.62,2.30 and 2.36,respectively.The predicted models based on SDb and RDVI produced better estimation for LAI at full-bloom stage and pod stage with R2 of 0.87 and 0.74(P＜0.01),RMSEP of 0.34 and 0.57,and RPD of 2.57 and 1.36.The unified validation of models(R2＜0.75,RMSEP＞0.65,RPD＜1.4) showed that there was low prediction precision with the unified spectral variables or vegetation indices monitoring LAI at the whole stages of growth.The prediction accuracy of monitoring model based on the appropriate spectral variables and vegetation indices to estimate LAI at different stages of the winter rapeseed growth was high.