玉米LAI和FPAR的高光谱遥感预测模型研究

Hysperspectral Estimation of Maize LAI and FPAR

摘要:

运用高光谱探测技术,获取不同密度下两个不同品种春玉米冠层不同时期的光谱特征反射参数,对高光谱原始反射率、反射率对数、反射率一阶导数、反射率二阶导数以及DVI、RVI、NDVI、SAVI 4种植被指数与玉米的LAI和FPAR进行相关分析,并建立两者之间的预测方程,通过对模型检验选出最佳预测模型。研究表明,一阶导数和4种植被指数都能很好地预测LAI和FPAR,其中,对LAI预测最好的是利用光谱一阶导数建立的指数函数模型,对FPAR预测最好的是利用NDVI建立的二次多项式模型。851 nm处的一阶导数对玉米的整个时期变化都比较敏感,LAI的最佳预测模型在玉米的整个时期都具有较好的预测性;855、758 nm波段组合的NDVI对FPAR的变化比较敏感,但当FPAR较大时,模型对其预测能力降低。

Abstract:

Hyperspectral data of two spring maize varieties in different densities at different stages were obtained with hyperspectral detection technology, correlation analysis were made and prediction equations were established between LAI/FPAR and the original hyperspectral reflectance, reflectance logarithm, reflectance first derivative and second derivative else with DVI/RVI/NDVI/SAVI four vegetation indexes. The result showed that first derivative and four vegetation indexes all can predict LAI/FPAR very well, the best prediction model of LAI was the reflectance first derivative and the best prediction model of FPAR was NDVI. The first derivative of 851 nm was sensitive to changes of the whole period of maize, the best prediction model of LAI had a good predictability in the whole period of maize. NDVI of band combination(855, 758) was sensitive to the change of FPAR, but when FPAR become larger, the lower the prediction ability was.