尘埃影响下的植被指数修正研究

植被指数是表征植被覆盖，生长状况简单有效的度量参数。本文以城市绿化主要植被大叶黄杨为例，研究叶片滞尘对植被指数的影响，并构建植被指数修正模型对植被指数进行修正优化，提高植被指数的测量精度。研究选取北京城区为研究区，采集20个采样点的200个叶片样本，利用电子分析天平、ASD高光谱辐射仪及Win FOLIA叶面积仪，分别获取叶片尘埃量、光谱信息、叶面积等数据。通过对比分析样本叶片除尘前、后光谱特征及NDVI、NDWI、NDNI、NDII、CAI、PRI植被指数分布特征差异，结合单位滞尘量与光谱数据，构建植被指数修正模型，并对修正模型进行精度检验。结果表明：大叶黄杨叶片在除尘前与除尘后的光谱曲线均表现出典型的植被光谱特征，且蓝边、红边均出现在520和705 nm处，然而在350～700，750～1 350，1 500～1 850，1 900～2 100 nm波段范围内，滞尘对叶片光谱反射率影响显著，同时对植被指数也有较大影响;通过对滞尘量定量的研究分析发现，当尘埃质量增加时, NDVI和PRI植被指数与尘埃量的线性关系变弱，而NDWI，NDII，CAI植被指数与尘埃量依然保持明显的线性关系。修正模型NDVI，NDII，CAI，PRI精度验证决定系数(R2)分别为0.547，0.430，0.653，0.960，RMSE分别为0.035，0.020，0.112，0.009。研究结果表明对以后利用植被指数进行大面积植被反演、评估时，根据滞尘量影响进行修正优化，提高反演精度有一定参考意义。

The Study Based on Rectification of Vegetation Indices with Dust Impact

Vegetation indicesarethe simplest and most effective metric parameters representing the features of vegetation cover and growth condition. This paper used Euonymus japonicas Thunb as a study case and collected 200 leaf samples in 20 locations. Using electronic analytical balance and ASD hyperspectral radiometer with Win FOLIA leaf area meter obtainedthe data of the amount of dust, spectral information and leaf area. Through comparative analysis between dust and clean leaves, differences of spectral curve and vegetation indices were apparent. Then, combined with dust weight and spectral data, dust correction modelsfor vegetation indices were built. The analysis results showedthat the spectral curve between clean and dust leaves havetypical characteristics: blue edge and red edge were at 520 and 705 nm; however, dust influenced leaf reflectance significantly in range of 350～700, 750～1 350, 1 500～1 850, 1 900～2 100 nm wavelength, and had a greater impact on vegetation indices. With dust weight increasing, the linear correlation of dust with NDVI and PRI was non-significant, but that with NDWI, NDII and CAI was still significant. The verification of correction models showedthat coefficient of determination (R2) of NDVI, NDII, CAI and PRI were 0.547, 0.430, 0.653 and 0.96 and their root mean square error (RMSE) was 0.035, 0.020, 0.112 and 0.009 respectively. Furthermore, itshowed that applyingdust correction models can improve the accuracy of vegetation indices calculation.