遥感是开展地面/近地面、 航空及航天层次无损伤探测植物叶绿素信息的主要手段。 目前多波段计算光谱指数方法已被广泛地应用于植被冠层叶绿素含量的经验/半经验反演及应用中。 考虑不同作物及同种作物不同品种间存在着一定的植被叶倾角分布(LAD)特征差异, 针对叶倾角分布对光谱指数反演冠层叶绿素含量(CCC)的影响进行分析, 并开展针对叶倾角分布变化不敏感的叶绿素相关光谱指数优选和冠层叶绿素反演建模研究。 基于PROSAIL辐射传输模型模拟了不同叶片叶绿素含量(LCC)、 叶面积指数(LAI)和LAD对应的冠层反射率数据。 模拟结果显示, 在相同LAI和LCC条件下, 不同LAD对应的冠层反射率有明显差异, 冠层反射率随着平均叶倾角的增加而降低。 通过计算12个常用的叶绿素相关光谱指数与CCC的相关性指标, 来评估光谱指数在不同LAD下反演叶绿素含量的敏感性差异, 并依次优选出MTCI, MNDVI8, MNDVI1和CIred-edge4个对LAD变化较不敏感的叶绿素相关光谱指数。 利用玉米实测数据对光谱指数进行冠层叶绿素估测的建模和模型检验, 模型的建立和验证结果显示, MNDVI8对LAD变化最不敏感, 反演模型的精度最高, 决定系数R2=0.70, 均方根误差RMSE=22.47 μg·cm-2。 CIred-edge(R2=0.63, RMSE=24.06 μg·cm-2), MNDVI(R2=0.66, RMSE=24.07 μg·cm-2)和MTCI(R2=0.65, RMSE=26.76 μg·cm-2)反演模型的精度较为接近并稍弱于MNDVI8。 通过对反演结果分析得出结论, 不同的光谱指数对LAD变化的敏感性不同, 优选的光谱指数普遍对叶绿素含量具有较好的相关性和敏感性, 其中MNDVI8受LAD影响最小, 能较高精度的反演LAD变化下的玉米冠层叶绿素含量。 优选的其他光谱指数MTCI, CIred-edge和MNDVI1反演能力虽然稍弱于MNDVI8, 但受LAD影响较小, 同样具有较好的反演能力。 该工作开展LAD对光谱指数叶绿素反演的敏感性分析和光谱指数优选研究, 其实测数据的检验结果和模拟数据的分析结果一致; 基于优选光谱指数的冠层叶绿素含量反演建模结果及精度分析结论, 对开展缺乏叶倾角分布差异先验知识下的大范围作物叶绿素含量遥感估测和应用具有借鉴意义。

Remote sensing is the main approach to carry out non-invasive detection of plant chlorophyll information on the ground/near ground, airborne and spaceborne levels. At present, spectral index for multi-band calculation has been widely used in empirical/semi-empirical estimation of canopy chlorophyll content. Taking the difference of leaf inclination angle distribution (LAD) between different crops and different varieties of homogeneous crop into consideration, this study analyzed the influence of LAD on retrieving chlorophyll content, and selected chlorophyll-related spectral indices that are insensitive to the variation of LAD and researched canopy chlorophyll retrieval model. PROSAIL radiative transfer model was used for simulating the canopy reflectance corresponding to different leaf chlorophyll content (LCC), leaf area index (LAI) and LAD. The simulation results showed that under the same LAI and LCC conditions, the canopy reflectance corresponding to different LAD was significantly different, and the canopy reflectance decreased with the increase of the average LAD. By calculating the correlation coefficient of 12 common chlorophyll-related spectral indices with CCC, the sensitivity of spectral indices in retrieval of chlorophyll content under different LAD was evaluated, consequently, four spectral indices that are insensitive to variation of LAD were selected: MTCI, MNDVI1, CIred-edge and MNDVI8.80 measured corn samples were utilized to model and validate the estimation models of CCC. Model establishment and verification results showed MNDVI8 was the most insensitive to variation of LAD, thus it was the best spectral index for estimating chlorophyll content with the coefficient of determination (R2) of 0.70 and the root mean square error (RMSE) of 22.47 μg·cm-2. The precision of CIred-edge (R2=0.63, RMSE=24.06 μg·cm-2), MNDVI (R2=0.66, RMSE=24.07 μg·cm-2) and MTCI (R2=0.65, RMSE=26.76 μg·cm-2) retrieval model was relatively close but was weaker than that of MNDVI8. Through analyzing the retrieval results, it was concluded that different spectral indices had different sensitivities to variation of LAD. The preferred spectral indices generally had the best correlation and highest sensitivity to chlorophyll content, among which MNDVI8 was least affected by LAD and can be used to retrieve CCC in corn under different LAD types. Although retrieval capabilities of MTCI, CIred-edge and MNDVI1 were slightly weaker than MNDVI8, they were less affected by variation of LAD and also had good retrieval capabilities. This paper researched the influence of LAD on retrieving chlorophyll content based on spectral indices, and the results from measured corn data were in accordance with those from simulated data. Based on the sensitivity analysis and validation results of canopy chlorophyll content retrieval models under different LAD types, this paper has a certain reference significance for remote sensing application of estimation of chlorophyll content in crop without prior knowledge of LAD in large scale.