基于SEVI的复杂地形山区植被FPAR遥感反演与地形效应评估

植物吸收性光合有效辐射分量（FPAR）的遥感反演是生态环境领域的核心研究内容之一,但在复杂地形山区,其估算精度严重受到地形效应的影响（包括本影与落影）。本文利用能够消除地形阴影影响的阴影消除植被指数（SEVI）对山区遥感影像进行FPAR反演,并分别与基于不同影像预处理程度计算的归一化植被指数（NDVI）、比值型植被指数（RVI）反演的FPAR做对比分析,以评估复杂山区反演FPAR存在的地形效应。结果表明：在不做地形校正的情况下,基于NDVI与RVI反演FPAR会使得本影及落影区域的值远小于非阴影区域的值,它们的相对误差均大于70%;基于C校正后的NDVI与RVI反演FPAR可以较好地校正本影区域,相对误差降至约6.974%,但落影处的校正效果不明显,相对误差约为48.133 %;而基于SEVI反演FPAR无需DEM数据的支持,可以达到经FLAASH+C组合校正后NDVI与RVI反演FPAR相似的结果,且能改善落影区域的地形校正效果,相对误差降至约2.730%。

Vegetation FPAR Retrieval based on SEVI in Rugged Terrain and Terrain Effects Assessment

The retrival of Fraction of Absorbed Photosynthetically Active Radiation (FPAR) by remote sensing is one of the major research fields in ecological environment. However, in mountainous areas with rugged terrain, the estimation accuracy is seriously affected by terrain effect, including the influence of self and cast shadow. In this paper, Shadow-Eliminated Vegetation Index (SEVI), which can effectually remove the influence of terrain shadow, was used to conduct FPAR inversion in mountainous areas from remote sensing data. The inversion result based on SEVI was compared and analyzed with the inversion results based on Normalized Difference Vegetation Index (NDVI) and Ratio Vegetation Index (RVI) with different pre-processing degrees and evaluate the topographic effect of FPAR inversion based on different indexes in complex mountainous areas from remote sensing images. The results show that the FPAR inversion based on NDVI and RVI have much smaller values in self and cast shadow areas than that in non-shadow area without terrain correction using DEM data, their relative error are both greater than 70%. C correction can be better used in the pre-processing of NDVI and RVI deriving and effectually corrected the FPAR inversion results based on these two indexes, its relative error dropped to about 6.974%. But the results after C correction not performed well in cast shadow areas, its relative error is about 48.133%. The FPAR inversion based on the SEVI without DEM data can achieve similar results with the FPAR inversion based on NDVI and RVI after the atmospheric correction of the FLAASH and C combination, and the result shows a better terrain correction effect in shadow area where relative error dropped to about 2.730%.