Hyper-spectral Estimation of Forest Leaf Area Index from Earth Observing 1 (EO-1) Hyperion Imagery Based on Empirical-Statistical Approach and Grey Relational Analysis

Zhaoming Zhang,Tengfei Long,Hao Yang,Mengmeng Wang,Guizhou Wang,Guojin He,Xiaomei Zhang

doi:10.1145/3121360.3121361

Abstract

Leaf Area Index (LAI) is an important surface biophysical parameter as an input to many process-oriented ecosystem models. In the past two decades, much work has been done to estimate forest LAI using multi-spectral remotely sensed satellite imagery. However, LAI studies based on hyper-spectral satellite data are scarcely reported due to the difficulty to acquire high quality space-borne hyper-spectral data, especially in the rainy tropical and subtropical region. The aim of this paper is to perform LAI retrieval studies based on EO-1 Hyperion hyper-spectral satellite imagery in Yongan city, Fujian province, located in the Asian subtropical monsoon climate region. Hyperion imagery acquired on May 22, 2012 was employed in this study. Ground LAI measurements were collected using the Plant Canopy Analyzer (PCA), LAI-2000 in July, 2012. The empirical--statistical approach was mainly performed, and different modeling parameters, including different kinds of vegetation indices and the same vegetation index constructed from different combinations of Near InfraRed (NIR) and red bands, were evaluated against ground based LAI measurements. Totally seven typical vegetation indices were employed in this study, including the Normalized Difference Vegetation Index (NDVI), Simple Ratio (SR), Soil Adjusted Vegetation Index (SAVI), Modified Simple Ratio (MSR), Perpendicular Vegetation Index (PVI), Global Environment Monitoring Index (GEMI), and Non-Linear Index (NLI). Grey Relational Analysis (GRA) was also utilized to determine the sensitivity of these typical vegetation indices to LAI. Performance of the different modeling parameters were comprehensively compared, and the result shows that MSR and SR, constructed with bands 53 and 30, are the best predictors for LAI estimation in this study area, with the highest R2 (coefficient of determination) value of 0.63.

Full Text