Hyperspectral Vegetation Indices for Estimation of Leaf Area Index

Spectral vegetation indices are frequently used to estimate vegetation biophysical/biochemical characteristics. In general they have been proposed to reduce spectral effects caused by external factors such as the atmosphere and the soil background. This study evaluated narrow band vegetation indices for estimation of vegetation leaf area index. The study takes advantage of using a dataset collected during a laboratory experiment. The spectral measurements have been carried out using a GER spectroradiometer. Leaf area indices were destructively acquired at the same time. Vegetation types sampled included four different types and sizes of leaves. For predicting leaf area index, five widely used vegetation indices were investigated. Narrow band vegetation indices involving all possible two band combinations of RVI, NDVI, PVI, TSAVI, and SAVI2 were computed. Cross-validation procedures were used to assess the predictive power of the regression model. We observed a significant relationship between the narrow band SAVI2 and the leaf area index (R =0.78, RMSE=0.57). All other narrow band indices respectively: RVI, NDVI, PVI and TSAVI had relatively lower R values (0.65≤ R ≤0.75) and higher RMSE values compared to SAVI2. Our results showed that bands from the SWIR region contain relevant information regarding to canopy LAI and are important for LAI estimation. The study demonstrates that hyperspectral data can be used to quantify leaf area index with a high accuracy.