Comparison among normalized vegetation indices for the determination of LAI

Abstract

Airborne hyperspectral images collected over San Rossore Natural Park (Pisa, Italy) by the Multispectral Infrared and Visible Imaging Spectrometer (MIVIS) on June 21st, 2000 were analyzed in order to assess the best indices for forest LAI estimation. Hemispherical photography was used for ground truth measurements, simultaneously with the overflights, in hardwood and conifer stands characterized by a LAI ranging between 1.2 and 4.5. All band combinations expressed as simple ratios and normalized indices (a total of 89 single bands, and 7832 and 3916 indices, respectively) were linearly correlated to LAI in order to detect the best correlations. Determination coefficients were analyzed by means of a graphical matrix to highlight relevant spectral regions. Normalized indices composed by the red chlorophyll absorption wavelength (680 nm) and the wavelengths after the green reflectance peak (580-640 nm) in the orange region were strongly correlated to LAI. Best results were obtained with the newly proposed Orange Slope Vegetation Index [OSVI=(ρ 620 -ρ 680 )/(ρ 620+ ρ 680 ), R 2 =0.88, RMSE=0.5). The index performed better than the normalized difference vegetation index (NDVI=(ρ 780 -ρ 680 )/(ρ 780+ ρ 680 ), R 2 =0.47) Using SAIL radiative transfer model 1 , canopy reflectance at different viewing angles and a wide range of LAI was simulated in order to verify the sensitivity of OSVI and NDVI. For LAI between 0.25 and 8 both indices resulted stable for viewing zenith angles between -60° and +60°. OSVI, being saturated with values greater than 4, could be used to estimate a wider range of LAI than NDVI. Application of GeoSail model 2 resulted in a good agreement between simulated and measured OSVI.

Coupled topographic and vegetation patterns in coastal dunes: Remote sensing observations and ecomorphodynamic implications

“…To this end, we first compare vegetation/topographic characterizations obtained from satellite optical remote sensing (such as hyperspectral‐derived normalized difference vegetation index, NDVI), airborne laser altimetry (such as lidar‐derived leaf area index, LAI), and field observations. Although the simultaneous computation of NDVI and LAI, and their comparison, has been explored in other ecosystems [e.g., Bernasconi et al ., ; Jin and Eklundh , ], this comparison had not yet been performed for short vegetation species typical of coastal systems. Next, we use the resulting spatial distributions of ecological and topographic properties to identify and study the outcome of coupled ecological and sediment transport processes.…”