RED EDGE BAND EFFECT ON ESTIMATING LEAF AREA INDEX IN TREE CANOPY

摘要

Leaf Area Index (LAI) is one of important biophysical parameters used in agriculture, forestry, ecology, hydrology and meteorology. Remote sensing estimation of LAI has been primarily relied on the relationship between canopy reflectance and LAI. Spectral indices, such as normalized difference vegetation index (NDVI) has been known to have positive correlation with LAI and have been widely used to estimate LAI over wide geographical area. Several studies have reported the saturation problem of canopy reflectance in high LAI, in which canopy reflectance or NDVI does not show the positive correlation under highly dense canopy situations. This study was designed to find alternative spectral region to solve this problem. The red edge band refers to the wavelength region where the spectral reflectance of vegetation increases between 680 nm and 750 nm, as the middle region between red and NIR wavelength. Red edge spectral region has been known to be useful to deliver additional information regarding vegetation characteristics. In recent years, there are a few satellite multispectral sensors that comprise red edge band. The RapidEye and WorldView-2 multispectral satellites contain a red edge band while the Sentinel-2A satellite has two or more red-edge bands of rather narrower spectral region. There were several studies showing that red edge band was useful to enhance the relationship with leaf chlorophyll concentration, vegetation stress and to classify vegetation types. In estimating LAI of several agricultural crops, red edge band was helpful to improve the relationship. In this study, we tried to analyze the relationship between red edge band reflectance and LAI of two tree species. A white oak (Quercus aliena) tree and a pitch pine (Pinus rigida) tree locating in Incheon city were targets of this study, which are some of dominant species in the Korean peninsula. Time series spectral measurements were performed on ground from May to October 2009. Spectral reflectance was measured using field spectro-radiometer (FieldSpec-3, ASD) to nadir that has 10 FOV and lnm bandwidth from 350nm to 2500nm. Spectral reflectance of canopy was measured 40m above ground using mobile crane which is about 20m height from top of the canopy. Spectral reflectance, canopy coverage, and LAI were measured simultaneously. LAI was measured using LAI-2000 (LICOR) optical instrument that measure LAI through ratio of irradiance between open area and under canopy. Continuous spectral reflectance were merged to simulate the red and near infrared (NIR) band reflectance of the wavelength region of the RapidEye spectral bands. Several red edge bands were simulated from 680nm to 750nm with the 40nm bandwidth. Using these band reflectance, NDVI and normalized difference red edge (NDRE) index were obtained. Simulated band reflectance and spectral indices were then compared with the time-series LAI. Since the two trees were older than 30 years, they show dense canopy closure and high LAI (3.0 to 8.4) even in early growing season in May. The correlation coefficients between the LAI and NDVI were relatively low (0.23 in white oak and 0.57 in pitch pine). The LAI relationships with the NDRE were greatly improved to 0.8 in both white oak and pitch pine. Among several red edge bands, the red edge band from about 700nm to 740nm showed the highest correlations. Although the results obtained from this study were premature and need further analysis, the use of red-edge band were effective to estimate LAI over highly dense canopy situation.