Forest leaf area index determination: A multiyear satellite-independent method based on within-stand normalized difference vegetation index spatial variability

The Leaf Area Index (LAI) and its spatial distribution are key features to describe the forest ecophysiological processes. A stable and reproducible relationship is obtained between the LAI and the standard deviation σNDVI of the pixel‐based satellite‐derived normalized difference vegetation indices (NDVI) of forest stands. In situ measurements of LAI have been performed with the LAI‐2000 Plant Canopy Analyser over 8 years in the managed Fontainebleau forest (France) on about 31 stands each year, including oak, beech, and mixed oak‐beech stands. Simultaneous satellite images have been acquired, atmospherically and geometrically corrected, and included into a geographic information system to get the mean NDVI and the σNDVI for each stand. A total of six different satellites with a 20‐ to 30‐m spatial resolution have been considered over the eight studied years: SPOT1, SPOT2, SPOT4, LANDSAT ETM+, IKONOS, and HYPERION. The mean LAI of a stand is linked to the σNDVI with a unique relationship that appears to be mostly year‐ and satellite‐independent, because the σNDVI is nearly insensitive to additive or proportional shifts on NDVI. The theoretical bases of the σNDVI‐LAI relationship are investigated. The results show the combined importance of the shape of the within‐stand LAI distribution (following a Weibull probability density function) and the shape of the within‐stand LAI‐NDVI curves (showing a saturation). The root mean square error of the predicted LAI over the 259 samples is 1.14 m2/m2 when all years and satellites are considered, using the following equation: LAI = −2.45 ln(σNDVI) − 5.58 (r2 = 0.63).