Estimation of Overstory and Understory Leaf Area Index by Combining Hyperion and Panchromatic QuickBird Data Using Neural Network Method

This paper presented a neural network method for combining Hyperion and panchromatic QuickBird data to retrieve the Overstory and Understory Leaf Area Index (OU-LAI) of forest stands in the Longmenhe forest nature reserve in China. A field survey was firstly carried out to collect thirty sampling sites located in typical forest stands in the study area. Then a Multi-Layer Perception artificial neural network model was used to integrate hyperspectral domain fusion and high spatial domain fusion techniques so as to deal with the non-linear canopy scattering between overstory and understory vegetation. Various combination of selected twenty-one optimal Hyperion bands and panchromatic QuickBird data were tested to evaluate the OU-LAI retrieval accuracy. The results show that nine Hyperion bands (3 VIS, 3 NIR and 3 SWIR bands) and standard deviation (SD) within each 50 by 50 mobile window from panchromatic QuickBird data have the best retrieval results for OU-LAI. This study also indicates that the non-linear artificial neural network can be utilized to retrieve OU-LAI parameters in the forest stands by combining Hyperion data and panchromatic QuickBird data. Improvements of neural network method using additional remote sensing information for retrieval OU-LAI are discussed.