Change detection of forest crown closure using an inverted geometric-optical model and scaling

Abstract

The physical based Li-Strahler geometric-optical model can be inverted to retrieve forest canopy structural variables. One of the main input variables of the inverted model is the fractional component of sunlit background (Kg). Kg is calculated by using pure reflectance spectra (endmembers) of the viewed surface components. To detect the forest canopy changes by the inverted geometricoptical model combined with the scaling approach in large areas at high temporal frequency is the main objective of this paper. We firstly extract the viewed surface components endmembers from the low spatial resolution MODIS data in two different years by a regional scaling-based linear unmixing model using the corresponding medium spatial resolution Landsat TM images; secondly, derive and map one of the forest structural properties, crown closure (CC), by inverting the Li-Strahler geometric-optical model based on the extracted endmembers; thirdly, complement prediction deficiencies of the inverted Li-Strahler model CC by using a spatial interpolation algorithm (regression kriging) in the infeasible regions; finally produce the spatially continuous CC maps and representing CC changes in two years. This methodology is illustrated with an example application in the Three Gorges region of China.