Unmixing a Variable Endmember Linear Mixture Model for Estimation of Heat Island Mitigation by Green Vegetation

Leaf area index within vegetated surface and its fractional area to a target pixel (fraction of vegetation cover) are two important factors for estimation of cooling potential by green vegetation in the context of mitigating heat island effects. This paper explains a technique to retrieve these two parameters from satellite images utilizing three bands of visible, near-infrared, and shortwave infrared for better estimation of mitigation potential. The technique assumes linear mixture model with variable end member approach in which end member spectra of vegetation and non-vegetation targets are varied continuously by representing these spectra as a function of leaf area index and soil brightness, respectively. A numerical algorithm is examplained to unmix the variable endmember LMM. The proposed technique was applied to the data acquired by LANDSAT7-ETM+ to discuss its validity on retrievals of the two biophysical parameters.