Comparative assessment of Two Source of Landsat TM vegetative fraction coverage for modeling urban latent heat fluxes

Traditional methods for the extraction of VFC using vegetation indices were found to have large uncertainty due to its sensitive to the surface heterogeneous characteristic. This study presents an improved Spectral Mixture Analysis (SMA) approach of Landsat TM data to map the VFC for modeling of urban heat fluxes, in the case of Beijing, China. Two widely used models (Two-Source model (TSEB) and Pixel Component Arranging and Comparing Algorithm (PCACA)) were adopted for model evaluation. A comparative analysis between NDVI-derived and SMA-derived urban VFC showed that the latter achieved more accurate VFC values for complex urban regions, with an better accuracy of 1.61% in Root mean squared error (RMSE) and 1.03% in Mean absolute error (MAE). Moreover, the SMA-derived urban VFC could be utilized to produce a higher precision in urban latent fluxes relative to the NDVI-derived urban VFC when used as input to both TSEB and PCACA model.