Multiresolution analysis of urban reflectance

Quantitative analysis of urban reflectance characteristics can provide insights into the spatial and temporal variations in urban reflectance that modulate solar energy fluxes through the urban environment. Characterization of urban reflectance may also facilitate development of algorithms for classification of urban landcover. The characteristic scale of urban reflectance variations is generally between 10 and 20 m and therefore comparable to the ground instantaneous field of view of most operational sensors. This results in spectral heterogeneity that limits the accuracy of pixel-based classifications of urban imagery. Multisensor analyses of urban reflectance indicate that many urban areas can be described as mixtures of three or four spectral endmembers. Most urban building materials lie along a continuum between high and low albedo endmembers but vegetation is a persistent and distinct component of the urban mosaic. Nonlinear mixing generally diminishes with increasing vegetation cover such that linear mixing models may be generally appropriate for urban vegetation estimation. Preliminary analyses of Landsat imagery in a number of urban areas worldwide suggests that linear mixing models with three and four endmembers may be generally applicable to urban areas. Spectral heterogeneity at scales of 10s of meters may be a more consistent characteristic of urban reflectance than any single reflectance spectrum.