Global and local endmembers spectral patterns to estimate the urban structure in Porto Alegre (Brazil) in Landsat images using the Linear Spectral Mixture Model procedure

Abstract. This study investigates the spectral patterns of global and local endmembers to assess urban structure estimation within Porto Alegre city, Brazil. Two distinct sets of endmembers were examined: one derived from global patterns acquired from 100 Landsat image subsets and the other from local patterns gathered across 27 analyzed images. Results reveal that the utilization of global endmember patterns tends to overestimate the presence of urban shadow in all analyzed images compared to estimates derived from local patterns. Consequently, this overestimation impacts the relative proportions of other endmembers. Despite the disparities in magnitude between fractions estimated with each set of endmembers, temporal variations exhibit similar interannual trends. The findings of this study suggest that, beyond the choice between standardized or localized models, the features that must be represented in the resulting maps should also be considered in the selection between global and local endmembers.