Abstract
Most official population censuses release a total count of residents by census area, despite the fact that the distribution of population varies widely in space and time. Increasingly, better spatial representation of population distributions are required for many applications, especially at the local scale. Combining available statistical and physiographic data sets, a dasymetric interpolation approach was developed and tested to produce 2001 spatio-temporal representations of population distributions for the municipalities of Cascais and Oeiras in Portugal. This model was implemented within a geographic information system using streets as a spatial basis to allocate population. For each municipality, digital raster maps of nighttime, daytime, daytime residential, daytime worker and student, total daytime, and ambient population densities were produced at high spatial resolution. Quality assessment procedures confirmed that the method suits the objectives. However, the accuracy of results is mostly dependent on the quality of input data sets.