Relatively new accurate commercial point databases incorporating population distribution are now available which could enhance areal interpolation estimates required for an underlying population. This paper explores the level to which results will be dependent on how well the underlying population is represented. Using a number of different levels of detail to represent the population distribution (postcode units, enumeration districts (ED) population centroids and no population information), this paper shows the differences that can occur in computing average ED dose levels due to the degree to which the population is represented.
Although generally the differing methodologies gave similar overall patterns, there were substantial disparities between specific ED results obtained. As anticipated, the sensitivity of results were influenced by the degree to which the method's assumptions were adhered to. In the case of the ED population centroid approach, the extent to which the summarising points reflect the underlying spatial objects is critical to the estimated result. For areal weighting, the results were significantly influenced by the level of homogeneity within the variable of interest (population distribution). Generally, greater variation within the pollution surface exacerbates any breach in the methods underlying assumptions.