Intelligent spatial disaggregative interpolation of socio-economic data

Stan Openshaw				            Andy Turner
   stan@geog.leeds.ac.uk			           andy@geog.leeds.ac.uk 

Centre for Computational Geography
School of Geography
University of Leeds
LEEDS LS2 9JT


Abstract

The paper reviews the Disaggregative Spatial Interpolation Problem (DSIP) which concerns how best to transform spatial variable values for a specific source geography into values for a different target geography which has a much higher general level of spatial resolution.  The DSIP is a distinct form of Spatial Interpolation Problem (SIP) because of the significant additional difficulty which arises because the disparity in spatial scale between source and target data zones is extreme.  Various contemporary Areal Interpolation Methods (AIMs) are described and criticised in relation to the DSIP gradually introducing the need for a more integrating "intelligent" approach.  Intelligent Interpolation Methods (IIMs) find and use patterns in geographical data at and between different scales are outlined and some experiments IIMs based on Neural Networks (NNs) designed to disaggregate population are reported.