Improving population estimates using remotely sensed and ordnance survey datasets

The accuracy of population data is of critical importance in supporting the design of public and private-sector facilities. Demographic data are usually supplied by national census organisations at pre-defined census output levels. However, demographic datasets may be required at user-defined spatial units that can be different from the initial census output levels. A number of population estimation techniques have been developed to address these problems. This thesis is one of those attempts aimed at improving small-area population estimates by using spatial disaggregation models of: 1) binary mapping, 2) address-weighted dasymetric and 3) volumetric estimation models. These interpolation approaches employs high-resolution aerial imagery, LiDAR-derived building volumes and the integration of building address points and occupancy information sourced from the Ordnance Survey © and Airbus Defence and Space. Census wards and output areas were used as source zones and target zones respectively, to estimate population counts in Leicester City and the Borough of Kensington and Chelsea, London where the population is distributed both horizontally and vertically. The predicted population values were compared with 2011 census of actual population datasets. Each method employed in the study generated different population estimates depending on their assumptions and required datasets. The accuracy appears to be mainly influenced by the type and quality of the ancillary datasets and also the interpolation method adopted. Based on the disaggregation models adopted in this study, the address-weighted model produced the best population estimates with Root Mean Square Error (RMSE) value of 0.64 and R2 score of 0.998 for the City of Leicester and RMSE value of 0.236 and R2 score of 0.997 for the Borough of Kensington and Chelsea. This estimation is an indication that building address point datasets that contain information on occupancy can be used within Dasymetric mapping approaches to improve population estimates over a range of urban areas.