Abstract
Dasymetric-mapping and pycnophylactic-interpolation methods have solid theoretical foundations and empirical supports in population-estimation research. Each of the methods has its own strengths, but also suffers obvious shortcomings. Dasymetric mapping makes good use of ancillary information to infer most likely population distribution, whereas it suffers from the unfounded assumption of uniform distribution of population among all eligible locations. Pycnophylactic interpolation warrants a smooth population surface in the study area without any presumption of uniform distribution. However, the method does not draw on information about real population distribution, so that its estimation accuracy cannot benefit from such useful information. In this paper, we develop a hybrid approach that takes advantage of the strengths and remedies the flaws of both methods. The hybrid method is tested with a case study. To evaluate the performance of the proposed hybrid method, this study compares its estimation accuracy with those of other popular methods including areal-weighting interpolation, binary dasymetric mapping and the pycnophylactic-interpolation method. The comparison results prove that the proposed hybrid method significantly outperforms the other methods. In addition, the study conducts a sensitivity analysis to examine the effect of search-radius size, which is the key parameter of the hybrid method, on estimation accuracy. The analysis result shows that the hybrid method can be further improved with appropriate choice of search radius.