A Novel Method to Model Spatial Distribution of Population

Abstract

This study proposes a simulation approach to model the spatial distribution of population (i.e., population density) in a given regioninto a great deal of detail. The approach is based on two mathematical notions: graph theory and fractal geometry. Accordingly, the approach takesself-similarity of road accessibility into account when simulating the spatial distribution of population. The study has followed three steps; (a) formulatea conceptual framework to model the spatial-distribution of population in a given area (i.e. a city or a region) by employingthe network centrality-based fractal dimension, (b) develop a modeling framework, and (c) calibrate the model utilizing empirical data at for five-selected case study areas: Colombo-Sri Lanka, Hanoi-Vietnam, Kolkata -India, Wuhu-China and Singapore. Network centrality-based fractal dimensions were computed by employing open-data and open-source GIS tools. The study applied threetypes of regression techniques: Robust Regression (RR), Ordinary-Least-Squares Regression (OLSR) and Poisson Regression (PR) for derivingthe most appropriate mathematical model. Then, the study validated the model by testing its prediction accuracy. Results revealed that the developed model to simulate spatial distribution of populationin a given area recorded an accepted level of accuracy (R2 >0.75) and predictability (MdAPE< 10%) on a par with the internationalspatial modeling standards. The proposed approach can be adoptedto simulate the spatial-distribution ofpopulation, particularly as a decision-making aid in the domain of urban & regional planning