No abstract

This paper examines and updates the rank-size distribution of cities and municipalities in Bangladesh between 1990 and 2019 based on two criteria: (1) built-up urban areas; and (2) population. The distribution of built-up urban areas and population are compared to provide a robust theoretical underpinning of Zipf’s law for future urban developmental planning framework. The data on built-up urban areas is extracted from land cover classification using Google Earth Engine and the population data is obtained from the decennial censuses. The comparison of the conformity to Zipf’s law indicated contradictory results. While a greater proportion of the population has been increasingly concentrated in the smaller and midsized cities over the last three decades, built-up urban areas, on the other hand, have been mostly clustered in two largest cities— Dhaka and Chittagong—accounting for 50 to nearly 60 percent of the total built-up urban areas. These results shed light on the magnitude of continued spatial inequalities in urban development amongst cities and municipalities in Bangladesh despite there being an overall increase of evenness in the distribution of population over time. These results imply an unsustainable rate of urban expansion in Bangladesh and reinforce the need for the exploration of policies and regulations targeted at guiding the rate and direction of evenness in urban expansion.

With the rapid and unregulated nature of urban expansion occurring in Chattogram, Bangladesh, the adoption of urban growth restriction mechanisms such as the urban growth boundary (UGB) can provide a robust framework necessary to direct the development of built-up areas in a way that curtails the growth in environmentally sensitive areas of the city. Using a support vector machine (SVM)-based urban growth simulation model, this paper examines the areas of future contiguous expansion of the city to aid in the delineation of the UGB. Utilizing landcover, topographic, and population density data from a variety of sources for the past twenty years, the SVM method with the radial basis function (RBF) kernel is used to develop a model based on fourteen predictor variables. A grid-search is used to tune the hyperparameters and determine the best performance combination of the hyperparameters for the RBF kernel function used in the SVM. The final SVM model using the best performance combination of the hyperparameters indicates a high percentage agreement of 91.79% and a substantial agreement for the Kappa coefficient of 0.7699. The developed SVM simulation model identifies potential areas that are more likely to undergo urban expansion in Chattogram in the next twenty years and provides aids for a stringent and strict delineation of UGB for this region.