Effect of neighbourhood and its configurations on urban growth prediction of an unplanned metropolitan region

Abstract

Rapid urbanisation, especially in developing countries like India, has resulted in unplanned and haphazard urban expansion. With saturated urban cores, growth is observed in the peri-urban areas, resulting in severe challenges for urban planners. The present study aims to study the urban growth patterns of the fast-growing Mumbai Metropolitan Region (MMR) using the Landsat data from 1999 to 2019 and to evaluate the neighbourhood configurations’ effect on urban growth prediction. The urban area maps are classified using a maximum likelihood algorithm and are used along with the potential drivers to test three levels of neighbourhood considerations. The first model assumes no neighbourhood effect, the second incorporates the built-up pixels in the neighbourhood as an additional potential driver variable, and the third uses a Cellular Automata (CA). The CA model explores variations in neighbourhood types and sizes, distance decay and iterations to identify the optimal configuration. The results show an 89.44% increase in built-up areas over two decades (1999-2019). The urban growth prediction model testing reveals the importance of neighbourhood, with the first model without neighbourhood consideration giving the least accuracy (67%) while the inbuilt neighbourhood model gives better results (71%). However, the CA-based model with a 9 × 9 Moore neighbourhood, distance exponent β = 2 and two iterations give the highest accuracy (76%). The growth prediction shows a new wave of peri-urban growth in MMR, with overall urban areas increasing by 25% between 2019 and 2029 and 20% between 2029 and 2039. The results provide urban planners with a valuable tool for informed decision-making and promoting sustainable development.