Kofi Bonsu

Titre de la thèse : “Urban hierarchy and the analysis of spatial patterns: towards explicit fractal modeling”

Directeur de thèse :BONIN Olivier       

Financement : Bourse de l’Etat du Ghana. Professeur assistant de l’Université des Mines et Technologies UMaT – Ghana.

Urban hierarchy and the analysis of spatial patterns: towards explicit fractal modeling

Landuse/landcover changes resulting from rapid urban growth is increasingly putting pressure on the natural environment as well as human welfare; and this has become a global concern. The high rate of urbanization which is being experienced all over the world has given rise to increased mobile travel and congestion, elevated levels of pollution, loss of farmland, duplicative infrastructure at high cost to society, limited employment accessibility and concentrated poverty (Kaim et al., 2018). It has therefore become necessary to monitor and control urban growth.

Fortunately, Remote Sensing provides a rich source of spatially consistent data that covers large areas with both high spatial detail and temporal frequency. Consequently, remote sensing has been used extensively in mapping urban areas and as a data source for the analysis and modelling of urban growth and land use change (Herold et al., 2003). Geographic Information System (GIS) and remote sensing have been successfully incorporated in urban modelling processes and have improved upon the analytical capabilities of the GIS techniques as well as provided modellers with a platform for data management and visualization (Liu, 2009).

Using GIS and Remote Sensing techniques, various land uses such as settlement, industrial, recreational and agricultural, the challenge, however, at the city scale, it becomes quite difficult to use classical image classification techniques to identify explicitly different urban structures (Bonin et al., 2015). Identification and analysis of the patterns of urban structures is very important for the modelling and classification of urban geographical development and landscape configuration as well as forecasting economic activities. Furthermore, building patterns convey important cognitive information and structural knowledge that facilitate map navigation and spatial reasoning (Yan et al., 2019).

Several researches have been conducted to aid the identification of urban structures and evolution of urban patterns. Bonin et al., 2015 used spatial calculus to identify urban structures on regular square grids. The identified urban structures were then organised using a graphical modelling language, chorems. Albert et al., 2018 simulated hyper-realistic urban patterns using generative adversarial networks trained with a global land-use inventory. Yan et al., 2019 used convolutional neural network to classify building patterns using spatial vector data.

This research aims to come up with novel techniques to explicitly classify remotely sensed data at a city scale for the purposes of urban modelling. Particular attention will be paid to the analysis of the urban hierarchy in centres and sub-centres, as well as to the spatial organization of these elements. The semantic analysis of urban patterns will eventually make it possible to model them with multi-scale primitives such as fractals.