Detecting city-hinterland pairs with a community detection algorithm

Dissertation submitted in partial fulfilment of the requirements for the Degree of Master of Science in Geospatial Technologies With the increasing urbanization, cities tend to grow together with the infrastructure networks beyond their administrative boundaries. Identifying the right dimension of the city is thus challenging for the geographers. One way to represent a city could be identifying the city core and its hinterland as a part of the city itself. Since the transport network is used to deliver the resources between the city and its hinterland, we can hypothesize that the connectedness between the city and its hinterland is represented by the transport network topology. Hence, it would be justifiable to consider the transport network topology as a foundation for identifying hinterlands. One way to achieve this task of identifying hinterland utilizing the transport network topology is by using community detection algorithms. This thesis presents and evaluates a methodology for defining city-hinterland pairs at a regional level using community detection algorithms applied to the transport network. Easily accessible OpenStreetMap(OSM) road network data was used as the data for our research. We used three modularity optimization based community detection algorithms to identify communities in the road network. The identified communities were assigned to city core areas and hinterland based on the proposed criteria. Results from different algorithms were compared among each other and with the OECD Functional Urban Areas(FUAs). Similarity among the results between the three algorithms was acceptable with the average Goodness-of-fit (GOF) scores of 0.60 and 0.65 for core urban area and hinterland respectively. With OECD FUAs results showed less similarity with average GOF scores of 0.40 and 0.31 for core urban area and hinterland respectively. The disimilarity of our result with OECD data is justifiable as the OECD FUAs use population data and commuting data of administrative units without considering transport connectivity but we use only road network as the basis for identifying city and hinterland communities.