Clustering Urban Multifunctional Landscapes Using the Self-Organizing Feature Map Neural Network Model
Yang Gao,Zhe Feng,Yang Wang,Jin-Long Liu,Shuang-Cheng Li,Yu-Kun Zhu
DOI: https://doi.org/10.1061/(asce)up.1943-5444.0000170
IF: 2.361
2014-01-01
Journal of Urban Planning and Development
Abstract:Multifunctionality in urban ecosystems has received much attention in the last decade from researchers and policy makers. This paper provides research on urban multifunctional landscape clustering, using the city of Shenzhen, China, as a case study. Utilizing the self-organizing feature map (SOFM) neural network model, six different landscape functional indices were identified, and urban multifunctional landscape regionalization produced five major units. According to SOFM clustering results, each region had its respective primary function, such as gas regulation, water supply, human nature regulation, soil environmental regulation, economy, and cultural priority. The gas regulation ecological supporting region (Zone I) covers 490.5km2, with long coastline form a nature-dominated, less human-influenced physical environment; the water supply ecological supporting region (Zone II) is 25.8km2, and river network density reaches 0.986km/km2, supporting function of water conservation and water supply; the mountain forest environmental regulating region is Zone III, 377.7km2 with substantial forest cover; covering the largest area of 547.8km2, zone type (IV) represents soil regulation region; the fifth zone type (V), on the west coast, is definitely a human-dominated region. The results show that the human and nature interfaced peri-urban region is the most affected and threatened area in the city. Under the control of urban sprawling local regulation, the urban population growth would slow down, but there is no convincing evidence that the limitation of build up land have negative influence on the urban economy. Thereafter, the authors analyzed the functions of each unit and compared the SOFM clustering technique with the traditional K-means clustering method. The result revealed that both methods are effective and appropriate for regionalization of urban multifunctional landscapes, but SOFM has advantages in identifying spatial patterns. Finally, approaches to achieve sustainable urban development were illustrated and their importance highlighted for policymakers and stakeholders.