Measuring urban landscapes for urban function classification using spatial metrics. (January 2020)

• Record Type: Journal Article
• Authors: Xing, Hanfa; Meng, Yuan
• Is Part Of: Ecological indicators. Volume 108(2020)
• Publication Date: 2020-01
• Subjects: Urban function -- Classification -- Urban landscape -- Spatial metric -- Urban form; Environmental monitoring -- Periodicals; Environmental management -- Periodicals; Environmental impact analysis -- Periodicals; Environmental risk assessment -- Periodicals; Sustainable development -- Periodicals; 333.71405
• DOI: 10.1016/j.ecolind.2019.105722
• Languages: English
• ISSNs: 1470-160X

Abstract

Highlights: A novel framework for urban function classification is presented by measuring urban landscapes using spatial metrics. 8 Spatial metrics are presented based on 6 types of landscape elements to measure urban landscapes. Spatial metrics are evaluated based on their sufficiency and effectiveness using partial analysis and analysis of variance. A conditional inference random forest approach is proposed to build an automatic urban function classification model. Abstract: The rapid extension of urban landscapes has substantially contributed to the widespread use of methodologies for the classification of urban functions. While previous research has focused solely on measuring the spatial arrangement of restricted urban landscape elements using spatial metrics, to date, the detailed landscape characteristics that enable different functions to be distinguished have not been discussed. This paper presents a novel framework for urban function classification by measuring urban landscapes using spatial metrics. To capture features related to urban functions rather than measuring limited physical configurations, spatial metrics that quantify multiple urban landscape elements and their interactions are proposed. Then, these metrics are evaluated based on their sufficiency and effectiveness in classifying urban functions using a partial analysis and analysis of variance. Finally, a conditional inference random forest approach is proposed to build an automatic urban function classification model with spatial metrics. The model was implemented in Futian District, Shenzhen, Guangdong Province. Using 8 spatial metrics based on 6 types of urban landscape elements, the results show a satisfactory testing accuracy of 0.757 and an overall classification accuracy of 0.813. The proposed method provides a novel framework for effectively delineating urban landscape characteristics and accurately classifying urban functions.