An artificial intelligence approach for identifying efficient urban forest indicators on ecosystem service assessment

Abstract

Urban trees provide multiple ecosystem services (ES) to city residents and are used as environmentally friendly solutions to ameliorate problems in cities worldwide. Effective urban forestry management is essential for enhancing ES, but challenging to develop in densely populated cities where tradeoffs between high ES provision and issues of periodic disaster-caused risks or maintenance costs must be balanced. With the aim of providing practical guidelines to promote green cities, this study developed an AI-based analytical approach to systematically evaluate tree conditions and detect management problems. By using a self-organizing map technique with a big dataset of Taipei street trees, we integrated the ES values estimated by i-Tree Eco to tree attributes of DBH, height, leaf area, and leaf area index (LAI) to comprehensively assess their complex relationship and interlinkage. We found that DBH and leaf area are good indicators for the provision of ES, allowing us to quantify the potential loss and tradeoffs by cross-checking with tree height and the correspondent ES values. In contrast, LAI is less effective in estimating ES than DBH and leaf area, but is useful as a supplementary one. We developed a detailed lookup table by compiling the tree datasets to assist the practitioners with a rapid assessment of tree conditions and associated loss of ES values. This analytical approach provides accessible, science-based information to appraise the right species, criteria, and place for landscape design. It gives explicit references and guidelines to help detect problems and guide directions for improving the ES and the sustainability of urban forests.