Artificial intelligence in urban forestry—A systematic review

• Published in: Urban forestry & urban greening 2021-12, Vol.66, p.127410, Article 127410
• Main Authors: César de Lima Araújo, Henrique, Silva Martins, Fellipe, Tucunduva Philippi Cortese, Tatiana, Locosselli, Giuliano Maselli
• Format: Article
• Language: English
• Subjects: Deep learning; Governance; Green infrastructure; Machine learning; Nature-based solutions; Urban planning; Urban trees
• ISSN: 1618-8667; 1610-8167
• DOI: 10.1016/j.ufug.2021.127410

•Urban forestry is becoming increasingly complex in the urbanizing world.•New methods are required for planning and managing the natural capital of cities.•Artificial Intelligence (AI) in Smart Cities points to new research avenues.•AI supported different aspects of urban forestry research leading to robust results.•This review points to a bright future of AI in urban forestry and decision making. Environmental quality and the citizens' well-being largely depend on the urban forests. But managing this natural capital is challenging for its biological complexity and interactions with other environmental, social, and economic aspects of the cities. In line with the current digital revolution with the rise of Smart Cities, the use of Artificial Intelligence (AI) is becoming more common, including in urban forestry. In this systematic review, we evaluated 67 studies on the interplay between AI and urban forestry surveyed on Science Direct and Scopus to provide an overview of the state of the art and identify new research avenues. The sample includes studies in 23 countries and 85 cities, including 5 megacities, comprising the remote assessment of canopy cover and species distribution; ecosystem services assessment; management practices; and socioeconomic aspects of urban forestry. Most studies focused on extant urban forests, with few examples evaluating temporal trends, and only one focused on future scenarios despite the predictive potential of AI. A total of 22 AI methods were employed in these studies. Only half of them point to clear advantages of the chosen methods, such as robustness against missing data, overfitting, collinearity, non-linearity, non-normality, the combination of discrete and continuous variables, and higher accuracy. The choice of these methods depends on the various combinations of aim, timescale, data type, and data source. The application of AI in urban forestry is in full growth and will support decision making to improve livability in the cities.