Investigation of the likelihood of green infrastructure (GI) enhancement along linear waterways or on derelict sites (DS) using machine learning

Abstract

Studies evaluating the potential for green infrastructure (GI) development using traditional Boolean logic-based multi-criteria analysis methods are not capable of predicting future GI development in dynamic urbanscapes. This study evaluated both artificial neural network (ANN) and adaptive, network-based fuzzy inference system (ANFIS) algorithms in conjunction with statistical modelling to predict green or grey transformation likelihoods for derelict sites (DS) and vacant sites along waterway corridors (WWC) in Manchester based on ecological, environmental, and social criteria. The soft-computing algorithms had better predictive capacity at 72% accuracy versus the 65% of logistic models. Site sizes, population coverage, and air pollution were identified as the main influencers in the potential for site transformation. In Manchester, the likelihood of GI transformation was higher for WWC than derelict sites at 80% versus 60% likelihood, respectively. Furthermore, DS were more likely to transform into grey development based on current trends and urban planning practice.