Towards adaptive green infrastructure: artificial intelligence and machine learning-based modelling for green roofs

In the pursuit of sustainable urban development, integrating Artificial Intelligence (AI) and Machine Learning (ML) with Vegetated Roofs (VRs)—including Green Roofs (GRs) and Purple Roofs (PRs) modelling offers a significant opportunity. This approach not only helps mitigate environmental challenges but also provides a strategy for optimizing urban ecosystems. In environmental engineering, AI and ML have enhanced traditional hydraulic and hydrological modelling, although significant research gaps remain. These gaps are particularly notable in our understanding of the interactions among vegetation, different VR layers, and microclimates, which offers substantial potential for further exploration and refinement. Most existing studies have focused on deterministic or empirical modelling approaches, lacking the adaptability and predictive accuracy offered by ML algorithms. Incorporating ML into VR hydraulic modelling could provide more robust and flexible tools for optimizing VR design and management strategies. This study aimed to explore the application of AI and ML techniques in VRs modelling with the goal of enhancing our understanding of VR performance dynamics, optimizing design parameters, and fostering more sustainable urban environments. To harness the continuous data stream from our experimental green roof setups in Western Sydney, we propose using ML algorithms for real-time analysis. This approach will facilitate the predictive modelling of VR performance metrics, focusing on stormwater quantity and improving temperature regulation, moisture retention, and plant growth. The application of Long Short-Term Memory (LSTM) on rainfall-runoff analysis for a particular day showed a significant correlation with the training data set, while a correlation matrix and K value clustering revealed meaningful correlations between several underlying meteorological factors and the rainfall event. By integrating AI and ML techniques with traditional hydraulic modelling methods, this study seeks to develop innovative approaches to facilitate the design, optimization, and management of VRs for enhanced stormwater management and urban sustainability, which could be a game changer in stormwater management.