The impacts on living conditions and natural habitats deriving from planning decisions require complex analysis of cross-acting factors, which in turn require interdisciplinary data. At the municipal level, both data collection and the knowledge needed to interpret it are often lacking. Additionally, climate change and species extinction demand rapid and effective policies in order to preserve soil resources for future generations. Ex-ante evaluation of planning measures is insufficient owing to a lack of data and linear models capable of simulating the impacts of complex systemic relationships. Integrating machine learning (ML) into systemic planning increases awareness of impacts by providing decision-makers with predictive analysis and risk mitigation tools. ML can predict future scenarios beyond rigid linear models, identifying patterns, trends, and correlations within complex systems and depicting hidden relationships. This article focuses on a case study of single-family houses in Upper Austria, chosen for its transferability to other regions. It critically reflects on an ML approach, linking data on past and current planning regulations and decisions to the physical environment. We create an inventory of categories of areas with different features to inform nature-based solutions and backcasting planning decisions and build a training dataset for ML models. Our model predicts the effects of planning decisions on soil sealing. We discuss how ML can support local planning by providing area assessments in soil sealing within the case study. The article presents a working approach to planning and demonstrates that more data is needed to achieve well-founded planning statements.
Read full abstract