Abstract This paper explores the application of space–time cubes (S–T cubes) and geospatial artificial intelligence (GeoAI) for monitoring vertical settlements in structures. Although S–T cubes are not commonly employed in this type of structural analysis, they enable the storage and visualization of multi-temporal monitoring data, offering an effective framework to represent differential settlement over time within a defined spatial domain, including at the scale of individual structural elements. The study further integrates GeoAI techniques for predictive analysis aimed at detecting discontinuities, leveraging the temporal datasets organized within S–T cubes. We employ an index that combines accuracy across multiple prediction steps with least-squares statistics of adjusted data. This provides users with a rapid diagnostic to verify the effectiveness of forecasts before deeper analysis. Results from three datasets—each representing a monitoring project with distinct characteristics—demonstrate that machine-learning-based forecasts remain reliable for at least three prediction steps ahead. This level of consistency is sufficient to support the detection of discontinuities in new monitoring campaigns, with a precision on the order of ± 0.2–0.3 mm.

The construction industry is a major contributor to global carbon emissions; therefore, finding a way to reduce emissions to meet climate-mitigation goals is vital. The challenges that impede emissions reductions are numerous and reliant on implemented policies. This study proposes an explainable machine learning framework to predict and interpret carbon emission reduction performance in building projects by integrating project characteristics, policy intervention indicators, certification levels, lifecycle information, and emission-related variables. Three machine learning models were created-Ridge Regression, Random Forest, and Gradient Boosting-and tested on a split of the data to assess their performance. To explain emissions reductions, the models were evaluated for predictive performance using a set of statistical measures and the explainable machine learning metric, Shapley Additive Explanations (SHAP). The models demonstrated robust predictive performance, with a coefficient of determination of over 0.81 for each model, and all models had similar performance despite using different statistical techniques. Shapley values attributed most of the focus to high-level green certifications and policy measures, whereas project size had little influence on reducing emissions. The results reinforce the idea that policies have the largest impact on emissions reductions in the building sector. The suggested explainable framework provides clarity and relevant insights into policies that aid evidence-based decision-making and strategic planning for decarbonising the building sector.