Fine-resolution spatial analytics are essential for urban planning and policy-making, yet traditional small-area estimation often struggles with sparse, hierarchical, or imbalanced data. This paper introduces a Spatially Regularised Bayesian Heterogeneous Graph Neural Network (SR-BHGNN) that integrates multiple census tract levels within a unified framework. The model builds a heterogeneous graph where nodes represent spatial units at different scales, edges encode adjacency or membership, and Bayesian inference quantifies uncertainty in parameters and predictions. A spatial regularisation term, inspired by Tobler’s First Law of Geography, penalises large discrepancies between neighbouring nodes, reducing errors in imbalanced datasets and ensuring coherent local estimates. We evaluate SR-BHGNN through two London case studies, population estimation and PM 2.5 prediction, comparing it against random forests, single-level GNNs, and spatial hierarchical Bayesian estimation. SR-BHGNN achieves strong performance gains, with classification accuracies of 0.85 for population estimation and 0.81 for PM 2.5 prediction. Its Bayesian design produces posterior distributions that capture uncertainty, enabling policy-relevant insights into vulnerable neighbourhoods or priority intervention zones (e.g. low-emission areas). These results demonstrate that SR-BHGNN advances the state of the art in small-area estimation, offering a flexible, uncertainty-aware framework for diverse urban analytics applications.

Geospatial artificial intelligence (GeoAI) is reshaping our understanding of Earth and urban systems by integrating advanced artificial intelligence techniques with diverse geospatial data and methodologies. This backstory highlights recent GeoAI advances and applications as presented in the 11 articles in the iScience special issue, “GeoAI shaping earth and cities: Advances, opportunities, and challenges.” Guest editors share perspectives on GeoAI’s advances in explainability, adaptability, and sustainability, demonstrating that GeoAI’s applications extend beyond traditional mapping functions. These 11 case studies illustrate four types of explainability, three levels of adaptability and three thematic areas of sustainability, showing the methodological diversity and practical relevance of GeoAI for Earth and urban systems. Here, interactions among these dimensions are mapped to support the evaluation and design of future GeoAI solutions. We also outline future research directions for GeoAI to address complex challenges across the sciences relating to the Earth and its cities.

Urban Digital Twins (UDTs) integrate multilayered spatial data to support urban planning and climate adaptation efforts. Although UDTs have advanced significantly in data integration and predictive modeling, their usability remains underexplored. This study evaluates the Baselining, Evaluating, Action, and Monitoring (BEAM) platform, a web-based UDT developed by the National University of Singapore (NUS) for thermal comfort analysis. Through usability testing with ten urban planners and researchers, five from Singapore and five from the United States, this research assesses navigation, data interpretation, and integration into real-world workflows. Participants completed guided tasks as well as usability surveys, revealing key challenges in data navigation, interface clarity, and analytical flexibility. User Experience Questionnaire (UEQ) results showed high scores for attractiveness (1.67) and stimulation (1.70), but lower ratings for perspicuity (0.75). The findings highlight the need to improve affordances, contextual information, and the ability to interpret complex data sets. Minor regional differences in usability preferences also emerged, particularly in seasonal analysis and measurement units. By bridging technical advances with practical usability insights, this study contributes to the development of more accessible and effective UDT platforms. The findings inform future design improvements to enhance the adoption of UDTs, ensuring that these tools better support urban planners, policy makers, and researchers in climate adaptation and decision-making processes.

The prediction of thermal and noise-based preferences in the urban context is valuable in characterizing interventions to mitigate the challenges of health, productivity, and satisfaction of urban dwellers. The growth of crowd-sourced data and data-driven techniques provides an opportunity to increase the understanding of which machine learning models are most accurate and applicable for this context. This paper outlines the results of a machine learning competition aiming to enhance the accuracy of predicting human comfort in the city context. The competition asks contestants to use contextual data to predict noise distraction and thermal preference in various indoor and outdoor spaces. This competition included the city-scale collection of 9,808 smartwatch-driven micro-survey responses that were collected alongside 2,659,764 physiological and environmental measurements from 98 people using an open-source watch-based platform combined with geolocation-driven urban digital twin metrics. This paper explains the two best solutions to this competition and provides a discussion of the factors that may have contributed to their accuracy of more than 0.7 in multiclass tasks. These solutions notably included the use of LightGBM, XGBoost, CatBoost, and simple Neural Networks while avoiding overly complex solutions such as deep learning or recurrent architectures, which offer limited advantages for structured data classifications.

High-resolution microclimate maps are critical for advancing urban climate resilience strategies by providing detailed spatial insights into environmental variables. This study evaluates the performance of using a random forest regressor with single and multiple features to fit interpolated environmental data, including solar radiation, temperature, humidity, and wind speed. Using regression kriging on observations from 42 weather stations distributed on the campus of the National University of Singapore (NUS), spatial environmental variables were mapped at a 5×5 meter resolution. The interpolation process uses various geospatial layers to refine the results and the fine-scale spatial resolution. As a result, this model captures the localized variability of environmental variables. This work contributes to urban climate modeling by advancing the methodological frameworks for microclimate mapping and addressing the growing need for reliable high-resolution environmental data to inform thermal comfort assessments and resilience strategies.