Digital twin-driven insights into microclimate in tropical urban environments

Abstract

Urban areas in tropical regions face rising thermal stress due to increasing temperatures and urbanization. This paper introduces a novel Digital Twin (DT) framework developed as part of a campus sustainability initiative designed to improve microclimate conditions and outdoor thermal comfort in the tropical environment. This DT integrates high-resolution LiDAR-derived urban morphology, detailed vegetation surveys, and data from 40 weather stations into a real-time microclimate monitoring and prediction system. Several machine learning models were evaluated, with Extreme Gradient Boosting (XGB) achieving the best performance on predicting outdoor air temperature, yielding root mean square errors (RMSE) below 0.3°C at night and 0.6°C during daytime. The final model was deployed via Amazon Web Services (AWS), enabling interactive temperature prediction in DT and scenario-based planning such as shading and vegetation adjustments. Compared to traditional static modeling approaches, this DT platform offers improved spatial and temporal resolution, scalability, and responsiveness to dynamic environmental conditions. The proposed framework provides a practical solution for supporting climate-resilient urban design and has the potential for broader applications across tropical cities.