Working Package 3
Advancing Modelling Techniques and Utilization of Emerging Datasets
Introduction
Over the past two decades, urban physical processes have been increasingly integrated into numerical weather prediction (NWP) models. However, their performance varies significantly across cities, especially struggling in tropical urban regions due to uncertainties in model initialization and physics. Moving from kilometer-scale to hectometer-scale (100 m) NWP offers many benefits for urban hazard modelling, including the ability to provide information at sub-neighborhood scales and a more accurate representation of convective precipitation. However, this transition also introduces new challenges, primarily due to the partial resolution of atmospheric processes such as turbulence. To fully realize the potential of hectometric models, significant work is still required to improve physics parameterizations and effectively incorporate data at this finer scale.
The Paris 2024 RDP intercomparison of urbanized hectometric NWP models revealed substantial model variability, particularly in simulating Urban Heat Islands (UHIs) and their associated heat plumes. Emerging technologies such as AI, machine learning, digital twins, and new data sources including IoT-enabled sensors, crowdsourced observations, and high-resolution satellite imagery provide opportunities to address these challenges. WP3 focuses on harnessing these advanced modelling techniques and integrating unconventional datasets to enhance forecasting capabilities for urban hazards, while testing their applicability in diverse geographical and data-scarce contexts.
Activities
- Develop hybrid AI–NWP models that can combine the strengths of traditional physics-based approaches with machine learning techniques, improving both timeliness and accuracy of hazard forecasts.
- Explore the development and use of city-scale digital twins to simulate hazards such as urban heat, air pollution, and flooding, while providing platforms for testing adaptation and mitigation strategies before implementation in the real world.
- Incorporate new and unconventional datasets into urban forecasting frameworks, including IoT-based environmental sensor networks, crowdsourced data from citizens, and satellite imagery, to improve real-time monitoring and model initialization.
- Apply advanced data assimilation methods to integrate diverse observational datasets into high-resolution urban models, addressing challenges related to turbulence, boundary layer processes, and heterogeneous urban morphology.
- Coordinate with WP2 to validate models against benchmarks of hazard predictability at multiple spatial and temporal scales, ensuring that improvements are measurable, transferable, and globally relevant.
- Assess the transferability and applicability of advanced modelling methods across different cities, with special focus on data-scarce regions, to ensure that innovations benefit not only resource-rich contexts but also vulnerable urban areas where forecasting capacity is most urgently needed.
