Cities need fast, trustworthy tools that show how transport choices ripple into air quality and day-to-day operations. High-fidelity simulators capture the science but are too slow for rapid “what-if” analysis or on-the-fly decision support. This project builds physics-aware, data-driven AI surrogates that aim to reproduce key simulator outputs while running in seconds. The first chapter focuses on a small-area NOx pilot in Birmingham using ADMS to prototype the pipeline, metrics, and uncertainty handling. Later chapters scale to CMAQ/WRF-Chem and multi-pollutant fields (e.g., NO₂, O₃, PM₂.₅) at broader scales. The overarching question is how to preserve essential physical structure and uncertainty credibility while integrating behaviour, infrastructure, and meteorology into decision-useful models.

• Curate traffic activity, fleet mix, network/street-canyon descriptors, and meteorology; generate diverse ADMS scenarios for NOx across typical and stress conditions for the City Of Birmingham in the United Kingdom.

• Train compact surrogates to emulate ADMS outputs at street to neighbourhood scales, preserving hotspots, diurnal patterns, and gradients.

• Validate with multi-metric criteria: errors for levels, structural similarity for spatial fields, peak recall for hotspots, temporal skill for rush hour dynamics, and basic physical checks (non negativity, simple mass/flow consistency).

• Generalise the pipeline to regional episodes from CMAQ/WRF-Chem and to multi-pollutant fields, reusing the same evaluation playbook.
• Integrate behaviour and infrastructure (adoption, policies, charging availability) to support realistic scenario analysis.
• Embed the surrogate in a carbon-aware routing module that weighs travel time, emissions outcomes, and reliability under uncertain scenarios (charger outages) and adverse weather.
• Provide a simple API that returns maps and time series with uncertainty for rapid “what-if” exploration by planners and operators.
• Document compute budgets, configurations, and datasets for reproducibility, enabling smooth hand-off to collaborators.

• Rapid scenario screening: explore policy and operational options in minutes, not days.

• Hotspot-aware planning: preserve street-scale detail so interventions target the right places and times.

• Operational decision support: robust, carbon-aware routing with quantified risk under outages and weather.

• Scalable blueprint: a validated NOx pilot that facilitates expansion to multi-pollutant, regional decision tools.

ClimBridge connects transport behaviour and infrastructure to air quality outcomes and routing choices through fast, physics aware surrogates. It starts small using an ADMS-based NOx pilot in Birmingham, then scales to CMAQ/WRF-Chem and multi-pollutant fields across larger regions. The result is a decision tool that preserves essential physics, quantifies uncertainty, and supports both planning and operations.

• Position/magazine article (in preparation): ClimBridge: fast physics-aware surrogates for transport–air quality decisions.
• Survey Paper (ACM Computing Surveys, in preparation): AI surrogates for environmental simulations: from city-scale dispersion to decision-ready digital twins.

You’ll receive the latest updates on CLEETS research and programming.