Dispersion of mobile emissions at sub-kilometer scales for the Aburra Valley in Colombia: emission disaggregation, identification of pollution hotspots, and scenario evaluation.

The dispersion of traffic emissions in the Aburra Valley, the second most populated urban area in Colombia, is examined by means of a passive tracer emulating carbon monoxide (CO) emissions with two types of dispersion models (Lagrangian and Eulerian). A simple top-down approach was implemented to disaggregate mobile emissions in space and time from the total annual emissions of CO in the entire area. The resulting emissions are fed to both dispersion models: the eulerian system consists of a WRF-CHEM configuration that treats CO emissions as a passive tracer (online); the lagrangian model is a passive particle stochastic tracking system run (offline) with the resulting high-resolution meteorology of the eulerian system. The area of interest is located in a narrow, highly-urbanized inter-Andean valley that requires a high spatial resolution to represent adequately the complex topography of the valley. The model set-up consists of 5 nested domains with horizontal resolutions ranging from 24.3 km in the outer domain to 300 m in the innermost domain, where the dispersion analysis is performed. The performance of the WRF model, including simulations with different boundary layer parameterizations, was evaluated using surface observation stations and a Microwave Profiler. We examined the potential of these model configurations in the identification of pollution hotspots, and studied the importance of the spatial resolution in the dispersion of pollutants. We further use the models to address the following questions:

- How are mobile emissions distributed in the valley?

- What is the role of the regional and local airflows in the dispersion of pollutants?

- What is the (simulated) impact of different air quality mitigation efforts?

Overall, we found that the lagrangian model is useful for source/sink analysis and to economically evaluate different air quality mitigation scenarios/measures. Although computationally more expensive, the WRF-CHEM set-up produces more realistic dispersion and can be useful for prognostic integrations of air quality episodes. We show that the methodology and results provide important insights for the distribution of pollutants in areas with limited air quality observation networks, and can be used to test the impact of different mitigation strategies.