Abstract: A multi-scale model evaluation has been conducted in Latin America in 2015 (January and July) with the purpose to assess the performance of global and regional air quality models and set a pathway for an analysis and forecasting system in the region. Two global and three regional models were implemented and compared over a shared domain (120W-28W, 60S-30N) by different research groups in Latin America and Europe as part of the PAPILA (Prediction of Air Pollutants in Latin America) project (https://papila-h2020.eu/papila). Observations from local air quality monitoring networks in Mexico, Colombia, Ecuador, Perú, Chile, Brazil, and Uruguay were used for model evaluation. An ensemble, created from the median value of the individual models, was assessed on its performance. Most models were able to adequately reproduce air quality observations even with the adopted coarse resolution (0.2 x 0.2°) used in this work. However, no single model or ensemble had the best performance among pollutants and sites available. In general, ozone achieved high scores across sites while SO2 had the lowest performance. The median ensemble showed superior results for NO2 than individual models. Large urban areas (> 3500 km2 ) outperformed smaller urban zones (area between 500 and 1000 km2 ), suggesting an accurate portrayal of the temporal and spatial variability in large cities and the need for a finer resolution in smaller cities. This study identified certain limitations in global emissions inventories that should be addressed with the involvement and experience of local researchers. For example, NO2 and CO were substantially underestimated by all models in Bogota due to the lack of local emission sources in global inventories, especially from on-road sources. Similarly, PM2.5 levels were overestimated in Sao Paulo due to overestimation of biomass burning activity. This work will present the results of the first multiscale model intercomparison in Latin America, highlighting the main outcomes and challenges for the setting of an analysis and forecasting system for the region.

Keywords: model intercomparison, ensemble, Latin America, evaluation scores, air quality forecasting