From Global to local: A multi-scale air quality modeling study over the Metropolitan Area of São Paulo

• Ano de defesa: 2020
• Autor(a) principal: Calderón, Mario Eduardo Gavidia
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Biblioteca Digitais de Teses e Dissertações da USP
• Palavras-chave em Português: Air pollution; Air quality models; CAM-Chem; Metropolitan Area of São Paulo; Modelos de qualidade do ar; MUNICH; Poluição do ar; Região metropolitana de São Paulo; WRF-Chem
• Link de acesso: https://www.teses.usp.br/teses/disponiveis/14/14133/tde-15042021-115527/

Air pollution is a multi-scale environmental problem that starts at the local scale, but its effects surpass the limits of cities in different scales of space and time. Measurements from air quality and meteorological stations are the main source of information on the state of the atmosphere. However, they are restricted in spatial coverage, limited in data interpretation, and are expensive to implement. Air quality models, by solving atmospheric motion equations and chemical reactions, offer an alternative approach to study air pollution by providing high temporal and spatial information of air pollutant concentration and meteorology. We used air quality models and emission inventories at different scales to study air quality in the Metropolitan Area of São Paulo (MASP). Output from the Community Atmosphere Model with Chemistry (CAM-Chem) global model is downloaded to be applied as dynamic chemical boundary conditions (CBC) for the Weather Research and Forecasting with Chemistry (WRF-Chem) community model. Then, WRF-Chem is used to simulate air quality at a regional and urban scale, considering Southeast Brazil and the MASP as simulation domains. Finally, the Model of Urban Network of Intersecting Canyons and Highways (MUNICH) is used to perform the first air quality simulation inside São Paulo urban canyons. During this process, we developed a new methodology to spatially disaggregate vehicular emissions based on total emissions and road length; and created a new tool to build WRF-Chem anthropogenic emission files. We also coded an R package to download and get high-quality data ready for analysis from the São Paulo State Environmental Agency air quality network that allows the automatization of model evaluation using all the available information. Results showed that CAM-Chem is suitable as CBC for WRF-Chem and can simulate coherently O3 and PM2.5 over the whole MASP with correlation coefficients greater than 0.7, but highly underestimates and fails to simulate primary pollutants. Both regional and urban WRF-Chem simulations achieved the meteorological benchmark of performance (e.g. ± 0.5 K mean bias of temperature, ± 10% mean bias of relative humidity, and ± 1.5 ms1 mean bias of wind speed). WRF-Chem presents an underestimation of primary pollutant with normalized mean bias (NMB) lower than -35 %, while O3 is best simulated achieving goal benchmarks with correlation coefficient of 0.83 and NMB of -5 %. MUNICH air quality simulation using WRF-Chem urban domain results as input improves NOX simulations with a NMB of -20 %. These simulations are an example of the capabilities that models have to address different scientific questions and how they can work to establish a multi-scale modeling system for air quality forecast. These tools allow the evaluation of air quality at different scales, the assessment of the efficacy of air pollution control policies, and the study of pollution health impact of pollutant exposure, even at street level.