Despite the levels of air pollution in Macao continuing to improve over recent years, there are still days with high-pollution episodes that cause great health concerns to the local community. Therefore, it is very important to accurately forecast air quality in Macao. Machine learning methods such as random forest (RF), gradient boosting (GB), support vector regression (SVR), and multiple linear regression (MLR) were applied to predict the levels of particulate matter (PM₁₀ and PM_2.5) concentrations in Macao. The forecast models were built and trained using the meteorological and air quality data from 2013 to 2018, and the air quality data from 2019 to 2021 were used for validation. Our results show that there is no significant difference between the performance of the four methods in predicting the air quality data for 2019 (before the COVID-19 pandemic) and 2021 (the new normal period). However, RF performed significantly better than the other methods for 2020 (amid the pandemic) with a higher coefficient of determination (R²) and lower RMSE, MAE, and BIAS. The reduced performance of the statistical MLR and other ML models was presumably due to the unprecedented low levels of PM₁₀ and PM_2.5 concentrations in 2020. Therefore, this study suggests that RF is the most reliable prediction method for pollutant concentrations, especially in the event of drastic air quality changes due to unexpected circumstances, such as a lockdown caused by a widespread infectious disease.

Air pollution is a major concern issue for most countries in the world. In Portugal and Macao, the values of nitrogen dioxide (NO₂), particulate matter (PM) and ozone (O₃) are frequently above the concentration thresholds accepted as “good air quality.” Portugal follows the European Union (EU) legislation (Directive 2008/50/EC) on air quality and Macao the air quality guidelines (AQG) from the WHO. Air quality forecasts are very important mitigation tools because of their ability to anticipate pollution events, and issue early warnings, allowing to take preventive measures and reduce impacts, by avoiding exposure. The work presented here refers to the statistical forecast of air pollutants for three regions: Greater Lisbon Area, Madeira Autonomous Region (both located in Portugal), and Macao Special Administrative Region (in Southern China). The presented statistical approach combines Classification and Regression Tree (CART) and multiple regression (MR) analysis to obtain optimized regression models. This consolidated methodology is now in operation for more than a decade in Portugal, and is subject to regular updates that reflect the ongoing research and the changes in the air quality monitoring network. Recently, the same methodology was applied to Macao in collaboration with the Macao Meteorological and Geophysical Bureau (SMG). Here, a statistical approach for air quality forecasting is described that has been proven to be successful, being able to forecast PM₁₀, PM_2.5, NO₂, and O₃ concentrations, for the next day, with a good performance. In general, all the models have shown a good agreement between the observed and forecasted concentrations (with R² from 0.50 to 0.89), and were able to follow the concentration evolution trend. For some cases, there is a slight delay in the prediction trend. Moreover, the results obtained for pollution episodes have proven that statistical forecast can be an effective way of protecting public health.

Statistical methods such as multiple linear regression (MLR) and classification and regression tree (CART) analysis were used to build prediction models for the levels of pollutant concentrations in Macao using meteorological and air quality historical data to three periods: (i) from 2013 to 2016, (ii) from 2015 to 2018, and (iii) from 2013 to 2018. The variables retained by the models were identical for nitrogen dioxide (NO₂), particulate matter (PM₁₀), PM_2.5, but not for ozone (O₃) Air pollution data from 2019 was used for validation purposes. The model for the 2013 to 2018 period was the one that performed best in prediction of the next-day concentrations levels in 2019, with high coefficient of determination (R²), between predicted and observed daily average concentrations (between 0.78 and 0.89 for all pollutants), and low root mean square error (RMSE), mean absolute error (MAE), and biases (BIAS). To understand if the prediction model was robust to extreme variations in pollutants concentration, a test was performed under the circumstances of a high pollution episode for PM_2.5 and O₃ during 2019, and the low pollution episode during the period of implementation of the preventive measures for COVID-19 pandemic. Regarding the high pollution episode, the period of the Chinese National Holiday of 2019 was selected, in which high concentration levels were identified for PM_2.5 and O₃, with peaks of daily concentration exceeding 55 μg/m³ and 400 μg/m³, respectively. The 2013 to 2018 model successfully predicted this high pollution episode with high coefficients of determination (of 0.92 for PM_2.5 and 0.82 for O₃). The low pollution episode for PM_2.5 and O₃ was identified during the 2020 COVID-19 pandemic period, with a low record of daily concentration for PM_2.5 levels at 2 μg/m³ and O₃ levels at 50 μg/m³, respectively. The 2013 to 2018 model successfully predicted the low pollution episode for PM_2.5 and O₃ with a high coefficient of determination (0.86 and 0.84, respectively). Overall, the results demonstrate that the statistical forecast model is robust and able to correctly reproduce extreme air pollution events of both high and low concentration levels.

The levels of air pollution in Macao often exceeded the levels recommended by WHO. In order for the population to take precautionary measures and avoid further health risks under high pollutant exposure, it is important to develop a reliable air quality forecast. Statistical models based on multiple regression (MR) analysis were developed successfully for Macao to predict the next day concentrations of PM₁₀, PM_2.5, and NO₂. All the developed models were statistically significantly valid with a 95% confidence level with high coefficients of determination (from 0.89 to 0.92) for all pollutants. The models utilized meteorological and air quality variables based on five years of historical data, from 2013 to 2017. The data from 2013 to 2016 were used to develop the statistical models and data from 2017 were used for validation purposes. A wide range of meteorological and air quality variables were identified, and only some were selected as significant dependent variables. Meteorological variables were selected from an extensive list of variables, including geopotential height, relative humidity, atmospheric stability, and air temperature at different vertical levels. Air quality variables translate the resilience of the recent past concentrations of each pollutant and usually are maximum and/or the average of latest 24-hour levels. The models were applied in forecasting the next day average daily concentrations for PM₁₀, PM_2.5, and NO₂ for the air quality monitoring stations. The results are expected to be an operational air quality forecast for Macao.