In recent years Southeast Asia has seen a significant increase in the intensity and frequency of haze events, or days in which visibility falls below 10 kilometers. Caused by airborne particulates known as aerosols, such low-visibility days reduce air quality and endanger human health. The main sources of the pollution are human activities that produce fires—biomass burning chief among them—and those that do not, such as fossil fuel combustion, construction and road dust. Air pollution mitigation measures are urgently needed to address the problem, but will only be effective through a better understanding of the relative contributions of fire and non-fire aerosols to the region’s air quality.
Toward that end, a team of researchers from the Massachusetts Institute of Technology and collaborating institutions in Singapore and Hong Kong has developed a model that could provide decision-makers with a useful breakdown of air quality impact by emissions source. Using the model to simulate fire emissions, non-fire emissions and a combination of both over an area covering the Association of Southeast Asian Nations (ASEAN) 50 cities during the years 2002-2008, the researchers assessed the relative contributions of different emissions sources to low-visibility days (LVDs) throughout that period.
Their study, which appears in the journal Atmospheric Chemistry and Physics, produced the following results: 39 percent of LVDs in the region were attributable to either fire or non-fire aerosols alone (i.e., on those days, the atmospheric concentration of either aerosol was sufficient to reduce visibility to below 10 kilometers); a further 20 percent of LVDs were attributable to non-fire aerosols alone; another 8 percent to fire aerosols alone; and another 5 percent to a combination of non-fire and fire aerosols. The model was unable to determine the emissions sources for a remaining 28 percent of LVDs.
These results show that the main driver of observed LVDs in 50 ASEAN cities is more likely aerosols from non-fire anthropogenic sources. The majority of the region’s population is thus exposed to air pollution that is primarily attributable to sources other than biomass burning.
Nonetheless, biomass burning remains an important factor in determining overall air quality. Using the 24-hour PM2.5 Air Quality Index (AQI) as a standard measure of air quality, the researchers also found that above and beyond the simulated result of the standalone non-fire emissions case, the combined fire and non-fire case can significantly increase the likelihood of a “moderate or unhealthy pollution level” AQI from 23% to 34%. In other words, under certain conditions biomass burning can singlehandedly tip the scales toward health-hazardous air quality.
“Our study indicates that while biomass burning should be minimized, emissions control of non-fire sources such as fossil fuel combustion must be a key component of Southeast Asia air pollution mitigation policy,” says Hsiang-He Lee, lead author of the study and a research scientist in the research group of supervising author Chien Wang, a senior research scientist at the MIT Joint Program and Department of Earth, Atmospheric and Planetary Sciences.
The model used in the study consists of three components—the Weather Research and Forecasting (WRF) model, an atmospheric chemistry/aerosol component (WRF-Chem), and observational data including surface visibility from the Global Surface Summary of the Day (GSOD) and aerosol measurement data from the Surface PARTiculate mAtter Network (SPARTAN). It builds upon a previous study led by Wang’s research group that did not include a full chemistry and aerosol modeling component. This added capability enabled the researchers to perform more precise assessments of the relative contributions of fire and non-fire aerosols to the region’s air quality, and to perform the AQI analysis.
Among other things, this new analysis revealed that in major Southeast Asian cities, premature mortality resulting from particulate pollutant-driven air quality degradation increased from about 4110 per year in 2002 to about 6540 per year in 2008.
Looking ahead, the study showed how machine learning algorithms can be used to forecast the occurrence of haze events and thereby guide preventative air quality mitigation measures that could minimize economic losses from such events. Applying six different machine learning algorithms, the researchers were able to predict severe haze events (visibility less than 7 kilometers) with an accuracy of more than 80 percent. The algorithms achieved this level of forecast reliability by correlating historical observational data of meteorological conditions and fire activity with past severe haze events.
“It is also almost impossible to forecast haze events with typical weather models due to the lack of timely emissions data as well as high computational cost and significant model uncertainty,” says Wang. “Machine learning avoids these drawbacks by going directly to historical, observational data to search for potential patterns or features that can be correlated with haze events and then exploited to predict future occurrences of such events.”
The study was funded through the Singapore-MIT Alliance for Research and Technology (SMART) Center for Environmental Sensing and Modeling (CENSAM) by the National Research Foundation of Singapore; and the U.S. National Science Foundation and Department of Energy.
Photo: Southeast Asia shrouded by smoke (Source: NASA Earth Observatory)