New research from Texas A&M University has used artificial intelligence to track pollution caused by accidental chemical emissions during weather events such as heavy rain or lightning strikes.
According to the Understanding the effects of natural hazards on chemical emission incidents using machine learning techniques study, an unplanned emissions event or unscheduled maintenance that results in unauthorized emissions of air pollutants is known as a chemical emissions incident. Extreme weather events frequently cause chemical emissions incidents, with some weather conditions causing more incidents than others.
Researchers used AI to analyze chemical emissions incident reports and weather data collected from the Houston area over the past 20 years.
“In this study, we pursued a data-driven understanding of how climate extremes elevate the likelihood of excessive industrial emissions,” said Dr Qingsheng Wang, a professor of chemical engineering at Texas A&M. “This understanding is laying the groundwork for predictive tools allowing regulators and operators to anticipate natural hazard-triggered technological accidents.”
Dominant predictors of pollution
The results indicate that precipitation and lightning are the two dominant predictors of chemical emissions incidents, and therefore of increased pollution. Flooding associated with heavy rain or hurricanes drives equipment failure, while lightning-induced power loss forces emergency flaring, the report found.
“Lightning and rainfall aren’t just weather forecast items; they’re leading indicators of pollution spikes,” commented Haoyu Yang, a chemical engineering PhD student at Texas A&M.
Equipped with the knowledge that lightning and precipitation are the most likely weather variables to lead to a chemical emissions incident, researchers and local officials can be more prepared for these conditions and even find ways to prevent them.
Forecasting ‘high-risk’ days allows agencies to issue early warnings and reduce public exposure to carcinogens and smog precursors released during chemical emissions incidents, the study suggests.
Improving resilience
The researchers also found that identifying trends can support the implementation of evidence-based policies with the quantification of lightning and precipitation-driven risks supporting targeted upgrades, like backup power and flood-proofing at facilities located near neighborhoods.
This can bolster emergency preparedness and community resilience, which may already be occurring according to researchers’ observations.
“We discovered that the strength of climate-incident correlations drifts over time, hinting at improving industrial resilience, potentially due to post-[Hurricane] Harvey upgrades,” said Yang. “This weakens previously strong climate signals in the later years of the study. Identifying changes in the data over time is now a priority for future research.”
The research was produced through a collaboration between the Artie McFerrin Department of Chemical Engineering and the Texas A&M University Department of Geography.
In related news, forecasters from NOAA’s National Weather Service have reported the number of expected named storms in the Atlantic with winds of 39mph to be 13-18 in this year’s hurricane season (June 1-November 30), with 5-9 expected to become hurricanes and 2-5 major hurricanes. Read the full story here