Commercial air travel is a much-maligned miracle. Amid complaints over cramped seats and delayed flights, it has safely and extraordinarily connected the world. However, there are certain clouds on the horizon for aviation. This was exemplified by Singapore Airlines flight SQ321.
On May 21, 2024, the flight hit severe turbulence over Myanmar, injuring dozens of passengers and killing one.
Turbulence is caused by unstable air. It can be associated with storm clouds or physical barriers, such as mountain ranges, that force air upward. Flight data showed thunderstorms around the turbulent area encountered by flight SQ321. The most dangerous type of turbulence, however, is clear-air turbulence (CAT). These pockets of unstable air occur without a visible warning, making them harder for pilots to predict and prepare for.
As flight SQ321 tragically demonstrated, turbulence can be more than a minor inconvenience. It is already the most common cause of injury on U.S. commercial flights. These injuries, combined with wear and tear to airframes, cost the U.S. aviation industry approximately $200 million per year. Furthermore, this issue is expected to worsen with regional and global change in climate and associated atmospheric conditions, which are projected to increase the frequency of CAT over U.S. airspace.
Since this problem has been around for so long, there are already regulations and practices to address it. Turbulence is one of the reasons, for example, why the Federal Aviation Administration (FAA) mandates that passengers obey the seatbelt sign. Many turbulent conditions are observable, allowing pilots to take steps to avoid them in the air. Even though pilot observations are not entirely accurate, modern software in the cockpit complements their perception. Ever-improving weather forecasts can guide pilots to routes that avoid turbulence altogether.
However, the prospect of increased turbulence, including unobservable CAT, necessitates further-improved countermeasures. It is here that generative artificial intelligence, a subset of AI, can play a key role. Evidence suggests that this technology can produce accurate weather forecasts more efficiently than current models. Weather forecasting company Tomorrow.io has already released Gale, a generative AI that can produce forecasts given user inputs about weather conditions. Given the rapid pace of advancements in this technology, generative AI could eventually play a key role in helping airlines plan their routes to avoid more frequent turbulence.
There are several ways this type of weather forecasting could become a feature of the U.S. airline industry. Commercial airlines are certified under 14 CFR Part 119 and must use weather services from the FAA and National Weather Service (NWS), both of which could incorporate generative AI into their forecasts. The FAA is already using other types of AI for weather-related applications. The National Oceanic and Atmospheric Administration (NOAA), the parent agency of the NWS, has previously used other AI models for weather forecasts. Depending on the terms of their certification, Part 119 carriers can also use in-house and third-party weather forecasting services, which could include those utilizing generative AI.
As of now, the fate of generative AI in U.S. aviation largely depends on two bodies: the FAA and the National Institute of Standards and Technology (NIST), which is currently finalizing a Risk Management Framework (RMF) for generative AI. Although this RMF is not a binding regulation, it will influence how the FAA adopts generative AI and likely inform future federal regulations of this technology. Thus, the NIST should ensure that the RMF focuses on specifying desired outcomes regarding potential issues with generative AI, rather than micromanaging the processes to develop and implement it.
This decision would help the FAA and other agencies find the best and safest generative AI tools in the most efficient way possible. It would also allow airlines to do the same by guiding the FAA towards a more productive regulatory approach, precluding restrictive certifications that prevent airlines from using other weather services powered by generative AI. Given the potential productivity gains, airlines are likely to adopt this technology in some form regardless. Performance-based regulations would optimize their doing so.
The passenger death on flight SQ321 has been rightly mourned as a tragedy. What is surprising, however, is that it was the first turbulence-related fatality on a commercial flight in over 25 years. Innovations, effective regulations, and sound practices have taken the miracle of flight to unprecedented heights. Today, despite the threat of increased turbulence due to climate change, they retain the potential to do so again.
Written by Isaac Oh, Public Policy Intern
The Alliance for Innovation and Infrastructure (Aii) is an independent, national research and educational organization. An innovative think tank, Aii explores the intersection of economics, law, and public policy in the areas of climate, damage prevention, energy, infrastructure, innovation, technology, and transportation.