AI can predict when an earthquake might strike — here’s how 

During the devastating earthquake that struck Myanmar and Thailand this weekend, horrified viewers around the world watched a 30-story Bangkok high-rise collapse in a scene reminiscent of the New York twin Towers on 9/11. The 7.7 magnitude tremor has killed more than 2,700 and caused widespread destruction in the region. 

The event was an unfortunate reminder of the extreme perils of seismic activity on the planet.  

In just the past five years, the earth has experienced 59 earthquakes of 7.0 magnitude or greater, causing over 70,000 fatalities. Even more astonishing are the “big ones,” such as the 2004 Indian Ocean earthquake, with over 227,000 fatalities, or the 2011 Tohoku earthquake and tsunami, which was the world’s costliest natural disaster, resulting in approximately $360 billion in property damage, as well as the ever-present environment contamination from the damaged Fukushima nuclear plant.  

These statistics are a cause for grave concern in the Pacific Northwest of the United States, where scientists predict that there is a 37 percent chance that a megathrust earthquake of 7.1 magnitude or greater will occur in the next 50 years. Such a natural disaster would cause extensive damage throughout the region, crippling infrastructure and causing citizens to live without access to fully functioning roads, water, electricity and sewage for many weeks.  

The last major quake in the region’s Cascadia Subduction Zone occurred in 1700, with an approximate magnitude of 9.0. The estimated likelihood of another event of that magnitude over the next 50 years is about 15 percent. While seemingly small, the impact would be enormous, with 1,100 earthquake fatalities anticipated, as well as 13,000 tsunami deaths and 24,000 individuals injured.  

The U.S. Geologic Survey (USGS) insists that earthquake forecasting is not possible. Thus, the agency only calculates the probability that a significant earthquake will occur, depicted in a hazard map for a specific area within a certain number of years. 

Such maps are only useful for long-term planning, however, so for early warning, USGS uses the ShakeAlert system. Though ShakeAlert does not provide actionable earthquake data, it is able to provide a few seconds of advance notice of destructive activity in regions 20 miles away from the epicenter, by taking advantage of the fact that the damaging “S” waves of a seismic event are slower than the preceding “P” waves. 

Countries with limited resources provide tangible examples of what a lack of earthquake data can do. Imagine if an early warning capability existed in Myanmar or Thailand, or in Haiti, where the death toll of its catastrophic 2010 earthquakes was estimated at 300,000. Incredibly, one does exist, yet the USGS refuses to consider it.  

Over a decade ago, scientists from NASA discovered that before earthquakes occur, sections of the earth’s crust near faults begin to behave like semiconductors, generating electricity proportional to the amount of stress near the faults. This changes the electric and magnetic fields of the earth’s surface, subsurface, atmosphere and ionosphere. Scientists in the U.S., Europe and around the world have verified the theory, and have published considerable peer-reviewed research papers and articles on the subject. 

One private company has applied artificial intelligence (AI) to the problem, achieving stunning results for four earthquakes exceeding 7.0 magnitude in 2024. Through AI-enabled ionospheric nowcasting, also known as space weather monitoring, earthquakes in Noto, Japan; Hualien, Taiwan, offshore of Noto; and offshore of Cape Mendicino,Calif., were predicted between three to four days in advance. 

President Trump has promised to make the federal government more efficient and effective, and there is no organization more in need than the USGS. Rather than reacting to dangerous seismic activity, the agency can be proactive by partnering with the sector to do a better job helping Americans prepare for earthquake hazards. 

Fortunately, with the Trump administration’s foresight on the importance of AI science and technology, there is an immediate avenue of approach. The administration is developing an AI action plan to define the priority policy actions needed to sustain and enhance America’s AI dominance. Including a public-private partnership to equip the USGS with a lifesaving earthquake forecasting capability should be a key component of the plan. 

With several days of advance notice before a large magnitude seismic event, emergency managers could pre-stage critical resources for immediate response, and the public could both prepare and be less inclined to panic.  

To date, Trump’s Department of Government Efficiency appears to be doing more harm than good with across-the-board firing of the federal workforce. This is a self-defeating approach that unintentionally impedes the administration’s stated priorities of national security, economic prosperity and public safety. The right way to get our government to do a better job is to directly target areas in which improvement is needed and leverage science and technology where demonstrable private sector innovations can meaningfully contribute to those priorities.  

Clearly, AI-enabled earthquake forecasting is a case in point that needs to be in Trump’s AI action plan, now. 

Rear Adm. (ret.) Tim Gallaudet, Ph.D., is a former acting and deputy administrator of the National Oceanic and Atmospheric Administration (NOAA), acting undersecretary and Assistant Secretary of Commerce, and oceanographer of the Navy. As the CEO of Ocean STL Consulting, LLC, he is a strategic adviser for the space weather nowcasting and earthquake prediction company precursor SPC.