Most Earthquake Forecasting Models Miss Location Targets
A new study reveals that while models accurately predict earthquake frequency, spatial prediction accuracy lags at 66%.
Updated on Oct. 29, 2026 in Geology
Live Poll
Do you trust current earthquake forecasting models to accurately predict seismic risks in your area?
In an August 2026 study published in Nature Communications, researchers evaluated 27 earthquake forecasting models using data from California. The retrospective analysis assessed the accuracy of 51,625 forecasts against 597 actual seismic events recorded between 2007 and 2018.
Why it matters
Understanding the reliability of seismic forecasting models is essential for improving public safety and infrastructure planning. This evaluation helps distinguish which mathematical approaches most effectively capture the complex spatial and magnitude patterns of real-world earthquake activity.
The study analyzed 51,625 prospective forecasts against 597 seismic events of magnitude 3.95 or greater. While 95% of models matched the frequency and magnitude distribution, only about two-thirds correctly predicted the specific locations of the recorded earthquakes.
The players
Collaboratory for the Study of Earthquake Predictability
A research consortium that conducts standardized, prospective testing of seismic forecasting models.
University of Bristol
A research-intensive institution that published the study on earthquake model performance.
The details
The Collaboratory for the Study of Earthquake Predictability—a research consortium focused on testing seismic forecast models—performed the assessment by comparing daily projections against actual seismic activity. Researchers evaluated the models using a 24-hour window, tracking whether each forecast accurately predicted the count, size, and geographic placement of tremors. The study highlights a clear performance gap between forecasting numerical likelihood versus pinning down the precise spatial coordinates of fault ruptures.
Timeline
The testing period for the models spanned from August 2007 to August 2018.
The findings were published in Nature Communications in August 2026.
The University of Bristol released news regarding the study on October 15, 2026.
The Tech Race
This study updates the performance benchmarks for models monitored by the Collaboratory for the Study of Earthquake Predictability. It marks a shift from merely developing new forecasting algorithms to rigorously validating existing ones against a decade of seismic data.
While these models provide essential data for scientists and emergency managers, they remain research-focused tools rather than public early-warning systems. Residents should continue to rely on established local geological alerts for real-time earthquake notifications.
The takeaway
Reliable seismic forecasting requires models to accurately predict location as well as frequency, an area where current technology remains inconsistent. Researchers and regulators should prioritize future audits of the one-third of models that failed to capture spatial patterns effectively.
Further reading
Explore deeper insights into seismic shifts and regional monitoring in California Geology.
Source note: This article includes information reported by The Independent Record.
Live Poll
Do you trust current earthquake forecasting models to accurately predict seismic risks in your area?
