Generalized Conditional Intensity Measure and Multivariate Return Period Theories Considering Joint Occurrence of Horizontal and Vertical Ground Motions and Their Applications in Joint Selection of Three-Dimensional Ground Motions
This paper proposes a vector generalized conditional intensity measure (VGCIM) framework and Kendall-based multivariate return period theories to address the joint occurrence of horizontal and vertical ground motions, demonstrating through a case study that these methods provide more accurate return period matching and robust ground motion selection compared to existing approaches.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Earthquakes do not shake the ground in a single direction; they push and pull in three dimensions, sending waves of energy horizontally across the landscape and vertically through the soil. For decades, engineers designing critical structures like nuclear power plants, massive dams, and long-span bridges have focused primarily on the horizontal shaking, assuming it carries the most destructive force. However, in many real-world events, the vertical jolt can be just as severe, or even more damaging, than the side-to-side motion. To ensure these structures remain safe, scientists must understand how these horizontal and vertical forces happen together. The challenge lies in predicting exactly how strong these combined forces will be at a specific location and then finding real earthquake records from history that match those predictions to test building designs. If the records chosen for testing do not accurately reflect the complex, joint nature of the shaking, the safety assessment of the building could be dangerously wrong.
A team of researchers from Shenyang Jianzhu University and the Harbin Institute of Technology has developed a new way to solve this problem. They focused on a concept called the "return period," which is essentially the average amount of time one might wait for an earthquake of a certain intensity to occur. For example, a 2,475-year return period is a standard benchmark used for designing structures that must survive very rare, catastrophic events. The researchers wanted to know: if we are waiting for an earthquake that happens once every 2,475 years, what does the combination of horizontal and vertical shaking actually look like? They tested three different mathematical approaches to define this "joint return period." One approach looked for the moment when both horizontal and vertical shaking exceeded a certain level at the exact same time. Another looked for the moment when either the horizontal or the vertical shaking exceeded the level, regardless of the other. The third approach, based on a statistical method known as the Kendall function, sought a middle ground that accounts for the complex relationship between the two directions without being too strict or too loose.
To test these ideas, the team created a detailed simulation of a specific site in California, using a massive database of historical earthquake data. They generated millions of virtual earthquake scenarios to see how the different methods performed. They found that the method looking for simultaneous exceedance and the method looking for any exceedance produced results that were too extreme. The simultaneous method predicted shaking that was too weak to be realistic for a 2,475-year event, while the "any" method predicted shaking that was far too strong. The Kendall-based method, however, landed right on target. When the researchers used this method to calculate the expected shaking, the results matched the target 2,475-year return period almost perfectly. This was a crucial discovery because it meant that the Kendall approach could reliably describe the true hazard of combined horizontal and vertical ground motions.
The study then moved from theory to practice by trying to select actual earthquake records from a global database to use as test cases for these predictions. They compared their new Kendall-based selection method against two other popular techniques: one that looks at a single "vector" of shaking conditions and another that averages multiple conditions together. The researchers discovered that the method based on averaging multiple conditions was highly sensitive to the specific choices made at the start of the process. If the engineers changed which specific periods of shaking they focused on, the average method produced wildly different results, making it unreliable. In contrast, the new Kendall-based method and the vector-based method were stable; they produced consistent results regardless of the specific starting choices. Furthermore, the records selected by the Kendall method showed a moderate and realistic spread of shaking intensity, avoiding the extreme highs and lows that can distort safety tests.
The researchers also found that the relationship between horizontal and vertical shaking is not simply about how closely they correlate with each other. Instead, the key to matching the correct return period lies in how the earthquake's energy is broken down or "decomposed" across different frequencies. When the way the horizontal shaking breaks down matched the way the vertical shaking broke down, the predictions were accurate. When they did not match, the predictions drifted away from the target. This insight helps explain why some older methods failed to capture the true danger of combined shaking. By using the Kendall-based approach, engineers can now select earthquake records that are not only strong enough to test a building's limits but are also representative of the specific, joint nature of the threat.
Ultimately, this work provides a more reliable toolkit for protecting society's most vital infrastructure. By proving that the Kendall-based method offers the most accurate picture of combined horizontal and vertical earthquake risks, the study gives engineers a clearer path forward. They can now select ground motion records that truly reflect the 2,475-year hazard, ensuring that the buildings designed to withstand these forces are tested against the right kind of shaking. The findings suggest that moving away from simpler, single-direction models or unstable averaging techniques toward this more robust statistical method will lead to safer, more resilient structures capable of withstanding the full complexity of an earthquake.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.