Long waves and tall buildings: Site effect considerations in catastrophe models following the Myanmar earthquake

01

Are distant seismic sources included — even those beyond the model domain?

Long-period motions can originate from earthquakes far outside the local hazard zone or even national borders. To avoid underestimating risk, it’s important to ensure that distant rupture sources, including those outside a model’s core geographic domain but still capable of causing losses in your region of interest, are properly considered. For places such as Bangkok, accounting for these distant sources is crucial. The 2025 Myanmar earthquake generated ground motions in Bangkok in the 0.8 to 3.4 second period range, which corresponds to the natural resonant frequencies of buildings between 13 and 55 stories tall. The 30-story tower collapsed because its natural vibration period aligned with the earthquake's dominant shaking frequency, which amplified the building’s movement.

So ask your model vendor: Which distant sources are included that could contribute to long-period risk? Which are excluded — and why? Sometimes, the faults that matter most aren’t the ones immediately beneath your exposure but the ones across the border.

02

Does the model align with observed impacts in past long-period events — and if not, why?

Catastrophe models are designed to capture average behavior, but key insights often come from how their simulated events measure up to real earthquakes. In cities such as Bangkok or Mexico City, where long-period shaking is known to cause significant damage, it’s worth asking: Does the model reproduce the character of observed ground motions — or do simplifications obscure important local effects?

In the case of the 2025 Myanmar earthquake, the rupture on the Sagaing Fault channeled seismic energy southward toward Thailand. This channeling, coupled with a long rupture length and supershear propagation, generated long-period waves that were significantly stronger than those predicted by the median Ground Motion Prediction Equations commonly used in catastrophe models.

Additionally, while catastrophe models often include generic basin effects, each sedimentary basin has its own resonance characteristics that is shaped by depth, layering and soil properties. If a model relies on global averages and isn’t locally calibrated with site-specific research, it may fail to reproduce the true character of long-period shaking.

These modeling limitations don't mean catastrophe models aren't useful; rather, they highlight the importance of understanding their constraints and using them appropriately. For (re)insurers with significant exposure in sedimentary basins, the management of the risk from long-period motions triggered by rare but plausible earthquakes can be enhanced by supplementing catastrophe model outputs with detailed scenario-based analyses.

03

How has high-rise vulnerability been validated
for local building stock and long-period shaking?

Given the strong amplification of long-period ground motions in sedimentary basins, understanding how mid- and high-rise building vulnerability has been calibrated is essential. Are the model’s damage functions based on local construction data, or have they been adapted from other regions?

The collapse of the under-construction high-rise building in Bangkok highlights the special vulnerability of buildings during the interim construction phases. Bangkok's construction industry commonly uses flat slab construction, post-tensioning and unusually slender columns — design approaches that create vulnerabilities to lateral movement during earthquakes, especially from long-period ground motion.^[1],^[2]

Thailand's late adoption of uniform seismic codes (Bangkok was excluded until 2007) means most high-rises were built without seismic provisions. Models should incorporate findings from recent studies of Thai buildings that revealed striking differences between how buildings were theoretically supposed to vibrate during seismic events versus their actual behavior. Although only one under-construction building collapsed completely, over 14,000 damage reports were registered across Bangkok. If the earthquake had been more severe or closer, vulnerability models suggest more widespread failures could have occurred in buildings with these design characteristics.

If local construction practices and vulnerabilities are not explicitly reflected in modelling, there’s a risk that modelled loss estimates will not match actual damages from long-period shaking events. For catastrophe model users, this gap demands the review of how vulnerability functions were developed, and where necessary, working with engineering experts to build region-specific adjustments that better reflect actual exposures.