What Does a Site-Specific PSHA Actually Do Differently?

In my previous article, I discussed whether the introduction of Multi-Period Response Spectra (MPRS) in ASCE 7-22 eliminates the need for site-specific seismic hazard analyses.

The short answer was no.

MPRS represents a substantial improvement over the historical two-period approach. It preserves more of the spectral shape, introduces intermediate site classes, and computes hazard directly at the representative VS30 values associated with those site classes.

But that naturally leads to another question:

If the USGS hazard model has become so sophisticated, what exactly does a site-specific probabilistic seismic hazard analysis (PSHA) do differently?

The answer is surprisingly simple.

A site-specific PSHA attempts to answer the same question as the national hazard model:

"What level of ground shaking is expected at this location?"

The difference lies in the amount of detail and project-specific information used to answer that question.

Every PSHA Requires Several Key Ingredients

Whether you are performing a national-scale hazard study or a project-specific study, the basic ingredients are the same:

• Earthquake sources

• Earthquake recurrence rates

• Ground motion prediction models

• Logic Tree (epistemic uncertainty)

• Site conditions

The difference is not the methodology. The difference is the level of detail. In many ways, a site-specific PSHA can be thought of as zooming in on a single location, while the national hazard model must provide a consistent solution for an entire country.

Earthquake Sources

The national USGS hazard model must represent seismicity across the entire United States.

This is an extraordinary undertaking. The model includes hundreds of faults, background seismicity zones, and multiple tectonic environments.

However, because the model is national in scope, it cannot always represent every local fault system with the level of detail that may be available for an important project.

For critical facilities, major bridges, dams, power infrastructure, and large developments, project teams often develop source characterizations specifically focused on the region surrounding the site.

Additional geologic investigations, paleoseismic studies, fault mapping, and local expert judgment may all contribute to a more detailed representation of the seismic sources that matter most for that project.

Earthquake Recurrence

Not all faults behave the same way. Some produce frequent moderate earthquakes, while others remain quiet for centuries before generating large events.

A PSHA must estimate the probability of future earthquakes occurring on each source.

The USGS model uses carefully developed national recurrence relationships, but site-specific studies may incorporate more recent research, local investigations, or alternative interpretations of fault behavior.

This is particularly common for projects where the consequences of failure justify a more detailed evaluation of uncertainty.

Ground Motion Prediction Models

Once an earthquake occurs, we need a way to estimate the shaking that reaches the site.

Ground motion prediction equations (GMPEs) provide that link.

Modern GMPEs account for factors such as:

• Magnitude

• Distance

• Fault mechanism

• Site conditions

• Regional attenuation characteristics

National hazard models typically use logic trees that combine multiple GMPEs to represent epistemic uncertainty—the uncertainty arising not from the randomness of earthquakes themselves, but from our incomplete knowledge of how to model them.

Site-specific studies often use similar approaches, but they may adjust model selections, weighting factors, or applicability ranges to better represent the tectonic environment surrounding the project.

Logic Trees and Epistemic Uncertainty

One of the less visible but most important components of a PSHA is the treatment of epistemic uncertainty. Unlike aleatory variability, which reflects the inherent randomness of earthquakes, epistemic uncertainty reflects uncertainty in our understanding of the seismic hazard itself.

For example, experts may disagree on fault geometries, recurrence rates, maximum magnitudes, or the ground-motion models that best represent a region. Rather than selecting a single "correct" model, PSHA commonly uses logic trees to evaluate multiple plausible interpretations and combine their results in a rational and transparent manner.

National hazard models employ logic trees to represent uncertainty at a broad scale, but site-specific studies may develop project-specific logic trees with alternative branch selections or weighting factors tailored to the characteristics of the site and the objectives of the project. In some cases, advanced studies may also incorporate regional or site-specific adjustments that move beyond the assumptions inherent in traditional ergodic ground-motion models.

Logic-Tree Weighting

Each branch of a logic tree is assigned a weight based on expert judgment regarding the credibility, applicability, and technical defensibility of that particular interpretation of the hazard. The weights do not imply that one model is "correct" and the others are "incorrect." Rather, they acknowledge that multiple scientifically defensible interpretations may exist simultaneously.

Logic trees also provide a practical mechanism for incorporating new scientific knowledge without requiring an abrupt transition from one model to another. For example, suppose a new ground-motion model has been developed and is viewed as promising, but the engineering community has not yet accumulated enough experience to fully replace the existing model. Rather than choosing one model and discarding the other, a logic tree might assign a weight of 60% to the established model and 40% to the new model. As additional research, validation studies, and practical experience accumulate, those weights may gradually shift to 30% and 70%, and eventually the older model may be removed altogether. In this way, logic trees provide a transparent framework for managing scientific uncertainty while allowing hazard assessments to evolve as our understanding improves.

Site Conditions

This is where many engineers first think of site-specific studies.

The site itself influences the shaking that ultimately reaches the structure.

Factors such as:

• VS30

• Soil layering

• Depth to bedrock

• Basin geometry

• Nonlinear soil behavior

can all influence both the amplitude and shape of the response spectrum.

The MPRS methodology significantly improved how site conditions are represented within the national hazard model, but it still relies on representative site classes.

A site-specific study can evaluate the actual measured properties of the site.

For sites located within deep sedimentary basins, additional analyses may also capture basin effects that are difficult to represent fully within a national hazard model. Deep basins can trap and reverberate seismic waves, often amplifying long-period ground motions and extending the duration of shaking. In some cases, these effects can be more significant than differences associated with site class alone. Basin effects are particularly important for tall buildings, long-span bridges, and other structures that are sensitive to long-period response.

Because basin effects depend on the geometry and geology of a specific basin, they are often evaluated more accurately through site-specific studies than through generalized national models.

The Often Overlooked Difference: Location

Perhaps the most important difference between a site-specific PSHA and a map-based approach is also the easiest to overlook: location. A site-specific PSHA evaluates hazard at the exact coordinates of the project, whereas the USGS maps do not perform a hazard analysis at your site. Instead, they perform hazard analyses at a network of grid points and interpolate the results to estimate the hazard at your location.

For much of the country, this distinction has little practical impact. The hazard field changes gradually, and interpolation provides an excellent estimate of the seismic hazard. However, in regions where hazard changes rapidly over relatively short distances, interpolation can smooth out important features of the hazard field. This is particularly relevant near major faults, where small changes in location can produce noticeable differences in predicted ground motions.

The issue is not that the USGS maps are wrong. They are designed to provide a practical and consistent national solution for millions of sites. A site-specific PSHA, by contrast, is focused on a single location and evaluates the hazard directly at that site. Most of the time the two approaches will produce very similar results. In some situations, however, the differences can become important.

So Which One Is Better?

That is actually the wrong question.

A site-specific PSHA is not a different type of hazard analysis than the one used to develop the national maps. It is fundamentally the same process. The difference lies in the level of detail, the project-specific information incorporated into the analysis, and the fact that the hazard is evaluated directly at the site of interest.

Whether that additional detail matters depends on the project. For many structures, the national hazard model provides an excellent representation of the seismic hazard. For others, the additional insight provided by a site-specific study can meaningfully influence both the resulting spectrum and the engineer's understanding of the uncertainty behind it.

In the next article, we'll explore when that additional effort is justified. When does a site-specific PSHA provide meaningful engineering value, and when is the modern USGS MPRS likely sufficient for the task at hand?

TL;DR

The introduction of Multi-Period Response Spectra (MPRS) in ASCE 7-22 represents one of the most significant advances in seismic design provisions in decades. By preserving more of the true spectral shape, introducing intermediate site classes, and incorporating site effects directly within the hazard calculations, MPRS addresses many of the limitations of the historical two-period approach.

As a result, an important question naturally arises: if the national hazard model has become so sophisticated, do we still need site-specific seismic hazard analyses?

To answer that question, it is first necessary to understand what a site-specific probabilistic seismic hazard analysis (PSHA) actually does. Contrary to a common misconception, a site-specific PSHA is not a fundamentally different type of analysis than the one used to develop the national hazard maps. Both seek to answer the same question: what level of ground shaking is expected at a particular location?

The difference lies in the amount of detail and project-specific information incorporated into the analysis. Site-specific studies may employ more detailed source characterizations, alternative recurrence relationships, project-specific logic trees, refined ground-motion models, measured site conditions, and basin-specific effects. Perhaps most importantly, they evaluate the hazard directly at the project location rather than relying on values interpolated from a national grid.

In this article, I explore the key ingredients of a PSHA, explain the role of logic trees and epistemic uncertainty, discuss the importance of site conditions and basin effects, and examine how site-specific studies differ from the national hazard model that underlies the USGS maps. Understanding these distinctions is essential for evaluating when a site-specific study may provide meaningful engineering value and when the mapped spectrum is likely sufficient.