A common language for reporting earthquake intensities
Crossing Europe by train used to be far more challenging than it is today. Travelers were required to pass through sometimes complicated and confusing passport checks at each international border and carry cash in various currencies (or—ugh—traveler’s checks). The expansion of the European Union (specifically, the Schengen Area) and the creation of the Eurozone largely resolved these challenges by eliminating barriers to travel across borders and adopting the euro as a common currency among many countries.
In 1998, European nations accomplished another feat of harmonization by developing and adopting the European Macroseismic Scale (EMS-98) [Grünthal, 1998]. This scale established a common basis for assigning intensities to seismic events across Europe, overcoming differences among existing intensity scales and producing more consistent and useful assessments of earthquake shaking and damage.
An important challenge now for the seismological and earthquake engineering communities is to eliminate barriers to sharing earthquake intensity data around the world and develop a globally accepted macroseismic scale to further extend the value of macroseismic observations. Combining the standardization of such a scale with the benefits of innovations in global Internet availability, artificial intelligence, and other technologies, we have an opportunity to gain a far more comprehensive understanding of seismic risks and safety around the world.
What is macroseismology?
Macroseismology involves determining the intensity of earthquake shaking on the basis of human perceptions and observed effects to structures, primarily buildings. Systematic approaches for assigning seismic intensities connect analyses of our collective seismological past with the present and the present to the future. They also provide a means of sharing and archiving assigned intensities in a reproducible form and of effectively communicating the effects of shaking to different audiences.
Intensity observers include members of the public who self-report their perceptions of shaking, as well as seismologists and civil engineers sent out on traditional surveys to interview individuals and record building damage after an earthquake. The use of such macroseismic intensity data in research and operational settings has grown with time, as illustrated by the millions of community observers who have participated in reporting their experiences and observations to the U.S. Geological Survey’s (USGS) Internet-based Did You Feel It? (DYFI) system [Quitoriano and Wald, 2020] (Figure 1) and to analogous systems worldwide [e.g., Bossu et al., 2017; Tosi et al., 2015; Goded et al., 2018].
Applications of intensity assignments from both approaches are far-reaching. For example, traditional macroseismic surveys illuminate critical aspects of earthquakes, such as the extents and intensities of shaking needed to constrain the locations and magnitudes of historical events, and their impacts on society. Internet-based macroseismic data, meanwhile, are extremely important for constraining real-time shaking estimates, which afford increased situational awareness to many affected populations and emergency managers and help prioritize postearthquake responses. In combination, traditional and crowdsourced macroseismic intensity data contribute to scientific analyses of earthquake behavior as well as to earthquake engineers’ loss and risk analyses related to built structures (see Quitoriano and Wald [2020] for many such examples).
Maintaining these essential applications of macroseismic observations requires that we revisit whether traditional macroseismic surveys remain well suited to modern environments and standardize Internet-based collection strategies. We must also ensure compatibility between traditional and Internet-based approaches to macroseismic data collection and intensity assignments.
Shortcomings of modern practices
Even when observers follow best practices, there are several limitations to modern macroseismic data collection approaches. For example, although crowdsourced Internet intensity reports are rapid and abundant, they are insufficient for describing higher, damaging intensity levels because accurate assignments require expert knowledge of buildings’ structural systems and damage characteristics. In addition, limited access to Internet and mobile phones in some countries and communities hampers equitable access to Internet-based intensity reporting systems and data [e.g., Hough and Martin, 2021]. So despite being costly in terms of time and finances, traditional field approaches to assigning shaking intensities are still required to assign higher intensities and fill gaps in Internet reporting.
Several macroseismic scales in use have been implemented independently, and intensity values from each cannot be directly compared.
Because traditional approaches vary worldwide, though, they can produce inconsistent results. This situation necessitates a truly international macroseismic scale (IMS) that standardizes procedures and accounts for regional differences in, for example, how building damage from similar seismic shaking can vary starkly because of different conditions and construction practices. Several macroseismic scales in use have been implemented independently, and intensity values from each cannot be directly compared. Scales used in highly seismically active nations like Japan, China, and India are particularly at odds. Informal adaptations of EMS-98 outside Europe, especially in Central and South America, present additional inconsistencies that could be improved. Furthermore, the United States, New Zealand, and many other countries still use the Modified Mercalli Intensity (MMI) Scale, which was developed in 1931. The MMI Scale is consistent with—yet inferior to—the more recent EMS-98 because the latter requires detailed building vulnerability assignments, well-specified damage grades, and statistical analyses of building damage.
Collecting postearthquake shaking and damage data is critical to archiving and documenting earthquake experiences and impacts. Incompatibilities in collection approaches and among the data themselves become limiting factors for many seismological and earthquake engineering analyses. The development and implementation of EMS-98 provide several lessons and guidelines for developing a worldwide standard to overcome these limitations.
Raising the bar for intensity assignments
The rollout of EMS-98 was an important innovation that substantially raised the standards of traditional intensity assignments. To assign higher intensities, EMS-98 requires detailed field reconnaissance observations gathered using strict data collection protocols. The EMS-98 methodology characterizes the seismic vulnerability of individual buildings and assigns each to an established vulnerability class (A–F, with A indicating most vulnerable), and it describes damage to structures using predefined damage grades
With sufficient postearthquake observations in a town or neighborhood, one can assign the intensity level at a particular location on the basis of the fraction of buildings in each damage state at that location. For example, intensity VIII on the EMS-98 scale is defined as “many buildings of vulnerability class B suffer damage of grade 3; a few of grade 4” [Grünthal, 1998, p. 19], with “many” meaning 15%–55% and “a few” meaning 0%–15%. Intensity IX requires that “many buildings of vulnerability class A sustain damage of grade 5” or “many buildings of vulnerability class B suffer damage of grade 4; a few of grade 5” and so on [Grünthal, 1998, p. 19].
Earlier macroseismic intensity scales are often ambiguous in how they define structural vulnerabilities, damage grades, and damage level fractions.
EMS-98’s stringent requirements ensure that quality building damage data are collected and archived, allowing shaking intensities at different locations to be assigned statistically and objectively. Indeed, EMS-98 raised the bar for the expected quality of macroseismic data used in macroseismology. In doing so, it brought to light limitations of earlier practices. Earlier macroseismic intensity scales—most of which were developed 50–150 years ago—are often ambiguous in how they define the essential EMS-98 ingredients: structural vulnerabilities, damage grades, and damage level fractions.
Moreover, building vulnerability classes of the past were, in some cases, quite different than they are now. At the same time, many scales have not been sufficiently updated to account for today’s broader range of structures and the presence of buildings with earthquake-resistant designs. For example, most older brick chimneys—damage to which was a key indicator for assigning intensities in California—have either been replaced or retrofitted, confounding assignments that would have been straightforward in the past.
When properly recognized, such limitations can, in principle, be accommodated by providing uncertainty estimates or by weighting data accordingly. These approaches are particularly helpful for characterizing historical earthquakes that occurred before EMS-98 standards existed and for which archival observations are less detailed.
Still, what is needed is a worldwide standardization of intensity assignment practices and associated data collection for both Internet-based and traditional, ground survey–based observations, especially for high-intensity, damaging levels of shaking. With the rapid development of new crowdsourcing technologies, vast improvements in damage data collection practices are already here or are on the horizon. So how can we ensure that professional damage data assessors in the field have the mandate and tools they need to collect sufficient observations (including of building types and damage percentages) for assigning EMS-98-like intensities?
The benefits of building an IMS
In October 2022, experts making up the International Macroseismic Scale Working Group (IMSWG) gathered at the USGS John Wesley Powell Center for Analysis and Synthesis in Fort Collins, Colo., to address these issues. Attendees focused on how to revise the MMI Scale in countries where it is in use, such as New Zealand and the United States, to be compatible with EMS-98, with the idea that these revisions will add momentum toward creating a global IMS. They also focused on improving strategies for rapid macroseismic assignments, especially for higher intensities, and producing recommendations for updating EMS-98 into a global IMS.
Our current effort can, fortunately, benefit from past efforts, and several of the luminaries behind the development of EMS-98 who participated in the October workshop had already begun devising strategies to move toward an IMS [e.g., Spence and Foulser-Piggott, 2014, 2015; Abrahamczyk et al., 2017]. These strategies involve adapting building vulnerability classes and illustrating damage grades so they comprehensively reflect the wide, heterogeneous array of structures and building practices outside Europe that were not considered in the original EMS-98 (Figure 2).
Moving to an international macroseismic scale would require and motivate collection of more dependable and accessible postearthquake building damage data sets.
However, challenges to updating EMS-98 remain. In particular, structural engineering experts from around the world must be consulted to guide vulnerability class assignments for missing endemic building types and validate these assignments with damage data from past earthquakes in the region. New approaches for evaluating evolving earthquake-resistant building designs must also be taken into consideration.
The benefits of moving to an IMS would be considerable. It would require and motivate collection of more dependable and accessible postearthquake building damage data sets. Consistent building damage data would facilitate a wide range of modeling and scientific studies by seismologists and engineers looking to better understand earthquake behaviors, hazards, and risks. These data would also facilitate more uniform messaging and communication of earthquake effects worldwide, and they would illuminate trends globally in building damage and losses caused by seismic activity, serving broader policymaking related to urban planning and community resilience, for example. Furthermore, they would serve other decisionmakers, such as the Federal Emergency Management Agency and other emergency management agencies, who depend on this impact information as a basis for equitable allocations of resources and financial adjudications during earthquake response and recovery.
Roughly 80% of DYFI responses are received within an hour of earthquakes occurring [Quitoriano and Wald, 2020], making them timely enough to constrain and improve the accuracy of ShakeMaps. Rapid, consistent shaking intensity assignments across the globe would be especially beneficial in areas that face substantial seismic hazards but lack real-time monitoring instrumentation. More accurate maps, in turn, would enable more detailed examinations of geotechnical forensics and the design of more resilient structures, as well as improved loss and risk estimates. These capabilities would also improve correlations between assigned macroseismic intensities and instrumentally recorded ground shaking measurements, which would inform relationships between shaking levels and earthquake effects.
The use of macroseismic intensity data is expanding widely as the metric by which shaking hazards and risks are depicted within essential real-time earthquake information products, including earthquake early warning systems (e.g., ShakeAlert), ShakeMaps, DYFI, and the USGS Prompt Assessment of Global Earthquakes for Response (PAGER) system [e.g., Quitoriano and Wald, 2020]. We even use it for presenting long-term probabilistic seismic hazard maps in a friendlier format to nontechnical users.
Taking steps to tackle challenges
Artificial intelligence could greatly augment and accelerate traditional intensity assignments while collecting valuable damage data that could be used for other humanitarian and scientific purposes.
With the adoption of standardized damage and intensity scales, new approaches and technologies could prove valuable for macroseismology. Community science–based observations reported online already account for 95% of the macroseismic data now collected, in part because lower intensities of shaking are much more common than higher intensities that currently require field observations to assign properly. What if a portion of the building damage data needed to assign higher intensities could also be gathered through crowdsourced or automated image recognition of photos collected on the ground or from satellite imagery? With proper training, artificial intelligence (AI) and other nontraditional methods could greatly augment and accelerate traditional intensity assignments while collecting valuable damage data that could be used for other humanitarian and scientific purposes.
Promising machine learning approaches for rapidly updating modeled loss estimates with satellite imagery analysis will benefit from any immediate on-the-ground damage observations that are available [e.g., Xu et al., 2022]. And AI-assisted intensity assignments will improve ShakeMaps and help constrain overall impact assessments for disaster response and recovery.
While those advances are still on the horizon, we are considering strategies for addressing challenges to integrating traditional and Internet-based intensity assignments by standardizing building damage data collection. As long as accurate assignments of higher intensities require more expertise than is available via community-based reports, we must ensure that professional damage assessors and building inspectors collect data that provide a statistically sufficient sampling of buildings in a given area.
During a second IMSWG workshop at the Powell Center this coming fall, we plan to bring in experts on postearthquake building damage data collection. Working with these experts, we can facilitate modifications to their inspection protocols to ensure they provide the needed sampling of buildings, along with accurate information about building types and damage—such modifications will provide the standardization required by an EMS-98-like IMS. We will also facilitate standardization of Internet-based intensity data collection practices.
Integrating traditional field-based assignments with crowdsourced assignments (e.g., from DYFI and similar systems), as well as those for past events based on historical sources like newspaper accounts and diaries, also requires accounting for the different uncertainties among these approaches. Importantly, these uncertainties vary with intensity—assignments of higher intensity are more uncertain than lower ones, particularly when assigned via crowdsourcing [Quitoriano and Wald, 2022].
It is essential that we improve our default approaches for estimating uncertainties, particularly when they are used quantitatively (e.g., in ShakeMaps and downstream earthquake loss models like the PAGER system). Recognizing variations in uncertainties in crowdsourced assignments and ensuring that traditional assignments are based on adequate samplings of building damage are steps that will help. For assignments of shaking intensity during past earthquakes from historical observations, we anticipate that experts—informed by knowledge of data sources, the scale used, and even the expertise of assigners—will help provide accurate uncertainties.
Tackling the challenges of improving access to the DYFI system and to more systematic shaking and damage reconnaissance protocols worldwide should help address problems of inequitable access to reporting systems. Improved access also will be on the agenda at the upcoming second workshop, along with the challenges mentioned above. Initial strategies include adding more language options into DYFI (beyond the currently provided English and Spanish), expanding regional efforts similar to DYFI in countries where participation in global systems is lower, and improving outreach to remote locales by working with local experts and media to publicize the availability of DYFI immediately after significant earthquakes.
Moving in the right direction
Developing a global IMS will not solve entirely all the challenges of collecting and analyzing macroseismic data after earthquakes. However, it will lead to more orderly macroseismic data collection across much of the planet. It also will help harmonize Internet-based crowdsourced observations with traditional damage data gathered in the field, leading to more systematic cross-border analyses of seismic activity and thus a better understanding of the hazards that populations face and the risks to which they are exposed.