The RMS RiskLink® Version 22 release scheduled for the end of June 2022 will include an update to the RMS® China Earthquake Model, which will contain significant improvements across all its model components to provide a robust, comprehensive representation of seismic risk.
The updated model now accounts for damage due to ground shaking and ground deformation associated with liquefaction and landslides. It also includes new stochastic event sets that represent the latest understanding of earthquake processes in China, incorporating fault-specific data as well as crustal strain rate data to delineate seismicity rates.
Earthquake risk in China varies greatly from region to region. East of Chengdu, China consists of more tectonically “stable” regions (South East China Plains, for example, Hunan, and North East Inner Mongolia and Heilongjiang) and low deformation active crust (the rim of the Ordos Basin, Northeast China Plain, for example, Beijing, and the South East Seaboard, for example, Fujian).
All have lower seismicity and lower crustal deformation rates when compared to western China, making the comprehensive assessment of seismic hazard (and risk) more challenging.
Seismicity across the country is very episodic in time and variable in space – particularly in eastern China (see Figure 1). For example, in the northeast China Plains, the 1668 M8.5 Tancheng event in the southeast of the region was followed by the 1679 M8 Sanhe M8 event in the north. The 1975 M7.3 Haicheng event in the northeast of the region was followed by the 1976 M7.9 Tangshan event to the southeast of Beijing.
This illustrates that knowing past mainshocks is not necessarily sufficient to infer where future mainshocks will occur. Additionally, knowledge of seventeenth-century events would not have prepared the country for 1970s events, for example.
China’s Historical Earthquake Catalog
China benefits from one of the longest documented catalogs of earthquakes of any country, based on over 2,000 years of documented history. Importantly for eastern China, the historical record of the larger earthquakes can be considered complete, beginning around the fifteenth and sixteenth centuries (Ming Dynasty, 1368–1644).
A standard seismic hazard modeling approach sees the development of characteristic or multisegment events on faults above approximately Mw6.5, and a grid of seismicity rates using the historic earthquake catalog to assign the rate for events lower than Mw6.5 near past mainshocks events.
These methods require very explicit information about the faults if one wants to include them, or else they are not used – such as the explicit geometry, explicit connectors between faults for jumping potential, explicit slip rate, etc.
Large events are then confined to the mapped faults and have rates coming from earthquake geology or inferred slip rate from geodesy, while lower magnitude events are spread over wider areas and with rates from catalogs. The idea behind seismicity-based gridded seismicity is that future events are likely to happen near past events.
To generate the new stochastic event set for China, RMS researchers needed to delineate a long-term model of seismic activity. This required accounting for the fact that the instrumental catalog does not represent the full extent of what could happen in the future, particularly in eastern China.
The seismicity catalog covering recent years, though high quality, may not include enough of the moderate-to-large events that drive risk to infer robust statistics. Using centuries’ worth of earthquake reports helps complement the modeling for moderate-to-large event rates, and to define where these events may occur – but is not useful enough. Additionally, the level of spatial resolution in the fault geometry varies, and the amount of information on slip rate or date of past events is variable and often incomplete.
New Approach to Quantify Seismicity Rates
Given these constraints, a new approach was required for the development of our stochastic event set. The updated RMS China Earthquake Model combines spatial distributions of three different data sources. One source is based on the location of mainshocks in the catalog, the second uses normalized surface deformation rate (see Figure 2), and the third is informed by the most recent fault maps for China.
Measuring Earthquake Potential: Strain Rate in and Around Mainland China
The Version 22 model does not simply weight these sources of information at the country scale, but rather utilizes an approach that best captures the spatially varying quality of the underlying datasets by breaking down China into dozens of seismotectonic regions.
As part of their analysis, RMS researchers performed retrospective forecast tests. They used an earlier part of the earthquake catalog to build the spatial distributions. This was followed by the use of a later part of the catalog to check which combination of catalog-, surface deformation-, and fault-based spatial distributions is the best predictor for earthquake locations for several magnitude ranges. Retrospective forecast tests enabled RMS to assess the forecasting power of the model using already-available data without having to wait for future seismicity to occur.
This method avoids prescribing future events only to the vicinity of past events or precisely on those mapped faults for which slip rate is known, as seen in traditional modeling approaches. This minimizes the risk of missing the potential for large events in regions that have not experienced any events in recorded history. The method shows that large loss-causing events are expected to happen around known faults – not only in locations near past events.
Including proximity to mapped faults improves the forecast performance for large events, even when the faults in question have no known slip rate and have an individual recurrence for large events longer than 1-in-5,000 years. It also counteracts the simplification of the fault geometry and allows the event rate to spread within the breadth of the fault zones. Together with a regional maximum magnitude, this allows the modeling of event magnitudes that would represent multisegment events.
By factoring in surface deformation rates in the model, it avoids missing events where faults are not known yet. The retrospective tests also showed the need for all three data sources to achieve the quantifiable best performance with respect to accurate earthquake forecasts.
Latest View of China Earthquake Risk
The new stochastic event set used by the RMS China Earthquake Model represents the most up-to-date datasets and the application of cutting-edge scientific methods.
RMS researchers applied a similar level of rigor and dedication to the development of the other components of the model. Ground motion footprints have been updated to reflect the latest science and observations from recent events in China, including the 2008 Wenchuan Earthquake. Reevaluated geotechnical data enhances the quantification of site-specific factors, such as ground motion amplification, liquefaction-driven ground deformation, and landslide susceptibility.
The updated assessment of building performance better reflects the latest construction practices as well as lessons learned during recent events. As with all RMS models, comprehensive documentation will be available at release, examining the innovations and datasets utilized in the development of this model.
Delphine Fitzenz works on earthquake source modeling for risk products, with a particular emphasis on spatio-temporal patterns of large earthquakes. Since joining RMS in 2012 after 10+ years in academia, she has strived to bring the risk and the earthquake science communities closer together through her articles and by organizing special sessions at conferences.
Delphine holds an Eng. MSc and a research MSc in geophysics from University Louis Pasteur, Strasbourg, France, and a Ph.D. from ETH Zurich, Switzerland. Her research has been on two major themes: the role of fluids and fault structure on the behavior of fault systems, and constraints on earthquake recurrence probabilistic models from individual segments to system-wide behavior. She is currently on the Technical Advisory Group for the National Seismic Hazard update project in New Zealand and on the Editorial Board of the Bulletin of the Seismological Society of America (BSSA).