Thousand-Year Earthquake Records Miss the Full Hazard of Slow Faults

Seismic hazard maps for southeastern Spain may depend heavily on an arbitrary choice: how long a stretch of time the earthquake catalogue used to build them happens to cover. A new study in Natural Hazards and Earth System Sciences shows that when a single million-year-long simulated earthquake record for the Eastern Betic Shear Zone is chopped into thousand-year windows, each window yields a noticeably different estimate of ground shaking hazard. The variability in return periods ranges from 11 to 21 percent at a peak ground acceleration of 0.04 g, and widens to between 25 and 58 percent at 1 g, the authors report.

Probabilistic seismic hazard assessment, the statistical framework that underpins building codes worldwide, traditionally relies on historical and instrumental earthquake catalogues. From such a catalogue, seismologists derive a Gutenberg-Richter relationship, a log-linear equation linking the number of earthquakes to their magnitude, along with a maximum magnitude that caps the distribution. The slope and activity rate of that relationship, together with ground motion prediction equations, determine how likely a given level of shaking is over a given return period.

The problem is that in slow-deforming regions, the observational record is short compared with the fault system’s own rhythm. Southeastern Spain is the most seismically active part of the Iberian Peninsula, driven by the convergence of the African and Eurasian plates, yet its seismicity remains low to moderate. The Eastern Betic Shear Zone, a system of left-lateral strike-slip faults including the Carboneras, Palomares, Alhama de Murcia, Los Tollos, Carrascoy and Bajo Segura faults, deforms at a mean slip rate of 1.5 plus or minus 0.3 millimetres per year. Paleoseismological studies indicate that intervals between major earthquakes on individual faults run to several thousands of years; for the Carrascoy fault, the estimated recurrence of large events is about three thousand years.

That means the historical and instrumental record for the region, which stretches back to 1048 and therefore spans nearly a thousand years, covers only about a third of the Carrascoy fault’s seismic cycle. A catalogue that samples such a narrow slice of the cycle may never register the largest events a fault can produce, and it may miss changes in fault behaviour over time. The 2011 Lorca earthquake, at magnitude 5.1, which killed nine people and caused heavy economic losses, illustrates that even moderate events in this region carry real consequences.

To test how much this limited sampling matters, Elena Pascual-Sánchez of the Universidad Politécnica de Madrid and colleagues turned to a synthetic earthquake catalogue one million years long, generated in earlier work with the RSQSim earthquake simulator. That simulator applies rate- and state-dependent friction physics to a fully interacting three-dimensional model of the Eastern Betic Shear Zone built from triangular cells one square kilometre in resolution, loading the faults through the backslip method using their mean slip rates. The resulting catalogue reproduces the general features of the region’s observed seismicity.

From the million-year record, the team randomly drew ten thousand sub-catalogues, each exactly as long as the historical record. Each sub-catalogue represents one possible realisation of the long-term seismic process, equivalent to a different thousand-year observation window. The magnitude-frequency distributions of these windows varied substantially: the Gutenberg-Richter activity parameter showed two maxima between 3.0 and 3.5 and between 4.5 and 5.5, the slope parameter peaked at 0.8 to 0.9 and 1.2 to 1.3, and the maximum magnitude followed a bimodal pattern with a main maximum at 7.3 to 7.4 and a secondary maximum above 6.3 to 6.4. Treated statistically, the ten thousand windows centre on a slope of 1.1, close to the value of 1.0 derived from the real historical and instrumental record, which also yields an annual rate of 0.63 earthquakes of magnitude 4.0 or larger.

The fault-by-fault breakdown proved revealing. The Alhama de Murcia and Carboneras faults kept the same bimodal maximum-magnitude distribution as the whole system, while the Carrascoy and Bajo Segura faults showed only a normal distribution centred on a maximum magnitude of 5.6, with no recorded nucleation of events above magnitude 7.1 in their sub-catalogues. The authors interpret the recurrent high-magnitude maxima as the signature of multi-fault ruptures, in which several kinematically compatible segments break together, and suggest such ruptures may be more common than previously thought. They also note that the pattern is compatible with the fault system behaving as an intermittently critical dynamic system, though these large events do not necessarily dominate hazard at the engineering return periods used in classical assessments.

A hundred of the sub-catalogues were then run through individual probabilistic hazard calculations using R-CRISIS software, with a single ground motion model from Campbell and Bozorgnia chosen deliberately to avoid adding other sources of variability. Hazard curves were computed for six of the region’s most populous cities: Murcia, Lorca, Alicante, Vera, Torrevieja and Almería. Murcia showed the shortest return periods for all ground motion levels, Alicante the longest, while the other four cities clustered together. Variability was modest at short return periods but grew markedly beyond 5000 years, with an asymmetry favouring higher ground motions. At a peak ground acceleration of 1 g, the coefficient of variability reached 58 percent in Torrevieja, 57 percent in Murcia, 54 percent in Vera, 53 percent in Lorca, 40 percent in Alicante and 25 percent in Almería. Notably, Torrevieja’s variability exceeded Murcia’s despite Murcia’s higher shaking levels, which the authors link to the influence of the Bajo Segura fault.

Comparing the synthetic results with hazard curves built from the updated historical-instrumental catalogue and from the ZESIS database used in Spain’s national hazard work, the observed records produced the less hazardous scenarios for almost all ground motion levels. Return periods from synthetic seismicity were generally shorter, suggesting the historical record, having sampled few large earthquakes, represents a comparatively benign scenario. A check on cumulative seismic moment reinforced the picture: the historical record released a total of 1.779 times ten to the nineteen newton metres, while the hundred synthetic windows ranged from 9.8 times ten to the eighteen to 4.1 times ten to the twenty, with a mean of 8.09 times ten to the nineteen and a median of 3.2 times ten to the nineteen, close to the observed value. The authors caution that assuming fully coupled faults may inflate the synthetic moment release, and that the synthetic model’s overestimation of seismic activity cannot be ruled out.

The study’s conclusions echo earlier findings that short observation windows can significantly underestimate seismic hazard, and they carry a practical warning for building codes: even at the 500-year return period commonly used for ordinary structures, the spread of plausible ground motions is substantial. The authors stress that their approach deliberately isolates the effect of the observation window, setting aside other epistemic uncertainties already explored in the underlying simulation work, and that the model’s main limitations are the absence of secondary faults, the lack of dynamic interactions and uncertainties in input data. Because the method depends on well-characterised fault geometry and behaviour, it is best suited to regions where geological knowledge is strong but deformation is slow, and the team proposes applying the same analysis to faster-moving fault systems as a natural next comparison.

Subject of Research: How the length of the earthquake observation window affects probabilistic seismic hazard estimates for the Eastern Betic Shear Zone in southeastern Spain, tested with a million-year synthetic catalogue.

Article Title: Dependence of seismic hazard assessment on the observation time interval: insights from a synthetic earthquake catalogue in southeastern Spain

Article References: Pascual-Sánchez, E., Álvarez-Gómez, J. A., García-Mayordomo, J., & Herrero-Barbero, P. (2026). Dependence of seismic hazard assessment on the observation time interval: insights from a synthetic earthquake catalogue in southeastern Spain. Natural Hazards and Earth System Sciences, 26(10), 5281-5298. https://doi.org/10.5194/nhess-26-5281-2026

Image Credits: AI Generated

DOI: 10.5194/nhess-26-5281-2026

Keywords: seismic hazard, Eastern Betic Shear Zone, synthetic seismicity, RSQSim, Gutenberg-Richter, probabilistic seismic hazard assessment, southeastern Spain, earthquake catalogue, return period, peak ground acceleration, slow-moving faults, R-CRISIS