Recurrence Interval Simulator: Earthquake Probability Forecasting

simulator intermediate ~10 min
Loading simulation...
P = 12% — probability in next 50 years

With a mean recurrence of 300 years, CV of 0.4, and 200 years elapsed since the last event, the conditional probability of a characteristic earthquake in the next 50 years is about 12% — compared to 15% from a simple Poisson model.

Formula

P(t, W) = [F(t+W) - F(t)] / [1 - F(t)] (conditional probability)
P_Poisson = 1 - exp(-W/μ) (time-independent probability)
h(t) = f(t) / [1 - F(t)] (hazard rate function)

Earthquake Clocks

Faults accumulate tectonic stress between earthquakes at a roughly constant rate, then release it suddenly during rupture. This cycle suggests quasi-periodic recurrence — not perfectly regular, but not random either. Paleoseismic data from trench excavations provides the dated earthquake sequence needed to estimate the mean recurrence interval and its variability, forming the foundation for time-dependent seismic hazard assessment.

Probability Models

The simplest model treats earthquakes as a Poisson process: each year has the same probability regardless of when the last earthquake occurred. But real fault data shows clustering around a mean interval, motivating time-dependent models. The Brownian Passage Time (BPT) distribution models the earthquake cycle as stress accumulation with random perturbations, producing a hazard rate that increases as elapsed time grows — capturing the intuition that a 'late' earthquake is overdue.

Conditional Probability

The key output for hazard planning is the conditional probability: given that the last earthquake was t years ago, what is the probability of another in the next W years? For a fault with 300-year mean recurrence and 200 years elapsed, this probability depends critically on the coefficient of variation. Regular faults (CV=0.3) show sharply increasing hazard as elapsed time approaches the mean; irregular faults (CV=0.8) behave nearly like Poisson processes.

From Probability to Policy

Recurrence statistics drive building codes, insurance rates, and emergency preparedness. The USGS earthquake probability maps for California combine recurrence intervals from dozens of faults, weighting time-dependent and Poisson models. This simulation lets you explore how the statistical parameters — mean interval, variability, and elapsed time — shape the probability landscape that guides billion-dollar infrastructure decisions.

FAQ

What is an earthquake recurrence interval?

The recurrence interval is the average time between repeated large earthquakes on the same fault. It is estimated from paleoseismic trench data, historical records, or geodetic strain accumulation rates. Recurrence intervals range from ~20 years (fast-moving plate boundaries) to >10,000 years (intraplate faults).

What does coefficient of variation mean for earthquakes?

The coefficient of variation (CV) measures how regular or irregular earthquake recurrence is. CV = 0 would mean perfectly periodic earthquakes. CV = 1 equals a Poisson (memoryless) process. Most faults show CV of 0.3–0.7, indicating quasi-periodic behavior that makes time-dependent forecasting meaningful.

What is conditional probability in seismic hazard?

Conditional probability is the chance of a characteristic earthquake occurring within a future time window, given that a specific time has already elapsed since the last event. For quasi-periodic faults, this probability increases as elapsed time approaches and exceeds the mean recurrence interval — unlike the constant Poisson rate.

Which probability model is used for earthquake forecasting?

Common models include the Poisson (time-independent), Brownian Passage Time (BPT), lognormal, and Weibull distributions. The BPT model is physically motivated by the earthquake cycle (stress accumulation with noise) and is used by the USGS Working Group on California Earthquake Probabilities.

Sources

Embed

<iframe src="https://homo-deus.com/lab/paleoseismology/recurrence-interval/embed" width="100%" height="400" frameborder="0"></iframe>
View source on GitHub