Aftershock Clock
On 6 February 2023 an M7.8 earthquake struck south-east Turkey near Pazarcık. Aftershocks always fade. The question is how fast.
How many days after the M7.8 did the chance of another M5+ quake within 250 km fall back to the region's normal rate?
"Normal" means the average rate of M5+ quakes in the same circle from 1990 to the day before, about one a year. The model's answer is the day the fitted aftershock rate falls to that level.
Fast at first, then a very long tail
Each dot is the observed rate of M4.5+ aftershocks per day in a time bin. Both axes are logarithmic, so a straight line means a power law. The fitted Omori-Utsu curve falls as roughly 1/tp. The dashed line converts it to M5+ quakes, and it meets the normal M5+ rate (amber) only after about two years. The other well-sampled fits put that day anywhere from 1.5 to 2.4 years.
The model is fitted on M4.5+ quakes and then extended to M5+ with the b-value. The table checks that step against real M5+ counts. They agree in the first month. In year two the model expects 1.4 and 3 happened, and after the window the rate was still above normal, so the real tail may be even slower. Windows with under 10 quakes are marked THIN.
How much the answer moves with the fitting choices
Run your own aftershock sequence
The sliders set a simplified Reasenberg-Jones model, the model that USGS aftershock forecasts grew out of. They start at the values fitted to Turkey. Move the start day to see the chance of an M5+ in the following week.
"About a week" is right for an M6, with the Turkey fit
The wait for normal grows by a fixed factor for every extra unit of magnitude. With the current sliders that factor is . With the current sliders an M6 settles in and an M7.8 in . People carry the small-quake intuition to the big one.
Model curve from the current sliders. Turkey fit: USGS ComCat M4.5+ aftershocks 2023-02-06 to 2025-02-06, fetched 2026-10-11. Productivity comes from the chosen preset.
The biggest aftershock is usually 1.2 smaller. Turkey's was 0.3 smaller.
Nine hours after the M7.8, an M7.5 struck near Elbistan on a different fault branch. Many seismologists call it a doublet, not an ordinary aftershock. Below, 2,000 simulated sequences with the current sliders show how often the largest aftershock comes that close to the mainshock. The answer leans hard on the b-value: .
Each run draws a Poisson count of M4.5+ aftershocks over two years and Gutenberg-Richter magnitudes. Real gaps: M7.5 (gap 0.3) and the next largest, M6.7 eleven minutes in (gap 1.1). Turkey fit from USGS ComCat, fetched 2026-10-11. The fit itself includes the M7.5's own aftershocks, so this test is partly circular.
Three laws to keep for the next big quake in the news
rate ∝ 1/(t + c)^pAftershocks fade as a power of time. With p near 1, day 10 has about a tenth of day 1's rate and day 100 about a hundredth. Turkey's early fit is steeper (p 1.23), but the tail is still long.
N(≥M) ∝ 10^(−b·M)Each step up in magnitude is about ten times rarer. That turns a count of small aftershocks into odds of a big one.
M_largest ≈ M_main − 1.2A rule of thumb, with a wide spread. Turkey's 0.3 gap is rare under this fit, and less rare if b is nearer 1.
Education only. This page fits one past sequence. For any live quake, use the official forecasts: the USGS publishes aftershock forecasts for large US quakes, and national agencies do so elsewhere.