Frequency-magnitude distribution
The frequency-magnitude distribution (FMD) counts how many earthquakes of each size a catalogue holds, and it is the chart from which both Mc and the b-value are read.
Interactive diagram · Real example from the atlas →Simple explanation
Sort the earthquakes of a place into size classes, 0.1 magnitude units wide, and count each class. That table is the FMD.
It can be drawn two ways. Non-cumulative: the count in each class. Cumulative: the count of events at least that big, N(≥ M). The cumulative curve is smoother, because every point includes all larger events.
On a normal count axis the curve drops off a cliff and the large events vanish into the baseline. On a logarithmic count axis, where each grid line is ten times the one below, the part above Mc becomes a straight line. That straight line is the Gutenberg-Richter law.
An everyday picture
It is like the count of coins in a jar by value: lots of small coins, fewer large ones. Plot how many coins are worth “at least this much”, on a scale that multiplies by ten each step, and a regular pattern shows up as a straight line.
Technical details
With magnitudes binned to ΔM = 0.1, the non-cumulative FMD is n(Mᵢ), the count with M in [Mᵢ − ΔM/2, Mᵢ + ΔM/2), and the cumulative FMD is N(≥ Mᵢ) = Σ_{j ≥ i} n(Mⱼ). Above Mc, both followlog₁₀ N(≥ M) = a − b · Mand the non-cumulative counts have the same slope with a lower intercept.
Fitting: the atlas never fits b by least squares on the plotted points (the points are not independent and the large-magnitude counts are tiny); it uses the maximum-likelihood estimator (see “Gutenberg-Richter b-value”).
Data: the cell detail files carry the non-cumulative FMD per depth class for ISC-GEM (the b sample, direct Mw since v0.2) and for ComCat as served. The cell panel draws the cumulative FMD on a log axis with the Mc line, and the CSV download lists counts and cumulative counts per bin.
Caution: with a few hundred events the upper tail is noisy, and single large events sit on a long flat step. Marzocchi et al. (2020) show that b from small samples is biased and correlates with the largest magnitude in the sample.
Further reading
- Mignan, A. & Woessner, J. (2012). Estimating the magnitude of completeness for earthquake catalogs. Community Online Resource for Statistical Seismicity Analysis (CORSSA). doi.org/10.5078/corssa-00180805
- Wiemer, S. & Wyss, M. (2002). Mapping spatial variability of the frequency-magnitude distribution of earthquakes. Advances in Geophysics, 45, 259–302. doi.org/10.1016/S0065-2687(02)80007-3
- Marzocchi, W., Spassiani, I., Stallone, A. & Taroni, M. (2020). How to be fooled searching for significant variations of the b-value. Geophysical Journal International, 220, 1845–1856. doi.org/10.1093/gji/ggz541