spatial density

Where the next week’s or month’s aftershocks fall. Every model outputs a density over the same 2 km grid, normalised to one; the score is the mean log density at the epicentres that followed. The reference is an adaptive kernel density estimate of past epicentres.

pooled over sequences

Per-event spatial information gain over the adaptive kernel estimate, 95% interval resampling sequences.

modelestimate [95% interval]
Cell GBM6 of 14 sequences above adaptive KDE+0.045 [−0.019, +0.106]
KDE (fixed bandwidth)1 of 14 sequences above adaptive KDE−0.101 [−0.143, −0.066]
ETAS kernel0 of 14 sequences above adaptive KDE−0.768 [−0.906, −0.612]
Uniform0 of 14 sequences above adaptive KDE−2.496 [−2.683, −2.316]

per sequence

sequenceUniformKDE (fixed bandwidth)ETAS kernelCell GBM
Landers 1992−2.192−0.156−0.772−0.062
Hector Mine 1999−2.802−0.085−1.192−0.020
Ridgecrest 2019−2.675−0.052−1.001−0.070
Northridge 1994−2.993−0.065−1.180−0.021
El Mayor–Cucapah 2010−2.026+0.018−1.267−0.027
Loma Prieta 1989−2.279−0.183−0.528+0.161
San Simeon 2003−3.268−0.097−0.596−0.021
South Napa 2014−1.765−0.731−0.375−0.094
Parkfield 2004−2.747−0.134−0.915+0.125
L’Aquila 2009−2.496−0.077−0.032+0.155
Amatrice 2016−2.645−0.034−0.752+0.046
Norcia 2016−2.230−0.065−0.776+0.187
Darfield 2010−2.522−0.145−0.684−0.053
Kaikōura 2016−1.410−0.064−1.384+0.078

landers 1992

Day 30, next 7 days; 100 epicentres followed.

KDE (adaptive)

ETAS kernel

Cell GBM

hector mine 1999

Day 30, next 7 days; 22 epicentres followed.

KDE (adaptive)

ETAS kernel

Cell GBM

ridgecrest 2019

Day 30, next 7 days; 108 epicentres followed.

KDE (adaptive)

ETAS kernel

Cell GBM