New Zealand Journal of Geology and Geophysics abstracts
Quantifying the M8 prediction algorithm: reduction to a single critical
variable and stability results
David Harte
Statistics Research Associates
P.O. Box 12 649
Wellington, New Zealand
Dong-Feng Li
Department of Mathematics
School of Mathematical Science
Beijing University
Beijing 100871, China
Maaike Vreede
Department of Statistics
Massey University
Private Bag 11 222
Palmerston North, New Zealand
David Vere-Jones
Statistics Research Associates
P.O. Box 12 649
Wellington, New Zealand
and
School of Mathematical & Computing Sciences
Victoria University of Wellington
Wellington, New Zealand
Abstract An implementation of the seven M8 functions
within a statistical seismology software library (SSLib) is described. An
algorithm is also developed for combining the seven series into a single
Critical Series such that a “Time of Increased Probability” (TIP) is declared
for the test region whenever the Critical Series passes its threshold level.
Values of the Critical Series are plotted for circles on a dense lattice
of points covering New Zealand, for each 6 monthly period. The resulting
maps can be used to obtain a geographical impression of the evolution of
regions close to or far from reaching TIP status according to the criteria
embodied in the original M8 algorithm. We examine the stability of the M8
series for small perturbations of the centres of the test regions and the
start date of the computations. It is inferred that such instability that
is observed in the declaration of TIPs is due more to the use of hard boundaries
in the definition of the TIP than to the instability of the component series.
Keywords earthquake prediction; probability forecasts;
M8 algorithm; pattern recognition; spatial-temporal predictor
G02006 Received 28 January 2002; accepted 22 October 2002; published 21
March 2003
New Zealand Journal of Geology & Geophysics, 2003, Vol. 46: 141-152
0028-8306/03/4601-0141 $7.00/0 © The Royal Society of New Zealand
2003
PDF file of entire paper (subscribers only): Print-quality (315K) |
screen-quality (263K)
This year's abstracts |
Journal home page |
All abstracts |
Publishing home page
