Smoothed Seismicity Rates Karen Felzer USGS
Decision points #1: Which smoothing algorithm to use? • National Hazard Map smoothing method (Frankel, 1996)? • Helmstetter et al. (2007) smoothing method down to M 2, back to 1981? • Helmstetter et al. (2007) smoothing method down to M 4, back to 1932? • Helmstetter et al. (2007) smoothing method down to M 2, back to 1932 or 1850, with extended planes for large historical sources? • Gaussian or power law smoothing kernel?
National Hazard Map smoothing method linear scale The catalog is declustered using Gardner and Knopoff (1975) The Weichert method is used to calculate rates in each bin from M≥4, M≥5, and M≥6 earthquakes from different periods. Rates are smoothed around each bin using a Gaussian kernel and a fixed 50 km smoothing constant. Map through 2010 created from automated part of algorithm
National Hazard Map smoothing method log scale Final 2008 map after manual adjustments, courtesy of Chuck Mueller
Helmstetter et al. (2007) smoothing method log10 scale • The catalog is declustered using Reasenberg (1985). Remaining catalog still has some clustering. • M≥2 earthquakes are used from >1981 only. • A Gaussian or power law kernel with an adaptive smoothing constant is expanded around each hypocenter. Map uses 1981-2005 catalog data
Approximated Helmstetter et al. (2007) method using M 4+ back to 1850 Normalized log10 scale 1850-2010 catalog data
Using the full Helmstettermethod would require using small earthquakes not in the UCERF catalog – okay?
Decision points #2: What declustering algorithm to use? • Gardner and Knopoff (1975): Traditional, good for removing aftershocks, but maybe not optimal for a smoothed forecast. • Reasenberg (1985): Arbitrarily chosen by Helmstetter et al. • One of the other methods from Andy’s Oxnard talk ? • Try different routines to find what works best for smoothed seismicity forecasting (My recommendation).
How we want the perfect declustering routine to work 2006-2010 smoothed seismicity /1932-2005 smoothed seismicity Decrease the Landers/Hector signal, but not too much! Decrease the Kern County signal, but not too much!
The different methods can be evaluated using the MLE Gaingiven in Helmstetter et al. (2007) G = Gain L = log likelihood of forecasting map Lunif = log likelihood of a uniform probability map N = Number of earthquakes Evaluation is performed only within the UCERF polygon
Summary • Do we have enough support to switch to Helmstetter et al. smoothing? • Do we have enough support to go down to M 2+ earthquakes? (And represent large historic earthquakes with planes?) • Do we have support to switch to a new declustering method?