The Gutenberg‐Richter frequency‐magnitude distribution of earthquakes has become well established in seismology. The slope of the relation between frequency and magnitude (b value) is typically 1, but it often shows variations around 1. Based on an analysis of seismicity prior to the 2011 Tohoku and 2004 Sumatra earthquakes (both in magnitude (M) 9 class), we show that the pronounced decade‐scale decrease inbvalue was a common precursor to both mega‐quakes around their hypocenters. This is the first report onM9‐class quakes to confirm a change inb value, which has been predicted based on the results of laboratory experiments. We propose that the b value is an important indicator of an impending great earthquake, and has great potential in terms of predicting a future large quake off the Pacific coast of Hokkaido, Japan.

Abstract.No proven method is currently available for the reliable short time prediction of earthquakes (minutes to months). However, it is possible to make probabilistic hazard assessments for earthquake risk. In this paper we discuss a new approach to earthquake forecasting based on a pattern informatics (PI) method which quantifies temporal variations in seismicity. The output, which is based on an association of small earthquakes with future large earthquakes, is a map of areas in a seismogenic region ("hotspots") where earthquakes are forecast to occur in a future 10-year time span. This approach has been successfully applied to California, to Japan, and on a worldwide basis. Because a sharp decision threshold is used, these forecasts are binary-an earthquake is forecast either to occur or to not occur. The standard approach to the evaluation of a binary forecast is the use of the relative (or receiver) operating characteristic (ROC) diagram, which is a more restrictive test and less subject to bias than maximum likelihood tests. To test our PI method, we made two types of retrospective forecasts for California. The first is the PI method and the second is a relative intensity (RI) forecast based on the hypothesis that future large earthquakes will occur where most smaller earthquakes have occurred in the recent past. While both retrospective forecasts are for the ten year period 1 January 2000 to 31 December 2009, we performed an interim analysis 5 years into the forecast. The PI method out performs the RI method under most circumstances.