Search

Currently, one of the best performing and most popular earthquake forecasting models rely on the working hypothesis that: "locations of past background earthquakes reveal the probable location of future seismicity". As an alternative, we present a class of smoothed seismicity models (SSMs) based on the principles of the Epidemic Type Aftershock Sequence (ETAS) model, which forecast the location, time and magnitude of all future earthquakes using the estimates of the background seismicity rate and the rates of future aftershocks of all generations. Using the Californian earthquake catalog, we formulate six controlled pseudo-prospective experiments with different combination of three target magnitude thresholds: 2.95, 3.95 or 4.95 and two forecasting time horizons: 1 or 5 year. In these experiments, we compare the performance of:(1) ETAS model with spatially homogenous parameters or GETAS (2) ETAS model with spatially variable parameters or SVETAS (3) three declustering based SSMs (4) a simple SSM based on undeclustered data and (5) a model based on strain rate data, in forecasting the location and magnitude of all (undeclustered) target earthquakes during many testing periods. In all conducted experiments, the SVETAS model comes out with consistent superiority compared to all the competing models. Consistently better performance of SVETAS model with respect to declustering based SSMs highlights the importance of forecasting the future aftershocks of all generations for developing better earthquake forecasting models. Among the two ETAS models themselves, accounting for the optimal spatial variation of the parameters leads to strong and statistically significant improvements in forecasting performance.

In seismic hazard studies, simulation techniques are exploited relatively less compared to other statistical methods. The main purposes of their use were merely scientific: either to reduce uncertainty in seismic hazard or to test the consistency of a forecast. Its limited exploitation in the field does not mean that the method could not be exploited further. Given sufficient data and with the right models, a full-scale simulation is possible, which in fact becomes a forecast with future timing, location and magnitude of future events are simulated. Within this context, the main purpose of this study is to layout a method to create a full scale simulation with all the parameters namely the time, location and magnitude of an earthquake is simulated. For that purpose, in addition to the main parameters defining past seismic patterns, namely the temporal, spatial, magnitude distributions, the combined spatio-magnitude distribution is also modeled. As another improvement compared to the current procedures, occurrence date is assigned to the simulated event according to the inter-event time distribution, which enhances the simulation procedure. The eventual temporal distribution and the associated earthquake rates are checked with the original earthquake rate for its consistency. In the end, a simulation procedure, which can also be used as a forecast model is developed to improve the existing procedures further.

Many studies have shown that b values tend to decrease prior to large earthquakes. To evaluate the forecast information in b value variations, we conduct a systematic assessment in Yunnan Province, China, where the seismicity is intense and moderate–large earthquakes occur frequently. The catalog in the past two decades is divided into four time periods (January 2000–December 2004, January 2005–December 2009, January 2010–December 2014, and January 2015–December 2019). The spatial b values are calculated for each 5-year span and then are used to forecast moderate-large earthquakes (M ≥ 5.0) in the subsequent period. As the fault systems in Yunnan Province are complex, to avoid possible biases in b value computation caused by different faulting regimes when using the grid search, the hierarchical space–time point-process models (HIST-PPM) proposed by Ogata are utilized to estimate spatial b values in this study. The forecast performance is tested by Molchan error diagram (MED) and the efficiency is quantified by probability gain (PG) and probability difference (PD). It is found that moderate–large earthquakes are more likely to occur in low b regions. The MED analysis shows that there is considerable precursory information in spatial b values and the forecast efficiency increases with magnitude in the Yunnan Province. These results suggest that the b value might be useful in middle- and long-term earthquake forecasts in the study area.

We, the ongoing Working Group on California Earthquake Probabilities, present a spatiotemporal clustering model for the Third Uniform California Earthquake Rupture Forecast (UCERF3), with the goal being to represent aftershocks, induced seismicity, and otherwise triggered events as a potential basis for operational earthquake forecasting (OEF). Specifically, we add an epidemic‐type aftershock sequence (ETAS) component to the previously published time‐independent and long‐term time‐dependent forecasts. This combined model, referred to as UCERF3‐ETAS, collectively represents a relaxation of segmentation assumptions, the inclusion of multifault ruptures, an elastic‐rebound model for fault‐based ruptures, and a state‐of‐the‐art spatiotemporal clustering component. It also represents an attempt to merge fault‐based forecasts with statistical seismology models, such that information on fault proximity, activity rate, and time since last event are considered in OEF. We describe several unanticipated challenges that were encountered, including a need for elastic rebound and characteristic magnitude–frequency distributions (MFDs) on faults, both of which are required to get realistic triggering behavior. UCERF3‐ETAS produces synthetic catalogs of M ≥2.5 events, conditioned on any prior M ≥2.5 events that are input to the model. We evaluate results with respect to both long‐term (1000 year) simulations as well as for 10‐year time periods following a variety of hypothetical scenario mainshocks. Although the results are very plausible, they are not always consistent with the simple notion that triggering probabilities should be greater if a mainshock is located near a fault. Important factors include whether the MFD near faults includes a significant characteristic earthquake component, as well as whether large triggered events can nucleate from within the rupture zone of the mainshock. Because UCERF3‐ETAS has many sources of uncertainty, as will any subsequent version or competing model, potential usefulness needs to be considered in the context of actual applications. [Electronic Supplement:][1] Figures showing discretization, verification of the DistanceDecayCubeSampler , average simulated participation rate, and average cumulative magnitude–frequency distributions (MFDs). [1]: http://www.bssaonline.org/lookup/suppl/doi:10.1785/0120160173/-/DC1

Voronoi residuals, deviances, super-thinning, and some other residual analysis methods are applied to a selection of earthquake forecast models in the Collaboratory for the Study of Earthquake Predictability (CSEP). Unlike simple numerical summaries such as the N-test, L-test, or R-test, graphical residual methods are proposed which can be useful for comparing multiple models and for highlighting when and where a given model does not agree closely with the observed seismicity. For gridded forecasts, deviances seem ideally suited for model comparison, and Voronoi residuals seem preferable to simple grid-based residuals for assessing one model individually. For models outputting an estimated conditional rate at any particular space–time location, Voronoi residuals and super-thinning can be especially useful at identifying departures from the data.

We introduce a method of resolving temporal incidence trends in earthquake sequences. We have developed a catalog partitioning method based on canonical earthquake scaling relationships, and have further developed a metric based on record-breaking interval (RBI) statistics to resolve increasing and decreasing seismicity in time series of earthquakes. We calculated the RBI metric over fixed-length sequences of earthquake intervals and showed that the length of those sequences is related to the magnitude of the earthquake to which the method is sensitive—longer sequences resolve large earthquakes, shorter sequences resolve small-magnitude events. This sequence length effectively constitutes a local temporal catalog constraint, and we show that spatial constraints can be defined from rupture length scaling. We have applied the method to several high-profile earthquakes and have shown that it consistently resolves aftershock sequences after a period of accelerating seismicity before the targeted mainshock. The method also suggests a minimum detectable (forecastable) mainshock magnitude on the basis of the catalog’s minimum completeness magnitude $$m_{\rm c}$$ .

The Chamoli earthquake of March 28, 1999 (M b 6.6, Origin time: 19:05:12, epicenter at 30.5° N 79.4° E and Focal depth 23 km) occurred in the Himalayan front arc which caused severe damage in the region. Anomalous seismic activity associated with this recent devastating earthquake in the Central Himalaya region in an area bounded by 30.0°-31.0° N and 79.0°-80.0° E have been studied using seismicity data from 1980-2000. The preparatory zone is delineated using the temporal and the spatial distribution of earthquakes, considering the events with cutoff magnitude m b ≥ 4.3. Daily number of events as well as cumulative number of earthquake with time within the preparatory zone has been considered as basis for identification of anomalous seismicity. Accordingly four anomalous episodes: Normal/ background (N); Anomalous/ swarm (A); Precursory gap (G) and Mainshock sequence (M) are identified. It is observed that the event was preceded by well defined patterns of anomalous seismicity/ precursory swarm which was lasted for about seven month and had started about three years and four months prior to mainshock. Key words : Anomalous seismicity; Central Himalaya; Precursory swarm; Earthquake forecast The Himalayan Physics Vol.2, No.2, May, 2011 Page: 11-15 Uploaded Date : 31 July, 2011

Earthquake activity is monitored in real time at the Koyna reservoir in western India, beginning from August 2005 and successful short term forecasts have been made of M ∼ 4 earthquakes. The basis of these forecasts is the observation of nucleation that precedes such earthquakes. Here we report that a total of 29 earthquakes in the magnitude range of 3.5 to 5.1 occurred in the region during the period of August 2005 through May 2010. These earthquakes could broadly be put in three zones. Zone-A has been most active accounting for 18 earthquakes, while 5 earthquakes in Zone-B and 6 in Zone-C have occurred. Earthquakes in Zone-A are preceded by well defined nucleation, while it is not the case with zones B and C. This indicates the complexity of the earthquakes processes and the fact that even in a small seismically active area of only 20 km × 30 km earthquake forecast is difficult.

Several earthquake forecast experiments in Italy have been initiated within the European testing center of the global Collaboratory for the Study of Earthquake Predictability. In preparation for these experiments, we developed space-rate-magnitude forecasts based on a simple model that incorporates the spatial clustering of the seismicity. This model, which we call the simple smoothed seismicity model (TripleS), has a minimal number of free parameters and is based on very few assumptions; therefore, it can be considered as a model of «least information» with which others can be compared. The fundamental TripleS parameter controls the spatial extent of the smoothing, and we selected its value based on an optimization procedure that was applied to retrospective forecast experiments. In this report, we present the motivation for developing TripleS, and describe the construction of forecasts for Italian seismicity. We also discuss the research questions that remain to be answered with respect to TripleS, and more generally, the smoothed seismicity approach to earthquake forecasting.

We investigated the seismicity patterns associated with the M∼5 earthquakes in the artificial water reservoir triggered zone of Koyna region situated near the west coast of India by means of the Region Time Length (RTL) algorithm. The RTL algorithm, which is widely used for investigating the precursory seismicity changes before large earthquakes, is being applied for the first time to a small region where triggered earthquakes have been occurring in an area of 30 km × 20 km, and there is no other seismic source within 50 km of the Koyna region. As this method is being implemented for the first time for such a small region, the validity of the characteristic coefficients r 0 , t 0 and d 0 was tested and found acceptable. We examined both the spatial and temporal characteristics of seismicity changes in the Koyna region. The RTL algorithm was applied to the earthquake catalogue of the Koyna region for the period 2006 through 2009, after removing the aftershocks of M ≥ 3.0 earthquakes. The catalogue is complete for M ≥ 1.8 events. We find the epicentral area showing anomalous seismic activation and quiescence before the main earthquakes. The temporal variation of the RTL values of all the four earthquakes of M 4.8 to 5.1, that occurred during 2006 through 2009, showed a seismic quiescence anomaly prior to these earthquakes. The seismicity changes inferred in this study may provide better understanding of the future M ∼ 5 earthquakes and increase the probability of earthquake forecast in the Koyna region.

We have initially developed a time-independent forecast for southern California by smoothing the locations of magnitude 2 and larger earthquakes. We show that using small m 2 earthquakes gives a reasonably good prediction of m 5 earthquakes. Our forecast outperforms other time-independent models (Kagan and Jackson, 1994; Frankel et al., 1997), mostly because it has higher spatial resolution. We have then developed a method to estimate daily earthquake probabilities in south- ern California by using the Epidemic Type Earthquake Sequence model (Kagan and Knopoff, 1987; Ogata, 1988; Kagan and Jackson, 2000). The forecasted seismicity rate is the sum of a constant background seismicity, proportional to our time- independent model, and of the aftershocks of all past earthquakes. Each earthquake triggers aftershocks with a rate that increases exponentially with its magnitude and decreases with time following Omori's law. We use an isotropic kernel to model the spatial distribution of aftershocks for small (m 5.5) mainshocks. For larger events, we smooth the density of early aftershocks to model the density of future aftershocks. The model also assumes that all earthquake magnitudes follow the Gutenberg-Richter law with a uniform b-value. We use a maximum likelihood method to estimate the model parameters and test the short-term and time-independent forecasts. A retro- spective test using a daily update of the forecasts between 1 January 1985 and 10 March 2004 shows that the short-term model increases the average probability of an earthquake occurrence by a factor 11.5 compared with the time-independent forecast.

Major objectives of the Japanese earthquake prediction research program for the period 2009–2013 are to create earthquake forecasting models and begin the prospective testing of these models against recorded seismicity. For this purpose, the Earthquake Research Institute of the University of Tokyo has joined an international partnership to create a Collaboratory for the Study of Earthquake Predictability (CSEP). Here, we describe a new infrastructure for developing and evaluating forecasting models—the CSEP Japan Testing Center—as well as some preliminary testing results. On 1 November 2009, the Testing Center started a prospective and competitive earthquake predictability experiment using the seismically active and well-instrumented region of Japan as a natural laboratory.

The most used and accepted models for daily forecasts are based on short-term space and time earthquake clustering for occurrence rates and on the Gutenberg-Richter law for the frequency-magnitude. These models have been demonstrated to produce reliable prospective space-time-magnitude forecasts during an aftershock sequence, but their skill in forecasting mainshocks is still under discussion. This paper studies the foreshock statistics of the Italian and Californian seismicity in two ways: i) we compare the foreshock activity observed in real seismic catalogs and in synthetic catalogs derived from a pure Epidemic-Type Aftershock Sequence (ETAS) model; ii) we analyze the triggering capability of earthquakes using different ETAS parameterizations, in order to check whether large events are triggered in the same way as regular earthquakes. The results indicate that the foreshock activity observed in the real catalogs is compatible with what is expected by the ETAS model. Moreover, we find that the empirical foreshock rates have an intrinsic variability due to limited sampling that may explain most of the differences found so far in different seismic catalogs. Copyright © 2011 by the American Geophysical Union.

On 1 August 2009, the global Collaboratory for the Study of Earthquake Predictability (CSEP) launched a prospective and comparative earthquake predictability experiment in Italy. The goal of the CSEP-Italy experiment is to test earthquake occurrence hypotheses that have been formalized as probabilistic earthquake forecasts over temporal scales that range from days to years. In the first round of forecast submissions, members of the CSEP-Italy Working Group presented eighteen five-year and ten-year earthquake forecasts to the European CSEP Testing Center at ETH Zurich. We considered the twelve time-independent earthquake forecasts among this set and evaluated them with respect to past seismicity data from two Italian earthquake catalogs. In this article, we present the results of tests that measure the consistency of the forecasts with the past observations. Besides being an evaluation of the submitted time-independent forecasts, this exercise provided insight into a number of important issues in predictability experiments with regard to the specification of the forecasts, the performance of the tests, and the trade-off between the robustness of results and experiment duration. We conclude with suggestions for the future design of earthquake predictability experiments., 43 pages, 8 figures, 4 tables

Forecast verification procedures for statistical events with binary outcomes typically rely on the use of contingency tables and Relative Operating Characteristic (ROC) diagrams. Originally developed for the statistical evaluation of tornado forecasts on a county-by-county basis, these methods can be adapted to the evaluation of competing earthquake forecasts. Here we apply these methods retrospectively to two forecasts for the M 7.3 1999 Chi-Chi, Taiwan, earthquake. We show that a previously proposed forecast method that is based on evaluating changes in seismic intensity on a regional basis is superior to a forecast based only on the magnitude of seismic intensity in the same region. Our results confirm earlier suggestions that the earthquake preparation process for events such as the Chi-Chi earthquake involves anomalous activation or quiescence, and that signatures of these processes can be detected in seismicity data using appropriate methods.

We present a RELM forecast of future earthquakes in California that is primarily based on the pattern informatics (PI) method. This method identifies regions that have systematic fluctuations in seismicity, and it has been demonstrated to be successful. A PI forecast map originally published on 19 February 2002 for southern California successfully forecast the locations of sixteen of eighteen M>5 earthquakes during the past three years. The method has also been successfully applied to Japan and on a worldwide basis. An alternative approach to earthquake forecasting is the relative intensity (RI) method. The RI forecast map is based on recent levels of seismic activity of small earthquakes. Recent advances in the PI method show considerable improvement, particularly when compared with the RI method using relative operating characteristic (ROC) diagrams for binary forecasts. The RELM application requires a probability for each location for a number of magnitude bins over a five year period. We have therefore constructed a hybrid forecast in which we combine the PI method with the RI method to compute a map of probabilities for events occurring at any location, rather than just the most probable locations. These probabilities are further converted, using Gutenberg-Richter scaling laws, to anticipated rates of future earthquakes that can be evaluated using the RELM test., 6 pages, 4 figures (color), 1 table

Pure and Applied Geophysics, 163 (11-12), ISSN:0033-4553, ISSN:1420-9136, ISSN:1557-7368

(1) The Parkfield-Cholame section of the San Andreas fault, site of an unfulfilled earthquake forecast in 1985, is the best monitored section of the world's most closely watched fault. In 1983, the M = 6.5 Coalinga and M = 6.0 Nunez events struck 25 km northeast of Parkfield. Seismicity rates climbed for 18 months along the creeping section of the San Andreas north of Parkfield and dropped for 6 years along the locked section to the south. Right-lateral creep also slowed or reversed from Parkfield south. Here we calculate that the Coalinga sequence increased the shear and Coulomb stress on the creeping section, causing the rate of small shocks to rise until the added stress was shed by additional slip. However, the 1983 events decreased the shear and Coulomb stress on the Parkfield segment, causing surface creep and seismicity rates to drop. We use these observations to cast the likelihood of a Parkfield earthquake into an interaction-based probability, which includes both the renewal of stress following the 1966 Parkfield earthquake and the stress transfer from the 1983 Coalinga events. We calculate that the 1983 shocks dropped the 10-year probability of a M � 6 Parkfield earthquake by 22% (from 54 ± 22% to 42 ± 23%) and that the probability did not recover until about 1991, when seismicity and creep resumed. Our analysis may thus explain why the Parkfield earthquake did not strike in the 1980s, but not why it was absent in the 1990s. We calculate a 58 ± 17% probability of a M � 6 Parkfield earthquake during 2001- 2011. INDEX TERMS: 7223 Seismology: Seismic hazard assessment and prediction; 7230 Seismology: Seismicity and seismotectonics; 7260 Seismology: Theory and modeling; KEYWORDS: Coalinga earthquake, Parkfield, stress change, earthquake probability, seismicity rate

The evaluation of any earthquake forecast hypothesis requires the application of rigorous statistical methods. It implies a univocal definition of the model characterising the concerned anomaly or precursor, so as it can be objectively recognised in any circumstance and by any observer.A valid forecast hypothesis is expected to maximise successes and minimise false alarms. The probability gain associated to a precursor is also a popular way to estimate the quality of the predictions based on such precursor. Some scientists make use of a statistical approach based on the computation of the likelihood of an observed realisation of seismic events, and on the comparison of the likelihood obtained under different hypotheses. This method can be extended to algorithms that allow the computation of the density distribution of the conditional probability of earthquake occurrence in space, time and magnitude. Whatever method is chosen for building up a new hypothesis, the final assessment of its validity should be carried out by a test on a new and independent set of observations. The implementation of this test could, however, be problematic for seismicity characterised by long-term recurrence intervals. Even using the historical record, that may span time windows extremely variable between a few centuries to a few millennia, we have a low probability to catch more than one or two events on the same fault. Extending the record of earthquakes of the past back in time up to several millennia, paleoseismology represents a great opportunity to study how earthquakes recur through time and thus provide innovative contributions to time-dependent seismic hazard assessment. Sets of paleoseimologically dated earthquakes have been established for some faults in the Mediterranean area: the Irpinia fault in Southern Italy, the Fucino fault in Central Italy, the El Asnam fault in Algeria and the Skinos fault in Central Greece. By using the age of the paleoearthquakes with their associated uncertainty we have computed, through a Montecarlo procedure, the probability that the observed inter-event times come from a uniform random distribution (null hypothesis). This probability is estimated approximately equal to 8.4% for the Irpinia fault, 0.5% for the Fucino fault, 49% for the El Asnam fault and 42% for the Skinos fault. So, the null Poisson hypothesis can be rejected with a confidence level of 99.5% for the Fucino fault, but it can be rejected only with a confidence level between 90% and 95% for the Irpinia fault, while it cannot be rejected for the other two cases. As discussed in the last section of this paper, whatever the scientific value of any prediction hypothesis, it should be considered effective only after evaluation of the balance between the costs and benefits introduced by its practical implementation.