New Zealand Earthquake Forecast Testing Centre

The New Zealand Earthquake Forecast Testing Centre is being established as one of several similar regional testing centres under the umbrella of the Collaboratory for the Study of Earthquake Predictability (CSEP). The Centre aims to encourage the development of testable models of time-varying earthquake occurrence in the New Zealand region, and to conduct verifiable prospective tests of their performance over a period of five or more years. The test region, data-collection region and requirements for testing are described herein. Models must specify in advance the expected number of earthquakes with epicentral depths h ≤ 40 km in bins of time, magnitude and location within the test region. Short-term models will be tested using 24-h time bins at magnitude M ≥ 4. Intermediate-term models and long-term models will be tested at M ≥ 5 using 3-month, 6-month and 5-year bins, respectively. The tests applied will be the same as at other CSEP testing centres: the so-called N test of the total number of earthquakes expected over the test period; the L test of the likelihood of the earthquake catalogue under the model; and the R test of the ratio of the likelihoods under alternative models. Four long-term, three intermediate-term and two short-term models have been installed to date in the testing centre, with tests of these models commencing on the New Zealand earthquake catalogue from the beginning of 2008. Submission of models is open to researchers worldwide. New models can be submitted at any time. The New Zealand testing centre makes extensive use of software produced by the CSEP testing centre in California. It is envisaged that, in time, the scope of the testing centre will be expanded to include new testing methods and differently-specified models, nonetheless that the New Zealand testing centre will develop in parallel with other regional testing centres through the CSEP international collaborative process.     

Earthquake Forecasting in Diverse Tectonic Zones of the Globe

We present a simple method for long- and short-term earthquake forecasting (estimating earthquake rate per unit area, time, and magnitude). For illustration we apply the method to the Pacific plate boundary region and the Mediterranean area surrounding Italy and Greece. Our ultimate goal is to develop forecasting and testing methods to validate or falsify common assumptions regarding earthquake potential. Our immediate purpose is to extend the forecasts we made starting in 1999 for the northwest and southwest Pacific to include somewhat smaller earthquakes and then adapt the methods to apply in other areas. The previous forecasts used the CMT earthquake catalog to forecast magnitude 5.8 and larger earthquakes. Like our previous forecasts, the new ones here are based on smoothed maps of past seismicity and assume spatial clustering. Our short-term forecasts also assume temporal clustering. An important adaptation in the new forecasts is to abandon the use of tensor focal mechanisms. This permits use of earthquake catalogs that reliably report many smaller quakes with no such mechanism estimates. The result is that we can forecast earthquakes at higher spatial resolution and down to a magnitude threshold of 4.7. The new forecasts can be tested far more quickly because smaller events are considerably more frequent. Also, our previous method used the focal mechanisms of past earthquakes to estimate the preferred directions of earthquake clustering, however the method made assumptions that generally hold in subduction zones only. The new approach escapes those assumptions. In the northwest Pacific the new method gives estimated earthquake rate density very similar to that of the previous forecast.     

Temporal distribution of earthquakes using renewal process in the Dasht-e-Bayaz region

Temporal distribution of earthquakes with M _w > 6 in the Dasht-e-Bayaz region, eastern Iran has been investigated using time-dependent models. Based on these types of models, it is assumed that the times between consecutive large earthquakes follow a certain statistical distribution. For this purpose, four time-dependent inter-event distributions including the Weibull, Gamma, Lognormal, and the Brownian Passage Time (BPT) are used in this study and the associated parameters are estimated using the method of maximum likelihood estimation. The suitable distribution is selected based on logarithm likelihood function and Bayesian Information Criterion. The probability of the occurrence of the next large earthquake during a specified interval of time was calculated for each model. Then, the concept of conditional probability has been applied to forecast the next major (M _w > 6) earthquake in the site of our interest. The emphasis is on statistical methods which attempt to quantify the probability of an earthquake occurring within a specified time, space, and magnitude windows. According to obtained results, the probability of occurrence of an earthquake with M _w > 6 in the near future is significantly high.     

Space- and Time-Dependent Probabilities for Earthquake Fault Systems from Numerical Simulations: Feasibility Study and First Results

In weather forecasting, current and past observational data are routinely assimilated into numerical simulations to produce ensemble forecasts of future events in a process termed “model steering”. Here we describe a similar approach that is motivated by analyses of previous forecasts of the Working Group on California Earthquake Probabilities (WGCEP). Our approach is adapted to the problem of earthquake forecasting using topologically realistic numerical simulations for the strike-slip fault system in California. By systematically comparing simulation data to observed paleoseismic data, a series of spatial probability density functions (PDFs) can be computed that describe the probable locations of future large earthquakes. We develop this approach and show examples of PDFs associated with magnitude M > 6.5 and M > 7.0 earthquakes in California.     

PI Forecast for the Sichuan-Yunnan Region: Retrospective Test after the May 12, 2008, Wenchuan Earthquake

The May 12, 2008, Wenchuan M _S 8.0/M _w 7.9 earthquake occurred in the middle part of the north–south seismic zone in central west China, being one of the greatest thrust events on land in recent years. To explore whether there were some indications of the increase of strong earthquake probabilities before the Wenchuan earthquake, we conducted a retrospective forecast test applying the Pattern Informatics (PI) algorithm to the earthquakes in the Sichuan-Yunnan region since 1992. A regional earthquake catalogue complete to M _L 3.0 from 01/01/1977 to 15/06/2008 was used. A 15-year long ‘sliding time window’ was used in the PI calculation, with ‘anomaly training time window’ and ‘forecast time window’ both set to 5 years. With a forecast target magnitude of M _S 5.5, the ROC test shows that the PI forecast outperforms not only random guess but also the simple number-counting approach based on the clustering hypothesis of earthquakes (the RI forecast). ‘Hotspots’ can be seen in the region of the northern Longmenshan fault which is responsible for the Wenchuan earthquake. However, when considering bigger grid size and higher cutoff magnitude, such ‘hotspots’ disappear and there is very little indication of an impending great earthquake.     

Conditional Probabilities for Large Events Estimated by Small Earthquake Rate

We examined forecasting quiescence and activation models to obtain the conditional probability that a large earthquake will occur in a specific time period on different scales in Taiwan. The basic idea of the quiescence and activation models is to use earthquakes that have magnitudes larger than the completeness magnitude to compute the expected properties of large earthquakes. We calculated the probability time series for the whole Taiwan region and for three subareas of Taiwan—the western, eastern, and northeastern Taiwan regions—using 40 years of data from the Central Weather Bureau catalog. In the probability time series for the eastern and northeastern Taiwan regions, a high probability value is usually yielded in cluster events such as events with foreshocks and events that all occur in a short time period. In addition to the time series, we produced probability maps by calculating the conditional probability for every grid point at the time just before a large earthquake. The probability maps show that high probability values are yielded around the epicenter before a large earthquake. The receiver operating characteristic (ROC) curves of the probability maps demonstrate that the probability maps are not random forecasts, but also suggest that lowering the magnitude of a forecasted large earthquake may not improve the forecast method itself. From both the probability time series and probability maps, it can be observed that the probability obtained from the quiescence model increases before a large earthquake and the probability obtained from the activation model increases as the large earthquakes occur. The results lead us to conclude that the quiescence model has better forecast potential than the activation model.     

Probabilistic Assessment of Earthquake Hazard in Gujarat and Adjoining Region of India

The Gujarat and adjoining region falls under all four seismic zones V, IV, III and II of the seismic zoning map of India, and is one of the most seismically prone intracontinental regions of the world. It has experienced two large earthquakes of magnitude M _w 7.8 and 7.7 in 1819 and 2001, respectively and several moderate earthquakes during the past two centuries. In the present study, the probability of occurrence of earthquakes of M ≥ 5.0 has been estimated during a specified time interval for different elapsed times on the basis of observed time intervals between earthquakes using three stochastic models namely, Weibull, Gamma and Lognormal. A complete earthquake catalogue has been used covering the time interval of 1819 to 2006. The whole region has been divided into three major seismic regions (Saurashtra, Mainland Gujarat and Kachchh) on the basis of seismotectonics and geomorphology of the region. The earthquake hazard parameters have been estimated using the method of maximum likelihood. The logarithmic of likelihood function (ln L) is estimated and used to test the suitability of models in three different regions. It was found that the Weibull model fits well with the actual data in Saurashtra and Kachchh regions, whereas Lognormal model fits well in Mainland Gujarat. The mean intervals of occurrence of earthquakes are estimated as 40.455, 20.249 and 13.338 years in the Saurashtra, Mainland Gujarat and Kachchh region, respectively. The estimated cumulative probability (probability that the next earthquake will occur at a time later than some specific time from the last earthquake) for the earthquakes of M ≥ 5.0 reaches 0.9 after about 64 years from the last earthquake (1993) in Saurashtra, about 49 years from the last earthquake (1969) in Mainland Gujarat and about 29 years from the last earthquake (2006) in the Kachchh region. The conditional probability (probability that the next earthquake will occur during some specific time interval after a certain elapsed time from last earthquake) is also estimated and it reaches about 0.8 to 0.9 during the time interval of about 57 to 66 years from the last earthquake (1993) in Saurashtra region, 31 to 51 years from the last earthquake (1969) in Mainland Gujarat and about 21 to 28 years from the last earthquake (2006) in Kachchh region.     

Long-Term RST Analysis of Anomalous TIR Sequences in Relation with Earthquakes Occurred in Greece in the Period 2004–2013

Real-time integration of multi-parametric observations is expected to accelerate the process toward improved, and operationally more effective, systems for time-Dependent Assessment of Seismic Hazard (t-DASH) and earthquake short-term (from days to weeks) forecast. However, a very preliminary step in this direction is the identification of those parameters (chemical, physical, biological, etc.) whose anomalous variations can be, to some extent, associated with the complex process of preparation for major earthquakes. In this paper one of these parameters (the Earth’s emitted radiation in the Thermal InfraRed spectral region) is considered for its possible correlation with M ≥ 4 earthquakes occurred in Greece in between 2004 and 2013. The Robust Satellite Technique (RST) data analysis approach and Robust Estimator of TIR Anomalies (RETIRA) index were used to preliminarily define, and then to identify, significant sequences of TIR anomalies (SSTAs) in 10 years (2004–2013) of daily TIR images acquired by the Spinning Enhanced Visible and Infrared Imager on board the Meteosat Second Generation satellite. Taking into account the physical models proposed for justifying the existence of a correlation among TIR anomalies and earthquake occurrences, specific validation rules (in line with the ones used by the Collaboratory for the Study of Earthquake Predictability—CSEP—Project) have been defined to drive a retrospective correlation analysis process. The analysis shows that more than 93 % of all identified SSTAs occur in the prefixed space–time window around (M ≥ 4) earthquake's time and location of occurrence with a false positive rate smaller than 7 %. Molchan error diagram analysis shows that such a correlation is far to be achievable by chance notwithstanding the huge amount of missed events due to frequent space/time data gaps produced by the presence of clouds over the scene. Achieved results, and particularly the very low rate of false positives registered on a so long testing period, seems already sufficient (at least) to qualify TIR anomalies (identified by RST approach and RETIRA index) among the parameters to be considered in the framework of a multi-parametric approach to t-DASH.     

Multi-Methods Combined Analysis of Future Earthquake Potential

Prior to an earthquake, natural seismicity is correlated across multiple spatial and temporal scales. Many studies have indicated that an earthquake is hard to accurately predict by a single time-dependent precursory method. In this study, we attempt to combine four earthquake prediction methods, i.e. the Pattern Informatics (PI), Load/Unload Response Ratio (LURR), State Vector (SV), and Accelerating Moment Release (AMR) to estimate future earthquake potential. The PI technique is founded on the premise that the change in the seismicity rate is a proxy for the change in the underlying stress. We first use the PI method to quantify localized changes surrounding the epicenters of large earthquakes to objectively quantify the anomalous areas (hot spots) of the upcoming events. Next, we delineate the seismic hazard regions by integrating with regional active fault zones and small earthquake activities. Then, we further evaluate the earthquake potential in the seismic hazard regions using the LURR, SV and AMR methods. Retrospective tests of this new approach on the large earthquakes (M > 6.5) which have occurred in western China over the last 3 years show that the LURR and SV time series usually climb to an anomalously high peak months to years prior to occurrence of a large earthquake. And, the asymptote time, t _c, “predicted” by the AMR method correspond to the time of the actual events. The results may suggest that the multi-methods combined approach can be a useful tool to provide stronger constraints on forecasts of the time and location of future large events.     

Operational earthquake loss forecasting: a retrospective analysis of some recent Italian seismic sequences

Operational earthquake forecasting (OEF) relies on real-time monitoring of seismic activity in an area of interest to provide constant (e.g., daily) updates of the expected number of events exceeding a certain magnitude threshold in a given time window (e.g., 1 week). It has been demonstrated that the rates from OEF can be used to estimate expected values of the seismic losses in the same time interval OEF refers to. This is a procedure recently defined as operational earthquake loss forecasting (OELF), which may be the basis for rational short-term seismic risk assessment and management. In Italy, an experimental OELF system, named MANTIS-K, is currently under testing. It is based on weekly rates of earthquakes exceeding magnitude (M) 4, which are updated once a day or right after the occurrence in the country of an M 3.5+ earthquake. It also relies on large-scale structural vulnerability and exposure data, which serve to the system to provide continuously the weekly expected number of: (1) collapsed buildings, (2) displaced residents, and (3) casualties. While the probabilistic basis of MANTIS-K was described in previous work, in this study OELF is critically discussed with respect to three recent Italian seismic sequences. The aim is threefold: (1) illustrating all the features of the OELF system in place; (2) providing insights to evaluate whether if it would have been a useful additional tool for short-term management; (3) recognizing common features, if any, among the losses computed for different sequences.     

汶川8.0级地震前地震趋势分析意见的回顾

该文在系统整理自2001年昆仑山口西8.1级地震到汶川8.0级地震前年度地震趋势预测中提出的地震活动异常和分析意见的基础上, 研究汶川地震未能做出中长期预测的原因。 结果表明: 2001年昆仑山口西8.1级地震发生后, 判定中国大陆处于强震连发阶段, 仍有发生8级大震危险, 但此后的2002—2007年中国大陆周边接连发生大震, 而内部连续6年的7级地震平静, 以及中国大陆5、 6级地震相继出现的显著平静, 是导致2006年以后对中国大陆地震活动水平预测明显偏低的原因。 南北地震带中段一直是作为近几年可能发生强震的危险地区, 但2007年云南宁洱6.4级地震后, 对西南地区强震危险的紧迫性估计不足。 而汶川地震所在的龙门山地震带历史上没有7级以上地震记录, 也是该地震带未作为近几年地震重点危险区的原因之一。     

Updated earthquake catalogue for seismic hazard analysis in Pakistan

A reliable and homogenized earthquake catalogue is essential for seismic hazard assessment in any area. This article describes the compilation and processing of an updated earthquake catalogue for Pakistan. The earthquake catalogue compiled in this study for the region (quadrangle bounded by the geographical limits 40–83° N and 20–40° E) includes 36,563 earthquake events, which are reported as 4.0–8.3 moment magnitude (M_W) and span from 25 AD to 2016. Relationships are developed between the moment magnitude and body, and surface wave magnitude scales to unify the catalogue in terms of magnitude M_W. The catalogue includes earthquakes from Pakistan and neighbouring countries to minimize the effects of geopolitical boundaries in seismic hazard assessment studies. Earthquakes reported by local and international agencies as well as individual catalogues are included. The proposed catalogue is further used to obtain magnitude of completeness after removal of dependent events by using four different algorithms. Finally, seismicity parameters of the seismic sources are reported, and recommendations are made for seismic hazard assessment studies in Pakistan.     

Forecasting the Locations of Future Large Earthquakes: An Analysis and Verification

The objective of this paper is to quantify the use of past seismicity to forecast the locations of future large earthquakes and introduce optimization methods for the model parameters. To achieve this the binary forecast approach is used where the surface of the Earth is divided into l° × l° cells. The cumulative Benioff strain of m ≥ m _c earthquakes that occurred during the training period, ΔT _tr, is used to retrospectively forecast the locations of large target earthquakes with magnitudes ≥m _T during the forecast period, ΔT _for. The success of a forecast is measured in terms of hit rates (fraction of earthquakes forecast) and false alarm rates (fraction of alarms that do not forecast earthquakes). This binary forecast approach is quantified using a receiver operating characteristic diagram and an error diagram. An optimal forecast can be obtained by taking the maximum value of Pierce’s skill score.     

A Bayesian Assessment of Seismic Semi-Periodicity Forecasts

Among the schemes for earthquake forecasting, the search for semi-periodicity during large earthquakes in a given seismogenic region plays an important role. When considering earthquake forecasts based on semi-periodic sequence identification, the Bayesian formalism is a useful tool for: (1) assessing how well a given earthquake satisfies a previously made forecast; (2) re-evaluating the semi-periodic sequence probability; and (3) testing other prior estimations of the sequence probability. A comparison of Bayesian estimates with updated estimates of semi-periodic sequences that incorporate new data not used in the original estimates shows extremely good agreement, indicating that: (1) the probability that a semi-periodic sequence is not due to chance is an appropriate estimate for the prior sequence probability estimate; and (2) the Bayesian formalism does a very good job of estimating corrected semi-periodicity probabilities, using slightly less data than that used for updated estimates. The Bayesian approach is exemplified explicitly by its application to the Parkfield semi-periodic forecast, and results are given for its application to other forecasts in Japan and Venezuela.     

Retrospective Study on the Predictability of Pattern Informatics to the Wenchuan M8.0 and Yutian M7.3 Earthquakes

Two large earthquakes occurred in the western part of China in 2008, one of them being the Yutian (35.6°N, 81.6°E) M7.3 earthquake that occurred on March 21 (BJT) and the other the Wenchuan (31.0°N, 103.4°E) M8.0 earthquake that occurred on May 12 (BJT). In this paper, the West Continental China (included in 20.0°–50.0°N, 70.0°–110.0°E region) was the study region for verifyong the predictability of the pattern informatics (PI) method using the receiver-operating characteristic curve (ROC) test and R score test. Different forecasting maps with different calculating parameters were obtained. The calculating parameters were the grid size Δx, base time t _b, reference interval t _b to t ₁, change interval t ₁ to t ₂, and forecasting interval t ₂ to t ₃. In this paper, the base time t _b fixed to June 1, 1971, the ending forecast time t ₃ fixed to June 1, 2008, and the forecasting interval t ₂ to t ₃ changed from 1 to 10 years, and the grid sizes were chosen as 1° × 1° and 2° × 2°, respectively. The results show that the PI method could forecast the Yutian M7.3 and Wenchuan M8.0 earthquakes only using suitable parameters. Comparing the forecast results of grid sizes 1° × 1° and 2° × 2°, the models with 2° × 2° grids were better. Comparing the forecast results with different forecasting windows from 1 to 10 years, the models with forecasting windows of 4–8 years were better using the ROC test, and the models with forecasting windows of 7–10 years were better using the R score test. The forecast efficiency of the model with a grid size of 2° × 2° and forecast window of 8 years was the best one using either the ROC test or the R score test.     

"地震可预测性国际合作研究"——1.0阶段工作理念及成果

20世纪90年代由世界多个国家的地震学家围绕“地震可否预测”问题进行国际讨论后, 人们开始思考适用于地震预测研究的规则应该有哪些, 尤其是地震学家针对地震预测研究中所采取的途径和工作思路开始发生了变化。 2007年开始的“区域地震似然模型”(Regional Earthquake Likelihood Models, RELM)工作组和由此进一步而来的“地震可预测性国际合作研究”(Collaboratory for the Study of Earthquake Predictability, CSEP)计划开始之后, 一大批地震预测模型和与评估其预测效能有关的统计检验方法加入进来, 在设立相同的预测规则和使用统一的数据来源下, 通过全球设立不同测试中心的方式, 共同参与到对地震可预测性问题的系统研究中来。 当前, CSEP计划已由开始的1.0阶段发展至2.0阶段, 为使读者了解与这几项国际合作研究相关的工作主旨和发展历程, 本文总结了与CSEP工作1.0阶段相关的工作理念和工作成果以及存在的问题, 以期为下一步工作的开展提供参考。     

Memory effect in M ≥ 6 earthquakes of South-North Seismic Belt, Mainland China

The M?≥?6 earthquakes occurred in the South–North Seismic Belt, Mainland China, during 1901–2008 are taken to study the possible existence of memory effect in large earthquakes. The fluctuation analysis technique is applied to analyze the sequences of earthquake magnitude and inter-event time represented in the natural time domain. Calculated results show that the exponents of scaling law of fluctuation versus window length are less than 0.5 for the sequences of earthquake magnitude and inter-event time. The migration of earthquakes in study is taken to discuss the possible correlation between events. The phase portraits of two sequent magnitudes and two sequent inter-event times are also applied to explore if large (or small) earthquakes are followed by large (or small) events. Together with all kinds of given information, we conclude that the earthquakes in study is short-term correlated and thus the short-term memory effect would be operative.     

Temporal variation of b value associated with M ��4 earthquakes in the reservoir-triggered seismic environment of the Koyna�CWarna region, Western India

It is generally found that the b values associated with reservoir-triggered seismicity (RTS) are higher than the regional b values in the frequency magnitude relation of earthquakes. In the present study, temporal and spatial variation of b value is investigated using a catalog of 3,000 earthquakes from August 2005 through December 2010 for the Koyna?CWarna region in Western India, which is a classical site of RTS globally. It is an isolated (30?×?20?km²) zone of seismicity where earthquakes of up to M ??5 are found to occur during phases of loading and unloading of the Koyna and Warna reservoirs situated 25?km apart. For the Warna region, it is found that low b values of 0.6?C0.9 are associated with earthquakes of M ??4 during the loading phase. The percentage correlation of the occurrence of an M????4 earthquake with a low b value outside the 1?? or 2?? level is as high as 78?%. A drastic drop in the b value of about 50?% being reported for an RTS site may be an important precursory parameter for short-term earthquake forecast in the future.     

Trends in the aggregated rate of pre-eruptive volcano-tectonic seismicity at Kilauea volcano, Hawaii

Accelerating rates of volcano-tectonic (VT) earthquakes are commonly observed during volcanic unrest. Understanding the repeatability of their behaviour is essential to evaluating their potential to forecast eruptions. Quantitative eruption forecasts have focused on changes in precursors over intervals of weeks or less. Previous studies at basaltic volcanoes in frequent eruption, such as Kilauea in Hawaii and Piton de La Fournaise on Réunion, suggest that VT earthquake rates tend to follow a power-law acceleration with time about 2 weeks before eruption, but that this trend is often obscured by random fluctuations (or noise) in VT earthquake rate. These previous studies used a stacking procedure, in which precursory sequences for several eruptions are combined to enhance the signal from an underlying acceleration in VT earthquake rate. Such analyses assume a common precursory trend for all eruptions. This assumption is tested here for the 57 eruptions and intrusions recorded at Kilauea between 1959 and 1984. Applying rigorous criteria for selecting data (e.g. maximum depth; restricting magnitudes to be greater than the completeness magnitude, 2.1), we find a much less pronounced increase in the aggregate rate of earthquakes than previously reported. The stacked trend is also strongly controlled by the behaviour of one particular pre-eruptive sequence. In contrast, a robust signal emerges among stacked VT earthquake rates for a subset of the eruptions and intrusions. The results are consistent with two different precursory styles at Kilauea: (1) a small proportion of eruptions and intrusions that are preceded by accelerating rates of VT earthquakes over intervals of weeks to months and (2) a much larger number of eruptions that show no consistent increase until a few hours beforehand. The results also confirm the importance of testing precursory trends against data that have been filtered according to simple constraints on the spatial distribution and completeness magnitude of the VT earthquakes.