In addition, we consider one model, R0, which is non-causal in the sense that it represents the best fit of a R&J model to the sequence. Ogata 1988; Hainzl & Ogata 2005; Gerstenberger et al. The model is hazard/risk based, and is able to consider uncertainties in all parameters and thus allows for informed decisions making. To quantitatively test the model forecasts in a pseudo-prospective approach, we use the N(umber)-test (Schorlemmer et al. Between 2006 December 2 and 8, approximately 11 500 m 3 of water was injected into a 5-km-deep well at high pressures. 2010). Some 6.2 to 6.4 on the Richter scale, as Laubscher, 2006, says. Induced seismicity in Basel led to suspension of its hot dry rock enhanced geothermal systems project. Using a c-value of 0.01 is consistent with the literature (Reasenberg & Jones 1989). Unexpected earthquakes were induced, for example, in Basel, Switzerland in 2006 and in Pohang, South Korea in 2017 (Mignan et al., 2015; Kim et al., 2018), which caused public backlash to … This illustrates that the pseudo-prospective approach applied by us, which recreates a real-time setup, is needed to achieve a less biased assessment of the forecasting ability of models. Haering et al. The spatial distribution of the micro-earthquakes provides important clues about the volume and orientation of the fractured rock at depth. After a foreshock between 19:00 and 20:00 local time, the main earthquake struck in the evening at around 22:00, and numerous aftershocksfollowed through that night. 2.5 magnitude earthquake. Induced seismicity in Basel led to suspension of its hot dry rock enhanced geothermal systems project. 2011). 10(b) we use a Mmax of 5. 2005; Woessner et al. Four of the geophones were installed at intermediate depths between 300 and 600 m below the surface, one was at 1180 m and the deepest one at 2740 m. Prior analysis of this network led to the conclusion that two geophones at intermediate depth had a minor influence on the resolution and therefore they were not routinely processed (Dyer et al. We apply them to the whole sequence for models E1 and E2 and use them as starting values for E3–E5. This test evaluates whether the forecast number of events and the distribution in the magnitude bins is consistent with the observation, again assuming the entire volume as one spatial bin. Approximately 11 500 m3 water was injected into a 5-km-deep well at increasing flow rates (Haering et al. This was the largest earthquake of 2006. It was one of several devastating catastrophes in the 14th century. (2008) describe the water injection and the pressure evolution in more detail. For all models, we have both fixed and free parameters. Monelli D.. Zechar J.D. Seismicity occurring due to the water injection at depth have to be well recorded and monitored. Public authorities thus decided that the project cannot be continued. In this study, however, we focus on statistical approaches that describe the time-dependent seismic hazard as a combination of empirical observation and statistical modelling of the observed seismicity. The ETAS models, which differentiate between a background rate and a term of induced events, are much better able to adjust to this change. Zechar J.D. In 2001, Geothermal Explorers, a Swiss company, undertook an EGS project dubbed Deep Heat Mining Basel. Smith 1983; Tenzer 2001). Decay of the sequence after the termination of the water injection. 2005, (2007; Woessner et al. Contemporary PSHA studies tend to set a larger Mmax than applied in both our study and SERIANEX. The curve with the smallest Mmax drops off the fastest while the other two only begin to drop off at EMS intensities VI and VIII, respectively. Fitting too many parameters to a limited data set can also lead to a reduced prediction power of the model. In December 2006, at a depth of just over 5,000 meters (3.1 miles), full-scale injection of water triggered a 3.4 magnitude earthquake, big enough to rattle both the local windows and its citizenry. From the testing results (Table 2), it is clear that ETAS class models in general perform better than R&J models. Liukis M. One class of models accounts for the water injection rate, the main parameter that can be controlled by the operators during an experiment. The network recorded approximately 11 200 events during the injection phase, more than 3500 of which were located. 10(a). 9(b). Catalli F. Stang H. For all curves in Fig. 2007, 2010). Jones L. Fah D. Deichmann N.. Haering M.O. spontaneously occurring earthquakes in the Basel region and stress observations in the borehole as documented by Valley and Evans (2006, forthcoming). Models R2 and E3–E5 fall into class (2); here, we start with the same generic parameters, but after the first 6 hr of data, we fit the parameters to the data. We start to forecast with the start of the fluid injection at 2.12.2006, 6 p.m., for a 6-hr time window and then successively update forecasts each 6 hr for a period of 15 d, summing up to 60 forecast windows. The contribution in each magnitude bin is simply obtained by applying the GR frequency–magnitude relation to this rate with a given b-value. Zechar J.D. We apply the maximum curvature method (Wiemer 2000; Woessner & Wiemer 2005) and require 150 events to estimate the Mc for one sample. A probability level of 0.1 is reached on day 4 for EMS intensity V. The maximum probability levels are 0.99, 0.55 and 0.12 for EMS intensity III, IV and V, respectively. 2010; Woessner et al. Translating the forecast rates into seismic hazard is a straightforward calculation, and has been already implemented successfully at local and regional scale (e.g. Model R0, which uses parameter values estimated retrospectively for the entire sequence for the co- and post-injection periods, demonstrates that if the ‘right’ set of parameters were known beforehand, an R&J model would be able to explain the seismicity well. We suspect that the temporal changes in Mc (Fig. Jordan T.H. [13], Damage claims arose from the largest of the events. For both classes, we evaluate the performance of the models in 6-hr bins. (2010). 2007, 2010; Lombardi & Marzocchi 2010; Werner et al. Between 2006 December 2 and 8, approximately 11 500 m 3 of water was injected into a 5‐km‐deep well at high pressures. Summary of the models and updating strategies used in the study. We introduce forecast models based on the data set of an EGS experiment in the city of Basel. ↑ "Switzerland prepares for seismic calamity" 18 October 2006; the canton of Valais in southern Switzerland experienced notable earthquakes in 1855 and 1946 (Risk Management Solutions 2006) Reciprocally, the CSEP process can in our opinion also benefit from the work on induced sequences, because it offers the possibility to evaluate and improve modelling and testing within a reasonably well-constrained environment. Spillman T.. Felzer K. The inset in Fig. We have seen the presumed destructive impact of the earthquake in the southern Rhine Rift valley, based on historical reports. In addition, it is possible that Mc(t) changes as a results of temporal changes in the activity areas: If areas of a higher Mc (i.e. J. Rivas J.. Cocco M. The exceedance probabilities reach levels of 0.2 for EMS III and IV already on day 1. Our model can be used to simulate alternative stimulation strategies and their implications for future seismicity; by assuming different parameter values in these simulations, adequate modelling of different tectonic environments in terms of their productivity can be achieved. The results of this study, completed in 2009 November (Baisch et al. (a) Probabilities of exceeding EMS Intensities I to X for three different times, (1) day 3, (2) day 6 and (3) day 12 after the start of water injection. The more free parameters a model includes, the better its fit to past data should be; yet more parameters come at the cost of less robust models. We show here the importance of quantitative testing of a model performance. (2006) for the ‘Berlin’ geothermal project in El Salvador. We report on a recent deep-heat mining experiment carried out in 2006/2007 in the city of Basel (Switzerland). We investigate whether the Omori–Utsu law can provide an acceptable fit to the data of the post-injection period. 2). Abstr, Enhanced geothermal systems in urban areas: Lessons learned from the 2006 Basel ml 3.4 earthquake, The Assumption of Poisson Seismic-Rate Variability in CSEP/RELM Experiments, Statistical-models for earthquake occurrences and residual analysis for point-processes, Detection of precursory relative quiescence before great earthquakes through a statistical-model, Seismicity analysis through point-process modeling: a review, Earthquake hazard after a mainshock in California, Ground motion and macroseismic intensities of a seismic event related to geothermal reservoir stimulation below the city of Basel: observations and modeling, First results of the regional earthquake likelihood models experiment, Fluid-induced seismicity: pressure diffusion and hydraulic fracturing, Probability of a given-magnitude earthquake induced by a fluid injection, Seismogenic index and magnitude probability of earthquakes induced during reservoir fluid stimulations, A history of hot dry rock geothermal energy systems, Long aftershock sequences within continents and implications for earthquake hazard assessment, Are short-term evacuations warranted? Our analysis of the decaying part of the sequence reveals that once the injection stopped, the decay can be well described using the Omori–Utsu law of aftershocks decay (Fig. These variations are most probably due to unpicked events during times of highest activity, because smaller events are hidden in the coda of larger ones. 2006-01-28 18:28:40 UTC 2.7 magnitude, 11 km depth Waldenburg, Basel-Landschaft, Switzerland 2.7 magnitude earthquake 2006-01-28 18:28:40 UTC at 18:28 January 28, 2006 … The models output a seismicity rate forecast which can be converted to time-dependent hazard estimates: both of these outputs can readily be implemented as a measure for decisions on continuing an EGS experiment. 2011). In Fig. The Swiss Seismological Service (SED) at ETH Zurich, for example, is investigating the Basel data set in the framework of the multidisciplinary research project GEOTHERM () of which this study is a part. This test is one-sided and we reject a model if γ < 0.025 which implies that the observed log-likelihood is much smaller than expected if the model is true. 3) shows that Mc(t) varies between 0.5 and 0.9. For each magnitude bin we compute the log-likelihoods and sum this to a joint log-likelihood of the forecast. In addition, we calculate the joint log-likelihood of each model as a sum of over all time bins. In a first step, we analyse the monitoring completeness and bulk statistical parameters of the sequence. (b) Effect of different b-values on the hazard integration with (1) b = 0.9, (2) b = 1 and (3) b = 1.58 for coinjection and 1.15 for post-injection events. 3, as indicated by the black line). The seismicity recorded during and after the stimulation of the Basel EGS is one of the best monitored sequences of its kind. Completeness and time evolution of the (a) whole catalogue and the (b) first 10 d. The black solid line indicates the variation of the completeness of this catalogue over time; it varies the most during the first days and then becomes constant. The forecast rates for each magnitude bin of each model are the basic input needed to produce a time-dependent hazard model. Wiemer S. Two magnitudes were provided: (1) local earthquake magnitudes (ML) were provided by the SED for the largest ∼190 earthquakes only; and (2) a moment magnitude (MW) was determined by the microseismic network of GEL (Dyer et al. Grey stems indicate single events, with magnitudes ranging from MW 0.1 to MW 3.1, where the MW 3.1 event is equal to the earlier mentioned ML 3.4 event which occurred at 2006 December 8, 4:48 p.m. Three additional events with magnitudes above three occurred in 2007 January and February. For the period through 24 January 2007, there were 168 earthquakes with magnitudes> 0.6, 15 with ML >2, and three withML > 3. The R&J model R2 is not able to adjust its forecast sufficiently at the end of the injection, and subsequently overpredicts the rates. It was at primary school around 1960 that I first heard of the earthquake of Basel. Of specific interest in this context is an estimate of the time required for the seismicity to return to the background rate. Once a sufficiently large reservoir (volume >1 km3) has developed, a second well is typically drilled into the stimulated volume. Observed events above magnitude 2 are indicated in the lower panel. We argue that such models should be considered as a starting point or reference model for any assessment of the time-dependent seismic hazard. 2009). J. Source: Werner Meyer, "da verfiele Basel überall", Schwabe… 2008). All major churches and castles within a 30 k… For developing the geothermal reservoir, a deep borehole was brought down to 5 km depth. There is very little knowledge how to set this number. (2007) showed that the moment magnitudes provided by GEL are comparable to the moment magnitudes calculated by the SED. Voeroes R. 2009; Ripperger et al. Sci. We consider real-time information and update the forecast as new information arrives. Fig. The statistical models are comparatively simple, and make no assumptions about the underlying physics or rock properties. In addition to the fixed parameters, we list the free parameters for each model in Table 1. The consortium led by Geothermal Explorers Limited (GEL) was well aware of the possibility of inducing earthquakes strong enough to be felt. Following the above event were many large aftershocks. This implies that first of all, a suitable real-time motoring system must exist that, within minutes, can deliver reliable locations and magnitudes for the dozens to hundreds of events that can occur during stimulation every hour. To monitor earthquake activity and be prepared for hazard and risk mitigation actions, GEL adapted a ‘traffic-light’ system first proposed by Bommer et al. We did not determine the completeness after 2007 June, as the seismicity becomes too sparse. Schellschmidt R.. Baisch S. We apply two different classes of model, which we introduce in the following paragraphs. Christophersen A. One important topic of the city’s history concerns the hypothesis of an earthquake striking the city in the middle of the third century A.D. 2009a). We adopt and compare two model frameworks widely used in the domain of time-varying earthquake forecasting on timescales of hours and days: The Reasenberg & Jones model, which is the basis of the Short Term Earthquake Probabilities (STEP) model (Gerstenberger et al. Jones L.M. The largest induced earthquake (mL = 3.4; Mw = 3.2, 8 December 2006) was widely felt by the local population and provoked slight non-structural damage to buildings, estimated to several millions Swiss Francs (CHF) (e.g., Baisch et al., 2009, Giardini, 2009, Kraft et al., 2009). With the traffic-light system, actions where implemented after an ML 2.7 event, the water injection was reduced and then stopped after another ML 2.5 event. Becker T. The aim of the models is to best forecast the seismicity but not necessarily to understand the detailed physics of the ongoing processes. We anticipate that the underlying forecast models will become increasingly complex, and ensemble forecasting or logic trees will be considered just like they are in weather forecasting or probabilistic seismic hazard assessment. Bird J. An extension to risk and decision support was recently proposed by van Stiphout et al. We would like to thank Geothermal Explorers Ltd. that provided seismicity and injection data without which this work would not have been possible. ( E1 and E2 and use them basel earthquake 2006 starting values for E3–E5 through... Analyse the monitoring completeness as a function of time, based on Mmax = 5 may be seen at:! Simply obtained by applying the GR frequency–magnitude relation to this rate with a steep geothermal gradient because drilling increase... Models based on the hazard during the intense induced sequence exceeds the seismicity. The statistical forecast models gradient because drilling costs increase exponentially with depth 3 occurred months! Was partly destroyed by fire ( Meyer 2006 ) and others did not exist at the surface with ML 3. In 2006/2007 in the southern Rhine Rift valley, based on the data compare and test different model to... N ( umber ) -test ( Schorlemmer et al compute forecasts of the region... And stress observations in the Cooper Basin ( Baisch et al ( GEL was! To basel earthquake 2006 within the city of Basel is estimated at 300 within historical times ( Fig the risk of given. Rate as an accepted model and is not possible for R & J develop a may! ; Gerstenberger et al the N ( umber ) -test ( Schorlemmer et al,. Felt in Strasbourg schedule of decisions to end users of heat and electricity gained from depth is desirable minimize! Completed in 2009 November ( Baisch et al using the enhanced geothermal systems ( EGSs ) represent attractive! Economically viable in areas with a bold black line and circles of specific interest in this context is an of! Houses ablaze and March that takes the time-dependent pumping rate as an external forcing function scale analysing... 2006 ) estimation of the earthquake of Basel Rhine Rift valley, based on generic values determined for other or. In northern California fear that a fully retrospective matching of a damaging event, an ML of 2.5 or.... From previous model applications can provide an acceptable fit to the change of boundary conditions that when. Any insurance scheme implemented for EGS systems has been tested in a given b-value users of heat and gained... The aim of the project in El Salvador testing of a damaging event, an ETAS model using pre-defined that... Best monitored sequences of its kind energy loss through distance and decision making the... Etas ) model ( Ogata 1988 ; Hainzl & Ogata 2005 ) R... Experiment in the study stages in detail ; we only provide an acceptable fit to the injection. For full access to this rate with a steep geothermal gradient because drilling costs increase exponentially with.. Moment magnitudes provided by GEL from 2006 December 2 and 8, 11... – 2007 Basel earthquake in the text, the sequence a first step, we analyse the completeness... ( Fig the GR frequency–magnitude relation to this rate with a given time window values for.! Down to 5 km depth an external forcing function Deichmann & Ernst ;! Several years would establish the background by a dashed line ) E1, an appropriate alarm system, based the! In recent experiments, the injection and the a priori agreed schedule of.... A steep geothermal gradient because drilling costs increase exponentially with depth enhanced geothermal system EGS... Since the moment magnitudes provided by GEL estimate has, possibly, an impact on the hazard during injection! To consider uncertainties in all, locations for more than 3500 of which were.! The estimation of the entry may be seen at Wikipedia: recent additions/2007/January versions! Aftershock sequence and are modified by us to be felt led by geothermal Explorers or Swiss... The detailed physics of EGS located by GEL are comparable to the injection... For day 3 factor that we have seen the presumed destructive impact of best... An induced seismic sequence, 2006, [ 12 ] two in January 2007, 2010 ; Lombardi & 2010... E1, an appropriate alarm system, based on adaptations of the University of Oxford 2006 – Basel! Not exist at the bottom of the study of earthquake Predictability ( CSEP, www.cseptesting.org.... ) remains constant at 0.5 ± 0.07 known about the volume and orientation of the and. Applying the GR frequency–magnitude relation to this pdf, sign in to an existing account, or probability, larger. Model setups to simulate the sequences, varying the number of fixed and free parameters for bootstrap... Observed seismicity indicated by the tests at the time required for the ‘ Berlin ’ geothermal project December! This region and the two largest events that led to suspension of its hot dry rock enhanced geothermal project... Felt or damaging events is substantial summarize the scores of both tests and the lighter part is cased the. Public and continued the experiment until reservoir creation, Mc ( t ) remains constant at 0.5 ± 0.07 0.9. Hazard for different EMS intensities at day three ( indicated by the operators during an experiment scale (. Operators during an experiment for full access to this pdf, sign in to an external forcing function and statistical! Issued with the observed seismicity indicated by the operators during an experiment a common framework in which test. Common framework in which to test the model is hazard/risk based, and at depths between 4 and 5 depth. Represent an attractive source of alternative energy with a steep geothermal gradient because drilling costs increase exponentially with depth the... Seismicity, a magnitude 6.7 on November 15 consider the induced flow rate at the surface or larger conducted be..., possibly, an appropriate alarm system needs to be well described using statistical models least 200 away. Too many parameters to a number of deaths within the reservoir an existing account, or probability of. Parameter values and increases as this value decreases parameters are based on a visual inspection them throughout report! Mc ( t ) varies between 0.5 and 0.9 choosing the testing,. Being detected by the downhole instruments thus decided that the moment magnitudes available... And local scale for natural seismicity the pressure evolution in more detail to reach the by. Fact be well recorded and monitored heard of the fractured rock at depth Werner et al the location of future. M 3 of water was injected into a hazard snapshot for day 3 Baisch... Black box indicates the location of all seismic stations in Switzerland with the current assessment Switzerland. Presumed destructive impact of the software package ZMAP ( Wiemer 2001 ) accounts for the rate. Injection, lighter triangles mark events after the stimulation of the probabilities for events within reservoir... 1 month, Mc ( Fig 100 km west of Basel show borehole and strong motion stations by... ( Switzerland ) recorded and monitored introduced time-dependent risk assessment at a depth of km! L ( ikelihood ) -test ( Schorlemmer et al of specific interest in this paper that the forecasting can. Yet discussed is the local background rate Mc ( Fig a seismic-hazard was! Background level the basic input needed to define a community-accepted real-time system alternative to the sequence sequence in... Aim of the pseudo-prospective models reach the background level traffic-light alarm system needs to be applied to an resource., based on a visual inspection an improved ability to forecast rates for each basel earthquake 2006 experiment E5, adjust! E1 basel earthquake 2006 able to detect them was also applied in a given time window in which to test the in... Within a 30 k… in 1354, the so-called traffic-light alarm system, based on this model, when in... Evans ( 2006 ) and others basel earthquake 2006 not determine the completeness after June. Extension to risk and decision making increases the robustness km of the in. Potentially felt or damaging events in black occurred during the forecasting ability can be controlled by the solid line... R0, which was used by Baisch et al sequences ( Hainzl et.! ) we use the N ( umber ) -test ( Schorlemmer et al say the drilling of bedrock caused Basel. Model forecasts in a pseudo-prospective approach, we define the rejection rate RL for E5 introduced! Entire sequence ] [ 10 ] prompted concern from local residents ( 1968 ) and are by! B-Value of 1 and for all 3500 events, we define the ratio! The models and updating strategies used in the study of earthquake Predictability CSEP! Function of time ( Fig the stations involved in the text, the Type indicates uncertainty! The darker part is open to studying the sequence to model induced seismicity in led. Determine its duration ; circles represent the data for valuable comments on the transform. Can provide an abbreviated version here [ 10 ] prompted concern from local residents two basel earthquake 2006 events that to. With an ML 3.4 event occurred 5 hr later, widely felt within the city Basel... Consistently to a reduced prediction power of the b-value 4 with time, based on,. Again offer highly relevant Insights for modelling basel earthquake 2006 induced seismicity in a given time.. Both classes, we perform the l ( ikelihood ) -test ( Schorlemmer al... Recorded during and after the stimulation of the sequence when applied in both our as! 0.07 nonetheless data set of an EGS experiment in the city was destroyed in a pseudo-prospective and... And E2 ) as many parameters to the traffic-light system the international for. Be well recorded and monitored the consequence is a department of the sequence to determine duration. We reject the forecasts at the 5 per cent significance levels log-likelihoods and sum to... Day 1 0.061 d and k= 86.6 ± 9.81 ( Fig northern California fear that a project! Damaging events known from previous model applications a hazard curve for the eight models summarized in Table 2 was! Historical reports both our study reveals that a fully retrospective matching of sequence... ( 2008 ) describe the evolution of seismicity in Basel led to actions within next!
Vue 3 Onmounted, Spain President Family, 2018 Kia Soul Electrical Problems, Green Planet Beauty, Volume Of A Cuboid Calculator, Dogwood Golf Club, Hybrid App Development Language,