Interview on the occasion of the scientific discussion about which mechanisms led to the seismicity in eastern Spain in 2013 and which hypocentral depth can be assumed.
The so-called "Castor sequence" off the coast of Spain near Valencia marks one of the most important cases of induced seismicity in Europe. It was a rare example of seismicity triggered by gas injection into a depleted oil field. There are hundreds of such underground gas reservoirs around the world, although they are usually never a seismic source.
However, the 2013 seismic sequence at the Castor offshore injection platform included three earthquakes exceeding magnitude 4 and occurred during the initial filling of the proposed underground gas storage facility. The processes underground are still being further investigated and there is scientific discussion to determine the exact cause. The filling of underground gas storage facilities remains a highly topical issue in order to reduce dependence on potentially fluctuating supplies from abroad. Precise understanding of the process is also important for other contexts, as the injection of fluids is necessary in the context of wastewater disposal, or in conventional hydrocarbon extraction and geothermal use.
Simone Cesca and his colleagues presented their new study on the Castor sequence to Nature Communications in 2021. Using very advanced seismological techniques, they were able to describe the triggering processes of the Castor sequence much more precisely than before and they were able to better explain previous scientific inconsistencies, so that a clearer overall picture emerges. In particular, the geometry of the activated fault zones was better understood.
If underground gas reservoirs are located in areas with geological zones of weakness (faults), then seismicity should always be closely monitored, as even earthquakes of medium magnitude but shallow depth could cause local damage to buildings or in the reservoir.
Determining the exact depth of origin of the earthquake is therefore of utmost importance. In 2021, the Spanish-Italian-German team applied the latest seismological techniques to an extended waveform dataset in order to identify the geometry of the faults more precisely, as well as to increase the seismicity catalogue, which is so important for analysis, and also to record details of the rupture kinematics.
According to their research, the sequence occurred through progressive failure and exposure of the fault, with seismicity first migrating away from the injection points, triggered by pore pressure diffusion, and then back again. In the process, larger asperities loaded with higher stress ruptured and produced the largest earthquakes. According to their research, the seismicity occurred almost exclusively on a secondary fault, which is located below the reservoir and dips opposite the fault bounding the reservoir.
Recently, a group of scientists challenged these results scientifically and Simone Cesca's team responded to the comment in the journal Nature Communication. In this interview, Dr. Simone Cesca from the GFZ, who prepared the scientific response to this.
Why, nearly nine years after the seismic events on the coast of Spain, the Castor sequence is still important and of public interest?
Dr. Simone Cesca: The Castor sequence remains to date the most outstanding case of seismicity induced by gas storage worldwide. Gas storage operations rarely produce seismicity and there are no other reported cases, besides Castor, where this type of operations triggered earthquakes with magnitude above 4 which are felt by the population. Therefore, there is a high interest in understanding why these particular earthquakes occurred as they were somehow unexpected.
There is, for the obvious reasons also a societal interest in the Castor sequence. At least in Spain, the issue raised a major interest, because of the very high associated costs, that society has to carry over the next thirty years and because of the lawsuit that has recently been filed against some of the involved operation managers. The latter were not found guilty cause they had taken all the precautions. Normally, the filling of geological gas storage facilities does not cause any noticeable earthquakes.
Were there other reasons?
Cesca: Indeed, after many years of research, the interpretation of induced seismicity at Castor remains debated. Unfortunately, this is not a good example where the understanding of the seismogenic processes improved over time. While a number of scientific papers were published since the seismic sequence, they only focused on specific aspects, provided partially inconsistent results, and finally a common model was not agreed upon.
Take us to the current scientific discussion, what is it about?
Cesca: Most of the publications on Castor aimed at describing, modeling and explaining the seismic sequence. Scientists didn’t have much choice: the only source of information at this site were seismic data. There are no other geophysical or geochemical data to support the interpretation since the Castor sequence occurred offshore. However, if at least the seismicity data can be accurately modeled, it will help to investigate which seismogenic process took place in the shallow underground in response to gas injection.
At the moment, there is an apparent disagreement among the location and depth of the seismicity, and thus on the fault which has been activated.
What did the most recent studies show?
Cesca: After some debate, the most recent studies agree on the overall fault geometry, which would be parallel to the coast and dipping towards South East. However, there is still an open controversy on the fault depth.
In your recent study, published in Nature Communications, you greatly improved the accuracy of the hypocentres and the parameters of the earthquakes. Your aim was to bring together previous results and to some extent also to discuss why they were not fully consistent. There was some scientific controversy about your study. Could you briefly outline what that particular discussion was about?
Cesca: The discussion concerned the hypocentral depths and the mechanisms leading to seismicity. We showed that the source depth is about 3-4 km, which is just below the reservoir by applying a dedicated depth estimation method. The controversy concerns a larger depth of 4-10 km, estimated in another publication (Villasenor et al. 2020). However, the depth range was suggested for a set of earthquakes, including some weak ones, where depth was presumably poorly resolved; looking to larger events only (M above or equal 4), the same reference reports 4-7 km, which is in better agreement with our estimate.
You also suggested a triggering mechanism?
Cesca: We suggested a combination of pore pressure diffusion, controlling the weak seismicity in the first phase of the sequence, and the failure of loaded asperities, controlling the occurrence of the later, largest earthquakes. Our hypothesis is based on a clear spatiotemporal migration pattern of the seismicity. Here the discussion invoked the presence of other triggering processes, such as buoyancy, which we do not exclude, but for which we found no evidence.
Your pioneering study also showed that a detailed look at the dynamics of small earthquakes is possible even in the absence of a dense, local measurement network. How does this work, how did you proceed?
Cesca: Well, we should mention that a dedicated local monitoring is always desired. However, in many cases this is not available. Our study provides an example of successful analysis without a dense local network. This was possible thanks to the usage of advanced waveform-based seismological methods, such as probabilistic moment tensor inversion, template matching and waveform similarity analysis, which we have continuously developed in the past years. Part of the success is also to be attributed to our attempt to estimate source uncertainties, which were not fully discussed in previous studies on Castor.
What tools did you use to estimate depths since a lot of controversy has stemmed from that point?
Cesca: Since we do not have very local data, we took advantage of seismic arrays at large distance. Seismic arrays are dense installations in a small region. We used data at the GERES array in Germany, which has 25 measuring stations. It is the most sensitive array in Central Europe. Each station records the largest events at Castor, but the quality of single recordings is poor, because of the large distance to the source. However, since we have many stations with similar poor recordings, we can stack their signals to obtain a high quality one. We can then model this signal, which is very sensitive to the earthquake source. In particular we can model the delay among two seismic phases: the first propagate down from the earthquake focus to the far distance receiver, the second propagate upward to the seafloor, where it is reflected downward and again to the receiver. The delay of these two phases is only controlled by the propagation between the seafloor and the earthquake depth, and thus provide valuable information on the hypocentral depth. We tested different velocity models, including those suggested by the authors of the comment, and found evidence for a shallow focus.
Could you briefly outline which monitoring solutions for offshore industrial operations have been proposed? What did we learn from this case at the coast of Spain?
Cesca: Industrial operations typically induce micro-seismicity, which can only be detected if seismic sensors are located close to the operations.
In the case of offshore operations, such local monitoring requires the installation of ocean bottom seismometers and seafloor geodesy instruments, e.g. as those developed at GEOMAR. In the case of Castor, the closest seismic sensor was located at about 20 km distance, thus quite far away on land and there was only one ocean bottom seismometer at a larger distance. Although we were able to model the largest earthquakes of the sequence, such local monitoring system would have been important. Therefore, we agree that potentially seismogenic operations should have a dedicated monitored in the near field.
On land it’s different: In the case of deep drilling (> 400m), for example in Germany, the monitoring of potential vibrations is carried out according to the specifications of the respective state mining authorities, which implement the regulations of the Federal Mining Act (BBergG).
Gas extraction is an issue in the north of Germany and also in Netherlands. During these operations there is scientific monitoring of micro-seismicity. In other cases, like hydraulic stimulations where you inject water, you actually want to improve the permeability so you can exploit this energy source. Having a seismological network helps to know whether we’ve created a network of tiny fractures that conduct the heat. These seismological networks would thus help us in two ways: On the one hand, it can be used to check whether a project is efficient, and on the other hand, it could be used to detect at an early stage whether geological weak zones (or faults) are present.
Scientific discussions are complex: could you nevertheless highlight again the main point why your team highly disagreed with this particular scientific comment that you just got on your study.
Cesca: The comment makes a number of unjustified statements plus we also found out just now, that it was based on locations that were incorrect. The comment claims for example that large earthquakes do not nucleate at shallow depth where we locate the earthquake.
Earthquakes at Castor with a Magnitude of 4 cannot be considered large earthquakes, but rather as moderate earthquakes. Having said that, moderate earthquakes (and sometimes even larger ones) can well nucleate at shallow depth. We provided many examples for that in our paper. Shallow seismicity is, most importantly, very typical for induced seismicity. So we see no argument against our depth estimate. Furthermore, we tested our approach with different data and velocity models and found quite consistent results.
Another statement where we disagreed on concerns the lack of evidence for a hydraulic connection between the storage formation and the earthquakes depth, which is used to question our proposed triggering mechanism. However, local faults are known and they could well provide a connection. Moreover, pressure measurements which were done during the exploitation of the former oil field also suggested a hydraulic connection.
Having said all that, we recently also showed that some of their data was simply wrong. All these discussions though lead to very verified results and help us to prove again that the one big picture we were able to show is consistent.
What is next? Will the issue be discussed at the IUGG General assembly meeting next year in Berlin?
Cesca: Induced seismicity will be an important topic at IUGG in Berlin, but I hope that by then the earthquake sequence in Spain will no longer be controversial. With each new example, like the one off the coast in Spain, we learn. With the war in Ukraine, gas storage is an increasingly important issue, and induced seismicity in gas storage is something we don't want to have. But if we cannot avoid it in some cases, we should at least be able to control it.