 So, we continue with the caldera processes of caldera collapse with my presentation here. I want to give the seismological viewpoint to this phenomena. We have seen a very nice lecture this morning showing analog experiments, field evidence how calderas look like, and we have worked a bit on the Parada-Bunga caldera collapse and studied moment tensors and source mechanisms accompanying the subsidence and the collapse itself. And I want to present this, but also use this as an example for the practical. The aim of the practical is to perform, so you should perform your own moment tensor inversion. So, we have prepared data from one of these events or actually a whole day of data and I will introduce also a bit the concept of moment tensor inversion and the method we have used. So, the aim is that you apply this and try to estimate your own moment tensor inversion. So, the key question as you will see in the lecture is also how we interpret these earthquakes accompanying the collapse. And you will see this is not so trivial because there are large non-double couple of components in these mechanisms and it's really not really solved how to interpret this for such a complex process. And an ultimate aim would be whether we can learn something from the mechanism on the magma body itself. I think we are not yet on the stage because our models maybe are not sufficiently developed for this. I think you all know this event. I have one slide here giving an overview but there are two or three posters even on this process where you have maybe even nicer figures and most of the studies in the posters concentrate on the seismicity accompanying the lateral dyke intrusion and the eruption, but at the same time in for this 2014-15 Barabunga sequence there was the subsidence of 65 meter here of the caldera which is below the glacial cover and there was an unusual sequence of magnitude larger four events which are plotted here in red. And so it was interesting from a seismological point of view to look to these. Overall there were 35,000 events detected mostly smaller events these larger events range up to depending which magnitude scale you use maybe 350 or 400 events that can be started with magnitude larger than 3.5 and where momentans inversion can be performed. So we use broadband stations from Iceland and this shows the network so it's the EMO broadband seismic network. The station coverage is not perfect as often because here is a gap of stations but overall there are quite a lot of broadband stations. Unfortunately there were no close by stations which would have been very helpful so we have to rely on these data sets I had to rely in our study using stations which are a bit more far away. So overall the distance up to which stations could be used was 185 kilometers and if you low pass filter the waveforms that have been recorded for instance here between 0.01 and 0.08 hertz this would roughly correspond to large wavelengths larger than 40 kilometers up to 300 kilometers. You have seen this red cluster of events all these more or less can be represented by a point source coming from the same point. But then it's interesting that you can recognize easily that there are different clusters event types in the waveforms meaning the mechanism must be different the location is more or less the same on this range but the mechanism is different there's a blue and a red cluster maybe some more smaller ones but these are dominant clusters and they have been sorted here with a method very similar to what you have exercised in the practical yesterday. So using these techniques correlating waveforms you can find out the clusters and you can make a plot like this and what we see actually is rally waves or surface waves so therefore yeah long period or in this frequency range these surface waves. So we will see later in the or in the next slides the mechanism we found for these two clusters and it turned out that they also happened to be on slightly on different sides of the caldera rim faults. For the moment tense inversion we used a method that has just been developed is even not not published but already in use which is which has the name so Sebastian Heimann one is has developed this method the name is gront and it comes from the ring of fire story I think there's a there's this gront so maybe some of you know this so if it's a kind of brute force method if you want to have a solution if you want to enter the castle so you use this ram and put it and drive it as long as you can to destroy the the doors of the castle and it's a bit similar it's a what is done here it's a bootstrap inversion meaning the aim is to simply sample the whole map model space and to find the best solution this is not very efficient in terms of computing time but on the other hand the problem is not very big if you go for a point source all together you have 11 parameters you have to solve for the origin time of the center eat the centroid locations with three parameters the moment tensor components I will explain later what this means with six components and the duration so these 11 parameters are solved with this bootstrap approach and the advantage is that you can produce figures as you see here where you can investigate more in depth the uncertainties and the trade-offs between parameters I mean it's not easy to make a plot with 11 dimensions and look at once but you can make cross sections one parameter over the others and then you see the families of best solutions and you see whether there are trade-offs for instance here between two components the isotropic component and the clvd component you will also see later what this means but these are for instance very interesting components to interpret the the events in terms of a of collapse mechanics and unfortunately there's a trade-off between these two parameters and if you only have a single solution sometimes it's not so easy to understand really what what are the uncertainties so this was the motivation for this method on the and it's also from in terms of programming in with a modern language I would say embedded in a toolbox in a python toolbox and all the data processing including restitution all this is done during the inversion on the fly so that you can easily change the parameters and then you produce finally fits and you will also produce these fits and the other plots we have seen of waveforms of low pass filtered waveforms and this frequency range for instance on a set component vertical component rally waves and on the transversal waves so this is actually what was inverted for these relatively large distances in regional ranges so you see overlaying the observed in the synthetic seismograms in these plots you see also some other parameters that are given and we can explain this in more detail during the lecture overall the fit is very good as you see so you would think your solution your best solution is there must be a good solution but as I said also for the laugh waves on the transversal component but as I said there that this does not mean that all parameters are very resolved so this is the aim of the exercise and the the example here introduces you to this method in principle and what we have done for the bada bunga collapse so we we applied this method first simply running a centroid moment tensor inversion and then the figure looked like this here you see the caldera rim you see the centroid locations and the uncertainties and you already see two types of mechanisms plotted here in this lower hemispherical projection later we will discuss this a bit more but what you can also see is I mean there are some trends you can possibly observe that red events this red type events have a different mechanism as the blue ones and they are more shallow this is a depth section than blue ones in in the other few also the red ones are more in the south than blue ones but overall it is not really satisfying because the uncertainties in epicenter location for instance and in depth seem to be quite large so we thought a bit what can be done to improve this and it's not so trivial because if you work with these long wavelength actually you don't have high resolution so the but it's maybe an example what you can try if you only have such data these broadband data and you want to improve your resolution a bit but finally the processing scheme is then non-standard I would say and and a bit more complicated therefore we prepare this plot so you this is not for the practical but it shows you how we try to improve the location for instance together with the moment tensor inversion so we had in principle two branches of analysis the moment tensor inversion is this branch too but at the same time you can do what you have learned yesterday you can make these cross correlations you can for this purpose you can even go to higher frequencies and you can find this template matching there are 400 events could be studied there and then you can analyze pairs of events estimate relative times between the wave arrivals and associate them with clusters of families of earthquakes you can estimate a relative strength if you know all the events in one cluster have the similar waveforms you can estimate a relative strength if they all come from more or less the same point source so a relative moment can be estimated precisely and a relative location but this is but if you go to the moment tensor inversion you have to and you want to invert for 11 parameters you have to ensure that this is a point source meaning you need long wavelengths and then you you you can run your inversion at these longer wavelengths but you can introduce station corrections and this is what we tried so we tried from the ensemble of all the data coming from the same cluster to estimate systematic the residuals at single stations and introduce station corrections it's a standard to introduce station corrections for location so for the centroid location it's it would be standard to ask to to introduce phase shifts for station corrections but we also try to to consider amplitude corrections which is not not really standard but it seems that it worked and really improved the results so then we had good let's say the best moment tensor solution we had we could get but for locating the events we could use then the results from these studies high-frequency studies and and then we can we finally produce these plots in comparison to before if we add the stations corrections station corrections in our moment tensor version the figure already looks a bit better in the term in the sense that the events are more cluster that more confined to the caldera the the depth here in between 10 kilometers and two kilometers and the assumed reservoir depth is in around 10 kilometer and then if you add the the relative locations from the goscalation of waveforms and associate these with the type of events then the figure finally that comes out looks like this so this is let's say the best result as we interpreted and here you see now interesting patterns much more clearly you see that the both types of events are more or less occurring on the northern and southern flank of of this rim fault there seem no events or no large events to occur here this is not true for the very very small events but for the larger events they only occur here which already possibly indicates the highest strains strain rates should have occurred on these two sides and the depth of these clusters is is different as we have already seen before maybe even better seen here and the mechanism so this is an has a normal fault in component we will discuss is a bit more this also but the strike of of the nodal planes here would be north south while here it's if you if you try to interpret it is more in this direction or more subparallel to the caldera rim fault and what you and usually you would expect from what what we have seen this morning if this is related associate defaulting on the sides that the strike should be subparallel to the ring fault and not perpendicular so well they vary a bit but but there's no systematic there's no systematic time variation there I will show a figure where you see this in terms of the moment release there's a there's some systematics and it can be easily nicely explained so maybe this is a so we want to interpret this mechanism you have seen you've seen this projections I think most of you principle know what this means but I I inserted in the in this presentation some very basic slides we can go quickly over this but only for those that possibly are not not absolutely familiar with the seismological concepts the the radiation pattern finally this is what we want to estimate if you have a double couple we know that the radiation pattern for p waves look like this you have these four lobes and for s wave it's different but this is the radiation pattern of a double couple of a sheer greck poorly sheer greck and it radiates this p and s waves in different columns and usually you estimate the nodal planes this is what I just tried to interpret or for it was not a double couple component that was plotted so this is associated with the p wave radiation pattern and yeah one point of discussion is always what is the two fault plane because this point source solution cannot distinguish between the two fault plane and the auxiliary plane they both equivalent if you plot this in a in a yeah in a projection here it looks much more simpler and these are the plots you often see for p wave radiation these four lobes for s waves it just turned maybe it's interesting to see that in the far field and near field it looks different not the orientation of the black and red lobes but the relative size it already shows you that if you can consider broadband stations from far field and near field and the transient signal you measure you would really have additional resolution on the double couple component but often your stations are too far away as also in our case in terms of the broadband centers and then it's also well known that this can this radiation pattern is plotted in a lower hemisphere and can be associated to the faults I don't and the orientation and tip of the faults so but coming back now to the moment tensor the moment tensor is a general representation of a point source not only a double couple or a shear crack and it it's a three times three matrix and it represents force couples generalized force couples with arm and without arm and the diagonal diagonal element of diagonal with arm so it's a general representation of a point source that can be associated to dislocation sources like shear cracks but also to other types of sources it's symmetric this tensor and as I said a general representation overall since it's symmetric these are six components and these are the six components we want to invert for in principle it is also time dependent every component is time dependent but if you really go to this point source approximation it's sufficient to have these six components you will you do not need really to resolve details of the time process the the moment tensor that you get this is the result you want to interpret and usually you you you have to do a moment tensor decomposition you decompose we can decompose the moment tensor in a yeah unique way in an isotropic part and the rest this is the deviatoric part the isotropic part is associated with volume change during the earthquake process but the deviatoric part is there's no unique decomposition valid so there are several opportunities or possibilities how to decompose it most common is to decompose the deviatoric part in a best double couple and a clvd a clvd is a vector dipole that is compensated for the volumetric change compensated linear vector dipole but you can also decompose it in a mixed mode dislocation if you think the deviatoric part or the general source has to be represents like a shear crack under mixed mode dislocation shear and opening at the same time then you can also use such a decomposition but it's different and the result would also look different and the angles of the of the planes would look different so this is the the problem with the interpretation of moment tensors and the decomposition and as I said it's a non-unique problem so it depends always on the on the process you study you cannot say this is always valid in this or the other way if you want to learn more on moment tensors I mean there I want to point out that there is this new manual of seismic observatory practice it's an online lecture material with very many different chapters including also a chapter on moment tensor description and moment tensor decomposition so it can be quite useful and if you want to plot moment tensor solutions or you want to decompose them according to the standard decomposition you can I can recommend this program it's a python program you can download and you can visualize the radiation pattern of a general moment tensor of a shear of the double couple component you can plot this from different view points you could this is platform independent and you can also decompose the moment tensor for instance here you see I mean only quickly so you simply type in mopad describe and then your components of the moment tensor you have the coordinate system for which this has been derived and then you get some results on strike deep break and so on and or you can decompose it so you get numerical results for the decomposed parts of the double couple the isotropic part in the percentage of these so without going in the theory I at least show you on this decomposition like you have this tool available and you can also look to the paper associated associated with this and you can also see how how this looks if you then plot these mechanisms the radiation patterns of general of moment tensors in different projections most often either lumbered or stereographic projections are used okay but how what what does the maybe it's it's worth to look before we come then to the interpretation of the barabunga events a bit more to the different elementary sources the shear crack is well known I think in terms of a moment tensor it has only yeah two components non-zero and which are have to be the same because it's in if it's plotted like this because it's a symmetric tensor and this is for instance a shear crack where the slip is in one direction and the fault normal is in three direction and the strength of the source is is given by the this is the shear modulus in my notation by the average dislocation in the area on the fault so if you estimate and this is related to the seismic moment directly that you can estimate if you know the seismic moment you can try and you know for instance the area and the and the shear modulus you can estimate the average slip of a source here yeah it is plotted these two components that are non-zero for and the radiation pattern associated with the maybe more of interest is then how how does the radiation pattern look if you have a tensile crack so opening poor opening if it's horizontally you would have three components here in the diagonal element two are the first lame parameters and then this and then two times the shear modulus in the last component added so if you would add here the trace you would find that there's also isotropic component left a trace is non-zero here and this means there isn't a volumetric change associated with this shear crack and the strength is also related to the opening in this case and the and the this opening and the area so this can also be this can really be observed and we have a nice example from a mine collapse where we started this magnitude 3.7 in Poland here and it was a broadband moment tensile inversion for instance using the program you will also use so altogether 18 traces also inverting this in low at low frequencies and the result looks like this pretty much as what we have seen before for for a poor tensile crack and the interesting thing here so there's a significant implosion component of 60 percent but it was a mine collapse and therefore all this makes sense at one moment and you see that it was documented in this case that there was a rock burst occurring in the mine and this is well known because even workers have been trapped in the mine for some days and they could even see in principle about what has happened and it was investigated quite good interestingly this all all this mine collapse event was triggered by a small sheath crack that triggered in principle the collapse event yeah there was a question sorry yes then this is not tensile but the the not opening but closing depending on whether it's opening or closing the colors will change here so no sorry this is opening and this is closing this is motion the first motion is positive or negative it is positive it's colored and if it would be horizontally and it goes up then it would be I would expect directly above to have a positive ground motion but the color yeah tells you whether this is either explosive or implosive you can also look to the explosion moment tensor which is very simple and for instance this is an explosion in terms of volume expansion there's no radiation pattern at all the p waves are radiated with the same amplitude in all directions and in theory no s waves are radiated the moment tensor has only all components are the same on the diagonal element it's independent of the orientation of the tensile always this and the strength is given by this factor here and this can also be tried to to we can try to see some cases where explosions have occurred recently in North Korea you had two interesting explosions nuclear explosions in January and in June this year we also studied these events with broadband stations also arrays and then you can see the moment tensor solution it's not a poor explosive component isotropic component but the major portion is isotropic for both events so in this case nearly 60 percent or 50 percent has this explosive component but there as you see are also some other components involved the overall radiation pattern of the moment tensor looks like here this is simply the result we found and this is a decomposition in a standard way with the largest part the isotropic component but also a clvd part and a double coupled part even 33 percent and it's always debated what does this mean how could this how can this be interpreted but it's yeah and one option is that during the explosion they're also like for a caldera process but maybe then the other way there are ring faults or some faults activated for instance above or below this area that that is and affected from the explosion if you if you plot these components the different components with more part for instance here you see again there's a wooden arm wooden mine example and poor shear crack strike slip from the Gulf of California and you can also produce plots like this which is a different representation this is this Hudson type plot where you can better investigate how large is the isotropic component how large is the shear crack component in this plot if it's a poor shear crack like this earthquake here the the point or this should um is represented by a point on these two axes which is zero zero so this means a poor shear crack if it's a poor opening or closing crack it would be oriented or it would be positioned here which is um yeah nearly the case for the wooden arm mine and if it's isotropic it should be here if it's explosive or if it's an implosion it should be here so these plots are quite helpful if you want to understand the portions of the different components in one figure for instance for the Korea explosion it does not look so simple the two investigated are these thick points and if it would be only isotropic it would be here but you see there are some other components also involved interesting was that there had been some more nuclear explosion in Korea in the years before studied by others and they had solutions here and here and we were puzzled a lot on whether this is so different or not but using this grant what you will apply where you investigate the uncertainties and if we only plot the uncertainties we have with our solutions then we see that there's a large smearing in this type of plot of the different components and this covers more or less exactly the range that of solutions that had been found before so meaning maybe it was not so different it's always the same it's more problem of inversion and uncertainties it points a bit to this problem it's important to know how how good the parameters are resolved what they are the uncertainties and what are the tradeoffs so now coming back to Bada Bunga we had this result as you see with this mechanism I have to see whether and now only the double coupled component is plotted in the radiation pattern you see better the strike which is here north south and here yeah nearly you know well a bit inclined but more so parallel to the fault so we have these two types we have for both components as you will see later large clvd components often 50 percent or yeah and we have these two clusters A is shallow which is the red one and B is deep in the north which is the blue one and the centroids are not in the center so they are really clearly separated one by each other which also already tells you something on the on the process of principle I will discuss it later and the moment release rate is different for A and B and this was your question and here's now the figure if you plot the moment release of blue and red you see clearly it's it's different this is the cumulative moment this is the time over the whole period of of the eruption principle when these events occurred so nearly 200 days or a bit less you see that the blue ones start with high moment release rate the deep ones but then saturated while the red ones continued much longer and this tells us something maybe whether the faults grow from bottom top or not because you mentioned yesterday they always grow from top to bottom this would in principle show the opposite yeah okay it's it's uh but it's interesting here and we will see a simulation later on this so how can how can these events be interpreted if there's a large clvd component there had been many studies before on seismological studies on caldera events collapse or or even uplift events also and suggesting for instance that the seismic events are associated with slip on the ring faults and then it depends on whether you have outward dipping or inward dipping ring faults or pulley vertical so exactly what we've seen this morning it makes a difference in the mechanism you would expect in this case you would expect as it was mentioned thrust faulting apparent thrust faulting type and the if it's subsiding during a collapse event you would then have the black part in the middle correct yes and in this because you have yeah this thrust faulting and this would lead to a to just the opposite so we can look to our clvd component for instance and and try to see whether it fits to one or the other but there is another problem that this model if this really if really the whole ring fault would be would be active and only then the solutions are valid in principle then the center it should be exactly in the middle and you cannot explain that the that the center it is is once here and once here and this is something we do not observe another point is if the faults are vertical or sub vertical there is actually no radiation maybe it's wrong to say there's no radiation but at least no radiation of second order moment tenses meaning the low frequency content is highly attenuated but high frequency base would be radiated this is something that in principle we also do not observe and we don't think that during this call-ups the whole ring fault was active only parts of it which can be better described by the parts of the of the fault slip so then we we try to use a standard decomposition for instance for this mechanism b on the northern rim the deeper ones but we had this strange observation that the double coupled component had a north south orientation with a standard decomposition in a clvd and a and a double coupled component this is the full moment tensor how it is plotted but you can make a you can make an a simple experiment you can assume that you have an outward dipping fault in the north this is small as evident from from other observations and and say the mechanism i i observe is maybe a composite mechanism of two processes one process is the slip on the fault and the other could be a reaction of the magma reservoir below the magma reservoir would very likely have only a clvd component vertically oriented so a vertical clvd while the um thrust faulting event would have the strike so if you simply add these two components in your point source solution you would for long wavelength radiation see only both together and if you give different weights to them to the double couple and the clvd from these two different sources you can generate a a bundle of mechanisms you would finally find with your momentum and you see if you decompose them in a standard way you flip suddenly from a east west oriented double coupled component to a north south and also the sign from thrust to normal faulting flips this would this is simply a fact that the overall mechanism is here controlled by the clvd and this over prints or in principle the double couple radiation so and if it's controlled by the double couple then you would not find this result so in principle this tells us a standard decomposition in some cases can be can be difficult um especially if the double couple component is small if it's very large you would not risk to make a mistake but if it's smaller even smaller maybe than the clvd component you have to be very careful by interpreting the double couple component for the southern rim we can make the same game assuming that this is a really vertical fault and then um so this would be in this range for instance we would we can explain exactly what we observe with the standard decomposition so maybe the mechanism flip from the orientations um is is simply related to the standard decomposition and to the fact that this is maybe not appropriate in our case so i've not shown yet figure what what is thought about the shape of the caldera and this is from a paper recently published where we have also contributed uh on one side with this modeling here this dm modeling and with the moment answers you have just seen um so here you see again the the caldera rim you see here the micro earthquakes which could be located with the the local short period sensor network and you see in this cross section that there it's already indicated that that one fault is on this cross section here is sub vertical the southern one and the other is slightly outward dipping and this was also used in this model and then and this model could very well explain the subsidence this asymmetric subsidence uh but this is uh not an an elastic continuum model it's a distinct element model that considers also inelastic um processes um um and nice is to see that you here it is plotted uh in in this result the um strain rate so high strain rate during the collapse is in in yellow or red so you see there's clearly high strain rate along the ring faults which can be expected if all this is subsiding but there are also some parts in in the in the in the body that experience apparently high um strain rates and the idea was if we plot this again um here um maybe a bit clearer from the coloring and if we only look to these two samples if we say one of the events clusters we observe is in the south this would be here no here this is north sorry north is deep this is in the south shallow and if for some reasons the ring fault or if we assume that most of this ring fault is simply as seismic subsiding but here in this area where there's a lot of um um yeah friction and and high strain rates we have the cluster of deep events and in this part where there's a complexity we have the um the shallow events type a and b and we can also compare this with the rates we expect for the earthquakes so this was the observed rate for red and blue moment release rate if we count simply the moment release rate from the simulation uh for red and blue assuming these two clusters we at least can can more or less find or explain this pattern that the deep ones start so in principle the process of high strain rate starts from the bottom and grows to the top in this in this numerical model and this is confirmed also with with the rate of seismicity which can in principle also already be seen if you simply plot the depth of them of the events for this time by eye you would also say that there's a it looks like an upward migration for the largest event which would tell in principle this okay so this is uh in principle the interpretation of the events um what we think it's likely or it is possible at least that this is a uh a uh a source process consisting of two different sources one is related to the radiation of waves is related to shear faulting but also to maybe a response of the magma reservoir this would be very interesting but as I said it's not uh fully established this model and the problem is maybe also if you look here in this Hudson blot again to all solutions we found for red and blue you see these trade-offs and they are relatively large and the only thing you can definitely say is they all have a negative clvd but some of them have positive isotropics and other negative isotropic components um I personally speculate that this isotropic component may be related to the depth and to the gas content but this is uh not all of our co-authors agree to this therefore we have not find a common conclusion on this how to interpret this because simply the trade-off is large so it's it's really a question whether you should interpret the isotropic component at all or not so in this sketch in principle you see a possible mechanism um you have uh this calvary you have the subsidence the piston goes down during the depletion of the reservoir this and on this side this is mostly um a seismic and we find two clusters of events one would be here in the inner part and the other here where we have maybe due to the geometry of the ring faults the complexity in in the in the faults that start to grow and um if always if there is an event with a shear crack uh at the same time you would also possibly see a reaction cosizing reaction of this magma chamber leading to a vertical clvd component and this is in principle what is what is written here and I think I directly come to the last slides now um to make the yeah to to start in principle explaining a bit the exercise so what you can do is now to to check our results in principle with the momentum inversion and then we hope that you first learn how how a momentum inversion can be run and what the possible problems are you should also learn then investigate these trade-offs and what happens if you change some parameters if you include stations or exclude stations if you change the frequency range but it's clear that in one and a half hour you cannot do all but you can start with a prepared dataset and then you can play with at home and continue if you are interested or you can contact us I've not really said a lot on momentum inversion itself the idea is in principle simple you have your source in the sketch plotted here there's maybe a complex source process as we just discussed maybe even two sources but if you are low pass filter your events you really go to a point of solution this can be described more or less as a step function and you have your observation at the station and this is also in the near field for instance this would also be only a step if you go to the very long periods and in the far field this would then be a pulse of the wave going so a very simple wave form so if you know the radiation or if you know the wave propagation from the source to the station then you you and you have a representation equation representation theorem of moment tensors that states that the displacement you observe at the station is simply a function or a linear function of the moment tensor which describes the six independent component in the point source times the green functions it's derivatives of green functions but all this what's here is simply the green function term or a green tensor then you can if you have more than one observations so this is here a component of the observations so northeast or that component if you have many stations many components you can write one equation after the other you have a matrix equation and at some point it will become over determined you have more observations and unknowns and you can simply solve for this so this is in a in a nutshell very quickly in principle what you do so you need data from the stations and you need green functions and this is the most important point with this slide to make clear only with data you cannot invert you need the green functions and you have to calculate them and green functions in this case are simply synthetic seismograms for elementary sources and this package you will use is a is a whole toolkit package that partly can calculate green functions you can plot green functions you can look to data filter data you can restitute the data you can download the data from the internet directly and and start the inversion all done in in more or less one toolkit which is called pyroco and this is the page here and we have installed this on the machines because it would possibly be too slow if everyone wants to install it and yeah as you see there are standalone applications which have nothing really to do with grunt but we will partly maybe try to use snuffler to look at two waveforms so it's a waveform fewer and yeah actually maybe if I still have time I can try it because I opened it here so you will see this is snuffler if you if you open it you see the data set you have it's one day 24 hours if I look these broadband data and there are some seen in the screen but you can scroll up and down or with minus and plus you can reduce maybe the size and then you can but you don't see the seismograms here this is because it we want to have it efficient if you look to the whole data set but you can easily skew scroll with the mouse and you see here for instance there are three events in this whole data set magnitude 4.5 or this one we can we can check one of the other we can zoom in and at some point then if you have closed enough you see the waveforms so these are the waveforms you also see that some are marked already because they were clipped and you should during an inversion avoid to invert clip traces you you have to throw them out but it's easier to look to the waveforms to to pick them simply and then the program has there's a pick file you this is considered for clip traces and then they they will not use these traces simply you can also filter a bit maybe I make this larger and you can hear in this snuffler it's nice that you can you can simply filter a bit these waveforms so you see how they change low pass filter high pass filter and then you can try to investigate whether the signal noise ratio is sufficient if you go to low lower frequency ranges so you should investigate a bit how deep is it is it useful the filter we have used or can you even go to lower frequencies maybe for the largest events you can but if you possibly I mean if you look here in this data set you see there are many more events which have not been picked so you can also try to invert for a small event that is seen here they are all most of them are really from this caldera so you have many options but I I don't want to explain now all what is snuffler doing and there's one one small problem on my laptop actually my mac it's not working since I have not properly working and therefore I cannot show all these things since I have updated the the mac os x system but for instance you can map you can plot a map with these snufflings and you can do this on these machines there where you can see the stations and the sensors or you can then also try to plot simply the phases phase arrivals of pns waves to help to interpret whether this is a p-wave or an s-wave which is also a snuffling which is called cake phases and the other one is mad okay but now going back here and actually I have to check how much time I still have if at all I think we have still some minutes not okay so where have I been so this ground package relies on this pyroco python tools with all these things like snuffler cake we have seen for most of it's maybe of interest we will not try it here but because we have prepared the green function data set already but you can you can also calculate your calculated yourself green functions and therefore you can use this for most of package you can choose between different programs in this case it was a program for a layered earth model a reflectivity code but you can also work with spherical codes and actually the spherical code is very interesting for those working with also with static displacement because in it includes also gravity and static displacement and the atmosphere and ocean coupling all this is included and you can calculate the green functions with this it's this page you can look to some more information for the green function database that we have also provided so if you don't want to calculate your own green function but for instance if you work globally there's a web page where you can already download earth models and green functions or if you calculate a new one you can send it to us and we can upload it here so there's a database of green functions because this green function calculation can be time consuming depending on whether how what is the frequency range how large is the the model and so on so this is inner directory provided and this is in principle what you what I suggest to do and actually what I have a small presentation but I we also printed out this what you see here and some more information how to use this program and and these two pages belong together and one is an introduction in grunt and what is the config file what the parameters means and the other one is what is it yeah how to get started in principle on this system here because you have to do some copies before some small installations everyone has to do I will show this on the screen during the practical but maybe it's good if you also have this sheet because everyone has a different speed and then you can simply also go after these commands and try to to do this all so we prepared this dataset you've just seen there's a bundle of earthquakes inside but at least three actually it's wrong three larger than five and many more larger than four I think they here's another mistake they are not restricted this data as I said it will be done during the inversion it was simply my old old approach that I typed in here and so the directory with green functions is provided and the software has been installed already and the idea is that you I mean there's already a configure file and in principle there are examples for the commands you can use to run this program and you can in principle only run the program but the idea is you first look to the data you check the quality of the data you can define a blacklist of stations this is maybe a main work if you think this station is bad I don't want to invert it you can put it in the configure file and then it will not be considered and yes this is the principle what should be done and you should also then interpret the trade-offs and and try to evaluate your result whether you think this is a good result or not let me see I think this is the end yes yeah yeah actually because for the practical we we do not invert for the source time function we go to to so low long wavelength long low frequencies that we assume simply in principle a step function also in a simple pulse not not no detailed information actually it would be possible to do this and there are also you have even seen it here on the poster there are near field observations I mean high high-rate GPS data for instance they are also limited in with the sampling but there are also these short period stations and if this would be included if these dataset would come together I think you can learn something on the source time function but this has not yet been possible for some reasons yes