10 Years of the Lusi Mud Volcano Disaster - Part 2: Did an Earthquake Trigger the Disaster?

By Mark Tingay, @Mark.Tingay

The Lusi mud volcano erupting at peak rates on the 2nd of August 2006, the day after eruption rates suddenly jumped from <40000m3/day to well over 100000m3/day (and up to estimates of 170000m3/day).

Foreword

This is the second part of a three-part blog marking 10 years of the Lumpur Sidoarjo mud volcano disaster. Part 1, examined the background and geology of the Lusi mud volcano. Part 3 will examine the evidence for and against the hypothesis that Lusi was triggered by a drilling accident in the nearby Banjar Panji-1 well. Note also that more details on mud volcanos and Lusi can be found by following my twitter feed all through May 2016 (@Mark.Tingay, #lumpurlapindo).

Introduction and Background

The city of Yogyakarta was struck by a powerful M_w6.4 earthquake at 5:54am on the 27^th of May 2006. Even in Indonesia, where natural disasters are tragically common, this 15km deep strike-slip event ranks one of the most destructive earthquakes in the country’s history. Over 5700 people perished and over 37000 were injured, with the earthquake resulting in a damage bill estimated at over US$3.1 billion. The earthquake triggered a noticeable increase in the eruption rate of the already erupting Merapi volcano 48km away, and one study has even suggested that the quake also caused an increase in the eruption rates of the active Semaru volcano 273km away (Harris & Ripepe, 2007).

The Lusi mud volcano first started erupting approximately 48 hours after the Yogyakarta earthquake. It is not known for sure when, and who, first started to claim an association between the Lusi disaster and the Yogyakarta earthquake (the disaster was initially believed to have been triggered by a drilling accident in the Banjar Panji-1 well by just about everyone (including the company drilling the well). However, by January 2007 (6 months after the disaster began), Lapindo Brantas, the operator of the Banjar Panji-1 well, was publically claiming no association with the disaster and proposing that the mud volcano was triggered by the Yogyakarta earthquake. Indeed, the Indonesian government has also, controversially, ruled that the disaster is a natural event caused by the earthquake. This official ruling is quite confusing. Firstly, because the government has still ordered that Lapindo Brantas pay for a large part of the damages associated with the disaster. Second, the ‘logic’ of the formal decision is essentially along the lines of: because it cannot be proven to 100% certainty that this disaster was caused by drilling, we will therefore declare with 100% confidence that it must have been caused by the earthquake (without actually bothering to similarly test whether or not the earthquake caused the disaster)! Regardless, the purpose of this blog is to focus on the science, and not cover the many and tortuous political aspects of the disaster. Though, I do encourage people to have a look into these aspects, as another fascinating example of how politics and preconceived biases tend to obfuscate science (as seen in climate change, GMOs, vaccines, evolution, etc).

Whilst the earthquake had been blamed for the disaster by several groups, most notably Lapindo Brantas, the hypothesis for a natural trigger to the Lusi mud volcano first gained scientific traction with the publication of the EPSL paper ‘Triggering and dynamic evolution of the LUSI mud volcano, Indonesia’, in July 2007 (Mazzini et al., 2007). This classic paper, and some of the follow-up studies led by Adriano Mazzini and his colleagues, forms the foundations of models and arguments proposing an earthquake trigger for Lusi (Mazzini et al., 2009; Mazzini et al., 2012). In this post, I will first go through the models and then cover the eight different arguments commonly made in support of the earthquake trigger, and carefully discuss the pros and cons of each. Please note that I am well known to favour the drilling-trigger argument, but will endeavour to cover both sides of the debate herein – though it will likely become clear as to why I am so much in favour of the drilling-trigger hypothesis!

The Earthquake Triggering Model for Lusi

A common misconception surrounding the Lusi disaster is that the two main triggering hypotheses (quake vs drilling trigger) are fundamentally different. However, I want to start by highlighting that the two hypotheses are actually remarkably similar when distilled down. Both hypotheses propose that Lusi was formed when a reduction in effective stress (stress minus pore fluid pressure) initiated or reactivated a fault/fracture at the Lusi location (Tingay, 2015). The activation of this fracture created a pathway for highly overpressured water, mixed with Kalibeng clay, to escape to the surface as the Lusi mud volcano. The key differences between the two triggering hypotheses are simply the primary source of erupting water (see blog part 1) and their ideas for what caused the effective stress to suddenly drop, and subsequently activate faults and fractures (Tingay, 2015).

In the earthquake triggering hypothesis, it is proposed that the shaking from teleseismic waves from the Yogyakarta earthquake was strong enough to induced liquefaction of the Kalibeng clays (Mazzini et al., 2009; Lupi et al., 2013). Liquefaction of clays is always associated with significant exsolution of gas, and the release of large amounts of gas is argued to increase the pore pressure (and thus decrease the effective stress) within the Kalibeng clays, triggering fault reactivation (Lupi et al., 2013; Tingay et al., 2015). Some examples of earthquake triggering models for Lusi are provided below.

Earthquake triggering model by Tingay (2016), which attempts to summarise the slight differences and modifications made to the earthquake trigger hypothesis over the past 10 years.

Earthquake triggering model as proposed by Lupi et al., 2013, with liquefaction induced by earthquake shaking, which is proposed o be amplified by the subsurface structure and lithology at the Lusi site.

Quake-trigger Argument #1: Earthquakes are known to remotely trigger mud volcanoes

Mud volcano activity is very commonly associated with nearby seismicity. Indeed, there are dozens of known examples of mud volcanoes being initiated, reactivated or enhanced by distant earthquakes (Manga, 2007; Bonini et al., 2016). Some of the most famous examples include the occasional formation of new mud volcano islands off the coast of Pakistan following large earthquakes in 1945, 1999 and, most recently in September 2003. Indeed, it is known that mud volcanos, and other similar hydrodynamic features (e.g. geyser eruptions, increased spring discharges), can even be triggered by very large earthquakes up to approximately 1000km away (Bonini et al., 2016). Hence, the concept of the devastating Yogyakarta earthquake triggering the Lusi mud volcano eruption ~254km away, is not as far-fetched as it may initially seem. However, there are two key criticisms of this argument.

First, the argument that earthquakes can trigger mud volcano eruptions does not, in any way, demonstrate that an earthquake triggered the specific Lusi eruption. It is entirely valid to note the relationship between earthquakes and mud volcano eruptions as a means of highlighting precedent, but this evidence is circumstantial with respect to this specific instance. Similar arguments are often used to deny anthropogenic climate change – statements like ‘the Earth’s climate has always changed throughout geological time – hence, recent changes in climate must also be natural’. Yet, this is a clear logical fallacy. Whilst we know about many natural causes of climate change, we are aware that humans can also affect the climate. Similarly, as discussed above, both main hypotheses for the Lusi eruption argue for ‘something’ triggering a drop in effective stress – and we know that both natural and human processes can cause changes in effective stress and create or reactivate faults. The residents of Oklahoma have recently become very aware of the role humans can play in changing effective stress, and subsequently induce seismicity!

Second, the large dataset of known earthquake-triggered mud volcanoes has resulted in some clear relationships about just when earthquakes can trigger such eruptions (Manga, 2007). Michael Manga and his colleagues (most notably Max Rudolph and Marco Bonini) have empirically demonstrated a threshold of the minimum required ‘seismic energy density’ (SED) for a mud volcano to be initiated at a certain location. SED has the units Joules per cubic meter (energy normalised to volume), and is essentially a measure of the amount of earthquake energy (shaking) that is transmitted to a particular point in the Earth. The SED at any particular location is a function of earthquake magnitude and distance from the earthquake – with SED proportional to the magnitude of the quake and inversely proportional to the distance from the quake.

After compiling a dataset of >300 instances of earthquake-triggered mud volcanos and related hydrodynamic phenomena, Manga and his colleagues have noted that a minimum SED of 0.1 J/m³ appears to be required for remote triggering of these events, particularly at distances of >100km (see figure below). Furthermore, the vast majority of triggered eruptions involved SEDs of ≥1.0 J/m³, particularly for eruptions that started only a short time after the earthquake (as proposed for Lusi). Yet, the Yogyakarta earthquake only generated a SED of slightly less than 0.01 J/m³ at the Lusi location – less than a tenth (and generally less than 1/100^th) of the SED empirically seen to be normally required for an earthquake to initiate a mud eruption. It is important to note that not all quakes with SEDs >0.1J/m³ will trigger a mud volcano, but rather that there has never been any historically documented instance of when a MV has been initiated at SEDs lower than then 0.1 J/m3 threshold at distances of more than 100km.

Three different studies showing empirical datasets for the relationship between earthquake-triggered hydrodynamic and magmatic events and earthquake magnitude and distance. Note that there are often thresholds of minimum seismic energy density (SED, in J/m3) required for triggering to occur. The SED of the Yogyakarta earthquake was only ~0.01 J/m3, which is well below the normal SED ever seen to have triggered mud volcanism, liquefaction of hydrodynamic events at greater than 100km from an earthquake. Note also the red dots in panel (a), which show other earthquakes in the Lusi area prior to the Yogyakarta event. There are 13 prior quakes with higher SEDs that the Yogyakarta earthquake, including two that show SEDs over 10 times greater. One key failure of the earthquake triggering hypothesis for Lusi is that no argument has ever been made to explain why the Yogyakarta earthquake triggered Lusi when other larger earthquakes failed to have any effect.

Manga (2007) also noted that there had been a total of 13 earthquakes, prior to the Yogyakarta quake, that generated higher SEDs at the Lusi location – including two earthquakes that generated SEDs above the 0.1 J/m³ minimum threshold (red dots in figure part (a) above). Hence, when the Yogyakarta earthquake is examined in detail, it appears that the earthquake was too small or too far away to trigger Lusi (when compared to empirical data), and that there were numerous events that were more likely to trigger a mud eruption, but which did not.

The counter argument that the Yogyakarta earthquake was too small and too far away to trigger the Lusi eruption was supported by the analysis in Davies et al. (2008), Tingay et al. (2008) and Rudoplh and Manga (2012), whom looked at a range of methods by which earthquakes can remotely trigger mud volcanoes or fault reactivation (e.g. co-seismically induced stress changes (such as ΔCFS), post-seismic relaxation of stress, poroelastic rebound effects and dynamic stress shaking). In all instances, the Yogyakarta earthquake was found to be too small and too far away to have been able to trigger the Lusi mud volcano just 2 days later and 254km away. However, it should be noted that this does not preclude or demonstrate that the quake did not trigger Lusi. But, it does suggest that, in order to invoke an earthquake trigger for Lusi, scientists need to identify a, as yet unknown, mechanism in order to explain why the Yogyakarta quake caused the disaster (and why bigger quakes did not).

Argument 2: There are pre-existing natural mud volcanoes near Lusi

The first major scientific paper on the Lusi mud volcano stated that there were no other nearby mud volcanoes, and suggested this as evidence that Lusi was unlikely to be natural (Davies et al., 2007). However, this was incorrect, as there are at least six known mud volcanoes within 50km of Lusi (Mazzini et al., 2007). Even more interestingly, several of these (Lusi, Kalang Anyar, Pulungan, Gunung Anyar and Madura) lie along an approximate NNE-SSW trending narrow corridor, suggesting some potential linear zone in which mud volcanoes are more prone to be initiated.

Regardless of whether pre-existing natural mud volcanoes exist near Lusi, this does not provide evidence for or against the specific triggering of the Lusi disaster. Indeed, there have been claims that the nearby Gresik mud volcano was originally triggered by drilling several decades ago, and drilling near Gresik is known to have caused a small mud eruption in December 2008. However, the existence of nearby natural and man-made mud volcanoes near Lusi merely provide some precedence, and indicate that the area is prone to mud volcanism. This do not provide any sort of direct or conclusive evidence on what triggered the Lusi disaster.

Argument 3: Merpati and Semaru Volcanic eruptions possibly enhanced by earthquake

Both the Merpati and Semaru ‘real’ volcanoes were quite active in the weeks and months prior to the great Yogyakarta earthquake. However, Harris and Ripepe (2007) used satellite imaging data to suggest that both volcanoes showed a 2-3 times increase in eruption rate from 3-9 days after the Yogyakarta quake, and proposed that both volcanic eruptions were enhanced by the Yogyakarta earthquake. The hypothesis that Semaru, located some 273km away, could be affected by the quake has been used to argue that the same quake may also have been able to trigger Lusi 254 km away (Sawolo et al., 2009).

This is certainly an interesting observation, and remains one of the strongest arguments against the empirical seismic energy density triggering relationship proposed by Manga (2007). However, there are also a number of significant counter-arguments that cast some doubt on using these magmagtic volcanoes as evidence to support an earthquake trigger for Lusi. It should be noted that the data before and after the earthquake is highly scattered, and that the study only examined time periods ~18 days before and after the quake, and thus did not fully examine longer term eruptive values and variations. Secondly, Semaru and Merpati were already active and existing volcanic systems, whereas Lusi was a brand-new mud volcanic system. It is well established that it is easier to enhance and existing eruptive magmatic system than it is to create an entirely new system (Delle Donne et al., 2010). Indeed, empirical comparisons suggest that it is at least an order of magnitude harder to create a new eruptive system than it is to reactivate or enhance an existing one, and that the SED of the Yogyakarta quake was within the empirical range of enhancing magmatic volcanism, yet still outside the range of triggering mud volcanism or liquefaction (See empirical triggering figures above; Manga, 2007; Delle Donne et al., 2010). Furthermore, the Yogyakarta earthquake was unusual in that it was a strike-slip event, which means that earthquake effects can be highly directional. Analysis of slip motions does indeed suggest that Semaru was directionally well aligned for some possible enhancement from the Yogyakarta quake. However, the Lusi eruption site is poorly aligned with respect to earthquake directionality effects (Tingay et al., 2008). Hence, even if the Yogyakarta earthquake affected the Semaru volcano, this would not provide any direct or conclusive evidence that the same earthquake triggered Lusi.

Argument 4: Surface faulting after Lusi began erupting

A number of surface faults formed in the weeks and years following the commencement of the Lusi eruption. In particular, there has been clear strike-slip displacement of railway lines and embankments, in a number of places, as well as some extensional surface faults (Mazzini et al., 2007; Istadi et al., 2009). Yet, the occurrence of this clear faulting after the start of the Lusi eruption has also been cited as evidence that Lusi must have a natural trigger.

Example of faulting that occurred around the Lusi mud volcano in the months and years after the disaster commenced. This surface faulting is often used as an argument to support a earthquake-trigger for the Lusi disaster.

This is a utterly nonsensical argument for a number of reasons. Firstly, it suggests that faulting can only ever have natural causes – ignoring the well documented instances of induced seismicity in many parts of the world since the Rocky Mountain Arsenal seismicity in the 1960’s and possibly oil extraction activity in Texas in the 1920s. Indeed, both the quake and drilling-trigger hypotheses propose that fault reactivation may have occurred (earliest versions of the drilling-trigger argument examined only tensile failure, but most drilling-trigger models since 2010 have proposed that either shear or tensile failure occurred due to the drilling kick). Hence, even if there was clear faulting observed on the initial day of the eruption (there wasn’t – some surface cracks appeared, but no slip was observed), this would equally support both hypotheses. Second, the faulting all happened after Lusi commenced, and so there is a ‘chicken or the egg’ problem – it cannot be determined whether the mud volcano was triggered by faulting, or whether the later faulting is actually triggered by the mud volcano.

The occurrence of surface faulting after the Lusi eruption merely helps to confirm our prior knowledge that the present-day stress regime in the region is primarily strike-slip to normal (Tingay et al., 2010), and demonstrates that the area is now extremely active. This is hardly surprising given the large volumes of high pressure fluids being potentially pumped into faults, as well as the extreme rates of subsidence occurring in the area. But, the occurrence of these surface faults does not provide any evidence to support the earthquake trigger hypothesis, nor does it provide support for the drilling-trigger hypothesis, which also predicts such faulting to occur.

Argument 5: Reports of losses in BJP-1 coincident with earthquake

In a paper written by Lapindo Brantas drilling engineers and geologists, it is claimed that the BJP-1 suffered minor (20bbl) losses at approximately 6am on the 27^th of May 2006 (Sawolo et al., 2009). Losses at this time would roughly coincide with the arrival of earthquake waves from the Yogyakarta earthquake, and thus could indicate that the quake caused some sort of opening of fractures for drilling mud to flow into. This would provide some support for the argument that the earthquake may have triggered the Lusi disaster. However, there are a number of uncertainties surrounding this claim.

The 20 bbl losses are not mentioned at any time in any of the daily drilling or mud reports, nor in any of the subsequent contracted independent reports. The evidence presented for these losses is a short section of the surface mud pit volume chart, showing a drop in pit volume of ~20bbls (Sawolo et al., 2009). Yet, it is not actually clear whether the drop in mud pit is actually related to downhole losses, or whether it was just a routine transfer of some pit volume to the active system. Furthermore, the chart is not a great reproduction – it is blurry and it has no time scale. The original time marks maybe look like the drop occurred around 6am, but the numbers are blurry and could be either 6am and 5am, so it’s not really certain when the loss occurred. What is very clear to read are the depths they were drilling at when the loss occurred, which was 9274’ to 9275’. Yet, the daily drilling reports and other drilling data are very clear in highlighting that this was the depth at approximately 5am (DDR states 5am depth to be 9277’; Tingay, 2015), and does not coincide with the drilling depth at 6am (~9283’).

 

Mud pit volume chart from Sawolo et al. (2009) suggesting 20 barrels of downhole losses in BJP-1 approximately coincident with the Yogyakarta earthquake. Howeer, it is uncertain whether these losses actually occurred downhole. Furthermore, there are discrepancies between the time and depth that the losses occurred. The time stamp outlines in red looks like either a 6:00 or 5:00, while the drilling depth at the time of the losses on the left hand side of the chart (9274.2') was actually the drilling depth at slightly before 5am according to the daily drilling reports (digital appendices to Sawolo et al., 2009). More discussion on these losses, as well as a complete timeline of drilling events in BJP-1, is in Tingay (2015)

In summary, demonstrated downhole losses synchronous with the earthquake would be interesting. However, it is not certain whether these losses occurred downhole, and the only available data shows extremely odd inconsistencies with depth and timing. Hence, this evidence must be considered as inconclusive at best. Furthermore, losses downhole at this time, if true, are likely caused by a combination of both drilling and the earthquake. I recently presented a geomechanical model of BJP-1 in which the effects of the quake and drilling are examined to see whether they would be sufficient to induce shear or tensile failure (Tingay, 2016). Whilst this model highlights the much greater effect of the later drilling kick, it also notes that a very small section of the BJP-1 well, right near the shoe, may possibly have been fractured when the stated drilling equivalent circulating density (ECD, the increase in downhole mud pressure when pumping) is combined with the largest ever predicted effect of the earthquake. In this model, the quake on its own is insufficient to cause any fracturing and losses, but the much larger effect of ECD combined with the quake, may just have been enough to briefly cause losses. It should be noted that such an effect would be transitory, and so unlikely to trigger large scale fault reactivation. Furthermore, it should be noted that the well actually experienced a larger effective stress decrease every time the mud pumps were turned on to drill than was generated solely by the earthquake!

As a final note on this argument, Sawolo et al. (2009), also claim that “total losses occurred immediately after 2 major aftershocks” in the BJP-1 well.   Indeed, there were actually three Mw4.4 -4.8 aftershocks that morning, but these were 1.5, 2.5 and 4.75 hours prior to the total losses (at 12:50pm on the 27^th May 2006). Perhaps Sawolo et al. use a different definition of the word “Immediately”, but this claim would seem to be an exaggeration. Furthermore, it is hard to reliably link aftershocks (that regularly happen in the hours after a major quake) with drilling losses (that had also occurred at numerous times in the drilling of BJP-1), especially when there is a 90 to 285 minute gap between events!

Argument 6: ‘Hydrothermal’ gas signature from Lusi

Adriano Mazzini and colleagues have done exhaustive and exceptional work collecting and undertaking geochemical analysis of Lusi eruptive fluids and gases, which was published in EPSL in 2012 (Mazzini et al., 2012). One of the most interesting findings of this analysis was the geochemical evidence indicating a contribution of very deep fluids (thermogenic CO2, thermogenic hydrocarbons and mantle helium). This observation of deep fluid involvement was significant, as it was quite different to the originally proposed models for Lusi earthquake triggering, which had proposed that the eruptive fluids were almost exclusively sourced from the Kalibeng clays. In contrast, the geochemical data indicated a contribution from some sort of deep source, such as the carbonates or deeper Ngimbang Fm shales, as well as a potential hydrothermal aspect to the Lusi mud volcano (linkage to the nearby Arjuno magmatic volcano complex).

Schematic model of the proposed 'hydrothermal' plumbing system for Lusi, as proposed by Mazzini et al. (2012).

When I first saw this data presented (in 2011) I was excited, as such a contribution of deep fluids is as would be expected from the drilling-trigger model for Lusi triggering, with the bulk of fluids flowing from the deep Miocene carbonates (which would have been likely to contain geochemical inputs that had migrated from deeper formations). However, instead the geochemical data was used to promote a refined earthquake-trigger model, in which Mazzini et al. (2012) argue that these deep fluids had migrated up a fault zone and been pumped into the Kalibeng clay formation. The new model proposed that the upwards flow of these gases had ‘charged’ the clays and fault zone, essentially ‘priming’ the clays for liquefaction and fault reactivation.

‘Priming’ of the fault system by deep hydrothermal fluids is an interesting hypothesis, and certainly worth investigating further, as this may help explain why a fault might be reactivated at such low seismic energy densities when compared to empirical data (though still does not explain why other bigger earthquakes had no effect). However, there are some issues with this model that make it difficult to use as a reliable argument for Lusi triggering. Firstly, the existence of deep gases could, as stated earlier, also be an indication of support for the drilling-trigger model. Second, the samples were almost all taken many months to years after the onset of the Lusi eruption. In particular, only two samples (which did not have the detailed key isotope tests made on them) were taken before the 1^st of August 2006, which is when the Lusi eruption spectacularly increased in rate and is believed to mark a significant change in the subsurface plumbing system (see blog part 1). The lack of any baseline data (pre-eruption), coupled with the data being almost exclusively collected after a likely significant change in the subsurface fluid flow system, make it very difficult to have confidence that the measurements reflect the fluid chemistry in the initial days of the eruption (that are most critical data for examining the trigger to the disaster).

Late in 2014, while researching the initial pore pressures under Lusi (Tingay, 2015), I came across data that I, and many others had completely missed (and still kick myself for missing), and which was crucial for testing the idea that the Kalibeng clays were ‘primed’ by deep hydrothermal fluids. This data was Banjar Panji-1 daily drilling and daily mud reports, that (we hadn’t realised) were provided as a digital appendix in a paper by the Lapindo drilling engineers and geologists (Sawolo et al., 2009).

One of the interesting things about the initial Lusi eruption was that it was characterised by the release of significant amounts of H₂S. Indeed, H₂S was observed in the first days of the Lusi eruption, as well as at a few other key times, such as when Lusi significantly increased its rate on the 1^st of August 2006, as well as at the BJP-1 well, both shortly before the Yogyakarta quake and during the major kick on the 28^th of May 2006. The occurrence of H₂S from Lusi suggests that a small amount of H₂S is present in the primary source of Lusi fluids, and can thus be used to help identify where the initial fluids for the Lusi eruption came from.

The drilling records for BJP-1 include detailed gas geochemistry collected during drilling, with flammable and potentially toxic H₂S being a particularly important chemical to constantly measure, especially given its known common occurrence in the East Java Basin (Tingay, 2015). The records of BJP-1 make no mention of any H₂S being detected during drilling, other than early on the 27^th of May 2006 and during the kick the following day. In particular, there was no detectable H₂S from the >60m³ of Kalibeng clays removed during drilling of BJP-1 (concentrations of just a few ppm are detectable). Indeed, the only indications of H₂S come from the bottom 20m of BJP-1, which are believed to be the base of the carbonaceous volcaniclastics and, possibly, uppermost Miocene carbonates (Tingay et al., 2015). This is supported by H₂S being observed in other Oligo-Miocene carbonates in the East Java Basin (Tingay, 2015).

The occurrence of H₂S in initial (and greatly enhanced) Lusi eruptions, as well as the occurrence (and lack of occurrence) in different formations in BJP-1 leads to two key conclusions (Tingay et al., 2015):

         i.            The H₂S and initial fluids for the Lusi eruption (and drilling kick) are from a depth near the bottom of the BJP-1 well, and are most likely from the Miocene carbonates.

       ii.            H₂S is not present in detectable quantities in the Kalibeng clays, and thus there is no evidence to support the hypothesis that initial fluids for the Lusi eruption came from the Kalibeng clays, nor is there any evidence to support the claim by Mazzini et al. (2012) of deep hydrothermal fluids ‘priming’ the Kalibeng clays.

In summary, whilst there is indeed geochemical evidence of a possible deep ‘hydrothermal’ component to the Lusi eruption, there are several reasons why this does not provide any compelling evidence for Lusi to be triggered by the Yogyakarta earthquake.

         i.            The presence of deep ‘hydrothermal’ fluids is also in line with the drilling-trigger model for Lusi, and thus does cannot be used to distinguish between a quake or drilling trigger for the disaster.

       ii.            It is not certain when things such as mantle Helium commenced erupting from Lusi, as these were only measured after the significant increase in eruption from Lusi on the 1^st of August 2006, which likely represented a significant change in the mud volcano’s subsurface plumbing system.

     iii.            BJP-1 gas chemistry data indicates that initial fluids for the Lusi eruption did not come from the Kalibeng clays, and finds no indication that these clays had been ‘pre-charged’ by deep hydrothermal fluids.

     iv.            The BJP-1 gas chemistry data suggests that initial fluids for Lusi come from near the base of the BJP-1 well, most likely the Miocene carbonates, which is as predicted by the drilling-triggering model, and not by the earthquake triggering model.

Argument 7: Proposed amplification of earthquake waves by geological structures (aka the ‘Layer of Steel’ controversy!)

The Lusi triggering controversy was ‘re-ignited’ in August 2013 with the publication of a featured paper in Nature Geoscience by Lupi et al. (2013). This study undertook detailed numerical modelling of the effect of the Yogyakarta earthquake at the Lusi location. In particular, this study was the first to build a 2D velocity model under the Lusi location, and investigate, in detail, how seismic energy might be enhanced or dissipated by the lithology and structure under Lusi. A central part of this study was the repeated highlighting of a “parabolic seismic reflector”, with the primary thesis of the study being that this domed and fast layer acted a bit like a satellite dish, reflecting and focusing the incoming earthquake waves into the Kalibeng clays. The study proposed that this layer enhanced the earthquake shaking so much within the clays that it triggered liquefaction, which in turn caused widespread gas dissolution and a reduction in effective stress that was sufficient to induce fault reactivation and trigger Lusi. The study even claims that the model results are so conclusive as to “exonerate” a drilling accident as a trigger for Lusi (despite the results not actually being compared with, or testing, the drilling trigger in any way).

The prominent nature of this article in such a high-impact journal meant that this study received a lot of attention. However, there was an immediate obvious question – what on earth is that “parabolic seismic reflector”? What in the world is a rock of 6300m/s p-wave velocity (Vp) doing at just ~1000m depth in Pleistocene clays? A velocity of 6300m/s would be anomalously fast for any shale – but is a wildly fast and implausible velocity for shallow and young clays that have not experienced sufficient loading, heat nor time to undergo the compaction and diagenesis needed to get down to what would be close to zero porosity.

Velocity model used to estimate local influence of the Yogyakarta earthquake at the Lusi site in Lupi et al. (2013). Note the focus on the extremely fast "parabolic seismic reflector" at ~1000m depth, which is central to the thesis of their study. Note also the absence of any noticeable reflector at this depth in the 2D seismic section on the left.

So, where has this high velocity layer come from? Well, Lupi et al. (2013) have taken this velocity data straight from a figure in an earlier paper primarily by Lapindo geologists (Istadi et al., 2009), which says that the data is from wireline sonic log data. But, this still does not explain what the layer is. Surely such an anomalous layer would be visible in other data? Well, no, it is not. Other sonic velocity data, such as a vertical seismic profile (VSP) survey and even alternate sonic log plots, show absolutely no indication of this fast layer (including a later 2012 paper by Istadi et al., the author of the original velocity model). What about other data that correlates with velocity? Again, no sign of this layer there either. There is no indication of any anomalous layer at that depth in the density log, resistivity log, neutron porosity log, drill cuttings, nor the corrected D-exponent (effectively a drilling-parameter normalised rate of penetration).

Velocity model from Banjar Panji-1 in Istadi et al., 2009, and which is the source of the velocity data used in Lupi et al. (2013). Note also the calculated porosity plot on the left hand side, which is derived using an equation in which Vp is the only input parameter. It is interesting to note that, despite the velocity data above being acknowledged as wrong, and changed in Istadi et al., 2012, the porosity plot directly derived from the erroneous data has not been modified and was used to justify the revised velocity model in Lupi et al., 2014.

Final corrected petrophysical log data from Banjar Panji-1 and related offset wells within 7km of Lusi. Note the absence of any significant change in density, resistivity or porosity that should be expected if the Lupi et al. (2013) 'parabolic seismic reflector' were a real geological feature.

Most significantly, there is no indication of any such layer on any 2D reflection seismic in the region. One would think that a shallow super-fast layer would show up as an absolutely booming reflector on seismic – but nope, no obvious reflector on the seismic! This is particularly odd. The paper’s premise is that low frequency/long wavelength earthquake waves would reflect off of this layer, and yet even high frequency and much higher resolution reflection seismic (including even seismic section published in the Lupi et al. 2013 study) completely fails to detect this layer?

East-west oriented interpreted 2D reflection seismic line over Banjar Panji-1 and the Lusi location, collected prior to drilling and Lusi initiation. Note the absence of any significant seismic reflectors near or shortly below the Base Pucangan Sands, as would be expected if a huge velocity (and impedance) contrast existed.  

So, there are absolutely no confirmatory indications of this high velocity layer, and we are still left with the question of what in the world this high velocity layer is? Well, the answer is actually pretty simple. The bottom of the high velocity layer is precisely the same depth as the 13-3/8” and final casing point in BJP-1. The high velocity ‘parabolic seismic reflector’ is not a real geological layer, it is just the extremely fast velocity measured by wireline sonic velocity tools when they are inside the steel and cement wellbore casing (Tingay, 2015). Feel free to ‘face palm’ now.

This mistake can happen pretty easily. Wireline log data is collected in stages as the well is drilled. In BJP-1, they drilled the 14.5” hole section from ~650m to ~1100m depth. This open hole section was logged using wireline petrophysical tools before the 13-3/8” steel casing was lowered in place, and cemented, giving the well a steel and cement lining from the surface to 1090m depth. The 12.25” section of BJP-1 was then drilled over several weeks, until they reached a depth of ~2650m, at which point they again stopped and undertook wireline logging and a vertical seismic profile. Now, when they log wells, they typically lower the logging tool to the bottom of the hole, and then take the measurements while pulling the tool up the well. They continue to take measurements until the logging tool has passed beyond the open hole section and into the cased section of the well – and don’t usually stop collecting data until well within the cased hole (in this instance about 100m inside the casing). Later on, the log data from the 12.25” and 14.5” hole sections gets processed and stitched together into a single curve. At this stage, the obviously fast erroneous velocities within the casing from the 12.25” logging run are usually removed and the correct velocities from the 14.5” logging run are used. But, for some reason, Lapindo did not do this. Instead, Istadi et al. (2009) included the velocity data from inside the casing, and even used this to predict porosity and argue for an essentially zero porosity sealing layer at this depth (sealing high pressure fluids in clays that are claimed as the source of water and clay erupting from Lusi). Indeed, it should also be noted that the ‘layer of steel’ is not the only error in the velocity model – careful checking and processing reveals a lot of common acquisition and processing errors in the log data collected in BJP-1 (see Tingay, 2015 for a full description of the issues associated with BJP-1 petrophysical data and a corrected dataset).

Final carefully corrected velocity data from Tingay (2015) compared to the erroneous p-wave (Vp) and shear-wave (Vs) velocity data for BJP-1 used in Istadi et al., 2009; Istadi et al., 2012; Lupi et al., 2013 and Lupi et al., 2014. Note the numerous regions of Vp affected by aquistition and processing artefacts, and especially the extremely high casing velocities that represent the 'layer of steel' used in Lupi et al., 2013. Not also the dramatic difference in Vs models proposed by Lupi et al., 2014 compared to those predicted by four different petroleum industry methods (panel d).

At this stage, you might still be thinking “WTF? No way! You’re kidding right? How has this not been picked up somewhere? Come on – this is Nature Geoscience we are talking about here!” Yes, indeed - this error has gone through internal review by Istadi et al., then gone through peer-review in Marine and Petroleum Geology, sat there unnoticed for years (it is not a key part of Istadi et al. (2009), but I missed it then too) then gone through internal scrutiny by Lupi et al. and then gone through editorial scrutiny and peer-review in Nature Geoscience (not to mention a special feature commentary on the paper by Professor Paul Davis in the same issue) – and yet no one picked this up? From reports, it was queried, especially by the Nature Geoscience reviewers, but the argument that the erroneous velocity model ‘comes from a peer-reviewed source’ appears to have trumped the standard scientific practice of ‘sense-checking’ any data used in analysis.

Nature Geoscience did, eventually, acknowledge the error and make Lupi et al. write a corrigendum, though it is interesting to note that this does not actually state what the error(s) were (Lupi et al., 2014). Instead, the corrigendum states “we were subsequently alerted to artefacts in that velocity profile, so below we present revised simulation results, based on additional data”. In the corrigendum, the original p-wave models are thrown out in favour of a new shear-wave velocity (Vs) model, which yields essentially the same results as the original study. This is because the original high Vp layer is simply replaced by a high Vs layer, conveniently located precisely at the top of measured shear-wave velocity log data. The argument for this new layer is the claim that this depth represents an extremely sharp change from normally pressured and highly compacted clays to highly overpressured high porosity clays.

In summary, it is my opinion (shared by many others) that the modelling results undertaken by Lupi et al. (in both 2013 and 2014) are likely fatally flawed by significant errors in the velocity models central to their thesis. If there is one key lesson I hope scientists reading this will learn is to always remember that ‘being peer-reviewed does not make data, or a study, correct’, and that we must always make efforts to understand the data we use in our analysis (and how it is collected). No matter the source, pass your data and arguments through a ‘common sense test’, so that potentially highly embarrassing errors can hopefully be avoided.

The supporting evidence for this big Vs contrast is first a porosity versus depth plot from the original Istadi et al. (2009) paper, which was calculated directly from the exact same velocity model already highlighted as erroneous (in other words, while Lupi et al. acknowledge the Vp data is wrong, they still use other data calculated solely from this erroneous data!). Second, Lupi et al. (2014) use a plot of Vp and Vs versus effective stress data from Gulf of Mexico sands (Lee, 2010), and use this Vp and Vs data from BJP-1 to argue that a big change in Vp/Vs ratio (and, thus Vs, as Vp doesn’t change) must exist (claimed to be the result of a sharp 9 MPa increase in pore pressure) in shales at ~900m depth.

Neither of these arguments for a sudden change in Vs seem the most suitable pieces of evidence. In addition, the revised ‘corrected’ model seems to ignore the well-established correlation between Vp and Vs, in which these always show very similar variations (aside from in gas saturated sediments and almost zero effective stress conditions). Instead, the revised model argues for a doubling in Vs (a 370 m/s change), when Vp only changes by less than 10% (150 m/s change). Pore pressure data also demonstrates that there are no rapid changes in pore pressure at this depth, nor significant changes in vertical effective stress as claimed (Sawolo et al., 2009; Tingay, 2015), with VES gradually changing by ~0.6 MPa through the overburden, and not showing anything like the sharp 9.0 MPa change argued by Lupi et al. Nor are there any geological reasons for a big Vs contrast, with the only significant changes being two thin sands within the base of the Pucangan Fm that show an approximately 40 m/s Vp contrast. Indeed, the method used by Lupi et al. (2014) to create a Vs model is very odd, and not something ever seen before. Estimation of Vs is common practice in the petroleum industry, and a test of four different methods for estimating shallow Vs from BJP-1 data yields four quite consistent predictions, all of which suggest a maximum Vs contrast of ~35m/s, which is an entire order of magnitude lower than claimed in Lupi et al.’s corrigendum.

A paper in GRL was published last year that aimed to test the effect of different velocity models on the numerical modelling conducted by Lupi et al. (Rudolph et al., 2015). This paper attempted to reproduce the modelling undertaken by Lupi et al, but comparing the numerical model results when using the Lupi et al. velocity model (actually, a digitised version of their velocity model, as they would not provide their velocity data) and the updated velocity model I published in 2015 (Tingay, 2015). This study produced two interesting results. First, as expected, the Lupi et al. model, with its order of magnitude larger velocity contrast, predicts a far larger (1.5-2.0 times greater) influence of the Yogyakarta earthquake than using the Tingay (2015) model. Second, both models actually do propose that local geology enhances the effect of earthquake waves under Lusi – but not due to any shallow high velocity or domed zone, but rather due to the sharp velocity change from the Kalibeng clays to the deeper volcanics/volcaniclastics. Upward travelling earthquake waves are modelled to slow down and increase in amplitude as they pass from the very fast volcanics to the slow Kalibeng clays, and this is a far more significant increase in earthquake effects than any shallower layers have. Despite this geological amplification, Rudolph et al. (2015) model results still indicate that the effect of the Yogyakarta earthquake was too small to trigger liquefaction (and evidence in the next section further highlights this), and again, still would not explain why the Yogyakarta earthquake, and not any of the 13 bigger earlier quakes, triggered the Lusi eruption.

In conclusion, arguments claiming an enhancement of earthquake effects due to Lusi subsurface structure and lithology are primarily related to major errors in the velocity models used. Modelling by Rudolph et al. (2015) does suggest that lithology causes some enhancement of earthquake effects, but these are still seen as insufficient to induce liquefaction.

Argument 8: Earthquake waves were large enough to induce clay liquefaction

The final argument used to propose that an earthquake triggered the Lusi mud volcano is that, regardless of how it happened, the shaking from the Yogyakarta earthquake was big enough to induce liquefaction at the Lusi location (Mazzini et al., 2009; Lupi et al., 2013). Given the above discussions, there does not seem to be much evidence to support this claim. Indeed, there is significant evidence against it. For example, empirical compilations of earthquake-induced liquefaction suggest that the Yogyakarta earthquake did not produce sufficient seismic energy density to trigger liquefaction (Delle Donne et al., 2010).

More recently, analysis of the gas data from BJP-1 provided further evidence that liquefaction did not occur following the Yogyakarta earthquake (Tingay et al., 2015). We tested the hypothesis and arguments made by Lupi et al. (2013), in which it was proposed that liquefaction would be associated with a huge exsolution of gas, primarily CO₂ and methane, within the Kalibeng clays. The BJP-1 well was optimally placed to test this, as almost the entire Kalibeng clays were open when the earthquake occurred, and so any large dissolution of gas would have caused an observable increase in gases to flow into the BJP-1 wellbore, which would be detectable on the mud gas equipment. Yet, the gas records from BJP-1 are conclusive – there was absolutely no increase in any measured gas (including CO₂ and CH₄) in the approximately 24-hour period between the Yogyakarta earthquake and the drilling kick on the 28^th of May. Furthermore, it is interesting to note that all the downhole effects that are predicted to occur during mobilisation of the Kalibeng clays (e.g. high gas readings, clay cavings, stuck pipe) were observed in BJP-1, but only when the drilling kick occurred, and not at any time between the earthquake and the kick.

In conclusion, a big gas increase after the earthquake was proposed as a key part of the earthquake-triggering model, and would potentially indicate liquefaction of the Kalibeng clays. However, measurements from BJP-1 conclusively show that no gas increase occurred, and thus provides strong evidence that no liquefaction at the Lusi location was triggered by the Yogyakarta earthquake.

Summary

A total of eight different arguments have been proposed as being supporting evidence for the hypothesis that Lusi was triggered by the Yogyakarta earthquake. Several of these arguments appear, at first glance, to have some merit. However, detailed analysis suggests that all arguments are either circumstantial, based on erroneous data/assumptions, or are equally predicted by the drilling-triggering model for Lusi initiation. Furthermore, there is strong evidence against the earthquake-trigger hypothesis, such as the lack of any observable or reliable subsurface response to the earthquake, the low seismic energy density of the quake, and the lack of any explanation for why larger (and, in two instances, much larger) earthquakes failed to trigger the mud volcano. As such, I personally consider the earthquake triggering hypothesis to be extremely unlikely, and can almost be considered to be completely debunked. But, this does not mean that the disaster was triggered by drilling. Logically, each hypothesis needs to be tested separately – one should not make the mistake of suggesting that essentially dispelling one argument automatically validates a different one! Hence, Part 3 of this blog (next week), will make a similar detailed analysis of the arguments for and against the drilling-trigger hypothesis.

References

Bonini, M., M. L. Rudolph, and M. Manga, 2016, Long- and short-term triggering and modulation of mud volcano eruptions by earthquakes: Tectonophysics, 672-673, 190-211.
Davies, R. J., R. E. Swarbrick, R. J.  Evans, and M. Huuse, 2007, Birth of a mud volcano: East Java, 29 May 2006: GSA Today, 17, 4–9.
Davies, R., M. Brumm, M. Manga, R. Rubiandini, R. Swarbrick, and M. Tingay, 2008, The east Java mud volcano (2006 to present): an earthquake or drilling trigger?: Earth and Planetary Science Letters, 272, 627-638.
Delle Donne, D., A. J. L. Harris, M. Ripepe and R. Wright, 2010, Earthquake-induced thermal anomalies at active volcanoes: Geology 38, 771–774. http://dx.doi.org/10.1130/G30984.1.
Harris, A .J. L. and M. Ripepe, 2007, Regional earthquake as a trigger for enhanced volcanic activity: Evidence from MODIS thermal data: Geophysical Research Letters, 34, L02304, doi:10.1029/2006GL028251.
Istadi, B., G. Pramono, and P. Sumintadireja, 2009, Modeling study of growth and potential geohazard for LUSI mud volcano: East Java, Indonesia: Marine and Petroleum Geology, 26, 1724-1739.
Istadi, B. P., H. T. Wibowo, E.  Sunardi, S. Hadi, and N. Sawolo, 2012, Mud Volcano and Its Evolution, in I. A. Dar, ed., Earth Sciences: InTech, ISBN: 978-953-307-861-8, DOI: 10.5772/24944. Available from: http://www.intechopen.com/books/earth-sciences/mud-volcano-and-its-evolution
Lupi, N., E. H. Saenger, F. Fuchs, and S. A. Miller, 2013, Lusi mud eruption triggered by geometric focusing of seismic waves: Nature Geoscience, 6, 642-646.
Lupi, N., E. H. Saenger, F. Fuchs, and S. A. Miller, 2014. Corrigendum to Lusi mud eruption triggered by geometric focusing of seismic waves: Nature Geoscience, 7, 687-688.
Manga, M., 2007, Did an earthquake trigger the May 2006 eruption of the Lusi mud volcano?: EOS, 88, 201.

Mazzini, A., H. Svensen, G. Akhmanov, G. Aloisi, S. Planke, A. Malthe-Sørenssen, and B. Istadi, 2007, Triggering and dynamic evolution of Lusi mud volcano, Indonesia: Earth and Planetary Science Letters, 261, 375–388.

Mazzini, A., A. Nermoen, M. Krotkiewski, Y. Podladchikov, S. Planke, and H. Svensen, 2009, Strike-slip faulting as a trigger mechanism for overpressure release through piercement structures. Implications for the Lusi mud volcano, Indonesia: Marine and Petroleum Geology, 26, 1751-1765.

Mazzini, A., G. Etiope, and H. Svensen, 2012, A new hydrothermal scenario for the 2006 Lusi eruption, Indonesia. Insights from gas geochemistry: Earth and Planetary Science Letters, 317, 305-318.

Rudolph, M.L. and M. Manga, 2012, Frequency dependence of mud volcano response to earthquakes: Geophys. Res. Lett., 39, L14303. http://dx.doi.org/10.1029/2012GL052383.
Rudolph, M.L., M. Manga, M. Tingay and R. Davies, 2015, Influence of seismicity on the lusi mud eruption: Geophys. Res. Lett., 42, 7436–7443. http://dx.doi.org/10.1002/2015GL065310.
Sawolo, N., E. Sutriono, B. P. Istadi, and A. B. Darmoyo, 2009, The LUSI mud volcano triggering controversy: Was it caused by drilling?: Marine and Petroleum Geology, 26, 1766-1784.
Tingay, M., O. Heidbach, R. Davies, and R. E. Swarbrick, 2008, Triggering of the Lusi mud eruption: earthquake versus drilling initiation: Geology, 36, 639-642.
Tingay, M., 2015, Initial pore pressures under the Lusi mud volcano, Indonesia: Interpretation, 3(1), SE33–SE49, doi:10.1190/INT-2014-0092.1.
Tingay, M., M. L. Rudolph, M. Manga, R. J. Davies, and C.-Y. Wang, 2015, Initiation of the Lusi mudflow disaster: Nature Geoscience, doi:10.1038/ngeo2472.
Tingay, M., 2016, What caused the Lusi Mudflow Disaster in Indonesia? AAPG Search and Discovery Article #41791.