Bonn was the second stop for our tour of Europe. From Delft, we hopped onto a train first stopping in the Hague where we took a couple hours to go through the Escher Museum and have our minds messed with a bit!
From the Hague, we then went through Utrecht, caught a high-speed train to Köln and finally a regional train to Bonn. We had booked tickets at the station, and it was only once we got onto the train that we realized the ticket agent had misunderstood us and booked us on a set of trains 2 hours earlier than we planned (so we were on the wrong trains!). The ticket agent came by, gave us a stern look, but didn’t kick us off! On the second train, we ended up sitting next to a geologist and his wife, both of whom were from India; the age-old problem of getting geologists and geophysicists to communicate came up again.
This brought us to Bonn on Thursday night. On Friday, we spent the day with Andreas Kemna and Florian Wagner chatting about induced polarization, open source resources in geophysics and checking on the setup of the room for the DISC.
We spent the morning discussing current research in Bonn, including applications of Induced Polarization (IP) for:
- groundwater: the complex part of electrical conductivity can be used to obtain information about hydraulic permeability,
- landfills: both for delineating the extend of a landfills as well as for monitoring changes with remediation (there is a lot we still have to learn here! Under some circumstances, the bacteria that break down the waste produce sulphides as a byproduct, thus potentially increasing the chargeability of the landfill. While in other cases, as in the example we present in the DISC, the waste is the main source of chargeable material, and as it breaks down, the chargeability of the landfill is reduced. The controlling factors in each case is still not something we have our heads wrapped around),
- agriculture: the complexity of the root system of a crop is important for understanding its health. Typically, direct sampling, which requires that the roots be dug up and counted, is applied. However, the IP response can provide an information on the complexity of the root system. Maximilian Weigand and Andreas Kemna recently published a paper (which is open-access!) on the topic: Multi-frequency electrical impedance tomography as a non-invasive tool to characterize and monitor crop root systems.
We went for lunch and a coffee with Florian, talked about open source software and resources for geophysics education. Florian is teaching a course on environmental geophysics, there are a lot of similarities to the environmental, geotechnical and exploration geophysics course we teach at UBC. A similar course is being offered at a number of Universities worldwide and each instructor is currently developing all of their own resources; there isn’t a widely used textbook for such a course. A number of years ago, Doug Oldenburg started the GPG (Geophysics for Practicing Geoscientists), a web-textbook resource targeted at introductory geophysics. We recently refactored the site with the hope that it will continue to grown as an open-source, collaboratively built, resource for introductory geophysics. There are some opportunities for collaborating here!
Later in the afternoon, Florian gave us a presentation on pyGIMLi, an open source software project for Geophysical Inversion and Modelling championed by Carsten Rücker, Thomas Günther and Florian Wagner. The pyGIMLi and SimPEG projects share similar motivations and a drive to open up forward and inverse problems in geophysics: to make the science more accessible, reproducible, and transparent. We chatted about opportunities for collaboration, including on data i/o, mappings and petrophysical relationships as well as analytic solutions. As we continued bringing the DISC around Europe, pyGIMLi is clearly making an impact for researchers and scientists who need to perform simulations and inversions, particularly in the groundwater community. Related to his continuing work in pyGIMLi and examining process-based modelling and inversion, Florian was awarded the Water Monitoring Prize at the Water Research Horizon Conference in Hamburg— Congratulations Florian!!
Day 1: DISC Course
Andreas and Florian are among the most enthusiastic local hosts we have worked with to date! They worked to gather an audience from several universities in the region: Aachen, Bonn, Jülich, Liege. There is a strong EM group in Köln, but unfortunately the group was out in the field, so the timing did not work for them to attend.
The course started at 9am on Monday, and we began with introductions from the participants. The audience was primarily academic, consisting of mainly professors, graduate students, and post-docs in the geosciences. Of particular interest were applications of EM for: hydrogeophysics, landfills, agricultural applications, permafrost, and monitoring alpine environments.
From the participants
“Very valuable. Diverse. It comes back to the basics without removing more complex aspects.” — Thomas Kremer, Post-doc at the University of Liege
“It was really insightful. It provided a better understanding in EM and drives me to seriously consider them in my current project. Personally, [I am] a big fan of open source projects like SimPEG [and] hope to contribute to them in the future.” — Jorge Lopez, PhD student
“It was a good, interesting, valuable lecture, with well chosen examples. [It] was also good to meed other people who are using EM methods. I really like that you are sharing your knowledge and experience. Thank you!”
A few of the quotes also spoke to the challenge we face with the volume of material included in the course (we are trying to cover all of EM geophysics in one day!)
- “It was too long. Nobody can concentrate that long… I like the motivation of Doug Oldenburg.”
- “I had a good impression — The pace was a bit quick for me, but it was an interesting overview”
- “The overview about EM methods was really well done, although it might be difficult to assimilate all at once [because] of the huge amount of information received.”
All of the presentation material from the day is available at https://disc2017.geosci.xyz/delft.
Following the course, Andreas and Florian had arranged a reception with Kölsch beer and pretzels! Kölsch beer is served in small glasses (0.2 oz). The glasses are small so the beer sparkles, and a perfect pour has two fingers of foam — it took a bit of practice!
Day 2: DISC Lab
We started off DISC lab day with additional material of interest that we didn’t have a chance to cover in the first day. This included working through the landfill case history from Denmark (in the IP slides), as well as discussing advances in Induced Polarization and using EM data to obtain a conductivity model for IP inversions. Following these, we turned the floor over to the DISC participants to discuss their current work.
Andreas Kemna (slides) started off the presentations discussing the use of electrical methods for characterizing permafrost sites. The aim is to be able to distinguish between frozen and unfrozen states of a rock containing water in its pores. Electrical Resistivity Tomography (ERT) is applied by a number of groups for examining permafrost sites, Andreas showed us one example from France (on the right of the image above), where resistivity is used as an indicator for the thermal state of a rock. As the temperature drops, liquid water, along with all of the charge carriers it contains, freezes, thus increasing the resistivity. Characterizing the thermal state based only on resistivity can be challenging: factors such as the porosity of a rock also change resistivity, so a single resistivity value is not necessarily enough to tell us if we have a low-porosity rock or if we have a high-porosity rock but the water in its pores is frozen; this non-uniqueness was a motivating factor for examining the potential application of IP for distinguishing between frozen and thawed states. With the IP method, we are examining the frequency dependence of the conductivity (or alternatively, permittivity) of a material. The conductivity is described as a complex number, having real and imaginary components or, equivalently, an amplitude and a phase. The polarization of ice occurs at much lower frequencies than the polarization of water, and therefore, we might be able to use IP to distinguish between the two.
To examine both the resistivity and polarization responses of with temperature, they conducted lab measurements in which the temperature of a sandstone sample was varied and the resistivity and polarization responses were measured. With resistivity, there is an interesting hysteresis effect with freezing vs. thawing (see slide 7); as the sample is cooled, the resistivity steadily increases, beyond 0°C, until there is a drastic jump in resistivity as the sample is cooled past -2°C. When going the other direction and warming the sample, the change in resistivity happens at 0°C. . This has been documented and shown to be due to a super-cooling effect — as water is cooled it can remain in liquid form even when it is below the freezing temperature; nucleation is required for water to turn to ice, whereas when thawing ice, once we are above the freezing temperature, the ice will melt. The polarization response, which can be observed as a phase shift in the conductivity, shows a systematic shift as temperature decreases: the low frequency response decreases (there is less contribution from the water) while the high frequency response increases (due to the response of ice). Together resistivity and spectral induced polarization provide information on the thermal state, and there may even be potential for quantifying the content of ice.
Maximilian Weigand (slides), a Post-doc at the University of Bonn, is examining the use of spectral induced polarization and electrical resistivity tomography for characterizing crop root systems (we briefly mentioned this earlier in this blog). He showed an example where roots were put in tap-water and the extent of the root system coincided with the observed polarization response. He further showed that polarization can be an indicator of functional changes in a root system. If roots are left in tap-water which is deprived of nutrients for several days, the root system will start to decay as the plant dies. Over 5–6 days of a plant being in tap-water only, a decrease in chargeability was observed. In another experiment, they decapitated a plant, meaning that nutrients no longer need to be transported up the plant. As a result, the roots are doing less work and can function with less nutrients for longer. This was again reflected in the IP data, where the polarization response was retained longer than the intact plant under the same nutrient-deprived circumstances. Even subtle changes corresponding to the diurnal cycle could be observed with the IP response. At night, when physiological activity within the plant decreases, and nutrients are stored, the polarization increased. These experiments indicate that chargeability sensitive to physiological health of crop root system and IP could be a valuable monitoring tool. Taking the next steps and applying this methodology in soils presents the additional challenge that the soil, and water it contains will also have an IP response in the measured data. Using the IP for monitoring the crop health will require that these signals be teased apart.
Upon request from myself and Florian, Maxi made a last minute discussion-opener to his presentation regarding the development of a python library for working with electrical data from various systems. Importing, exporting and validating data is a tedious task that each group working with electrical data needs to do. Right now, everyone is writing their own libraries or scripts to do this. It is not interesting work, but needs to be done, and if we could develop open-source strategies to combine efforts on this, we would reduce a lot of duplication of work as well as promote more transparency in our workflows.
Jan van der Kruk (slides), the head of the Hydrogeophysical Imaging and Characterization Group at Jülich University, gave a presentation on Full Waveform Inversion (FWI) of Ground Penetrating Radar (GPR) data. GPR is an electromagnetic experiment conducted at high frequencies, where wave propagation is the dominant effect. Rather than just looking at first arrivals in the data, a full waveform inversion aims to explain all features in the observed GPR signal. FWI has been examined and in use for seismic data for the past 20 years; its application to GPR is more recent (within the last ~9 years). One of the challenges that need to be contended with is that the source wavelet is unknown and needs to be estimated from the data. Typically, the parameter that is estimated in an FWI is velocity; one of the developments that Jan and others in the group are working on is inverting for dielectric permittivity and electrical conductivity, similar to how both attenuation and velocities may be inverted for in a seismic full waveform inversion. Jan showed a field example that compared a ray-based inversion approach and a full waveform inversion for a tracer experiment conducted in an aquifer. A plume splitting was observed with monitor wells, and the GPR data were inverted with the aim of understanding why. The ray-based inversion did not have sufficient resolution to delineate distinct geologic units, while the full waveform inversion showed that a layer, which pinched out near the injection well was responsible for the splitting. In terms of future work, quantifying uncertainty, which requires a significant amount of computational power, is a topic of interest, as is simulating the effect of the borehole in cross-hole GPR experiments.
Christian von Hebel (slides), who is also at Jülich University, gave a presentation on fixed-boom electromagnetic induction systems. Fixed boom systems are often used in geotechnical or agricultural applications. They can be used to cover large areas of ground much more quickly than a DC resistivity experiment as they do not need to be in direct contact with the ground. Christian is interested in the calibration of and inversion of data from such systems. To calibrate, a short transect is chosen and both EM data and DC resistivity are measured along that transect. The DC resistivity data are inverted, providing a resistivity model of the subsurface. Using this resistivity model, a forward simulation using the EM instrument specifications is performed. The predicted response from the forward modelling is then compared to the measured data through a linear regression; the result of this is a scaling factor and a shift — these are the calibration values. Christian showed a field example where paleochannels were delineated for an agricultural application. The location of paleochannels influences groundwater flow and the delivery of nutrients. The EM data was compared with leaf-area-index measurements (larger leaf area indicates healthier plants) and it was found that the paleochannels mapped with the EM data correlated with regions having a larger leaf area index.
Frederic Nguyen (slides), the head of the GEO³ research division, presented an overview of their current research. GEO³ consists of three groups: (1) hydrogeology and environmental geology, (2) applied geophysics and (3) geomechanical and engineering geology. He focussed in on applied geophysics; the three topics of current interest are: inverse problems, data assimilation, environmental geophysics. Within the context of inverse problems, Frederic talked about time-lapse inverse problems. In a time lapse problem, we are aiming to image (often subtle) changes through time. A typical inversion strategy may smooth over the changes making them difficult to detect, however the changes that one is looking for in a time-lapse inversion are typically compact targets. In order to image such targets, the regularization functional in the inversion can be tailored, perhaps by using a different norm, to image compact targets. Frederic showed two examples of time-lapse inversions in action one synthetic and one field example.
The second topic he discussed is data assimilation; a process-oriented, probabilistic approach for making predictions about geology. Rather than using an inversion to estimate a parameter and then make a prediction, the aim of their approach is to bypass the inversion and go directly from data to prediction. In this approach, a bunch of geologic scenarios are built up (from some probabilistic distribution) and the geophysical data are used to falsify some of the models — a geophysical simulation is run for each of the geologic models and the models for which the simulated data do not sufficiently match the observed data are disregarded.
The final topic Frederic discussed was an application in environmental geophysics: using induced polarization to characterize aggregations of bacteria. Different aggregations of bacteria produce differing spectral IP responses. Understanding these signals has implications for bioremediation, for example in the biodegradation of hydrocarbons.
Thomas Kremer (slides) presented on the application of surface nuclear magnetic resonance (NMR) for groundwater. NMR has the benefit of being sensitive to the water content in the subsurface and can distinguish between free and bound water. Hydrogen atoms (eg in water: H2O), are like small magnets that will essentially line up with (actually, they precess around) any eternal field, like the earth’s magnetic field. In an NMR experiment, the goal is to flip the magnetic moment of the hydrogen atoms and observe as they return back to the initial state. In order to “kick” the magnetic moment of the hydrogen atoms off of the axis of the Earth’s magnetic field, a transmitter loop is laid out on the ground and a current passed through it to generate a magnetic field (in a different orientation from the Earth’s field). The transmitter is then shut-off and magnetic fields observed as the hydrogen atoms come back to steady state. The amplitude of the signal is related to the water content and the relaxation time depends on the subsurface hydraulic properties. Often, an NMR experiment will be conducted with multiple different transmitter moments to sample different depths. The signals measured are generally quite subtle, so signal processing, including addressing noise sources such as power lines, spherics (eg. from lightning strikes), and instrument noise, is an important aspect. Thomas presented examples of recent work on using adaptive noise cancellation to handle noise sources such as power lines.
Stefanie Keichel (slides) presented a short overview of her Masters work on examining the impact of snow cover on electrical resistivity tomography (ERT) measurements. When monitoring permafrost, we often have to contend with snow that covers the electrodes. If conducting an ERT experiment where snow covers the electrodes: does the current go into both the ground and the snow? Stefanie conducted synthetic modelling and inversion studies to examine the impact of a snow layer on our ability to image the subsurface using ERT. Her model consisted of a conductive target in a half-space covered by a snow layer (one model used a thin snow layer of ~1.5m , another used a thick layer of ~2.5 m). In her first example, she performed a forward simulation over the try model and inverted the synthetic data assuming that there is no snow-cover. In this scenario, the inversion introduced a near-surface, resistive artifact. In the second example, she included the snow layer in the inversion and allowed the inversion to update the resistivity values in the snow-cover. For a thin snow layer, the inversion results improved. For a thick snow layer, there is an ambiguity introduced — due to the geometry of the survey, it is unclear in the inversion if the conductive material is above or below the line of electrodes on the surface. Thus, the inversion tends to mirror the anomaly, that is, put conductive material both below the surface and within the snow layer. In future work, Stefanie is interested in updating the inversion algorithm to input a-priori information of where resistive snow is expected vs. where more conductive materials may exist subsurface.
A few adventures
In Bonn, our free days were the Saturday and Sunday prior to the DISC. We spent a bit of time each morning getting a bit of work done and prepping for the presentations. On Saturday afternoon, we went to the Deutsches Museum to see the exhibition for 100 years of Einstein. There were a number of interactive exhibits, but all were in German, so we took the chance to practice some yoga photos. From there, we took the train back to town and walked through the town square in Bonn, where there is a statue of Beethoven (he was born in Bonn!).
On Sunday, we started off the day at the Botanic Garden near the University. There was a live Jazz band playing in one of the halls, it was a nice way to start the day. From there, we took the train to Köln.
You step out of the main train station in Köln and the cathedral towers over you. It is one of Germany’s most visited landmarks. We wandered through the cathedral and took some photos. We were a bit too tight on time to climb the stairs up the spires as we wanted to be on time for a Bikram yoga class! The Bikram class we went to was instructed half in German and half in English (for us foreigners!).
We ate some things
We try to eat some of the food iconic to each country at each location we present. When in Germany…. bratwurst, schnitzel, potatoes and beer are on the menu (along with some delicious peas!). We didn’t need to eat the next day.
Restaurant recommendations from yoga studios have yet to fail us! Following our Bikram class in Köln, the instructor recommended a Vietnamese restaurant next door to the studio. We had a curry soup, papaya salad and a rice noodle bowl — the perfect post-yoga meal!
We are very grateful for all of the effort and enthusiasm Andreas Kemna and Florian Wagner poured into making the DISC in Bonn a success!