The earth is complex. What happens on it and in it is difficult to explain and can only be partially predicted. Yet, this would be important, especially in light of the increasing number of natural disasters that occur worldwide due to e.g. climate change. Devastating earthquakes, severe droughts, or massive flooding resulting from thawing - earth and environmental scientists want to better understand all of these phenomena. Important data, however, are often missing, either because they cannot be collected or their interplay can only be inadequately reconstructed. This is why more and more geoscientists are resorting to computer simulations. They can be used to check data, to map developments, and to understand, reconstruct, and verify relationships. It has become clear that scientific studies have to consider existing uncertainties in order to track down “uncertain reality” – from imprecise and missing data via indirectly determinable parameters, like soil properties in an area under review, to possible inaccuracies in mathematical equations of a computer model. This is exactly the goal of the Helmholtz Research School GeoSim, where young researchers bridge mathematics with earth and environmental sciences.
Models help to enable earthquake-proof construction
One such person is Sanjay Bora. The young Indian researcher has developed a model that can reconstruct and simulate ground motion for future earthquakes. Based on the results, seismic safety standards for buildings can be established, which can help in preventing the catastrophic destruction of critical or sensitive public facilities like dams and power stations. Many regions have not enough observed data that can be used to calibrate such models. “We simply often lack the respective data,” Bora says. “Here is where my model can help. In simple words, we transfer a model from one region for which we have sufficient data to another with incomplete data.” This is possible because it is a so-called filter (transfer function) based model. While the basic pattern remains the same, with this model one can easily adjust different segments of the model to different region specific seismological environments – “like the attenuation of seismic wave amplitudes with distance in California can be adjusted to that in the Lower Rhine Basin,” Bora explains. The researchers have been extremely successful: Their work has already received awards in international scientific conferences and “arrived” in practical applications. The model framework has already been used and validated in a recent seismic hazard assessment project carried out in a French earthquake region. For Bora, this is both, a jumping-off point and an incentive to continue his research. “Others use my model and give me feedback on its functionality, which in turn helps me to refine it. It has been mainly used in Europe but will be used globally in the future, for example in the US and Japan.”
Annabel Händel also studies ground motion. Like Bora, her studies are supervised by Frank Scherbaum, Professor of Geophysics and Seismology at the Institute of Earth and Environmental Sciences of the University of Potsdam. Händel addresses attenuation phenomena close to the earth surface, i.e. how much seismic energy is absorbed by the uppermost layers and how this happens. Attenuation parameters help to determine the amount of high-frequency energy reaching the surface during an earthquake. The parameter can provide information on the requisite concrete wall thickness of a dam or safety standards for piping systems in chemical plants. “The attenuation parameter is very important when you want to adapt a more general ground motion model to a certain region,” Händel explains. Unfortunately, we often lack the data necessary to determine a region’s attenuation. Händel tries to extract the desired parameter from ground noise data recorded by any seismometer together with seismic signals as a kind of “data garbage”. Ground noise also consists of seismic waves that travel through the ground and, thus, necessarily contain attenuation information – a suitable collection method, though, is not yet available. Händel is currently developing such a method and is working on the recordings from several sensors in Greece, a region whose attenuation properties are well known. Her aim is to find a method that can be easily used worldwide to unproblematically apply information about high-frequency attenuation when adjusting ground motion models.
Simulations bridge data gaps
For Scherbaum, this is just another chapter in the success story of the graduate research school. Yet another is the “arrival” of simulation as a research instrument. “Computer simulation has established itself in applied natural sciences alongside experimentation and theory. The research school has taken this into account, training a new generation of researchers who are experts at mathematics as well as environmental science and geoscience.”
It has become apparent that simulations can specifically help with very difficult and complex problems. “Data are often imprecise or inaccurate, so we need different approaches,” says Scherbaum. Earthquakes, for example, still cannot be predicted. “But you can approach the problem with the help of approximating models – by determining parameters or ground motion – and ultimately better understand the underlying patterns. You can then improve, for example, building regulations in affected areas.”
A few years ago researchers at Freie Universität Berlin, the GFZ German Research Center for Geosciences, and the University of Potsdam joined forces to bring mathematics and geosciences closer. The result is a graduate research school, where a geoscientist and a mathematician supervise each PhD student. One of the supervisors is the Potsdam Professor of Hydrology and Climatology Axel Bronstert. “Exploratory simulation helps us answer questions regarding our understanding of the system and simulating processes lacking data due to geographic inaccessibility or because it is simply too expensive.” Computer models cannot replace fieldwork, though. “Those who believe you can simulate everything are wrong,” says Bronstert. “The models actually reflect the current state of knowledge. We are hoping to collect the best possible empirical data for the individual processes and combine them mathematically in the model. Ideally, the simulation depicts the question – and is able to calculate a probable scenario based on this. The better the empirical data, the better the model.”
Large-scale hydrological cycles
The PhD students supervised by Bronstert also work on hydrological topics. While Klaus Vormoor studies the impact of climate change on flooding resulting from thawing in Norway, Filipina Catherine Abon examines if and how so-called flash floods in the tropics can be predicted with the help of rain radar data. Tobias Pilz focuses on regional water cycles, a highly complex topic. Water cycles are difficult to measure empirically because the factors determining how water seeps, for example, may differ significantly only a few meters apart. Regional hydrological models are accordingly difficult, especially for large territories – like in Tobias Pilz’s PhD project. He examines the Jaguaribe region in northeastern Brazil, a region twice the size of Brandenburg. Bronstert has been conducting research there for decades, collaborating with Brazilian partners. The local climate has long dry periods and a short, often intensive rainy season that may also fail, sometimes even for years at a time. Thousands of water reservoirs have thus been built in the region. Some of them are very large, like those for the water supply of Fortaleza, a city with millions of inhabitants, but the majority of these reservoirs are rather small. In “wet” years there is extensive flooding, with often devastating consequences due to uncontrolled, informal housing developments, also in flood areas. A flood forecast system is not available. “We want to help change this,” Pilz says. “We are, therefore, closely collaborating with the local hydro-metrological institute.” The hydrological model that Pilz is developing will allow for the forecasting of how and where water is allocated to a river, particularly when there is a dangerous abundance of it during the rainy season. The problem is that the processes cannot be determined to such a level of precision that they are able to capture the complete hydrological cycle and certainly not for the entire area. This is where mathematical modelling comes in. “Infiltration, retention, and runoff of water can be calculated with the help of equations,” Bronstert explains. “In the model, we combine them with additional information like soil properties, ground water, and precipitation.”
Not ignoring but considering uncertainty
Simplifying complex processes through mathematical simulations is advantageous for two reasons: “Models are not as precise as measured values and never exactly reflect reality but only in this way are we able to approach it in its entirety and describe it, because we are unable to exactly determine the processes. It would be much too complicated.” On the other hand, models can consider and comprise existing gaps, errors, and uncertainty. “You can, of course, try to minimize the uncertainty with better data,” Bronstert says, “but you can also ‘recognize’ it, integrate it into the model, and assess it statistically. This has been done in weather forecasting for a long time.”
This is exactly what Pilz does. “We feed all the data on hydrological cycles in the Jaguaribe region into the model,” he says. “For some parameters we have no empirical data, for example how much water is actually in the reservoirs.” The simulation calculates how much water flows in, but only to a certain degree of accuracy. “We try to close these gaps with other data like satellite images. This still involves uncertainty – and my model intends to consider, compare, and weigh it.”
The more uncertainty he is able to take into account, the more he approximates reality, even if it is only a probable one. Models can be “updated” as soon as more precise data are available for individual parameters. For Pilz and many other PhD students at GeoSim, “research” takes place in front of the computer screen. This is not a problem for Pilz. “It is a good mix of theory and application.” There is no need to jettison “fieldwork”. When the project started, he and Axel Bronstert visited Brazil to get acquainted with his research field first-hand. He is sure that this was not his last trip to Brazil and is now learning Portuguese in his free time. “English doesn’t get you very far there.”
Finding a common language was also a specific challenge at GeoSim, as Bronstert says with a smile. “All supervisors are enthusiastic about the GeoSim project, because the methodological combination of mathematics, earth science, and environmental science has resulted in new research approaches. Communication between the disciplines was not always easy at first. It took six months until the one knew what the other thought or what the problem was.” They have since come to understand each other.
Scherbaum sees GeoSim as “an absolute success. I wish this model would be transferred to the academic curriculum. We should start training the next generation of researchers who are working at the interface of mathematics and geosciences before the PhD phase.”
The “Helmholtz Research School for Explorative Simulation in Earth-Sciences”, or GeoSim, is funded by the Helmholtz Association as well as grants from the participating institutions - GFZ German Research Center for Geosciences, Freie Universität Berlin, and the University of Potsdam. Earth and environmental scientists as well as mathematicians from these institutions have joined forces to provide training and conduct research in the field of exploratory earth science simulation. Currently 39 PhD students from 10 countries are researching there.
The Researchers
Prof. Dr. Axel Bronstert studied civil engineering with focus on hydrology and water management and earned his doctorate in Karlsruhe. From 1995-1999 he worked in water research at the Potsdam Institute for Climate Impact Research (PIK). Since 2000, he has been Professor of Hydrology and Climatology at the University of Potsdam.
Contact
Universität Potsdam
Institut für Erd- und Umweltwissenschaften
Karl-Liebknecht-Str. 24–25, 14476 Potsdam
E-Mail: axel.bronstertuuni-potsdampde
Prof. Dr. Frank Scherbaum studied physics, geology, (and musicology) at the University of Tübingen. After a doctorate and habilitation in geophysics at the University of Stuttgart as well as spending several years in the US and Japan, he was appointed Professor of Seismology at Ludwig Maximilians Universität in Munich in 1989. Since 1997, he has been Professor of General Geophysics at the University of Potsdam.
Contact
Universität Potsdam
Institut für Erd- und Umweltwissenschaften
Karl-Liebknecht-Str. 24–25, 14476 Potsdam
E-Mail: fs@geo.uni-potsdam.de
Dr. Sanjay Bora studied physics at Kumaun University (India) and computational seismology at the Indian Institute of Technology Kharagpur. He successfully defended his doctoral thesis at the University of Potsdam in early 2016. Bora is currently working as a postdoc at the GFZ German Research Center for Geosciences.
Contact
Helmholtz-Zentrum Potsdam – Deutsches GeoForschungsZentrum
Helmholtzstraße 6, 14467 Potsdam
E-Mail: boraugfz-potsdampde
Annabel Händel studied geosciences and specialized in geophysics at the universities of Tübingen and Potsdam. Since 2012, she has been doing her doctorate at the University of Potsdam.
Contact
Universität Potsdam
Institut für Erd- und Umweltwissenschaften
Karl-Liebknecht-Str. 24–25, 14476 Potsdam
E-Mail: ahaendeluuni-potsdampde
Tobias Pilz studied at the University of Potsdam and earned a Master’s degree in geoecology. Since 2014, he has been doing his doctorate in hydrology.
Contact
Universität Potsdam
Institut für Erd- und Umweltwissenschaften
Karl-Liebknecht-Str. 24–25, 14476 Potsdam
E-Mail: tpilzuuni-potsdampde
Text: Matthias Zimmermann
Translation: Susanne Voigt
Published online by: Agnetha Lang
Contact for the editorial office: onlineredaktionuuni-potsdampde