Earth scientist Andrea Manconi and an international team develop algorithms within the scope of the EU’s BETTER Project that analyse satellite data for the early detection of natural hazards such as landslides.

It happened at the end of September 2016. The rock slopes on the Moosfluh, located at 2,333 metres above the sea and in the heart of the Aletsch Arena of the Swiss canton of Valais, started to slide. This did not only scare the cableway operators and the official authorities of the affected Riederalp municipality – ETH Zurich’s scientists, who had been studying this area for years, did not expect such a strong ground displacement coming either. «It was astonishing,» Andrea Manconi says. He, too, was taken aback by the speed at which the rocks on the Moosfluh moved. Unstable slopes are the field of expertise of the Sardinian geophysicist and expert for satellite data who has been conducting research at the Department of Earth Sciences at ETH Zurich for the past five years. Toppling slopes can be found everywhere on the planet. Causes are for example earthquakes, volcanic eruptions, long-lasting rainfalls – or global warming, responsible for the recession of the Great Aletsch Glacier.

The Moosfluh slides much faster than anticipated

We are sitting in a seminar room in the venerable building of the Department of Earth Sciences. On this day in mid-June, PhD candidate Nikhil Prakash sees his supervisor Andrea Manconi again for the first time in three months. During the lockdown the two of them communicated via videoconference. They are glad to meet again in person and greet each other with due caution, on the ground floor of the magnificent four-storied atrium, which radiates even more grandeur than usual due to its emptiness and tranquility.

«Nature does not take
its cue from mankind.»
(Andrea Manconi)

Andrea Manconi boots his computer and shows by means of a three-dimensional simulation how the Great Aletsch Glacier will recess according to the calculations of ETH Zurich’s glaciologists – until in 2100 (that is, in 80 years) there will be only a few tiny ice fields left. He uses a time-lapse film to illustrate how the Moosfluh slope slid down since September 2016. It looks as if the slope could no longer hold on to the mountain and therefore had to shed its outer stone layer.

It is known that the glacial melting partly pulls the literal rug from under the mountain. Earth scientists discovered in 2010, with the help of satellite data, ground deformation near the tongue of the longest glacier in Europe existing today. Simon Löw, Professor of Engineering Geology, and his team study the area in order to understand the interaction between glacier recession and slope instability and to attempt the prediction of future developments. Nine months prior to the spectacular landslide the latest version of the Gletscherbahn Moosfluh Combi-Lift was opened – a «superb technical achievement», according to the operators’ description on their website. The lift was built in such a way that the geological subsidence predicted by geologists as a result of glacial melting can be compensated in the next 25 years thanks to special constructions at the middle and upper stations: The upper station can be moved by eleven metres horizontally and by nine metres vertically. And then, shortly afterwards, there was this unexpected, heavy acceleration. «Nature does not take its cue from mankind,» Manconi states laconically. After this movement from the depth of the earth, which was torn open widely at several locations and saw boulders landing on hiking trails, the on-site field measurements as well as the satellite data analysis revealed the following: In 2016, the mountain did not only slide 20 to 30 centimetres per year as predicted – it slid 70 to 80 centimetres per day.

Each incident triggers a delayed reaction

The globe is in constant motion, it is bubbling within, the continents are shifting and the poles are moving. Each incident, be it an earthquake, a volcano or a melting glacier, triggers a reaction on the surface of the planet. The earth’s surface, in which mankind has settled, often shows delayed reactions to these events. «We do know that the recession of the Great Aletsch Glacier will trigger a reaction on the adjacent slopes, but we do not know when, where exactly and how big the ground movement will be,» Manconi explains. Earth scientists only see the famous tip of the iceberg: torn meadows, crumbling rocks, sliding slopes. «What is happening in the broad mass below is not easy to investigate and often remains in the dark.» This is why Manconi gets annoyed sometimes when earth scientists are being criticised for making inaccurate estimations or for not being able to deliver long-term predictions. As if a meteorologist could report a rain forecast for a specific day half a year from now. «Nature is intractable. That’s the way it is.»

Observing tectonic shifts with satellite data

Andrea Manconi tries to render earth displacement at least somewhat more predictable. He wants to help prevent unpleasant surprises such as the Moosfluh incident. In the BETTER Project, which is funded by the EU Research Programme Horizon 2020, ETH Zurich and five other partners develop a digital processing service which will, for example, help local authorities to identify risk areas in order to proactively set up a management for probable natural phenomena. The project partners focus on natural hazards, food security as well as the EU’s foreign and security policy. Satellite data is suitable for investigating natural phenomena at global scales.

In order to further evaluate the extent of the slope instability on the Moosfluh the researchers analysed, for example, radar images of the European Commission’s Earth Observation Programme Copernicus Sentinel-1 shortly before and after the incident with a virtual platform of the European Space Agency and thereby established a velocity map of the earth movement. «With this, we can monitor the spatial development of surface shifts and interpret them more accurately,» Andrea Manconi says.

«We do not know when,
where exactly and how big
the ground movement will be.»
(Andrea Manconi)

Satellite radar devices deliver special images: They measure the distance between the satellite and the surface of the earth also at night and through clouds. Earth displacements can thus be measured for any desired time period and the development of mountain slopes, for example, can be observed during a certain amount of time. The European satellites take a radar picture every six days. However, when the development of a ground is very fast – like on the Moosfluh – these pictures sometimes have to be combined with real-time measurements on ground in order to establish a more precise velocity map.

From satellite data to the dam project

If the local authorities had access to these maps, they could apply risk management and better plan delicate major projects. In Bhutan, for example, an important dam project was evaluated based on satellite data analysed and interpreted at ETH Zurich. The EU BETTER Project is aimed at granting access to analyses and maps of their «active» areas to countries with reduced specialist competencies in the processing of satellite data, so that they can react accordingly. Manconi likens this approach to the corona crisis: «The virus caught us unawares. But now we know how we can protect ourselves and are better prepared for a next wave.» The massive volume of Earth Observation Data (EO data) cannot be handled by a layperson; besides, one has to complete specialised training in satellite data evaluation to be able to analyse and interpret radar data. For this reason, Manconi places great emphasis on the education of students in the area of satellite data evaluation, because «EO data is increasingly used to help solve global societal challenges.» According to him, modern earth scientists should be able to handle and assess large volumes of diverse data sources within a short period, so that analyses are quickly available. He therefore established a new course at ETH Zurich pursuing this goal.

Platform for requested satellite evaluation

The BETTER Project offers a different solution for the laypersons of the local authorities in areas with hazardous zones: artificial intelligence. This is what engineer Nikhil Prakash is working on. Already during his master degree studies in India he dealt with digital mapping. The aim of the project is that, for example, a municipal authority makes a request and artificial intelligence then systematically analyses the satellite data of the corresponding region and establishes a map of landslides in this area. «We are developing an algorithm for a platform that offers satellite analysis on demand,» Manconi adds. Thus, artificial intelligence could analyse the satellite data of the entire area after an earthquake incident in order to map landslide hazards, Nikhil Prakash explains. «The machine is better than the human being because it delivers results about large surfaces in a short time, where a man-made analysis would take very long or would require many experts.» He developed an algorithm that is able to learn by itself. Like a face recognition algorithm learns to distinguish between human and non-human faces, Prakash’s algorithm is capable of learning, for example, to distinguish between sliding and non-sliding slopes based on satellite data and to reproduce this knowledge in a map.

«The machine is better
than the human being
because it delivers results
about large surfaces in a short time,
where a man-made analysis
would take very long.»
(Nikhil Prakash)

Both scientists are fascinated by the fact that this EU project, which will expire at the end of October, is not merely purely academic research but allows them to develop an instrument for practical implementation. «It has always been important to me that our research can be used in practice,» Manconi states.

For the same reason, it is one of his priorities that earth scientists learn to explain the analyses and interpretations of their research to the laypersons. His role models are the meteorologists on TV or on the radio who are able to explain weather phenomena in a non-technical language so that everyone can participate in the discussion afterwards. «Likewise, earth scientists are also dealing with natural phenomena that can affect us all. This is why it is important that everyone understands what our research is about and what we are finding,» Manconi says. Whenever he meets with local authorities for a presentation, he is equipped with pictures of landslides, simulations of melting glaciers or hazard maps. He is convinced that images help to better understand abstract concepts.

Interview mit Andrea Manconi