AI and corals

Based on images taken by cameras, an AI developed at the EPFL can reconstruct coral reefs in 3D in just a few minutes. A minor revolution for seabed study and preservation missions such as those of the Transnational Red Sea Center.

ECEO, Prof. Devis Tuia

Although they may appear in the background in the shots of shimmering fish taken by diving enthusiasts, corals are at the forefront of many scientists’ minds, given their fundamental ecological role. These animals with their limestone skeletons are among the most diverse ecosystems on the planet: although they cover less than 0.1% of the total surface area of the oceans, they are home to almost a third of known marine species. They also have a major impact on the lives of people in many reef-bordered countries: according to a study by the US Oceanic and Atmospheric Administration, up to half a billion people worldwide depend on coral reefs for their food security and tourism-related income. Endangered by rising ocean temperatures and local anthropogenic pollution, which cause coral bleaching and death, corals are the subject of in-depth studies, such as those carried out by the Transnational Red Sea Center (TRSC), which aims to unlock the secrets of Red Sea species that are particularly resistant to the stresses of climate change. DeepReefMap, an artificial intelligence developed by the Laboratory of Computational Science for the Environment and Earth Observation (ECEO) at EPFL, was tested as part of this EPFL-led mission*. It is capable of reconstructing several hundred metres of coral reefs in 3D on the basis of an underwater film taken by a commercial camera in just a few minutes. It can also recognise and quantify certain coral characteristics. “This method democratises the digital reconstruction of reefs and gives a major boost to their monitoring by reducing the work, equipment, logistics and IT costs involved,” points out Samuel Gardaz, TRSC project manager. This research has now been published in Methods in Ecology and Evolution.

Classify corals according to their state of health and shape

In order to further facilitate the work of their biologist colleagues working in the field, the scientists have included semantic segmentation algorithms to classify and quantify corals according to their state of health – from healthy, i.e. very colourful, to dead and covered in algae, via the white of bleaching – and to identify the forms of coral growth that are most common in the shallow reefs of the Red Sea according to an internationally recognised hierarchy – branching, massive, hard, soft, etc. -. “The aim was really to meet the needs of scientists and conservationists working in the field with a tool that can be deployed widely and very quickly: in Djibouti, for example, there are 400 km of coastline”, stresses Jonathan Sauder, who made the development of this AI the subject of his doctoral thesis. “Our method does not require a costly IT infrastructure: on a computer equipped with a simple graphics processing unit, semantic segmentation and 3D reconstruction can be obtained in the same amount of time as the video”.

Read the full article

Diversity and distribution of bacteria in glacial streams

Climate change is having a significant effect on glacial rivers, particularly on their microbial ecosystems. This study enabled the microbial population of 152 glacial rivers around the world to be surveyed.

RIVER, Prof. Tom Battin

Diversity and distribution of bacteria in glacial streams

The rapid melting of mountain glaciers and the disappearance of the rivers they feed are powerful symbols of climate change. Glacial rivers are cold, nutrient-poor and unstable ecosystems, dominated by microbial biofilms. However, current knowledge of the microbiome of these rivers remains limited, making it difficult to understand how it reacts to the shrinking of glaciers.

In this study, we used metabarcoding (a technique that rapidly identifies the species present in a sample by analysing small pieces of DNA characteristic of each species) and metagenomics (a method that consists of analysing all the DNA present in a given environment, in order to learn not only about the species that live there, but also about their biological functions) to draw up a global inventory of the bacteria present in 152 glacial rivers from the planet’s main mountain ranges.

Our results show that the bacterial microbiome of these rivers is taxonomically and functionally distinct from those observed in other icy environments. It is highly diverse, with more than half the species specific to a given mountain region, some to a single site, and only a few cosmopolitan and abundant.

We demonstrate that geographical isolation and environmental selection structure their biogeographical distribution, with microbial compositions clearly differentiated between mountain ranges and between hemispheres. Phylogenetic analyses (methods that make it possible to reconstruct the relationships between different microbial species or strains on the basis of their genes or DNA) also reveal the existence of micro-diverse lineages, shaped by environmental selection, which probably contribute to the functional resilience and overall diversity of the microbiome in these environments.

The shrinking of glaciers due to climate change is therefore threatening this unique microbial ecosystem. Our study provides a global reference for future research into the effects of climate change on disappearing glacier ecosystems.

Energy demonstrator

Swiss energy production from renewable sources is still very dependent on the seasons. In summer, there is a huge overproduction, while in winter we have to import several TWh. A demonstrator has been installed in Sion to convert solar energy into gas and store it for reuse in winter.

GEM, MER Jan Van Herle

A “Power-to-Gas” demonstrator has been installed on the Sedun campus. This is a unique experimental platform in Switzerland. This infrastructure enables the most promising technologies in the energy transition to be tested on a real scale, from the capture of CO₂ to the seasonal storage of energy in the form of hydrogen or methane. The result of collaboration between EPFL, HES-SO and several industrial players, this demonstrator places Valais at the heart of innovation in the production of cleaner, circular energy. While MER Van Herle’s laboratory is responsible for the reversible fuel cell needed for the process to run smoothly, several other EPFL professors are involved in the demonstrator: Prof. Wendy Queen, Prof. Kumar V. Agrawal and Prof. F. Maréchal.

The process can be summarised as follows:

Production of photovoltaic energy using panels installed on the roof of the HES-SO Valais Wallis campus
Decarbonisation and carbon storage for initial injection into methanisation plants
Hydrogen production using a reversible solid oxide fuel cell operating at 750°C
Transformation of the hydrogen produced into methane, in particular by adding_CO2 extracted in phase 2 (for the first cycle, the_CO2 is then reused in a closed circuit)
Methane (CH4) storage in the Sion gas network, in collaboration with Oïken
In winter, gas is extracted from the public grid as required and converted into electricity and heat using a reversible solid oxide fuel cell. The process recovers just under 50% of the initial solar energy.