The climate history of the Earth can be investigated by analysing ice cores – ice cylinders extracted from glaciers or ice sheets like Greenland and Antarctica. The ice stratification along the core depth provides a continuous record of the past climate: the deeper the ice is, the older. In ice cores it is possible to find particles that were once deposited and became trapped in the ice matrix. Amongst those are insoluble particles, just like volcanic glass particles injected into the atmosphere during volcanic explosions, mineral dust from deserts or particles of biological origin such as pollen and algae sourced from forests or from the surface of the oceans. Detecting these particles is crucial to understand the past conditions of atmosphere, the biosphere and the oceans in their past interactions. At present, revealing such particles is carried out during time-intensive manual microscopy sessions. The Marie-Curie EU-funded ICELEARNING project aims to develop an innovative technique for automatically revealing insoluble particles in ice cores, by combining automatic image analysis with Artificial Intelligence pattern recognition techniques. This powerful synergy can provide knowledge about the past climate over the last 1.5 million years.
Ice cores are among the most pristine record of past climate. The longest ice cores drilled in northern hemisphere have been drilled in Greenland and cover 125,000 years, while the oldest one of the southern hemisphere was recovered in Antarctica and cover the last 800,000 years. An ongoing-project aims to extend this record to 1.5 million years (see here). The technique used to perform ice core geochemical analyses depend on the object of the investigation, which can be generally classified in the chemical composition of the ice and of the gases and the types of particles trapped in the ice matrix. The ICELEARNING project is related to this last class: the particles.
Insoluble particles of terrestrial origin include mineral dust which originate from desertic regions, volcanic material injected into the atmosphere during volcanic explosions, biologic material and oceanic organisms such as pollen, marine diatoms and foraminifera living in the surface waters of the oceans. All these particles can be trasported through the atmosphere and consequentially deposited at polar latitudes. Following deposition onto ice sheets, they are buried by the snow and remain trapped in the ice. The photo on the left shows a 'dark' layer, indicative of the presence of terrestrial particles, in the Renland ice core, drilled in 2015 in East Greenland. These particles are some tens of thousands of years old.
Acquiring micro-scale images of liquid samples using microscopes is the standard analytical routine used in a wide range of research fields, from drinkable water monitoring, to phytoplankton and zooplankton analysis to several other aquatic applications. The ICELEARNING project aims to acquire images of aqueous ice core samples to investigate the content of insoluble particles using an image particle analyser. The information stored in these images is then analysed using Artificial Intelligence.
"The field of pattern recognition is concerned with the automatic discovery of regularities in data through the use of computer algorithms and with the use of these regularities to take actions such as classifying the data into different categories." In the same way an algorithm can be trained to recognize this hand-written digit as number '8', the goal of the ICELEARNING project is to develop Particle Recognition algorithms to automatically classify the different particles contained in ice cores based on their images. Ideally this approach can replace human intensive manual microscopy investigations.
Ca' Foscari University of Venice (IT)
Department of Environmental Science, Informatics and Statistics
+39 041 234 8504
+39 041 234 8549
University of Bergen (NO)
Bjerknes Centre for Climate Research (NO)
+47 55 58 98 66
Eivind Wilhelm Nagel Støren
University of Bergen (NO)
The ICELEARNING project has received funding from the European Union's Horizon H2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement no. 845115. See EU page.