[ad_1]
Volcanic ash under the microscope includes thousands of tiny particles with complex shapes. Credit: Shizuka Otsuki
Scientists led by Daigo Shoji of the Earth-Life Science Institute (Tokyo Institute of Technology) have shown that a type of artificial intelligence called convolutional neural network can be formed to categorize forms of volcanic ash particles. . Because volcanic particle shapes are related to the type of volcanic eruption, this categorization can provide information on eruptions and facilitate volcanic risk mitigation efforts.
Volcanic eruptions take many forms, since the explosive eruptions of the Icelandic Eyjafjallajökull in 2010, which disrupted European air transport for a week, up to the relatively quiet lava flows of May 2018 from the Hawaiian Islands. Similarly, these eruptions have different associated threats, from ash clouds to lava. Sometimes the mechanism of eruption (for example, the interaction of water and magma) is not obvious, and must be carefully evaluated by volcanologists to determine future threats and answers. The volcanologists closely examine the ash produced by the eruptions (for example, Fig. 1), because different eruptions produce ashes particles of various forms. But how do you objectively examine thousands of tiny samples to produce a consistent picture of the eruption? The classification at the eye is the usual method, but it is slow, subjective and limited by the availability of experienced volcanologists. Conventional computer programs quickly classify particles according to objective parameters, such as circularity, but the selection of parameters remains the task because simple forms classified by a single parameter are rarely found in nature.
Enter the convolutional neural network (CNN), an artificial intelligence designed to analyze imagery. Unlike other computer programs, CNN learns organically as a human, but thousands of times faster. The program can also be shared, eliminating the need for dozens of geologists trained in the field. For this experiment, the program was fed in images of hundreds of particles with one of four basal forms, which are created by different eruption mechanisms (examples are shown in Figure 2). Ash particles are clustered when the rocks are fragmented by vesicular eruptions when the lava is bubbly, elongated when the particles are melted and crushed and rounded to the surface tension of fluids, such as water droplets. The experiment allowed the program to classify basal forms with a 92% success rate and assign probability ratios to each particle, even for the uncertain form (Figure 3). This could allow greater complexity of the data in the future, providing scientists with better tools to determine the type of eruption as if an eruption was phreatomagmatic (as the second phase of the Eyjafjallajökull eruption in 2010) or magmatic (like the eruptions of Mount Etna).
Four idealized categories to simplify classification. Credit: Daigo Shoji
Dr. Shoji's study showed that CNNs can be trained to find useful and complex information on tiny particles of high geological value. To increase the scope of CNN, more advanced magnification techniques, such as electron microscopy, can add color and texture to the results. From collaboration with biologists, computer scientists and geologists, the research team hopes to use the CNN in new ways. The microcosmic world has always been complex, but thanks to some scientists studying volcanoes, the answers may be easier to find.
Results of the convoluted neural network. The ash particles were assigned a probability ratio for each of the four basal forms: Blocky (B), Vesicular (V), Elongated (E) and Rounded (R). Credit: Daigo Shoji
Explore further:
Lava, ash flows, mudslides and unpleasant gases – good reasons to respect volcanoes
More information:
Daigo Shoji et al., Classification of volcanic ash particles using a convolutional neural network and probability, Scientific reports (2018). DOI: 10.1038 / s41598-018-26200-2
Source link
