The hottest new application of artificial intellig

  • Detail

The new application of artificial intelligence can help predict volcanic eruptions

China] satellites can provide data on global active volcanoes, but researchers have been committed to using these data to predict the risk of volcanoes. This idea is likely to be realized soon, because the current algorithm can automatically calculate volcanic risk data signals, so as to help scientists establish a global volcanic early warning system within a few years

michaelpoland, a scientist from the Yellowstone Volcano Observatory in Washington, D.C. (a department of the self regulatory chaos survey led by the leading U.S. Geological polyurethane insulation enterprise), said that without these algorithms as auxiliary tools, geologists would not be able to keep up with the pace of satellite information. Andrew Hooper, a volcanologist at the University of Leeds in the United Kingdom, led the development of an algorithm. He said that this algorithm should benefit about 800million residents who live near the volcano, especially in the modified and high value-added use of engineering plastics such as cars. "About 1400 volcanoes may erupt on the sea," he said. "About 100 of them are under monitoring, but most of them are not monitored."

at the American Geophysical Union meeting (AGU) held in Washington, D.C., this week, the meeting is held every six months and shows two methods of predicting volcanic eruptions. In the past few years, with the launch of European Space Agency satellites sentinel 1a and sentinel 1b, the field of Volcanology has devoted itself to observing the movement of land around volcanoes. Sentinel 1 satellite uses a technology called radar interferometry, which can compare the radar signals sent to and reflected from the earth to track the changes of the planet's surface

volcanic eruption

this method is a commonplace, but it is worth mentioning that every six days, sentinel 1 satellite will re detect every location on the earth, and sentinel team can quickly receive these high-resolution observation results. In the UK, a research team called the center for structural observation and modeling of earthquakes and volcanoes (comet) has begun to build a snapshot database of ground motion called "interferograms" for the world's volcanoes. Hooper, who cooperates with comet, said that considering that the pattern detection of learning machines is quite successful in other fields, we naturally thought of using automatic detection coverage to cover this database

the change of ground motion can usually reflect the magma movement under the volcano, but it can not fully predict the volcanic eruption. Unlike hot spots or ash plumes that can be automatically detected by hot weather satellites, ground movements can help predict volcanic eruptions, not just indicate their occurrence. "Moving doesn't always mean that volcanoes will erupt," Hooper said. "But there are few cases where volcanoes erupt directly without moving."

in order to achieve this goal, the team must teach them b) divide the important, rare, obvious and pointer according to the instigation type at the end of measurement; The algorithm cannot easily confuse the atmospheric changes of ground motion. To this end, Hooper SMM: Recently, the team used a technology called independent component analysis, which can decompose the signal into different parts: such as stratified atmosphere or short-term turbulence, and the ground displacement of volcanic craters or flanks. This technology enables them to capture the latest ground movement or movement rate changes, both of which may be signs of future volcanic eruptions

at the same time, another comet team led by Juliet Biggs, a volcanologist at the University of Bristol, UK, used artificial intelligence to build a second algorithm, called convolutional neural network. The researchers first used the original interferogram from ENVISAT, the predecessor of sentinel, to train their neural networks. They have some examples of volcanic eruptions. Although the algorithm has made some progress in analyzing 30000 sentinel interferograms, its prediction results are still unsatisfactory. Fabin albino, another volcanologist in the group, said that at present, they have only a few research examples, and for learning machines, thousands of examples are needed

to solve this problem, Biggs and her colleagues created a synthetic data set simulating volcanic eruptions. Albino said that as more sentinel examples are uploaded to the algorithm, the prediction results will become more and more accurate

Copyright © 2011 JIN SHI