Natural Disaster Prediction: A Leap Forward with AIby Valentina Urbasova | 28-08-2024 23:58 |
|---|
|
Natural Disaster Prediction: A Leap Forward with AI  Credit: Unsplash Predicting natural disasters is a cornerstone of geosciences. The ability to foresee tsunamis, earthquakes, volcanic eruptions, hurricanes, and floods can save countless lives and significantly reduce material damage. Despite substantial technological advancements, achieving 100% accuracy in predicting impending natural disasters remains elusive. However, substantial strides are being made: scientists are harnessing AI and machine learning to forecast cataclysms and mitigate their impact. Earthquakes It's possible to sense an impending catastrophe in advance. For instance, scientists recorded the strained state of Japan's lithosphere before the catastrophic Tohoku earthquake in March 2011, which claimed nearly 20,000 lives. However, despite advancements in seismic monitoring networks and data analysis algorithms, it remains challenging to pinpoint the exact time and location of an earthquake, although some scientific attempts have been successful. In 2023, researchers at the University of Texas at Austin published the results of a seven-month trial in China of an AI algorithm trained to detect statistical anomalies in real-time seismic data and historical earthquake records. In the experiment, AI accurately predicted 70% of earthquakes a week in advance, with 14 forecasts within 200 miles of the predicted location and matching the expected magnitudes. However, it also issued eight false alarms and missed one earthquake. Tsunamis Scientists have been studying tsunamis for nearly a century. Since the 1960s, seismic stations and ocean-bottom pressure sensors have been detecting underwater earthquakes, the primary cause of these giant waves. This data is processed by specialized centers, such as the Pacific Tsunami Warning Center (PTWC). They can alert coastal cities of potential tsunami waves several hours before they arrive. These centers send warnings to meteorological services in threatened countries. The PTWC's area of responsibility encompasses the Pacific Ocean, including its southern part, and all adjacent seas such as the East China Sea, Yellow Sea, and Sea of Okhotsk. In 2011, Japanese authorities were given a few minutes' warning of an impending tsunami immediately after the earthquake was detected. Early tsunami warnings are complex due to the varying risk factors associated with the triggering underwater earthquake. Traditionally, deep-sea buoys measuring changes in water level were used to assess the magnitude of underwater tremors. However, analyzing this data takes time, leaving little time for evacuation in case of danger. But artificial intelligence (AI) is coming to the rescue.  The model delivers two potential fault orientations for each earthquake scenario, which are numerically modeled and compared. Credit: Bernabe Gomez and Usama Kadri In an article published in the Journal "Physics of Fluids" in April 2023, researchers from the University of California, Los Angeles, and Cardiff University in the UK described a new tsunami early warning system they developed. It combines advanced acoustic technology with AI. This system can instantly classify underwater earthquakes and determine their potential danger and whether they could generate a tsunami. Instead of relying on the displacement of deep-sea buoys, scientists propose measuring the acoustic radiation (sound) generated by the earthquake. This also carries information about the tectonic event and travels significantly faster than tsunami waves. Underwater microphones, called hydrophones, record acoustic waves and monitor tectonic activity in real time. AI algorithms classify the type of displacement and magnitude, then calculate critical parameters such as the length and width of the fault zone, the rate of uplift of the ocean floor, and the duration of the earthquake, which determine the size of the tsunami. "Acoustic radiation travels through the water column much faster than tsunami waves. It carries information about the originating source and its pressure field can be recorded at distant locations, even thousands of kilometers away from the source. The derivation of analytical solutions for the pressure field is a key factor in the real-time analysis" notes one of the study's authors, Usama Kadri. Thus, an evacuation signal can be issued almost immediately if an underwater earthquake is more powerful than usual. Volcanic Eruptions Unlike earthquakes, volcanic eruptions can be predicted with greater accuracy. Monitoring gas emissions, temperature, and volcanic activity allows scientists to detect signs of an impending eruption. There are projects that monitor volcanic activity and analyze it using AI. In 2019, a group of researchers from the Technical University of Berlin and the Helmholtz Centre Potsdam developed the MOUNTS platform, which collects satellite images of volcanoes worldwide – from Russia and Japan to Colombia and Papua New Guinea – and processes them using AI. Information is available for each volcano, including sulfur dioxide emissions, temperature changes, and changes in topography. Historical observation data since 2000 is also available.  Interface MOUNTS platform Another project is the Centre for Observation and Modelling of Earthquakes, Volcanoes and Tectonics (COMET), which unites scientists from 14 UK universities. Using satellites, ground observations, and geophysical models, it collects data on 95 volcanoes: the dynamics of tectonic movements and maps of the probability of changes in the shape of volcanoes and the surrounding land, formed using machine learning methods. In 2021, the meteorological company Weathernews launched a volcanic ash detection system. It analyzes satellite images and recognizes volcanic ash in cloud gaps by its shape and texture. The system helps the company inform its customers about volcanic eruptions and the speed and direction of ash spread. This is particularly important for airlines in Japan and Southeast Asia, where many active volcanoes are located. Smoke and ash from volcanic eruptions can endanger flights due to poor visibility and the risk of engine damage. For example, the eruption of the Eyjafjallajökull volcano in Iceland in 2010 led to the closure of airports in about 30 countries and the cancellation of many flights. Typically, airlines make flight decisions based on information published by government agencies, but Weathernews' service can also be used to change flight plans or alert aircraft en route. Hurricanes and Precipitation Hurricane forecasting has become much more accurate thanks to satellite observations and computer modeling. Meteorologists can predict the formation of hurricanes long before they develop and track their trajectories. In 2022, researchers from IRT AESE Saint Exupéry and Météo-France developed three neural networks to forecast future precipitation. With their help, meteorological, government, and other organizations can predict the occurrence of hurricanes and other extreme weather events from one to six hours in advance. The networks were trained on a dataset of 20,400 weather maps – high-resolution images taken using radar technology in France in 2017-2018 over an area of approximately 1000x1000 km©÷. "We propose the use of three popular deep learning models (U-net, ConvLSTM and SVG-LP) trained on two-dimensional precipitation maps for precipitation nowcasting," the researchers write in their article. These models accurately predict the future contour of precipitation. The work of French scientists could also be a step towards developing similar models for forecasting extreme weather events.  Two examples of precipitation predictions. Top row: Ground truth; second row: U-net model; third row: ConvLSTM model; bottom row: SVG-LP model. The results for the network represent 6 outputs from the model. Credit: Bakkay et al. The Future of Disaster Mitigation The development of artificial intelligence and machine learning promises to improve the analysis of large amounts of data and the identification of complex patterns that may indicate impending disasters. Improvements in sensor technology and the expansion of monitoring networks also contribute to earlier detection of hazardous natural phenomena. Russia is also using artificial intelligence in disaster prediction. For example, in 2022, the Emergency Situations Ministry began testing an AI system that identified potential wildfire and flood sites. In the first year of using the technology, the area of active fires in Yakutia decreased 16 times. According to expert forecasts, by 2027, systems that use artificial intelligence and big data to predict emergencies will appear in all major Russian regions. Although 100% accurate prediction of many natural disasters is still impossible, ongoing scientific and technological progress is improving humanity's ability to warn of natural disasters and mitigate their consequences. Sources: https://global.utexas.edu/news/jackson-school-researchers-employ-ai-predict-earthquakes-china#:~:text=A%20team%20of%20researchers%20at,one%20week%20before%20they%20happened. https://www.gt-crust.ru/jour/article/view/49/51 https://tsunami.ioc.unesco.org/en/pacific?option=com_content&view=category&id=1153&Itemid=1153 https://publishing.aip.org/publications/latest-content/creating-a-tsunami-early-warning-system-using-artificial-intelligence/ https://news.un.org/ru/story/2011/03/1179881 http://www.mounts-project.com/home https://comet.nerc.ac.uk/ https://global.weathernews.com/news/16070/ https://techxplore.com/news/2022-04-deep-imminent-precipitations.html https://arxiv.org/abs/2203.13263 https://rg.ru/2022/11/10/stihiia-po-prognozu.html https://nauka.tass.ru/nauka/16296671 |
