Mount Pinatubo in the Philippines serves as a reminder of the importance of forecasting volcanic eruptions. On June 15, 1991, the volcano erupted in a densely populated area but prior volcanic monitoring activities by the US Geological Survey and the Philippines Institute of Volcanology and Seismology saved lots of life and property. Eruption scientists highlight the difficulty of predicting volcanic eruptions due to nature’s unpredictability. However, recent trends in technology and innovation especially artificial intelligence offer valid hopes of accurate volcanic eruptions. The article explores the headways made in predicting volcanic eruptions for a safer world.
Remote Volcanic Monitoring
Scientists and eruption researchers acquire torrents of data from satellites on the world’s active volcanoes, but there is a disconnection on how to convert the data into a prediction model. Encouraging advancements in volcanic detection algorithms reinforce hopes for a global volcanic warning system. Relevant factors include a volcano’s geologic past, magnitude, frequency, and ages of any past eruptions form a basis for the prediction model. The US Volcano Observatories leads the global efforts in providing real data to forecast volcanic eruptions. For instance, using satellite images, scientists point out high-temperature spots and images of areas showing impending volcanic activity.
Overwhelming Volcanic Data
A general consensus exists among scientists and eruption scientists on the uncertainty and overwhelming task of interpreting data on potential volcanic activities. In 2021, the world’s most active volcano, Mt. Nyiragongo in the Democratic Republic of Congo erupted releasing lava, gas, and volcanic ash. Additionally, tens of people died and critical infrastructure such as schools, electricity, and transport were destroyed. Volcanologists term the volcano as chaotic yet no accurate volcanic eruptions forecast were released before the event.
The US Volcano Observatories say 1400 volcanoes hold eruption potential yet only 100 are monitored. It highlights the immense task of developing proper forecasting tools. Opinions from volcanologists show the key to predicting volcanic eruptions lies in consistently observing any abnormal changes in collected data during quiescence. The forecasting of volcanic eruption processes and models is difficult but using modern monitoring tools looks promising.
Radar Interferometry Sequencing
The European Space Agency uses satellites christened Sentinel 1A and Sentinel 1B to study and predict volcanic eruptions. While this is not a departure from any of the existing satellite imaging, it integrates the radar interferometry technique. Under this model, researchers receive images of ground shifts around the world’s volcanoes. The radar signals received reflect any changes to the planet’s surface as observed from Earth. Using the Sentinel technique, the satellites revisit each planet’s surface once every 6 days and rapidly release high-resolution observations to the eruption researchers.
In the UK, a reputable research group, The Centre for Observation and Modelling of Earthquakes, Volcanoes, and Tectonics (COMET) creates a database using these ground-movement images and snapshots, called interferograms, for all world’s volcanoes. Based on the success of machine learning in detecting patterns, the scientists overlay the data with automated selection to further simplify their work and make accurate inferences on potential volcanic eruptions in the UK and beyond.
Changes in ground motion indicate magma movement beneath the volcano thus an inaccurate way of predicting eruptions. Land shifts, on the other hand, help validate eruptions since very few cases exist of volcanic eruptions without land deformation. Ash plumes and thermal hotspots, though easily detectable from weather satellites don’t objectively forecast eruptions.
Independent component analysis was applied in developing algorithms to eliminate the risk of interferograms confusing atmospheric shifts to ground movement. The technique involves breaking apart any signal into various elements: short-term turbulence or stratified atmosphere and ground shifts along a volcano’s flank. With this model, the algorithms detect any rapid changes and massive ground shifts, which cumulatively predict volcanic eruptions.
Artificial Intelligence Forecasting Volcanic Eruptions
A key algorithm currently under trial is the use of artificial intelligence to make eruption forecasts. Researchers based at the University of Bristol in the United Kingdom apply the convolutional neural network. The underlying idea is using biologically inspired layers of ‘neurons’ in breaking apart Sentinel satellite images into most abstract entities, such as differentiating horses from zebras. Initially, interferograms from the Sentinel with multiple eruption case studies were applied but this proved inconclusive due to too many false positives.
The scientists keep pushing to refine the model. The target remains to attain an objective and consistent artificial intelligence model to predict volcanic eruptions. Computer-simulated eruptions helped decrease the false positives from 60% to 20% and the researchers remain hopeful with more Sentinel images poured into the algorithm system, the false positives will decrease and create a stable system. Encouragingly, the model accurately detected eruptions in the Sierra Negra and Wolf volcanoes in Galapagos Islands.
Measuring Volcanoes Gas Vapour
Since the 1950s, the volcanology community continues to innovate in its quest to find a unicorn model to forecast volcanic eruptions. Normally, the underground magma lies under immense pressure thus dissolving the vapour. As the magma rises, the pressure eases and gases such as carbon dioxide and sulfur dioxide start bubbling into the surface.
Theoretically, the excessive release of these gases indicates an impending volcanic eruption. Eruption activities in Mt. Pinatubo in the Philippines were successfully detected following a rise of sulfur dioxide levels to insane levels of 16,500 tonnes a day. Therefore, scientists developed infrared telescopes to monitor the concentration levels of gases escaping from the volcanic vents. An infrared spectrometer to capture other gases common during volcanic eruptions such as hydrochloric acid, methane, and carbon monoxide.
Predictions and Costs
No scientist or government agency wants to inaccurately announce a volcanic eruption. An estimated 800 million people live within areas marked at risk of seismic activity. Evacuation costs and societal disruption from false alarms could be catastrophic. The US military lost two of its largest bases in the Philippines following Mt. Pinatubo’s eruptions. The estimated 20 million tonnes of sulfur dioxide gases released from the eruption also led to a temporary global decrease in temperatures by about 1°F (0.5°C) from 1991 to around 1993. This indicates the potential of volcanic eruptions to cause climate change. Indigenous communities previously living around the mountain were permanently displaced.
Such is the cost of volcanic eruptions and resolves any scepticism on why developing an accurate, consistent, and unquestionable volcanic prediction model remains a priority.
The Next Big One
Volcanologists often joke, and compete, on who will correctly call out the next big volcanic eruption. The caldera-forming volcanic eruptions are rare but eruption experts are in frenzy. The need to save lives must override any professional gratification. Luckily, with the current advancements, the chances of an unheralded volcanic eruption occurring are minimal. The existing tools would potentially pick it out.
The future lies in consolidating existing volcanic eruption forecast methods and increasing the capacity to collect and interpret more data on potential eruption activities. Technology, artificial intelligence, and ongoing research give hope the solution is potentially close.