{"id":494,"date":"2026-10-27T10:52:22","date_gmt":"2026-10-27T09:52:22","guid":{"rendered":"https:\/\/www.ct.ingv.it\/VT-CLAIM\/?p=494"},"modified":"2026-10-27T10:54:50","modified_gmt":"2026-10-27T09:54:50","slug":"monitoring-volcanic-so2-every-few-minutes-artificial-intelligence-integrates-tropomi-and-seviri-satellite-data","status":"publish","type":"post","link":"https:\/\/www.ct.ingv.it\/VT-CLAIM\/monitoring-volcanic-so2-every-few-minutes-artificial-intelligence-integrates-tropomi-and-seviri-satellite-data\/","title":{"rendered":"Monitoring volcanic SO2 every few minutes: artificial intelligence integrates TROPOMI and SEVIRI satellite data\u00a0"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">A new INGV study explores how artificial intelligence can combine the high sensitivity of Sentinel-5P TROPOMI with the high temporal resolution of the geostationary MSG-SEVIRI satellite to monitor volcanic sulfur dioxide emissions in greater detail.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Monitoring a volcano\u2019s sulfur dioxide (SO<sub>2<\/sub>) emissions over time is important for studying its activity and understanding how it evolves. From space, this can be achieved using different satellite instruments, which offer complementary capabilities.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The European Sentinel-5P satellite, through its TROPOMI instrument, provides highly sensitive measurements of atmospheric SO<sub>2<\/sub>. However, because it is in a polar orbit, it observes a given area at a relatively limited temporal frequency, generally once a day.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The MSG-SEVIRI system, by contrast, is onboard a geostationary satellite and continuously observes the same area, providing images at intervals of just a few minutes, typically every 5-15 minutes. SEVIRI is already used to monitor the evolution of volcanic plumes and SO<sub>2<\/sub> emissions.&nbsp;<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The study, published by Dozzo and colleagues in the <em>Bulletin of Volcanology<\/em>, addresses a key question: can the high-frequency, multispectral observations provided by SEVIRI be used to obtain SO<sub>2<\/sub> estimates consistent with those measured by TROPOMI?<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Artificial Intelligence connects the two satellite instruments<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">To investigate this question, Dozzo and colleagues developed a multilayer perceptron, a type of artificial neural network trained using simultaneous observations from TROPOMI and SEVIRI over Mount Etna.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The approach is simple: TROPOMI provides a reference measurement of SO<sub>2<\/sub>, while SEVIRI observes the volcano much more frequently. The artificial intelligence model learns the relationship between the information contained in SEVIRI images and the SO<sub>2<\/sub> concentrations derived from TROPOMI.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The neural network uses information from 11 SEVIRI spectral channels, together with plume height and selected atmospheric parameters. These variables are important because the signal detected from space depends not only on the amount of gas emitted, but also on the altitude of the plume and on atmospheric conditions.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The study was carried out within the Pianeta Dinamico project, as part of the SAFARI and VT-CLAIM subprojects of the Istituto Nazionale di Geofisica e Vulcanologia (INGV).<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">From a single observation to a time series<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Once trained, the model can be applied to subsequent SEVIRI observations (Figure 1).<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The main advantage is therefore temporal. Rather than relying solely on the measurements collected during TROPOMI\u2019s overpass of the volcano, researchers can use the much denser sequence of SEVIRI images to reconstruct the evolution of SO<sub>2<\/sub> emissions throughout the day.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">This capability can prove particularly valuable during episodes of intense degassing or eruptive activity, when volcanic plumes can rapidly change in intensity, shape and direction.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The results obtained at Mount Etna show that the model can reproduce the spatial and temporal variability of SO<sub>2<\/sub> concentrations observed by TROPOMI. The coefficient of determination (R<sup>2<\/sup>) is generally around 0.8, reaching approximately 0.86 in the best-performing case analysed.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">\u201cThese results are encouraging, although model performance depends on the observation conditions and on the specific case considered\u201d, the researchers note.<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-large is-resized\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"735\" src=\"https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-content\/uploads\/2026\/10\/image-2-1024x735.png\" alt=\"\" class=\"wp-image-497\" style=\"aspect-ratio:1.3936170212765957;width:524px;height:auto\" srcset=\"https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-content\/uploads\/2026\/10\/image-2-1024x735.png 1024w, https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-content\/uploads\/2026\/10\/image-2-300x215.png 300w, https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-content\/uploads\/2026\/10\/image-2-768x551.png 768w, https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-content\/uploads\/2026\/10\/image-2.png 1140w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption class=\"wp-element-caption\"><strong>Figure 1<\/strong>. Examples of SO<sub>2<\/sub> plumes predicted by the neural network at four different times, spaced five minutes apart, on 14 July 2021 at Mount Etna.<\/figcaption><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Comparison with ground-based measurements<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The estimates obtained through the satellite-based method were also compared with data from ground-based UV measurements used to monitor volcanic emissions.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The comparison showed good agreement during several degassing and eruptive episodes, providing an additional assessment of the method\u2019s performance.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The integration of satellite observations and ground-based measurements is particularly valuable&nbsp; as the two approaches provide complementary views of the same phenomenon. Satellites can detect the distribution of a volcanic plume over a wide area, while ground-based networks provide continuous, localized measurements.&nbsp;<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">A tool for studying the past<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The methodology could also offer a way to investigate volcanic episodes that occurred before the launch of Sentinel-5P.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Since the model is trained using comparisons between TROPOMI and SEVIRI, it can subsequently be applied to historical SEVIRI observations. This opens up the possibility of using SEVIRI archives to reconstruct SO<sub>2<\/sub> emissions during volcanic episodes that occurred before 2017, when Sentinel-5P observations were not yet available.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Such an approach could contribute to the development of longer time series of volcanic emissions. Reconstructing past periods, however, requires careful consideration of differences in atmospheric conditions, eruption characteristics and the quality of the available observations.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The study also investigates the possibility of applying the method to nighttime SEVIRI observations, further extending the ability to monitor plume evolution over time (Figure 2).<\/p>\n\n\n\n<figure class=\"wp-block-image aligncenter size-large is-resized\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"720\" src=\"https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-content\/uploads\/2026\/10\/image-3-1024x720.png\" alt=\"\" class=\"wp-image-498\" style=\"aspect-ratio:1.420749279538905;width:493px;height:auto\" srcset=\"https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-content\/uploads\/2026\/10\/image-3-1024x720.png 1024w, https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-content\/uploads\/2026\/10\/image-3-300x211.png 300w, https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-content\/uploads\/2026\/10\/image-3-768x540.png 768w, https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-content\/uploads\/2026\/10\/image-3.png 1200w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><figcaption class=\"wp-element-caption\"><strong>Figure 2<\/strong>. Example of an SO<sub>2<\/sub> plume predicted by the neural network during evening hours (17:44 UTC) on 9 August 2021 at Mount Etna.<\/figcaption><\/figure>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Combining the strengths of two satellite systems<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">The key strength of the approach lies in combining two complementary sources of information.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">TROPOMI provides highly sensitive SO<sub>2<\/sub> measurements and represents the reference used to train the model. SEVIRI, thanks to its geostationary position, continuously observes the volcano and makes it possible to track changes in the plume at intervals of just a few minutes.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">Artificial intelligence is used to link these two capabilities, seeking to transfer the SO<sub>2<\/sub> concentration information derived from TROPOMI to the high-frequency SEVIRI observations.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The aim is therefore not to replace one satellite with the other, but to exploit their complementary characteristics.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\">The possibility of obtaining frequent SO<sub>2<\/sub> estimates could be useful both for studying volcanic processes and for monitoring the evolution of emissions during periods of more intense activity.<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong><em>The study was carried out as part of the Pianeta Dinamico project, within the SAFARI and VT-CLAIM subprojects of INGV, dedicated to the development and use of advanced technologies for volcanic monitoring.<\/em><\/strong><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><strong>Read the<\/strong> <a href=\"https:\/\/link.springer.com\/article\/10.1007\/s00445-026-02036-x\"><strong>study<\/strong><\/a><strong>:<\/strong><strong><br><\/strong>Dozzo, M., Aiuppa, A., Lo Bue Trisciuzzi, G., Ganci, G<em>.<\/em> (2026), <em>AI-driven volcanic SO\u2082 estimates using MSG-SEVIRI and Sentinel-5P TROPOMI<\/em>, <em>Bulletin of Volcanology<\/em>, 88, 114. doi: 10.1007\/s00445-026-02036-x<\/p>\n","protected":false},"excerpt":{"rendered":"<p>A new INGV study explores how artificial intelligence can combine the high sensitivity of Sentinel-5P TROPOMI with the high temporal resolution of the geostationary MSG-SEVIRI satellite to monitor volcanic sulfur dioxide (SO\u2082) emissions in greater detail. Using an artificial neural network trained on observations over Mount Etna, researchers can reconstruct the evolution of SO\u2082 emissions throughout the day, tracking changes in volcanic plumes at intervals of just a few minutes. The methodology could also contribute to reconstructing past volcanic emissions using historical SEVIRI observations.<\/p>\n","protected":false},"author":1,"featured_media":497,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"iawp_total_views":16,"footnotes":""},"categories":[1],"tags":[],"class_list":["post-494","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-senza-categoria"],"_links":{"self":[{"href":"https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-json\/wp\/v2\/posts\/494","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-json\/wp\/v2\/comments?post=494"}],"version-history":[{"count":1,"href":"https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-json\/wp\/v2\/posts\/494\/revisions"}],"predecessor-version":[{"id":499,"href":"https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-json\/wp\/v2\/posts\/494\/revisions\/499"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-json\/wp\/v2\/media\/497"}],"wp:attachment":[{"href":"https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-json\/wp\/v2\/media?parent=494"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-json\/wp\/v2\/categories?post=494"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.ct.ingv.it\/VT-CLAIM\/wp-json\/wp\/v2\/tags?post=494"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}