Next Article in Journal
Natural H2 Transfer in Soil: Insights from Soil Gas Measurements at Varying Depths
Previous Article in Journal
Unraveling Information from Seismic Signals Generated by Gravitational Mass Movements
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Geosciences
Volume 14
Issue 11
10.3390/geosciences14110295
geosciences-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Article

Machine Learning Insights into the Last 400 Years of Etna Lateral Eruptions from Historical Volcanological Data

by
Arianna Beatrice Malaguti
,
Claudia Corradino
,
Alessandro La Spina
,
Stefano Branca
and
Ciro Del Negro
*
Istituto Nazionale di Geofisica e Vulcanologia, Sezione di Catania, Osservatorio Etneo, 95125 Catania, Italy
*
Author to whom correspondence should be addressed.
Geosciences 2024, 14(11), 295; https://doi.org/10.3390/geosciences14110295
Submission received: 12 September 2024 / Revised: 25 October 2024 / Accepted: 1 November 2024 / Published: 3 November 2024
(This article belongs to the Section Natural Hazards)
Browse Figures
Versions Notes

Abstract

:
Volcanic hazard assessment is generally based on past eruptive behavior, assuming that previous activity is representative of future activity. Hazard assessment can be supported by Artificial Intelligence (AI) techniques, such as machine learning, which are used for data exploration to identify features of interest in the data. Here, we applied a machine learning technique to automate the analysis of these datasets, handling intricate patterns that are not easily captured by explicit commands. Using the k-means clustering algorithm, we classified effusive eruptions of Mount Etna over the past 400 years based on key parameters, including lava volume, Mean Output Rate (MOR), and eruption duration. Our analysis identified six distinct eruption clusters, each characterized by unique eruption dynamics. Furthermore, spatial analysis revealed significant sectoral variations in eruption activity across Etna’s flanks. These findings, derived through unsupervised learning, offer new insights into Etna’s eruptive behavior and contribute to the development of hazard maps that are essential for long-term spatial planning and risk mitigation.

1. Introduction

Volcanoes are complex dynamic systems, and understanding their activity at any given time poses an ongoing challenge, particularly for frequently erupting volcanoes in densely populated areas. One such case is Mount Etna (Sicily, Italy), where lava flows present the primary hazard. Although less frequent than explosive events, lava flows can have significant socio-economic impacts, especially given the densely populated areas surrounding the volcano. Over the past century, several large flank eruptions have caused extensive damage to agricultural land and disrupted towns and villages [1,2,3]. A marked increase in both eruption frequency and lava production has been observed since 1971 [4,5,6,7], with 20 major flank eruptions occurring along two prominent fracture zones trending south and northeast. This activity highlights the persistent threat posed by Etna’s eruptive behavior over the past 40 years [6,8].
Recent advances in Artificial Intelligence (AI) have greatly enhanced our ability to monitor, analyze, and predict volcanic processes, ultimately improving risk management and the protection of communities in hazard-prone regions. Machine Learning (ML), a key subset of AI, has transformed volcanology by enabling the analysis of vast and complex datasets from a wide range of sensors. These include data on ground deformation, seismic activity, geomagnetic and electromagnetic variations, and thermal anomalies, processed with high levels of accuracy [9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26]. Furthermore, ML algorithms can automatically analyze satellite imagery and aerial photographs to detect morphological changes in volcanoes, such as the formation of new vents, the collapse of crater walls, or the development of lava flow paths [13]. These capabilities allow for the real-time processing of data streams from monitoring stations, providing early warnings of impending eruptions and enabling real-time volcanic risk assessments [9,10,11,12,13,14].
In particular, machine learning techniques have proven invaluable for handling large volcanic datasets to uncover hidden patterns and extract meaningful insights [27,28,29]. Unsupervised learning models, which do not require explicit instructions, are particularly suited for analyzing the complex, multi-dimensional data associated with volcanic activity [18,30,31,32]. These models can learn the underlying structure of the data, helping to elucidate the dynamics of historical eruptions based on geophysical and volcanological observations (e.g., ground deformation and seismic events).
At the Etna Volcano Observatory (EVO), a continuous flow of geophysical and volcanological data is used to deepen our understanding of volcanic processes and improve our ability to identify signs of volcanic unrest, forecast eruptions, and assess hazards. Here, we apply Cluster Analysis (CA), an unsupervised learning approach, to classify Etna’s flank eruptions over the past 400 years. This technique groups similar eruptions and distinguishes those with different characteristics, providing new insights into the eruptive dynamics of Mount Etna’s volcanic system.

2. Etna Flank Eruptions

Mount Etna is a basaltic stratovolcano located along the Ionian coast of eastern Sicily, Italy (Figure 1), known for its persistent activity [33,34]. This activity includes continuous degassing, intermittent Strombolian eruptions, and more violent explosive events at its summit craters [35]. Frequent flank eruptions, characterized by lava flows and often accompanied by weak explosive activity, and rarer eccentric eruptions also define its eruptive behavior [2,35,36,37,38]. While summit eruptions are relatively frequent [2,36,37,39,40,41,42], lateral eruptions occur less often, with intervals of several years between events [2,37].
Etna’s eruptive history over the past three millennia is documented in historical sources, making it a unique case globally. However, comprehensive and accurate documentation of flank eruptions begins only after the mid-17th century [2,37,43], while the catalog of summit activity becomes reliable in the late 19th century [44]. Historical analyses, such as those by Branca and Del Carlo [2], indicate that following the major 1669 eruption, a period of reduced volcanic activity ensued. This pattern shifted after 1763, with a notable increase in flank activity, peaking between 1961 and 2003, when the frequency of eruptions quadrupled compared to earlier periods [37].
Geological data suggest that summit eruptions are generally more frequent than flank eruptions [12,45]. However, flank eruptions pose a greater hazard due to the dense population surrounding Etna [11,12]. In the last 400 years, the most destructive eruption occurred in 1669 [46]. In the 20th century, the town of Mascali was almost completely destroyed by lava flows in 1928 [47], and towns such as Fornazzo, Randazzo, and Zafferana Etnea were threatened in 1971, 1981, and 1992, respectively [48,49,50,51]. More recently, the 2001 and 2002–2003 eruptions caused significant damage to tourist facilities in the summit area [52,53,54,55,56].
Geosciences 14 00295 g001
Figure 1. Sketch map of Etna flank lava flows over the last 400 years from historical catalogs [2,38,57,58,59,60]. The shaded relief image was derived from the 10 m resolution TINITALY DEM [61].
Figure 1. Sketch map of Etna flank lava flows over the last 400 years from historical catalogs [2,38,57,58,59,60]. The shaded relief image was derived from the 10 m resolution TINITALY DEM [61].
Geosciences 14 00295 g001

3. Data Analysis

This study focuses on Etna’s flank eruptions from 1610 to 2019 (Figure 1 and Table 1), drawing primarily on historical catalogs that document 60 lateral eruptions [2,38,57,58,59,60]. These catalogs provide detailed information on eruption duration, lava volumes, and the spatial distribution of past effusive events.
We analyzed the temporal distribution of these lateral eruptions over the past 400 years, focusing on their duration, lava volume, and the position of effusive vents (or fissures) to identify distinct eruptive clusters. The Mean Output Rate (MOR) was also calculated by dividing the final volume of erupted lava by the total duration of each eruption (Table 1). The use of MOR allows for better discrimination of the subclasses previously considered in studies, e.g., [62], as it characterizes eruptions independently of nominal volume or duration, shedding light on the different effusive dynamics of eruptive episodes. Relying solely on duration and volume intervals does not fully capture these effusive dynamics. This approach allowed us to classify the eruptions and better understand the varying dynamics of Etna’s flank activity over the centuries.

4. Method

A detailed classification based on lava volumes, MOR, and the duration of these effusive eruptions was performed using the k-means function in MATLAB. Volcanic features are given as input to an unsupervised learning algorithm to group similar data, where similarity is evaluated quantitatively using a distance function in the space of feature vectors. We chose the non-hierarchical k-means algorithm because it is simple and efficient. It is an iterative procedure that partitions data into a predetermined number of clusters. Firstly, inside the feature space, cluster centroids are randomly initialized, and observations are assigned to the nearest centroid according to the cosine distance. The cosine distance is one minus the cosine of the included angle between feature vectors. Secondly, cluster centroids are updated with the mean location within their cluster. These steps are repeated until an optimal convergence is reached. A cluster analysis using the k-means algorithm was applied to the normalized input data, i.e., z-score normalization is needed to scale the data and reduce variability across arrays [63].
A critical parameter in the k-means clustering algorithm is the number of clusters, k. Estimating k can be challenging as it depends on the data’s complexity and the desired resolution. Too many clusters can complicate interpretation, while too few may overlook significant patterns. When the optimal number of clusters is not known beforehand, three common techniques can help determine k: (1) the elbow method, e.g., [64,65]; (2) the gap statistic, e.g., [31,66,67]; and (3) the silhouette method, e.g., [68,69]. These methods involve performing k-means clustering for a range of k values and computing different metrics to identify the optimal k.
We used the elbow method to find the right number of clusters. The elbow method is a graphical routine for finding the optimal K value in a k-means clustering algorithm. The elbow graph shows the within-cluster sum-of-square (WCSS) values on the y-axis corresponding to the different values of k (on the x-axis). The optimal k value is the point at which the graph forms an elbow.

5. Results

Using the elbow method, the optimal number of clusters to be set is six (Figure 2). Thus, a k-means clustering defining k = 6 is applied to the entire dataset. The centroids (duration, volume, MOR) of the randomly initialized clusters can be seen in Table 2. By combining these considerations on the distributions of the durations, volumes, and MOR, six eruptive classes can be defined (see also Table 2 and Table 3; Figure 3):
  • Cluster 1 is characterized by eruptions of short duration, low lava volume, and high MOR (Table 2 and Table 3; Figure 3a,b).
  • Cluster 2 is characterized by a very low MOR, high duration, and low lava volume (Table 2 and Table 3; Figure 3a).
  • Cluster 3 is characterized by low volume and low duration but medium to high MOR (Table 2 and Table 3; Figure 3a,b).
  • Cluster 4 is characterized by a high volume, high duration, and low MOR (Table 2 and Table 3; Figure 3a).
  • Cluster 5 is characterized by medium duration, high lava volume, and high MOR (Table 2 and Table 3; Figure 3a).
  • Cluster 6 is characterized by eruptions of short duration, low volume, and low MOR (Table 2 and Table 3; Figure 3a).
We also investigated the spatial distribution of these past effusive events. The boundary angles for each sector were determined using the zeros and minima of the fissure orientation distribution over the last 15,000 years [70]. Following the analysis by Duncan et al. [71], the sectors are defined as follows: Sector 1 spans from 355° to 115°, Sector 2 extends from 115° to 225°, and Sector 3 ranges from 225° to 355°. Thus, we divided the Etna area into three different sectors: the E-S, N-E, and W sectors (Figure 4):
  • Clusters 1 and 3 affect all three Etna sectors, with greater incidence in the E-S and N-E flanks of Etna (Figure 4a,c).
  • Cluster 2 includes events that occurred in the E-S sector, except for three episodes (Figure 4b).
  • Cluster 4 is characterized by eruptive events that occurred mostly in the E-S sector of Etna, except for one eruption in the N-NE sector and another one in the W sector (Figure 4d).
  • Cluster 5 includes events that occurred in the E-S and W sectors (Figure 4e).
  • Cluster 6 is characterized by two events located in the E-S sector and one eruption in the N-NE sector (Figure 4f).
In addition, the number of eruptive episodes within the six clusters across the three sectors was analyzed and is illustrated in Figure 5. This allows for the identification of the most recurrent clusters, which differ between the various sectors.

6. Discussion

The use of unsupervised learning enabled the classification of Etna’s flank eruptions over the last 400 years into six distinct clusters based on duration, lava volume, and MOR (Figure 3; Table 1 and Table 3). Clusters 1 and 3 share similarities in terms of short eruption durations (<35 days) and low lava volumes (<80 Mm3), but the inclusion of MOR and the clustering technique allowed us to distinguish between them (Figure 3b). Cluster 1 has a significantly higher MOR (2.8 Mm3/day) compared to Cluster 3 (0.5 Mm3/day), indicating different eruptive behaviors despite similar duration and volume. From a volcanological perspective, these clusters correspond to two types of activity defined by Branca and Del Carlo [37]. Cluster 1 is associated with Strombolian activity limited to the early stages of lava effusion, producing hornitos and small scoria cones, consistent with Class A eruptions. In contrast, Cluster 3 represents more explosive eruptions, often extending until the end of lava effusion, characteristic of Class B eruptions, which include vigorous strombolian and lava-fountaining activity.
The two eruptive classes reflect distinct mechanisms. Class A eruptions (Cluster 1) involve magmas that have undergone multi-stage decompression and volatile loss before reaching the surface [72,73]. Meanwhile, Class B eruptions (Cluster 3) are fed by magma rising from deeper levels, retaining higher volatile content [73]. Spatially, Cluster 1 events are predominantly located in the N-NE sector, with a few occurrences in the E-S and W sectors (Figure 4a). In contrast, Cluster 3 events are distributed across the E-S and W sectors (Figure 4c), with notable examples such as the formation of Mt. De Fiore (1974 [74,75]) and Mt. Nuovo (1763 [75]) in the W sector and the 1874 eruption in the N sector [76].
Cluster 2, comprising 12 eruptions, is defined by intermediate durations (90–187 days) and volumes (30–80 Mm3). Most of these events occurred in the E-S sector, with three exceptions in the NE sector (Figure 4b). Cluster 4 includes longer-duration eruptions (173 to 530 days, up to 3650 days in one case) and large volumes (120 to 1070 Mm3). These eruptions are mostly concentrated in the E-S sector, with a few exceptions in the NE and W sectors (Figure 4d). Cluster 5 includes three notable eruptions with volumes between 100 and 120 Mm3 and durations of 80–90 days, with the exception of the 1669 eruption, which lasted 122 days and produced 600 Mm3 of lava. These eruptions are primarily located in the E-S sector, except for the 1610 eruption, which occurred in the W sector (Figure 4e [62]). Finally, Cluster 6 represents three short-duration events (6–9 days) with low volumes (7–10 Mm3), with the exception of the 1942 eruption [2], which lasted just one day and had a volume of 2 Mm3. Two of these events occurred in the E-S sector, while one (the October 2002 eruption) took place in the NE sector (Figure 4f).
The differences between Clusters 2, 4, 5, and 6 are explained by variations in the driving and lithostatic pressures, as well as the thermal evolution of the magma and host rock. Bruce and Huppert [77] proposed that during an eruption, the driving pressure decreases over time, reducing the flow rate. High-volume eruptions (>100 Mm3) tend to have longer durations (over 170 days) and lower average flow rates (under 6 m3/s), as eruptions falling into Clusters 4 and 5, indicating a steady discharge maintained by the balance between internal dike pressure and lithostatic pressure, with cooling eventually closing the dike [78]. Shorter eruptions (<45 days; Clusters 2 and 6) likely result from insufficient internal pressure, while short-lived but high-volume eruptions are driven by high initial pressures.
A sectoral analysis of the volcano (Figure 5), considering three radial sectors (N-NE, E-S, and W), revealed that the N-NE sector is primarily characterized by Cluster 1 events. Both Clusters 2 and 3 also occur here, though Cluster 5 is absent. The E-S sector includes all six clusters, with Cluster 3 being the most prevalent, followed by Clusters 2 and 1. In the W sector, fewer events have occurred, and only Clusters 1, 3, 4, and 5 are present. The concentration of eruptions in the N-NE and E-S sectors suggests that these areas, particularly the densely inhabited eastern flanks, are at greater risk and should be the focus of hazard assessments.

7. Conclusions

The application of the k-means unsupervised machine learning algorithm, combined with MOR, lava volume, and eruption duration data, has yielded new insights into the eruptive behavior of Etna over the last 400 years. This clustering method identified, for the first time, two groups with similar eruption durations and volumes but differing MOR values. Cluster 1 is associated with Strombolian activity, primarily occurring during the early stages of lava effusion, leading to the formation of hornitos and small scoria cones (Class A of Branca and Del Carlo [37]). In contrast, Cluster 3 represents more explosive eruptions (Class B of Branca and Del Carlo [37]). The remaining clusters (2, 4, 5, and 6) exhibit variations in duration and volume, which are related to differences in eruptive pressures and magma evolution. This demonstrates the capability of unsupervised machine learning techniques to discern nuanced patterns in volcanic data that may not be immediately evident through traditional methods.
Sectoral analysis of the volcano indicates that the N-NE sector is predominantly represented by Cluster 1, with Clusters 2 and 3 also present, while Cluster 5 is absent. The E-S sector contains all six clusters, with Cluster 3 being the most frequent, followed by Clusters 2 and 1. In contrast, the W sector has fewer events, comprising only Clusters 1, 3, 4, and 5. This analysis reveals that the N-SE and S flanks have been more volcanically active compared to the N-SW flank. Notably, the E-S sector is characterized by eruptions with low duration, low lava volume, and medium to high MOR (Clusters 1 and 3), setting it apart from other sectors. Over the past 400 years, flank eruptions have primarily been concentrated in the densely populated eastern flanks, indicating a higher hazard for these areas. Consequently, these regions should be prioritized in volcanic risk assessments.
Because volcanic hazard assessment is generally based on past eruptive behavior, assuming that previous activity is representative of future activity, the characterization of duration, volume distributions, and MOR, as well as the spatial distributions of these past lateral eruptions, are essential to define the input for numerical simulations. These simulations can then be used to create more accurate volcanic hazard maps. Thus, the identification of these clusters not only enhances our understanding of Etna’s eruptive history but also provides a foundation for improving hazard maps. These maps are crucial for long-term planning and risk mitigation, especially in highly populated areas on Etna’s eastern slopes.
Our findings emphasize the importance of incorporating historical volcanological parameters—such as MOR, volume, and duration—into hazard assessment frameworks, providing a more comprehensive perspective of volcanic risk.

Author Contributions

Conceptualization, A.B.M. and C.C.; methodology, C.C.; software, A.B.M. and C.C.; validation, S.B.; formal analysis, A.B.M., C.C. and A.L.S.; data curation, A.B.M., C.C. and A.L.S.; writing—original draft preparation, A.B.M. and C.C.; writing—review and editing, A.L.S., S.B. and C.D.N.; visualization, A.B.M.; supervision, C.D.N.; project administration, C.D.N.; funding acquisition, C.D.N. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by the ATHOS Research Programme (INGV OB.FU. 0867.010), by the 2019 Strategic Project FIRST—Forecasting Eruptive activity at Stromboli volcano: timing, eruptive style, size, intensity, and duration—of the INGV Volcanoes Department (Delibera n. 144/2020), and by Project INGV Pianeta Dinamico VT_ORME 2023–2025 (INGV OB.FU. 1020.010).

Data Availability Statement

All data used in this article are available in the tables in this article.

Acknowledgments

This work was developed within the framework of the Laboratory of Technologies for Volcanology (TechnoLab) at the INGV in Catania (Italy).

Conflicts of Interest

The authors declare no conflicts of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

References

  1. Chester, D.K.; Duncan, A.M.; Dibben, C.; Guest, J.E.; Lister, P.H. Mascali, Mount Etna region Sicily: An example of fascist planning during the 1928 eruption and its continuing legacy. Nat. Hazards 1999, 19, 29–46. [Google Scholar] [CrossRef]
  2. Branca, S.; Del Carlo, P. Eruptions of Mt Etna during the past 3200 years: A revised compilation integrating the Historical and stratigraphic records. In Mt. Etna: Volcano Laboratory; American Geophysical Union: Washington, DC, USA, 2004; pp. 1–27. [Google Scholar]
  3. Bonaccorso, A.; Calvari, S.; Boschi, E. Hazard mitigation and crisis management during major flank eruptions at Etna volcano: Reporting on real experience. Geol. Soc. Lond. Spec. Publ. 2016, 426, 447–461. [Google Scholar] [CrossRef]
  4. Wadge, G.; Guest, J.E. Steady-state magma discharge at Etna 1971–1981. Nature 1981, 294, 548–550. [Google Scholar] [CrossRef]
  5. Smethurst, L.; James, M.R.; Pinkerton, H.; Tawn, J.A. A statistical analysis of eruptive activity on Mount Etna, Sicily. Geophys. J. Int. 2009, 179, 655–666. [Google Scholar] [CrossRef]
  6. Harris, A.; Steffke, A.; Calvari, S.; Spampinato, L. Thirty years of satellite-derived lava discharge rates at Etna: Implications for steady volumetric output. J. Geophys. Res. Solid Earth 2011, 116, B08204. [Google Scholar] [CrossRef]
  7. Andronico, D.; Lodato, L. Effusive activity at Mount Etna volcano (Italy) during the 20th century: A contribution to volcanic hazard assessment. Nat. Hazards 2005, 36, 407–443. [Google Scholar] [CrossRef]
  8. Bonaccorso, A.; Calvari, S. Major effusive eruptions and recent lava fountains: Balance between expected and erupted magma volumes at Etna volcano. Geophys. Res. Lett. 2013, 40, 6069–6073. [Google Scholar] [CrossRef]
  9. Currenti, G.; Del Negro, C.; Lapenna, V.; Telesca, L. Multifractality in local geomagnetic field at Etna volcano, Sicily (southern Italy). Nat. Hazards Earth Syst. Sci. 2005, 5, 555–559. [Google Scholar] [CrossRef]
  10. Cappello, A.; Neri, M.; Acocella, V.; Gallo, G.; Vicari, A.; Del Negro, C. Spatial vent opening probability map of Etna volcano (Sicily, Italy). Bull. Volcanol. 2012, 74, 2083–2094. [Google Scholar] [CrossRef]
  11. Cappello, A.; Bilotta, G.; Neri, M.; Negro, C.D. Probabilistic modeling of future volcanic eruptions at Mount Etna. J. Geophys. Res. Solid Earth 2013, 118, 1925–1935. [Google Scholar] [CrossRef]
  12. Del Negro, C.; Cappello, A.; Neri, M.; Bilotta, G.; Hérault, A.; Ganci, G. Lava flow hazards at Mount Etna: Constraints imposed by eruptive history and numerical simulations. Sci. Rep. 2013, 3, 3493. [Google Scholar] [CrossRef]
  13. Del Negro, C.; Cappello, A.; Ganci, G. Quantifying lava flow hazards in response to effusive eruption. Bulletin 2016, 128, 752–763. [Google Scholar] [CrossRef]
  14. Del Negro, C.; Cappello, A.; Bilotta, G.; Ganci, G.; Hérault, A.; Zago, V. Living at the edge of an active volcano: Risk from lava flows on Mt. Etna. Geol. Soc. Am. Bull. 2020, 132, 1615–1625. [Google Scholar] [CrossRef]
  15. Ebmeier, S.; Biggs, J.; Poland, M.; Pritchard, M.; Zoffoli, S.; Furtney, M.; Reath, K. Satellite geodesy for volcano monitoring in the Sentinel-1 and SAR constellation era. In Proceedings of the IGARSS 2019–2019 IEEE International Geoscience and Remote Sensing Symposium, Yokohama, Japan, 28 July–2 August 2019; pp. 5465–5467. [Google Scholar]
  16. Ramis, R.; Garcia, A.; Marrero, J.; la Cruz-Reyna, D.; Carniel, R.; Vila, J. Volcanic and volcano-tectonic activity forecasting: A review on seismic approaches. Ann. Geophys. 2019, 62. [Google Scholar] [CrossRef]
  17. Ren, C.X.; Peltier, A.; Ferrazzini, V.; Rouet-Leduc, B.; Johnson, P.A.; Brenguier, F. Machine learning reveals the seismic signature of eruptive behavior at Piton de la Fournaise volcano. Geophys. Res. Lett. 2020, 47, e2019GL085523. [Google Scholar] [CrossRef] [PubMed]
  18. Carniel, R.; Guzman, S.R. Machine learning in volcanology: A review. In Volcanoes—Updates in Volcanology; IntechOpen: Rijeka, Croatia, 2021. [Google Scholar]
  19. Amato, E.; Corradino, C.; Torrisi, F.; Del Negro, C. A Deep convolutional neural network for detecting volcanic thermal anomalies from satellite images. Remote Sens. 2023, 15, 3718. [Google Scholar] [CrossRef]
  20. Ramsey, M.S.; Corradino, C.; Thompson, J.O.; Leggett, T.N. Statistical retrieval of volcanic activity in long time series orbital data: Implications for forecasting future activity. Remote Sens. Environ. 2023, 295, 113704. [Google Scholar] [CrossRef]
  21. Corradino, C.; Malaguti, A.B.; Ramsey, M.S.; Del Negro, C. Quantitative Assessment of Volcanic Thermal Activity from Space Using an Isolation Forest Machine Learning Algorithm. Remote Sens. 2024, 16, 2001. [Google Scholar] [CrossRef]
  22. Marzban, P.; Bredemeyer, S.; Walter, T.R.; Kästner, F.; Müller, D.; Chabrillat, S. Hydrothermally altered deposits of 2014 Askja landslide, Iceland, identified by remote sensing imaging. Front. Earth Sci. 2023, 11, 1083043. [Google Scholar] [CrossRef]
  23. Anzieta, J.C. Application of Data Analysis and Machine Learning Techniques to Improve Baseline Volcano and Mountain Hazards Monitoring. Ph.D. Thesis, Simon Fraser University, Burnaby, BC, Canada, 2024. [Google Scholar]
  24. Andaru, R.; Rau, J.Y.; Syahbana, D.K.; Prayoga, A.S.; Purnamasari, H.D. The use of UAV remote sensing for observing lava dome emplacement and areas of potential lahar hazards: An example from the 2017–2019 eruption crisis at Mount Agung in Bali. J. Volcanol. Geotherm. Res. 2021, 415, 107255. [Google Scholar] [CrossRef]
  25. Anantrasirichai, N.; Biggs, J.; Albino, F.; Bull, D. A deep learning approach to detecting volcano deformation from satellite imagery using synthetic datasets. Remote Sens. Environ. 2019, 230, 111179. [Google Scholar] [CrossRef]
  26. Biggs, J.; Anantrasirichai, N.; Albino, F.; Lazecky, M.; Maghsoudi, Y. Large-scale demonstration of machine learning for the detection of volcanic deformation in Sentinel-1 satellite imagery. Bull. Volcanol. 2022, 84, 100. [Google Scholar] [CrossRef] [PubMed]
  27. Hajian, A.; Cannavò, F.; Greco, F.; Nunnari, G. Classification of Mt Etna (Italy) volcanic activity by machine learning approaches. Ann. Geophys. 2019, 62, VO231. [Google Scholar] [CrossRef]
  28. Manley, G.F.; Mather, T.A.; Pyle, D.M.; Clifton, D.A.; Rodgers, M.; Thompson, G.; Londono, J.M. A deep active learning approach to the automatic classification of volcano-seismic events. Front. Earth Sci. 2022, 10, 807926. [Google Scholar] [CrossRef]
  29. Falcin, A.; Métaxian, J.P.; Mars, J.; Stutzmann, É.; Komorowski, J.C.; Moretti, R.; Malfante, M.; Beauducel, F.; Saurel, J.-M.; Dessert, C.; et al. A machine-learning approach for automatic classification of volcanic seismicity at La Soufrière Volcano, Guadeloupe. J. Volcanol. Geotherm. Res. 2021, 411, 107151. [Google Scholar] [CrossRef]
  30. Messina, A.; Langer, H. Pattern recognition of volcanic tremor data on Mt. Etna (Italy) with KKAnalysis—A software program for unsupervised classification. Comput. Geosci. 2011, 37, 953–961. [Google Scholar] [CrossRef]
  31. Witsil, A.J.; Johnson, J.B. Analyzing continuous infrasound from Stromboli volcano, Italy using unsupervised machine learning. Comput. Geosci. 2020, 140, 104494. [Google Scholar] [CrossRef]
  32. Gaddes, M.E.; Hooper, A.; Bagnardi, M. Using machine learning to automatically detect volcanic unrest in a time series of interferograms. J. Geophys. Res. Solid Earth 2019, 124, 12304–12322. [Google Scholar] [CrossRef]
  33. Romano, R. Succession of the volcanic activity in the Etnean area. In Memorie della Società Geologica Italiana; Società geologica italiana: Rome, Italy, 1982; Volume 23, pp. 27–48. [Google Scholar]
  34. Chester, D.K.; Duncan, A.M.; Guest, J.E.; Kilburn, C.R.J. Mount Etna—The Anatomy of a Volcano; Chapman & Hall: London, UK, 1985. [Google Scholar]
  35. Rittman, A. Mount Etna and the 1971 eruption—Structure and evolution of Mount Etna. Philos. Trans. R. Soc. Lond. Ser. A Math. Phys. Sci. 1973, 274, 5–16. [Google Scholar]
  36. Guest, J.E. Styles of eruption and flow morphology on Mt Etna. In Memorie della Società Geologica Italiana; Società Geologica Italiana: Rome, Italy, 1982; Volume 23, pp. 49–73. [Google Scholar]
  37. Branca, S.; Del Carlo, P. Types of eruptions of Etna volcano AD 1670–2003: Implications for short-term eruptive behaviour. Bull. Volcanol. 2005, 67, 732–742. [Google Scholar] [CrossRef]
  38. Andronico, D.; Cannata, A.; Di Grazia, G.; Ferrari, F. The 1986–2021 paroxysmal episodes at the summit craters of Mt. Etna: Insights into volcano dynamics and hazard. Earth-Sci. Rev. 2021, 220, 103686. [Google Scholar] [CrossRef]
  39. Coltelli, M.; Pompilio, M.; Del Carlo, P.; Calvari, S.; Pannucci, S.; Scribano, V. Eruptive activity. Acta Vulcanol. 1998, 10, 141–148. [Google Scholar]
  40. Coltelli, M.; Del Carlo, P.; Pompilio, M. Eruptive activity. Acta Vulcanol. 2000, 12, 63–67. [Google Scholar]
  41. Calvari, S.; Neri, M.; Pinkerton, H. Effusion rate estimations during the 1999 summit eruption on Mount Etna, and growth of two distinct lava flow fields. J. Volcanol. Geotherm. Res. 2003, 119, 107–123. [Google Scholar] [CrossRef]
  42. Alparone, S.; Andronico, D.; Lodato, L.; Sgroi, T. Relationship between tremor and volcanic activity during the Southeast Crater eruption on Mount Etna in early 2000. J. Geophys. Res. Solid Earth 2003, 108, 2241. [Google Scholar] [CrossRef]
  43. Tanguy, J.C. Les éruptions historiques de l’Etna: Chronologie et localisation. Bull. Volcanol. 1981, 44, 585–640. [Google Scholar] [CrossRef]
  44. Salvi, F.; Scandone, R.; Palma, C. Statistical analysis of the historical activity of Mount Etna, aimed at the evaluation of volcanic hazard. J. Volcanol. Geotherm. Res. 2006, 154, 159–168. [Google Scholar] [CrossRef]
  45. Acocella, V.; Neri, M. What makes flank eruptions? The 2001 Etna eruption and its possible triggering mechanisms. Bull. Volcanol. 2003, 65, 517–529. [Google Scholar] [CrossRef]
  46. Branca, S.; De Beni, E.; Proietti, C. The large and destructive 1669 AD eruption at Etna volcano: Reconstruction of the lava flow field evolution and effusion rate trend. Bull. Volcanol. 2013, 75, 1–16. [Google Scholar] [CrossRef]
  47. Branca, S.; De Beni, E.; Chester, D.; Duncan, A.; Lotteri, A. The 1928 eruption of Mount Etna (Italy): Reconstructing lava flow evolution and the destruction and recovery of the town of Mascali. J. Volcanol. Geotherm. Res. 2017, 335, 54–70. [Google Scholar] [CrossRef]
  48. Barberi, F.; Carapezza, M.L.; Valenza, M.; Villari, L. The control of lava flow during the 1991–1992 eruption of Mt. Etna. J. Volcanol. Geotherm. Res. 1993, 56, 1–34. [Google Scholar] [CrossRef]
  49. Allard, P.; Behncke, B.; d’Amico, S.; Neri, M.; Gambino, S. Mount Etna 1993–2005: Anatomy of an evolving eruptive cycle. Earth-Sci. Rev. 2006, 78, 85–114. [Google Scholar] [CrossRef]
  50. Coltelli, M.; Marsella, M.; Proietti, C.; Scifoni, S. The case of the 1981 eruption of Mount Etna: An example of very fast moving lava flows. Geochem. Geophys. Geosystems 2012, 13. [Google Scholar] [CrossRef]
  51. Branca, S.; Privitera, F.; Palio, O.; Turco, M. Prehistoric human presence on Mount Etna (Sicily), in relation to the geological evolution. Ann. Geophys. 2021, 64, VO542. [Google Scholar]
  52. Acocella, V.; Behncke, B.; Neri, M.; D’Amico, S. Link between major flank slip and 2002–2003 eruption at Mt. Etna (Italy). Geophys. Res. Lett. 2003, 30. [Google Scholar] [CrossRef]
  53. Behncke, B.; Neri, M. The July–August 2001 eruption of Mt. Etna (Sicily). Bull. Volcanol. 2003, 65, 461–476. [Google Scholar] [CrossRef]
  54. Clocchiatti, R.; Condomines, M.; Guénot, N.; Tanguy, J.C. Magma changes at Mount Etna: The 2001 and 2002–2003 eruptions. Earth Planet. Sci. Lett. 2004, 226, 397–414. [Google Scholar] [CrossRef]
  55. Barberi, F.; Brondi, F.; Carapezza, M.L.; Cavarra, L.; Murgia, C. Earthen barriers to control lava flows in the 2001 eruption of Mt. Etna. J. Volcanol. Geotherm. Res. 2003, 123, 231–243. [Google Scholar] [CrossRef]
  56. Neri, M.; Acocella, V.; Behncke, B.; Maiolino, V.; Ursino, A.; Velardita, R. Contrasting triggering mechanisms of the 2001 and 2002–2003 eruptions of Mount Etna (Italy). J. Volcanol. Geotherm. Res. 2005, 144, 235–255. [Google Scholar] [CrossRef]
  57. Tanguy, J.C.; Condomines, M.; Le Goff, M.; Chillemi, V.; La Delfa, S.; Patanè, G. Mount Etna eruptions of the last 2750 years: Revised chronology and location through archeomagnetic and 226 Ra-230 Th dating. Bull. Volcanol. 2007, 70, 55–83. [Google Scholar] [CrossRef]
  58. Branca, S.; Vigliotti, L. Finding of an historical document describing an eruption in the NW flank of Etna in July 1643 AD: Timing, location and volcanic products. Bull. Volcanol. 2015, 77, 95. [Google Scholar] [CrossRef]
  59. Branca, S.; Abate, T. Current knowledge of Etna’s flank eruptions (Italy) occurring over the past 2500 years. From the iconographies of the XVII century to modern geological cartography. J. Volcanol. Geotherm. Res. 2019, 385, 159–178. [Google Scholar] [CrossRef]
  60. De Beni, E.; Cantarero, M.; Neri, M.; Messina, A. Lava flows of Mt Etna, Italy: The 2019 eruption within the context of the last two decades (1999–2019). J. Maps 2021, 17, 65–76. [Google Scholar] [CrossRef]
  61. Tarquini, S.; Nannipieri, L. The 10 m-resolution TINITALY DEM as a trans-disciplinary basis for the analysis of the Italian territory: Current trends and new perspectives. Geomorphology 2017, 281, 108–115. [Google Scholar] [CrossRef]
  62. Proietti, C.; De Beni, E.; Coltelli, M.; Branca, S. The flank eruption history of Etna (1610–2006) as a constraint on lava flow hazard. Ann. Geophys. 2011, 54, 480–490. [Google Scholar] [CrossRef]
  63. Corradino, C.; Amato, E.; Torrisi, F.; Calvari, S.; Del Negro, C. Classifying major explosions and paroxysms at Stromboli volcano (Italy) from space. Remote Sens. 2021, 13, 4080. [Google Scholar] [CrossRef]
  64. Anzieta, J.C.; Ortiz, H.D.; Arias, G.L.; Ruiz, M.C. Finding possible precursors for the 2015 Cotopaxi volcano eruption using unsupervised machine learning techniques. Int. J. Geophys. 2019, 2019, 6526898. [Google Scholar] [CrossRef]
  65. Syakur, M.A.; Khotimah, B.K.; Rochman, E.M.S.; Satoto, B.D. Integration k-means clustering method and elbow method for identification of the best customer profile cluster. In IOP Conference Series: Materials Science and Engineering; IOP Publishing: Bristol, UK, 2018; Volume 336, p. 012017. [Google Scholar]
  66. El-Mandouh, A.M.; Abd-Elmegid, L.A.; Mahmoud, H.A.; Haggag, M.H. Optimized K-means clustering model based on gap statistic. Int. J. Adv. Comput. Sci. Appl. 2019, 10. [Google Scholar] [CrossRef]
  67. Hasibuan, A.; Kembuan, D.R.; Manoppo, C.T.M.; Tinambunan, M.H. Optimization of K-Means algorithm in grouping data using the statistical gap method. J. Intell. Decis. Support Syst. 2023, 6, 112–120. [Google Scholar]
  68. Kaufman, L.; Rousseeuw, P.J. Finding Groups in Data: An Introduction to Cluster Analysis; John Wiley & Sons: Hoboken, NJ, USA, 2009. [Google Scholar]
  69. Mato, F.; Toulkeridis, T. An unsupervised K-means based clustering method for geophysical post-earthquake diagnosis. In Proceedings of the 2017 IEEE Symposium Series on Computational Intelligence (SSCI), Honolulu, HI, USA, 27 November–1 December 2017; pp. 1–8. [Google Scholar]
  70. Azzaro, R.; Branca, S.; Gwinner, K.; Coltelli, M. The volcano-tectonic map of Etna volcano, 1:100,000 scale: An integrated approach based on a morphotectonic analysis from high-resolution DEM constrained by geologic, active faulting and seismotectonic data. Ital. J. Geosci. 2012, 131, 153–170. [Google Scholar]
  71. Duncan, A.M.; Chester, D.K.; Guest, J.E. Mount Etna volcano: Environmental impact and problems of volcanic prediction. Geogr. J. 1981, 147, 164–178. [Google Scholar] [CrossRef]
  72. Armienti, P.; Innocenti, F.; Petrini, R.; Pompilio, M.; Villari, L. Sub-aphiric alkali basalts from Mt. Etna: Inferences on the depth and composition of the source magma. Rend. Soc. Ital. Mineral. Petrol. 1988, 43, 877–891. [Google Scholar]
  73. Corsaro, R.A.; Pompilio, M. Dynamics of magmas at Mount Etna. Geophys. Monogr.-Am. Geophys. Union 2004, 143, 91–110. [Google Scholar]
  74. Lormand, C.; Harris, A.J.; Chevrel, M.O.; Calvari, S.; Gurioli, L.; Favalli, M.; Fornaciai, A.; Nannipieri, L. The 1974 west flank eruption of Mount Etna: A data-driven model for a low elevation effusive event. Front. Earth Sci. 2020, 8, 590411. [Google Scholar] [CrossRef]
  75. Corsaro, R.A.; Branca, S.; De Beni, E.; Tanguy, J.C. Tales from Three 18th Century Eruptions to Understand Past and Present Behaviour of Etna. Front. Earth Sci. 2021, 9, 774361. [Google Scholar] [CrossRef]
  76. Silvestri, O. XVI. The eruption of etna on the 29th of August 1875. Lond. Edinb. Dublin Philos. Mag. J. Sci. 1874, 49, 126–134. [Google Scholar] [CrossRef]
  77. Bruce, P.M.; Huppert, H.E. Solidification and melting along dykes by the laminar flow of basaltic magma. In Magma Transport and Storage; John Wiley & Sons: Hoboken, NJ, USA, 1990; pp. 87–101. [Google Scholar]
  78. Wadge, G. The variation of magma discharge during basaltic eruptions. J. Volcanol. Geotherm. Res. 1981, 11, 139–168. [Google Scholar] [CrossRef]
Geosciences 14 00295 g002
Figure 2. Elbow graph showing the within-cluster sum-of-square (WCSS) values on the y-axis corresponding to the different values of K (on the x-axis).
Figure 2. Elbow graph showing the within-cluster sum-of-square (WCSS) values on the y-axis corresponding to the different values of K (on the x-axis).
Geosciences 14 00295 g002
Geosciences 14 00295 g003
Figure 3. (a) Distribution of durations, MOR, and volumes of the 60 lava flows and their subdivision into six clusters. (b) Detail of clusters 1, 3, and 6.
Figure 3. (a) Distribution of durations, MOR, and volumes of the 60 lava flows and their subdivision into six clusters. (b) Detail of clusters 1, 3, and 6.
Geosciences 14 00295 g003
Geosciences 14 00295 g004
Figure 4. Sketch map of Etna flank lava flows over the last 400 years from historical data [2,38,57,58,59,60], subdivided according to the six clusters identified. The shaded relief image was derived from the 10 m resolution TINITALY DEM [61]. (a) Cluster 1 events are mainly located in the N-NE sector, with some occurrences in the E-S and W sectors; (b) Most of the Cluster 2 events occurred in the E-S sector, with some exceptions in the NE sector; (c) Cluster 3 is distributed across the E-S and W sectors; (d) Cluster 4 eruptions are mostly concentrated in the E-S sector, with a few exceptions in the NE and W sectors; (e) Cluster 5 includes eruptions located in the E-S sector, except for one event in the W sector; (f) Cluster 6 includes events located in the E-S and NE sectors.
Figure 4. Sketch map of Etna flank lava flows over the last 400 years from historical data [2,38,57,58,59,60], subdivided according to the six clusters identified. The shaded relief image was derived from the 10 m resolution TINITALY DEM [61]. (a) Cluster 1 events are mainly located in the N-NE sector, with some occurrences in the E-S and W sectors; (b) Most of the Cluster 2 events occurred in the E-S sector, with some exceptions in the NE sector; (c) Cluster 3 is distributed across the E-S and W sectors; (d) Cluster 4 eruptions are mostly concentrated in the E-S sector, with a few exceptions in the NE and W sectors; (e) Cluster 5 includes eruptions located in the E-S sector, except for one event in the W sector; (f) Cluster 6 includes events located in the E-S and NE sectors.
Geosciences 14 00295 g004
Geosciences 14 00295 g005
Figure 5. Distributions of the six clusters in the three sectors into which Etna was divided.
Figure 5. Distributions of the six clusters in the three sectors into which Etna was divided.
Geosciences 14 00295 g005
Table 1. Summary of the main flank eruptions from 1669 to 2019 from historical catalogs [2,38,57,58,59,60].
Table 1. Summary of the main flank eruptions from 1669 to 2019 from historical catalogs [2,38,57,58,59,60].
Starting YearDuration (Days)Volume (Mm3)MOR (Mm3/day)
30 May 201984.40.55
24 December 201824.252.12
4 September 2008419770.18
7 September 2004182400.21
27 October 200293300.32
27 October 20028101.25
17 July 200123200.86
14 December 19914732350.49
27 October 198910262.6
30 October 1986122600.49
25 December 198530.80.26
12 March 1985125300.24
28 March 1983131790.60
17 March 19816223.66
3 August 197967.51.25
18 November 197812110.91
24 August 1978640.66
29 April 197837270.72
24 February 197518760.03
11 March 1974182.10.11
30 January 1974172.40.14
5 April 197169450.65
7 January 196811710.008
25 November 19503721510.40
2 December 19493124
30 June 194211.61.6
2 November 192818522.88
17 June 192331782.51
30 November 191821.20.6
10 September 191113554.23
23 March 191026652.5
29 April 1908122
9 July 18921731210.69
19 May 188620381.9
22 March 188330.20.06
26 May 187912221.83
29 August 187421.50.75
30 January 1865150300.2
20 August 1852280870.31
17 November 184311524.72
31 October 183223502.17
25 May 181970470.67
27 October 1811182510.28
27 March 180914362.57
15 November 18023103.33
26 May 1792380900.23
18 May 178010292.9
27 April 17661941350.69
18 June 1763841001.19
6 February 176332190.59
9 March 175564.70.78
8 March 170260170.28
11 March 16691226004.91
14 March 168930200.66
18 July 16431040.4
January 1651–December 165410954750.43
20 November 1646–1647581532.63
19 December 1634–June 16365302030.38
1 June 1614–1624365010700.29
3 May 1610-July 1610901201.33
Table 2. Centroids for each cluster.
Table 2. Centroids for each cluster.
ClustersDurationVolumeMOR
K115.341.52.8
K2197.348.40.3
K322.6 12.85 0.5
K4926.7341.40.5
K598.7273.32.5
K656.41.4
Table 3. Cluster subdivision of Etna’s lateral eruptions.
Table 3. Cluster subdivision of Etna’s lateral eruptions.
Starting YearClustersStarting YearClusters
24 December 2018117 July 20013
27 October 1989125 December 19853
17 March 1981118 November 19783
2 December 1949124 August 19783
2 November 1928129 April 19783
17 June 1923111 March 19743
10 September 1911130 January 19743
23 March 191015 April 19713
29 April 1908130 November 19183
19 May 1886122 March 18833
26 May 1879129 August 18743
17 November 1843125 May 18193
31 October 183216 February 17633
27 March 180919 March 17553
15 November 180218 March 17023
18 May 1780114 March 16893
20 November 1646–1647118 July 16433
4 September 2008214 December 19914
7 September 2004225 November 19504
27 October 200229 July 18924
30 October 1986227 April 17664
12 March 19852January 1651–December 16544
28 March 1983219 December 1634–June 16364
24 February 197521 June 1614–16244
7 January 1968218 June 17635
30 January 1865211 March 16695
20 August 185223 May 1610-July5
27 October 1811227 October 20026
26 May 179223 August 19796
30 May 2019330 June 19426
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Malaguti, A.B.; Corradino, C.; La Spina, A.; Branca, S.; Del Negro, C. Machine Learning Insights into the Last 400 Years of Etna Lateral Eruptions from Historical Volcanological Data. Geosciences 2024, 14, 295. https://doi.org/10.3390/geosciences14110295

AMA Style

Malaguti AB, Corradino C, La Spina A, Branca S, Del Negro C. Machine Learning Insights into the Last 400 Years of Etna Lateral Eruptions from Historical Volcanological Data. Geosciences. 2024; 14(11):295. https://doi.org/10.3390/geosciences14110295

Chicago/Turabian Style

Malaguti, Arianna Beatrice, Claudia Corradino, Alessandro La Spina, Stefano Branca, and Ciro Del Negro. 2024. "Machine Learning Insights into the Last 400 Years of Etna Lateral Eruptions from Historical Volcanological Data" Geosciences 14, no. 11: 295. https://doi.org/10.3390/geosciences14110295

APA Style

Malaguti, A. B., Corradino, C., La Spina, A., Branca, S., & Del Negro, C. (2024). Machine Learning Insights into the Last 400 Years of Etna Lateral Eruptions from Historical Volcanological Data. Geosciences, 14(11), 295. https://doi.org/10.3390/geosciences14110295

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop