A New Catalogue of Historical Eruptions in Santorini Volcano: Documentation and Completeness Analysis

Abstract

The Santorini volcano attracts global interest. The 17th century BCE Minoan eruption was likely the largest ever occurring in the Holocene. The evaluation of an enriched collection of documentary sources combined with geological observations showed that in historical times, 15 eruption episodes were recorded from the 2nd century BCE up to 1950 CE. Little-known episodes occurring in 1667 CE and 1773 CE were uncovered, analyzed and included in the catalogue. Due to many uncertainties and inconsistencies involved in the historical sources, a reliability score ranging from 1 (lowest) to 4 (highest) has been assigned to each one of the 15 episodes. This procedure was based on a Reliability Assessment Matrix, which is a novelty in historical volcanology. The size of an eruption was evaluated in terms of the 8-grade Volcanic Explosivity Index (VEI) by introducing a bimodal assignment form of “V or V + 1”, an approach used for the first time, instead of the traditional unimodal assignment of “V”. The VEI of 4–5 was assigned to the eruptions of 725/726 CE in the Santorini caldera and of 1650 CE in the Kolumbo submarine volcano. The rest of the eruptions were assigned with smaller VEIs. Completeness analysis of the catalogue with statistical tests and Monte Carlo simulation with random and non-random models as baselines showed a 10-fold increase in the eruption record in 450 years after the breakpoint of 1572 CE eruption compared to the 1747 years before that point implying that there are about 30 missing eruption events that escaped record in the pre-1572 CE historical interval. This statistically significant difference is interpreted as a bias discovery effect of anthropogenic origin, but further research is needed to show that it does not represent a drastic change of magma plumbing rate.

1. Introduction

The Santorini volcano in the south Aegean Sea, Greece (Figure 1), attracts global geoscientific interest. The 17th century BCE eruption, known as a Minoan event, was likely the largest ever experienced in the Holocene [1]. Geological and archaeological evidence supports that the Minoan eruption was associated with significant tephra falls, earthquakes, and tsunamis drastically influencing eastern Mediterranean civilizations [1,2,3,4,5,6]. Comparative studies indicate great similarity between Santorini and Krakatoa volcano in Sunda Strait [5], but the Minoan eruption exceeded the 1883 CE Krakatoa eruption in size.

During historical times, several small-to-moderate eruptions with magma emplacement signified the initiation of a new volcanic cycle in Santorini. As such, the so-called Kamenae (Burned) islands were shaped inside the caldera [2,6], exactly as in the Krakatoa caldera with the post-1883 generation of the Anak (Child) island. After several eruptive episodes documented from the 2nd century BCE up to the first half of the 20th century, Santorini has remained dormant since 1950 CE. In February–March 2025, a cluster consisting of thousands of earthquakes, with maximum magnitude 5.3 and sources located at distances of 20–40 km to the east of Santorini Island, caused extensive social anxiety but no eruptive activity was noted, e.g., [7,8].

Santorini constitutes a top tourist destination because of its unique volcanic landscape. The annual number of visitors has been estimated to exceed three millions in the last years (https://greektriplanner.me/insights/santorini-tourism-statistics#how-many-tourists-visit-santorini; last access 27 May 2026). Therefore, the assessment and public communication of the volcanic and associated hazards including earthquakes, tsunamis and landslides, is a need not only for local people, but also for visitors arriving οn the island all year round. The recorded history of previous eruptions and associated phenomena is one of the most important components involved in hazard assessments. Risk scenarios and risk mitigation plans heavily rely on hazard evaluations. However, the eruption records throughout the entire historical period is likely incomplete and, in several cases, uncertain.

The purpose of the present study is twofold. The first is to compile a new reliable catalogue of the historical volcanic eruptions in Santorini, while the second is to examine the completeness of the catalogue. To this aim, the compilation of a rich collection of documentary sources has been evaluated along with relevant geological observations. Novelties in compiling the new catalogue are the inclusion of the two little-known episodes of 1667 CE and 1773 CE, the introduction for the first time in historical volcanology of a Reliability Assessment Matrix, as well as the revision of the Volcanic Explosivity Index (VEI) [9] assigned to each eruption episode on a bimodal form instead of the traditional unimodal form. The completeness of the eruption time series has been examined on the basis of firm statistical tests, simulations, and probability models.

Figure 1. Geodynamic setting of the broad Aegean Sea region. The Mediterranean or Nubian lithosphere moves from about SW to NE, and subducts beneath the Aegean Sea at the southern Eurasian plate margin along the Hellenic Trench system, e.g., [10]; PTT, PLT and ST represent the Ptolemy, Pliny and Strabo trenches, respectively. Arrows show the directions of lithospheric plate motions. There are five main volcanic centers (triangles) situated along the South Aegean Volcanic Arc above the seismic isodepth of ~150 km: Methana (Me), Milos (Mi), Thera or Santorini (Th), Nisyros (Ni) and Kolumbo (for the last, see position in Figure 2). The b ox illustrates the study area shown in Figure 2. To the northeast of Santorini, the 9 July 1956 large tectonic earthquake of M_w = 7.7 ruptured the submarine trough between the islands of Amorgos (Am) and Astypalaia (As).

Figure 1. Geodynamic setting of the broad Aegean Sea region. The Mediterranean or Nubian lithosphere moves from about SW to NE, and subducts beneath the Aegean Sea at the southern Eurasian plate margin along the Hellenic Trench system, e.g., [10]; PTT, PLT and ST represent the Ptolemy, Pliny and Strabo trenches, respectively. Arrows show the directions of lithospheric plate motions. There are five main volcanic centers (triangles) situated along the South Aegean Volcanic Arc above the seismic isodepth of ~150 km: Methana (Me), Milos (Mi), Thera or Santorini (Th), Nisyros (Ni) and Kolumbo (for the last, see position in Figure 2). The b ox illustrates the study area shown in Figure 2. To the northeast of Santorini, the 9 July 1956 large tectonic earthquake of M_w = 7.7 ruptured the submarine trough between the islands of Amorgos (Am) and Astypalaia (As).

Figure 2. Volcanotectonic sketch map of Santorini volcano (redrafted from [11]). Dashed lines show tectonic lineaments and arrows illustrate sense of motion.

Figure 2. Volcanotectonic sketch map of Santorini volcano (redrafted from [11]). Dashed lines show tectonic lineaments and arrows illustrate sense of motion.

To better understand the phenomena described in the documentary and scientific sources, the volcanic and geographical setting of the area is first reviewed briefly.

2. Volcanic and Geographical Setting

In the Late Bronze Age era, a giant caldera-forming eruption of Plinian-type occurred in Santorini volcano at around 1613 BCE [2]. The size of the so called Minoan eruption was estimated as high as at least VEI = 7 [1] at an 8-grade scale [9]. The Santorini caldera formed after the Minoan event is of elliptical shape with diameter 11 km in the N–S direction and 7.5 km in the W–E direction [6] (Figure 2). The three islands of Santorini (or Thera), Therassia and Aspronisi, that remained after the volcanic cone collapse, surround the caldera as a ring (Figure 2). The caldera cliffs are quite steep and rise to a height of up to ~300 m above sea level, although the elevation varies across the island’s edge. Tephra layers that overtop the three islands consist mainly by pumice with thickness up to ~30 m.

Historical documentary sources in previous studies [2,6,10], indicated that in the caldera volcanic episodes occurred in 197 BCE, 19 CE, 46, 726, 1457 or 1458, 1572, 1707–1711 and 1866–1870. During the 20th century, the eruptions of 1925–1926, 1928, 1939–1941 and 1950 have been very well documented [2,6].

Historical eruptions generated several new volcanic islands in the caldera. The islands existing today are the Palaia (Old) Kameni and Nea (New) Kameni. Palaia Kameni, likely generated by eruptions preceding the 1572 CE one, was unified with an earlier islet named Hiera (Figure 3) that probably was formed by the 197 BCE eruption. During the 1572 CE eruption lava effusions shaped the island of Mikri (Lesser) Kameni, being of ~600 m in length and of ~250 m in width in the N–S and E–W directions, respectively [11]. The eruptive activity of 1707–1711 CE generated the island of Megali (Large) Kameni between Palaia Kameni and Mikri Kameni. Megali Kameni grew up during post-1711 eruptions and eventually unified with the Mikri Kameni. This way, the Nea Kameni was created, which, today, constitutes the largest island existing in the caldera. A map constructed in the field and published by Olivier [12], a French diplomat who visited the Greek islands, nicely illustrates the islands existing in the caldera by the late 18th century (Figure 3). This map is consistent with the one published by Comte de Choisel-Gouffier [13], who also visited Santorini by the end of the 18th century. The last map was recently republished by Papadopoulos [11].

In 1650 CE a powerful eruption, accompanied by earthquakes and tsunamis [2,6,14,15], occurred at the Kolumbo polygenetic submarine volcano, situated about 15 km from the Nea Kameni to the NE direction (Figure 2). This has been the only known historical eruption at Kolumbo.

From the geographical point of view, of special interest to our investigation are the historical settlements of Apanomeria, modern Oia, which is situated at the northern side of Santorini Island, as well as of the medieval town in the Castle of Skaros at the south of Apanomeria (Figure 3).

3. Research Methods and Information Sources

3.1. Research Methods

The post-Minoan eruptive history in the Santorini island complex is documented by a variety of classical texts, books, articles, letters, travelogues, maps, newspaper reports and eyewitnesse accounts referring to the eruptions that occurred before the 20th century. Collections and quotations of such documentary sources can be found in several publications of modern authors [2,6,11,15,16,17,18,19]. The examination also included little-known documents referring to the episodes of 1667 CE and 1773 CE, which have not been considered so far in the volcanological tradition. The eruptions occurring during the 20th century are very well documented by a series of scientific publications and, therefore, they are briefly reviewed.

3.1.1. Reliability of Sources

Of particular importance is the reliability of the documentary sources because contradictory information is frequently provided by different sources referring to a single eruptive episode. In addition, some sources cast doubts as to whether a reported episode was a real eruption or not. On the other hand, one should carefully discriminate between original documents based on eyewitnesses’ accounts and second-hand sources produced by authors who recollected material from previous sources. Important lessons about this issue have been learnt from the study of historical earthquakes, e.g., [20]. In the last years, our knowledge about some historical eruptions in Santorini were complemented by geological observations, which are helpful in better understanding at least two eruption episodes [21,22]. A field inspection performed by the author in Palaia Kameni proved helpful for the examination of the questionable eruption of 1457 CE.

Based on the evaluation of documentary and scientific sources, a supposed eruptive episode was characterized by a reliability score, R, as if it was a real eruption or not. The assignment of R has been done on a 4-degree reliability scale applied in the past for the reliability characterization of historical earthquakes and tsunamis in Japan and in the Mediterranean region [23,24,25,26]. Although the reliability assignment to an eruption is susceptible to subjectivity it has been done by following a couple of general rules. The first is that a set of more than one documentary sources, which are consistent but independent each other and referring to a single eruption, provides higher reliability than a single source. Also, the consistency of a source with the volcanological, chronological and geographical contexts of the area increase reliability. Based on these rules a Reliability Assessment Matrix was constructed, which considers three main attributes for the reliability characterization of an eruption: number of sources, credibility of sources, and data or information consistency (

Table 1. Reliability Assessment Matrix.

Reliability Score (R)	Number of Sources	Source Credibility	Data Consistency
4 Definite	n ≥ 2	Documentary and/or scientific sources are trustworthy.	Confirmed by independent sources. No evidence of error.
3 Probable	n ≥ 1	Reliable source(s); some minor doubt possible.	Consistent with other data; not yet fully verified. Minor possible discrepancies.
2 Questionable	n = 1	Questionable source with possible errors.	Inconsistent; contradicts other known data.
1 Improbable	n = 1	Untrustworthy source; high risk of misinformation.	No corroboration; highly improbable data.

). The reliability scoring of an eruption may range from improbable (R = 1) to questionable (R = 2), probable (R = 3) and definite (R = 4).

The five eruptions from 1866 up to 1950 are the best studied based on measurements and field observations performed by professionals. These eruptions have been assigned with reliability 4. This set of eruptions has been used as a useful benchmark to assign R values to the pre-1866 eruptions in the frame of the Reliability Assessment Matrix and the respective information available.

3.1.2. Assignment of VEI

Assignment of VEI to an eruption according to the 8-grade scale [9] is based on the ejecta volume, plume height and duration of the main eruptive phase. Secondary criteria may include the eruption type, e.g., Plinian, Strombolian etc., as well as eruption characterizations like small, large etc. However, for several of the eruptions examined no sufficient data are available for VEI assignments. A traditional practice followed by previous authors has been to assign a single VEI value to a single eruption. Because of the many uncertainties involved in the documentary sources collected, it was preferred to estimate VEI of an eruption in a bimodal form, i.e., “either VEI or VEI + 1”, e.g., 3 or 4, conventionally written as 3–4, instead in the traditional unimodal form. In terms of probability, VEI and VEI + 1 assigned to a single eruption are considered to share equal probability of 0.5 of being the true VEI of the eruption.

VEI assignments have been made possible from the evaluation of the documentary and scientific materials collected. However, previous assignments to Santorini eruptions by the Global Volcanism Program [27] (https://volcano.si.edu/volcano.cfm?vn=212040; accessed on 9 February 2026) as well as by other studies, e.g., [15,21,28], have also been considered. For the 20th century four eruptions, the VEIs provided by the Global Volcanism Program [27] were adopted.

3.1.3. Completeness of the Volcanic Record

The possible incompleteness of the volcanic record before the 20th century calls for careful examination. The first evidence of incompleteness comes from the discovery of documents that reveal episodes missing from the eruptions catalogue so far. The 1667 CE and 1773 CE episodes examined in this paper constitute an example. Another example regards the volcanic episode that supposedly occurred in 1570 or in 1573 CE, which was re-examined very recently by Papadopoulos (2025) [11] based on a little-known manuscript. The examination of this document not only showed that the eruption occurred in 1572 CE but also helped in better understanding the explosivity and duration of the eruption as well as the impacts on the population and of cultivation in Santorini.

For the completeness analysis of the new eruptions catalogue two statistical tests were used under the assumption that the eruption time series follows a steady-rate Poisson process. The first is the Kolmogorov–Smirnov (K-S) test and the second is the conditional binomial test. To further test incompleteness, sensitivity analysis was performed with a Monte Carlo simulation for the calculation of the chance that a stationary random Poisson process would produce clusters or gaps as the ones seen in our historical eruption time series. To understand if the results obtained are model-dependent, the completeness of record was independently examined with the assumption that the time series is represented either by a Poisson random model or by the time-dependent Weibull model.

3.2. Information Sources

3.2.1. Historical Eruptions

The list of documentary sources collected and examined is very long and several of them have already been cited by previous authors [2,6,11,14,16,17,18,19]. As a classification tier, primary (original) and second-hand documents are considered. The examination of some eruption cases is supported by scientific observations published by modern authors.

Dates after the 16th century CE are according to the New Style (N.S., Gregorian) calendar unless otherwise indicated, e.g., the Old Style (O.S., Julian) calendar. Excerpts of documentary sources have been inserted, if needed, as short explanations in the form [explanation].

• 197/196 BCE

Original and second-hand documentary sources for this eruption are quoted by several modern publications [2,6,17,19]. The earlier original source is a set of various fragments of the ancient Greek writer Posidonius (135 BCE–51 CE), which survived through subsequent classic authors including Strabo (64 BCE–24 CE) and Seneca (4 BCE–65 CE). Second-hand documents produced by Pliny the Elder (23–79 CE), Plutarch (c. 45–120 CE) and Justin (2nd century CE) provided relevant information. Summarizing those authors, we may conclude that an eruption occurred between Thera and Therassia during the fourth year of the 145th Olympiad, i.e., during 197/196 BCE. Fires broke from the sea and lasted for four days, so that the entire sea boiled and brazed. An island with an estimated circumference of ~2 km emerged from the sea as if by burning materials. Several sources confirmed that this island was called Hiera—The Holy—and was likely the first shaped in the caldera after the Minoan eruption (Figure 3).

• 19 CE

Pliny the Elder reported that the island Thia (Theia)—The Godly—appeared in the neighbourhood of Hiera in the year 19 CE [2,6]. However, this is not mentioned by other sources and probably is based on a transcription error [2]. For this reason, the 19 CE eruption has been considered as a discredited event [2,27].

• 46 July 6

This eruption was reported by several authors including Seneca, Dio Cassius (c. 155–235 CE) and Aurelius Victor (c. 320–c. 390 CE) [2,6,17,19]. Correlation with historical events and a moon eclipse makes 6 July 46 CE as the most likely eruption date [19]. During this eruption the island of Thia perhaps was born with a circumference of ~5.6 km. Fouqué (1879) [29] suggested that Thia is the present Palaia Kameni and that Hiera is the shallow Banco, which at that time emerged above sea level, but afterwards disappeared.

In his book Life of Apollonius of Tyana, Flavius Philostratus (160/170 CE–244/249 CE) writes that while Apollonius of Tyana (c.15 CE–c.100 CE) was at Leben, on the southern coast of Crete, an earthquake occurred accompanied by sea retreat. In his narration he goes on to say [17]): “…but a few days later some travellers arrived from Cydonia [modern Chania in NW Crete] and announced that on the very day on which this portent occurred and just at the same hour of midday, an island rose out of the sea in the firth between Thera and Crete.” After careful inspection and review of many documentary sources and geological evidence, Papadopoulos (2011) [26] concluded that there was a strong tsunamigenic earthquake of tectonic origin that occurred around 66 (±1 year) CE in southwest Crete having no association to activity in Santorini volcano.

• 725/726

A new eruption between Thera and Therassia was reported by Byzantine authors, like Theophanes (752–818), Patriarch Nicephoros (758–823) and Cedrenos (11th century), as well as by Michael the Syrian (1126–1199), Patriarch of Syria [2,6,19,30]. An evaluation of these sources indicates that the eruption occurred in the year a.M. (Anno Mundi) 6218, i.e., between September 725 CE and August 726 CE. The ejected lava likely caused enlargement of Hiera. The paroxysm lasted about three days, but the eruptive episode continued perhaps for more than a month. A great quantity of pumice erupted and was transported by the sea over a large area, reaching coastal sites in the North Aegean Sea, like Lesvos Island, Avydos near Dardanelles, and the coasts of Asia Minor and Macedonia. We may add the island of Samothraki where pumice was also observed in coastal sites [31].

Valuable results about this eruption were obtained by Preine et al. (2024) [21] based on high-resolution seismic reflection data with cored lithologies at four sites. Their shore-crossing analysis revealed the deposits of a submarine explosive eruption that produced up to 3.1 km³ of pumice and ash.

• 25 November 1457

An eruption likely occurred in the Santorini caldera, although the available documents do not help clearly understand what exactly happened. First evidence comes from the Latin text on an inscription carved beside the portal of Skaros castle [2,19,32,33,34]. The text says that the event happened on 25 November 1457. The translation of the rest part of the Latin text reads as next [2]: “…with a great rumble, immense Thererinus [the sea around Therassia] tore enormous rocks from Kameni, and with a groan, a crag appeared from the depths of the waves bringing with it a great and memorable portent.” A similar translation was adopted by Ambraseys [19] who dated the event on 25 November 1455. Friedrich [2] noted that this text is unclear but seems to state that an island either disappeared or appeared at that time.

The Florentine monk Cristoforo Buondelmonte travelled in Greek islands from 1414 to 1420 CE. Around 1420 CE he circulated his first travelogue manuscript titled Liber Insularum Archipelagi [32]. Maps of several islands, including Thera and Therassia, were also included in the travelogue. A later version of the manuscript, dated from the year 1465 to 1466 CE, contains also a short text saying that eight years earlier, i.e., in 1457 or 1458 CE, an island appeared there [32]. The length of the island was equal to that of one galley, i.e., of ~160 m. It is not clear if the short text was added by Buondelmonte himself or by one of the manuscript editors. We may not exclude that the text added in the map was based on the Latin text carved beside the portal of Skaros.

The emergence of an island is referred to by Suriano [18], a Venetian nobleman and traveler who writes that “…The following year [1463] in the archipelago, there was such a terrible earthquake that the islands of Chio [perhaps Scio=Ios] and Santorini and nearby islands shook so much that they could not remain standing; and the islanders say that they thought it was the end of the world. And in the morning, they saw that an island five miles long had been created nearby, and they called it Chaimeni [Kameni].” Ambraseys [19] believes that this episode is a duplication of the activity that occurred in 1455 CE. However, the reported length of five miles of the emerging island appears as an exaggeration if not completely erroneous.

In his pilgrimage to Jerusalem, the Italian nobleman Pietro Casola (1427–1507 CE), narrates that [18] “…very late on Saturday, the 27th of September [1494], the captain [of the galley] would gladly have approached an island called Santurin. And there was a great dispute in consequence between the captain and the comito and the pilot, who said to the captain that it was not a suitable place to stop at—that is, to cast the anchor.” The next paragraph of Casola’s narration is of particular interest to the examination of the 1457 CE eruption: “…The captain maintained that on other occasions he had anchored there; and he related that one time when he was in the Canal of Santurin with several galleys, a storm arose in the West which continued for the space of three days, and was not only violent but very terrible—great thunder, great flashes of lightning, and noises as if there had been battle chargers there; and all on board the galleys were so terrified that they did not know what world they were in. On the morning of the third day an island as black as coal made its appearance; and the aforesaid captain said that they made every effort to approach it, but could never discover the bottom, and that he had never been able to anchor there since.”

The last passage implies that in some unidentified time before 1494 CE a black island, very likely Palaia Kameni, emerged on the west side of the caldera. The captain’s narration does not make it clear whether an eruption occurred or a rock reef broke away from the cliffs of Palea Kameni, as discussed by Friedrich [2]. The significant time distance of 37 years between the events of 1457 CE and 1494 CE makes it hard to correlate the island in the captain’s narration with the one that supposedly emerged in 1457 CE, although one may not rule such a possibility.

The story about an eruption occurring in 1457 CE is complicated by the presence of a conspicuous rock failure on the northeastern side of Palaia Kameni. This is a fault scarp or rock detachment (Figure 4). One gets the impression that Palaia Kameni is the remaining island, the other part of which disappeared into the sea. However, the causative mechanism remains unknown. One possibility is that the rock failure was caused by dynamic shaking during an earthquake or a volcanic eruption.

Another explanation could be that a gravitative landslide of a fractured rock mass was produced without shaking. This scenario excludes the possibility of an eruption even of small size. There is no evidence about the event date, although it looks like it happened in historical times. According to Friedrich [2], a possibility would be that Palaia Kameni appeared in 1457 CE rather than in 726 CE and was enlarged by the younger blocky lava now seen on the flatter part of the island. Then, the occurrence of a possible eruption is assumed, and the rock failure could have happened during that eruption in spite of the fact that this episode is not included in the list of the Global Volcanism Program [27].

• 1572 spring–summer

Regarding this eruption, little information was recollected from local memory until recently and it was considered as occurring either in 1570 or in 1573 or from 1570 to 1573 CE [33,34]. Papadopoulos (2025) [11] brought to light a very little-known but reliable manuscript revealing that the eruption occurred in 1572 CE between Santorini and Palaia Kameni. The manuscript makes clear that “fire, smoke, and stones” were coming out between the two islands and a new volcanic island named Mikri Kameni—The Lesser—was born (Figure 3). Floating pumice was transported by the sea as far as to Thessaloniki and Constantinople, resembling pumice transportation in the North Aegean Sea after the 725/726 CE eruption. Smoke and heat destroyed the vineyards and the planting on Santorini, which implies that the eruption occurred in spring-summer season. Residents of Santorini were forced to move to nearby islands likely because of sulfurous gas release. The eruption duration was estimated at ~1 yr, but the fire and smoke disappeared suddenly.

• 9 October 1650

The only known historical eruption that occurred at Kolumbo volcano was the one during 1650 CE. From a series of original information sources [2,6,14,15,16,19,26], we learn that intense seismic activity was noted as early as March 1650 CE. The eruption occurred from September to December 1650 CE with the paroxysmal phase taking place on 29 September/9 October with a Plinian-style rhyolitic explosive episode. The eruption was accompanied by important phenomena such as damaging earthquakes, tephra falls, deadly gas discharge and a powerful tsunami that hit Santorini violently, nearby islands and north Crete. It has been estimated that about 50 deaths were caused in Santorini, nearly all due to suffocation caused by the release of poisonous gases rich in H₂S.

• February 1667

Monsignor Giuseppe di Santa Maria Sebastiani was a Carmelite monk who traveled around with Italian missions in the Greek islands in 1666–1667 CE and published an extensive travelogue [35]. Sebastiani arrived on the island of Syros on 19 January 1667. By the beginning of February, he sailed from the island of Ios to Santorini. Although his destination was the Castle of Skaros, he disembarked along with his company at Santorini’s headland called Apanomeria (Figure 3), modern Oia, at the northern side of Santorini.

The narration of Sebastiani goes on as follows: “We landed with our belongings. The shore at that point was all sand and flat but no wider than 10 yards. From it rise very high rocks, so high a straight that they look like built with plumb lines. While we were making a little collection, near a small opening in the middle of the height of the said rocks, a whirlwind or rather a furious wind came from a small mouth in the middle of the height of the said rocks, which vomited upon us a great quantity of white pumice stones. It stopped suddenly and after some time it made the same indentation and did the third time to our amazement and fear and it increased even more when in a time that when no light breeze was blowing, we saw the sand moving on that perfectly flat shore. However, we intended to leave that island. And a small boat came to use from a village called St. Nicolas [likely situated in Apanomeria].” Perhaps Sebastiani landed at what is today the Ammoudi shore, at the northwestern side of Santorini. In fact, it is a sandy beach beneath the steeply caldera cliffs and situated just after entering the caldera from the north.

Sebastiani’s narration is not clear as for the nature of the phenomenon described. Was it a gas discharge along with pumice material from an opening at the caldera cliff? One may not exclude the occurrence of a local landslide in three sequential episodes. Ambraseys (2009) [19] suggested that it was a minor eruption together with what may have been a slight earthquake. Then, we may not rule out the occurrence of a small phreatic eruption. However, one may wonder if it is possible for an even small eruption to occur at the caldera cliffs.

• May 1707–September 1711

The series of eruptions, which started in May 1707 and lasted until September 1711, are well documented by the eyewitnesses Jesuit Fathers Goree (1712) [36] and Turillon (1714) [37], as well as by the traveler Aubry de la Motraye’s (1723) [38]. A relevant map (Figure 5), which is included in the publication by Father Goree (1712) [36] helps to better understand the evolution of this episode. Testimonies about this eruption are summarized as follows. By the end of May 1707, a White island (Figure 5) consisting of pumice and some black lavas was seen uplifting to the west of Mikri Kameni. On the 5th of June fire was set, and a new Black island was formed north of the island of white pumice (Figure 5). By September 1707 the Black island was so large that it united with the White one; in this way the island of Nea Kameni was born. Explosions and ash eruptions continued until September 1711. Ross (1840) [30], visiting Santorini in 1836 and 1839 noted that fishermen had observed that not far from Nea Kameni a sharp reef had started to rise from the sea growing up every year.

• 10 January 1773

A correspondence of the Spanish newspaper El Mercurio Histórico y Político (1773) [39] says that “From a ship which left the Archipelago we learn that on Santorini and in adjacent islands, on 10 January of this year [1773], an earthquake was felt, accompanied by a low but continuous roar, and whirlwinds of smoke and flames. When this calmed down, it was noticed that the lands of Santorini had extended around half a mile northward. Between this island and Megali [the Grand], a strait was formed, which could provide a new route for ships to enter and leave the port of Santorini.”

According to the press correspondence, three main phenomena occurred on that date, (i) an earthquake that was presumably strong enough because it was felt not only in Santorini but also in other islands; (ii) accompanying roar, whirlwinds of smoke and flames, which imply the possible occurrence of an eruption; (iii) the lands of Santorini were extended half a mile northward, indicating the generation of a new island close to Megali, which was a predecessor of Nea Kameni.

But what was the new island that reportedly emerged in 1773? To respond to that question, we should have a clear picture of the landscape in the caldera after the supposed eruption of 1773. To this aim, detailed maps prepared in the field after 1773 and published a few years later are helpful. The first map, prepared by Comte de Choisel-Gouffier [13] and re-published recently [11], illustrates three islands existing in the caldera: (1) Grande Cameni–anciennem Hiera (Large Kameni–ancient Hiera), i.e., the Palaia Kameni, (2) Petite Cameni (Lesser or Mikri Kameni), and (3) L’ Ile Nouvelle (New Island or Nea Kameni). Similar is the landscape in the caldera illustrated in the map published by Olivier [12] (Figure 3) who traveled in Greek islands by the end of the 18th century. It is noticeable that Mikri Kameni was a smaller island generated by the 1572 eruption [11], while Nea Kameni was born nearby during the 1707–1711 eruption as described by eyewitnesses. According to the Spanish correspondence a strait was formed between the new island and Megali (Nea Kammeni). Bearing in mind the landscape illustrated in contemporary maps mentioned earlier, a realistic suggestion would be that the island that appeared during the 1773 episode could be a northward extension of Mikri Kameni. This is consistent with the narration that a strait was formed between the new island and Megali. That the lands were reportedly extended half a mile northward may imply lava flow to that direction.

The previous explanation is supported by the results of an investigation with a high-resolution merged LiDAR–bathymetry grid [22]. This enabled detailed mapping of both onshore and offshore historical lava flows of the Kameni islands. One of them, which is located north of Nea Kameni, appears to predate the 1925–1928 lava flows but was emplaced after the 1707–1711 lava flows. The extension of lands northwards, that reportedly happened at Mikri Kameni, could be the new lava flow mapped with LiDAR bathymetry techniques.

• January 1866–October 1870

This eruption is very well documented thanks to detailed eyewitness accounts, field observations, photos and drawings collected and published by many authors. Compilations of relevant documents can be found in review publications [2,6,29].

Volcanic activity started at the end of January and the beginning of February 1866 with submarine lava effusion. Explosions started a few days later. The volcanic activity continued with effusive and explosive episodes up to 15 October 1870. During this period several volcanic centers and domes were generated, thus increasing the size of the Nea Kameni Island [2,6]. After the end of the eruptions the activity remained at a solfatara stage.

3.2.2. Eruptions During the 20th Century

Four main eruption phases occurred during the 20th century, all taking place in Nea Kameni and between this island and Mikri Kameni. Many scientific publications and documentation by photos are available [2,6,16].

• August 1925–January 1926

This eruption phase started on 11 August 1925. After numerous explosions and submarine effusion of lava, the Mikri Kameni and Nea Kameni united on 12 August 1925. The activity lasted until mid-January 1926.

• January–March 1928

After the rest of 20 months a new activity phase started on 23 January and ended on 17 March 1928. Several explosions were accompanied by the effusion of lava, and a new small dome was created. Solfatara activity followed until the next phase.

• August 1939–July 1941

This eruption phase was subdivided into five episodes with eruptions, effusions and generation of new small domes. The activity started on 20 August 1939 and ended in the first days of July 1941.

• January–February 1950

An epigenetic eruption began on 10 January 1950. The activity lasted only 24 days, i.e., until 2 February, with explosions accompanied by effusions and the formation of a small dome. Afterward, the volcano was at a stage of solfatara and fumarolic activity, which lasts until now.

4. Results

4.1. Historical Catalogue

The examination of many documentary sources and scientific observations indicates that 15 post-Minoan eruption episodes have been reported in the Santorini island complex in historical times. The eruption history starts in 197/196 BCE and extends up to 1950 CE.

Table 2. Main features of the 15 historical eruptions in the Santorini area. Geographical coordinates of eruptions 1–6 and 9–15 have conventionally placed at the position of Nea Kameni. All eruptions occurred in the caldera except the one of 1650 CE, which occurred at the Kolumbo submarine volcano. R = Reliability score, VEI = Volcanic Explosivity Index, symbol ? means no suffiecient data are avaialble to estimate VEI.

No.	Date	Lat°N	Lon°E	R	VEI	Landscape Changes
1	197/196 BCE	36.404	25.396	3	2–3	Hiera Island emerged
2	19 CE	36.404	25.396	1	?	Thia appeared near Hiera?
3	46 July 6 CE	36.404	25.396	3	2–3	Thia (Palaia Kameni) emerged
4	725/726	36.404	25.396	4	4–5	Hiera increased
5	25 November 1457	36.404	25.396	2	?	An island emerged or a rock reef broke away?
6	spring–summer 1572	36.404	25.396	4	3–4	Mikri Island emerged
7	9 October 1650	36.524	25.483	4	4–5	Kolumbo cone collapsed
8	February 1667	36.462	25.370	2	?
9	July 1707–3 September 1711	36.404	25.396	4	3–4	White and Black Islands emerged; generation of Nea Kameni
10	10 January 1773	36.404	25.396	3	2–3	Mikri Kameni extented northward
11	January 1866–2 October 1870	36.404	25.396	4	2–3	Nea Kameni increased
12	August 1925–January 1926	36.404	25.396	4	2	Mikri and Nea Kameni united
13	January–March 1928	36.404	25.396	4	2	New small dome at Nea Kameni
14	August 1939–July 1941	36.404	25.396	4	2	New small domes at Nea Kameni
15	January–February 1950	36.404	25.396	4	2	New small dome at Nea Kameni

is a catalogue that summarizes dates, reliability, VEI and associated main volcanic landscape changes. In the next lines the Global Volcanism Program [27] is frequently quoted and therefore it is abbreviated as GVP.

197/196 BCE. The original description of this eruption is quite realistic and well suited in the geographical context of the area. However, no other independent original source is available beyond the surviving Posidonius fragments. Therefore, a reliability score of R = 3 was assigned. The GVP estimated VEI = 3. Allowing for some uncertainty to be involved, VEI 2–3 was assigned.

19 CE. This eruption has been considered as a discredited event [2,27]. Therefore, reliability R = 1 was assigned.

46 July 6 CE. Seneca appears as the only contemporary source. One may assume that Seneca collected information from various sources because he was very likely living in Rome at that time. The description of the eruption is well suited in the geographical context of the area. Given that the information is not verified by other independent source reliability score of R = 3 was assigned. According to the GVP the VEI is 3 but allowing for some uncertainty to be involved VEI of 2–3 was adopted.

725/726. Reliability score of R = 4 was estimated to this eruption because it is documented by reliable historical sources and geological observations. VEI = 5 or VEI = 4(?) was estimated by Preine et al. (2024) [21] and GVP, respectively. Eventually, VEI of 4–5 was assigned to this eruption.

25 November 1457. Documentation of this episode is puzzling and does not help us to understand what exactly happened in Santorini caldera in 1457 CE. A reliability score of 2 was estimated. Because of the many uncertainties involved no VEI was assigned.

1572. This eruption is supported by a reliable manuscript as well as by independent and consistent local memory. Therefore, a reliability score of R = 4 was assigned. The GVP assigned VEI = 3 but a VEI of 3–4 was estimated because of the emergence of Mikri Kameni Island, which was of ~600 m in length and of ~250 m in width [11].

9 October 1650. This eruption is documented by many independent historical sources and geological observations (R = 4). The GVP assigned VEI = 4(?) but VEI = 5 was estimated by others [15]. Therefore, a VEI of 4–5 was assigned to this eruption.

1667. This episode is mentioned by only one source, which does not help us to conclude if an eruption occurred or not. Therefore, reliability of R = 2 was estimated and no VEI was assigned.

1707–1711. Documentation of this episode comes from several independent sources (R = 4). GVP assigned VEI = 3. Allowing for some uncertainty to be involved, a VEI of 3–4 was assigned.

10 January 1773. This episode is documented by a historical source. Geological observations [22] fit well enough the historical information. Therefore, a reliability score of 3 was estimated and a VEI of 2–3 was assigned.

1866–1870. Documentation of this eruption comes from many independent sources (R = 4). GVP estimated VEI = 2. Allowing for some uncertainty to be involved, a VEI of 2–3 was assigned.

1925–1926, 1928, 1939–1941, 1950. Reliability score R = 4 was estimated for the four eruption episodes because they are very well documented from observations conducted and published by professional scientists [2,6,16]. VEI = 2 assigned by the GVP was adopted.

4.2. Completeness of the Catalogue

Identifying the most complete portion of an eruption record is essential before estimating eruption recurrence and probability [40]. A straightforward approach, which is known as break in-slope examination, identifies the point at which the cumulative number of eruptions increases most rapidly with time [41]. This is suitable particularly for records with a well-defined breakpoint. The selection of a breakpoint facilitates the application of the two-sample Kolmogorov–Smirnov (K–S) test [41], which can be combined with Monte Carlo simulations to identify breakpoints [42]. The Poissonian time distribution of eruptions is a standard baseline selected in such approaches for testing eruption record completeness. However, other probability models have also been proposed.

On our dataset (Table 2), a break-in-slope illustration (Figure 6) reveals a statistical signature of a possibly incomplete historical record before the breakpoint of 1572 CE because the eruption record was very sparse, only three events in ~1747 years. In this case the activity rate is λ₁~0.002 eruptions/yr. The year 1572 CE marks a significant shift given that after this point nine eruptions occurred in a time interval of ~450 years up to 2025, with activity rate λ₂~0.02 eruptions/yr, reflecting a 10-fold increase in the activity record. To test incompleteness in the eruption record, a set of 12 episodes that received reliability score R ≥ 3 was considered after excluding the less reliable episodes of 19 CE, 1457 CE and 1667 CE.

Completeness testing was performed by combining K-S test with Monte Carlo simulation and considering that the baseline is the Poisson (random) process for the eruptions time distribution. The conditional binomial test was executed as an additional testing tool. The Poisson model implies that if the mean activity rate is λ eruptions/yr, the probability of observing k eruptions in a time unit, here one year, is

P (X = x) = λ^x [exp (−λ)]/x!

(1)

4.2.1. Kolmogorov–Smirnov Test

With the K-S test the null hypothesis that the eruptions follow a steady-rate Poisson process was examined. The Empirical Cumulative Distribution Function (ECDF) of 12 events was compared against a theoretical Cumulative Distribution Function (CDF) of a steady-rate Poisson distribution in the entire time interval from 197 BCE to 2025 CE (Figure 6). Because in this time interval there are 12 events, each eruption adds 1/12 ≈ 0.083 to the cumulative probability. If the time distribution of eruptions were perfectly uniform, the probability would grow linearly at a rate of roughly 1/185 ≈ 0.0055 per year. Application of the test results in calculation of maximum vertical distance D-value 0.546, while for the significance level of α = 0.05 and n = 12 the threshold of the critical value is approximately 0.375. Since 0.546 > 0.375, the null hypothesis that the eruptions follow a steady-rate Poisson process is rejected.

4.2.2. Conditional Binomial Test of a Steady-Rate Poisson Process

The average rate in the entire 2197-year time interval is ~0.0055 eruptions/yr. Multiplying the average rate by the length of the time interval of 1747 years, it was found that the expected number of events is μ ≈ 9.54. With the use of the Poisson probability mass function (PMF, Formula (1)) the probability of observing three events in 1747 years is

P (x = 3) = μ^k [exp (−μ)]/3!

(2)

P (x = 3) as low as ~0.0104 (or 1.04%) was calculated (Figure 7).

A hypothesis test was performed to examine if the rate in 1747 years is significantly lower as compared to the rate in 450 years. The conditional binomial test was applied, which has been proposed as providing better results with respect to the classic C-test [43]. In practice, the null hypothesis “the event rate is constant across both time intervals” was compared against the alternative hypothesis “the rate in the recent 450 years is significantly higher than in the previous 1747 years”. The test logic is based on that if the rate were truly steady, an event would have probability 450/2197 = 0.205 of falling into the more recent interval. However, nine out of 12 total events occurred during those 450 years. The probability of observing nine or more events in that time interval by chance is very low, P = 7.6 × 10⁻⁵, which is significantly smaller than the significance level of α = 0.05 or even α = 0.01. Therefore, the null hypothesis is rejected indicating that the process is not a single steady-rate Poisson process.

The statistical tests strongly indicate that in the pre-1572 CE time interval the historical eruption record is significantly incomplete as compared to the time interval after that breakpoint. It is noteworthy that incompleteness before the 16th century CE has also been noted in the recorded history of earthquakes and tsunamis in Greece [44].

4.3. Sensitivity Analysis

4.3.1. Monte Carlo Simulation

Because of the small number of events in the data set and the relatively short time interval, which is bounded between 197 BCE and 2025, a Monte Carlo simulation was run by creating 100,000 random timelines (synthetic datasets) looking for how often a maximum deviation of K-S distance D = 0.546 or higher occurs by chance. Each trial placed exactly 12 events randomly distributed between the fixed time bounds. For every single simulated timeline, the maximum distance, D, was calculated. The average simulated value of D by pure chance was found 0.2415, with the 95th percentile D being 0.3698. The empirical probability calculated is 0.00018, which is far below the threshold of 0.01 and indicates that there is only a 0.00018 chance that a stationary Poisson process would produce the cluster observed after 1572 CE or the extreme gap before that year.

4.3.2. Testing the Weibull Distribution

Introducing a Weibull distribution shifts the analysis of the eruption time record from a memoryless Poisson model to a history-informed model. Instead of measuring the number of events in fixed intervals, a Weibull distribution models the repose time, i.e., the time interval between successive eruptions, and introduces a variable hazard rate. Poisson assumes a constant eruption rate, λ, while the Weibull Model introduces a shape parameter, k, that changes the hazard rate over time, and a scale parameter, λ. A standard Weibull model treats all data as a completely accurate physical reality. It lacks any parameter to represent human historical factors that may affect the completeness record.

For the time interval covered by our entire data set a Weibull distribution with shape parameter k = 0.6719 fits better as compared to the random distribution (Figure 8), which indicates a decreasing hazard rate. However, a standard Weibull distribution cannot explicitly pinpoint a historical calendar year where record completeness changes. Therefore, a Piecewise Weibull Model (PWM) was constructed and evaluated using a Likelihood Ratio Test (LRT). PWM is an advanced survival analysis tool that divides time into distinct intervals. Within each interval, it applies a separate Weibull distribution to model the hazard rate. This hybrid approach combines the strict, easy-to-interpret assumptions of a piecewise constant hazard model with the flexibility of changing shape and scale parameters over time. Splitting the eruption dataset at the 1572 CE boundary we separated the history into two distinct eras. Running independent Maximum Likelihood Estimations for each era we found k = 1.3835 and λ = 651.98 yrs for the pre-1572 era and k = 1.2290 and λ = 51.49 yrs for the post-1572 era. In this split model, both eras flip to k > 1 instead of k < 1 found for the standard Weibull model, which is strong evidence for completeness breakdown at 1572 CE. The actual drastic change occurred in the scale parameter λ, which plummeted from 652 yrs down to 51 yrs. This massive scale shift mathematically exposes historical under-reporting incompleteness before 1572 CE.

To verify if this split is statistically justified, the run of an LRT test and comparison of single Weibull model with PWM through log-likelihoods provided Chi-square p-value 0.0001 for two degrees of freedom. Because this p-value is far below the high standard confidence threshold of 0.01, the PWM overwhelmingly rejects the single Weibull model.

5. Discussion

Eruption catalogues are often incomplete [45], especially when derived from geological data [46]. The compilation of a new historical eruption catalogue for Santorini has been possible thanks to the joint evaluation of documentary sources and geological fingerprints left by the volcanic activity. The catalogue is new not only because it brought to light the 1667 CE and 1773 CE episodes but also because it provides for the first time the reliability characterization of each episode. This is a step that harmonizes the taxonomy of historical volcanological information with the compilation of catalogues and data bases for historical earthquakes and tsunamis. Refreshing estimations of VEI on a bimodal form is an additional new step characterizing the catalogue.

In spite of these advancements, incompleteness of the new catalogue compiled is a challenging issue. Statistically, if the “true” reporting rate is represented by the post-1572 data, there should have been approximately 30 eruption episodes between 197 BCE and 1572 CE that went unrecorded or were lost to history. Smaller eruptions are much less likely to be recorded in earlier centuries, creating a false impression of “lower” activity before 1572. On the other hand, larger events of VEI4+, like the ones of 725/726 CE and 1650 CE, which are separated by a repose time of 725 years, survived through the historical record, thus leaving a deeper imprint on both history and geology. In this sense, the larger events record tends to be “complete” over a much longer timeframe. These results were reached after statistics and simulations applied to the set of 12 reliable eruption episodes. The results did not change by adding to the less reliable events of 1457 CE and 1667 CE.

These findings for Santorini are consistent with the global eruption rate, which shows an apparent increase in volcanic activity over the past 200 years but is an artifact of an incomplete eruption record [45]. Indeed, although nearly half of the expected eruptions of VEI5 or larger are present in the Volcanoes of the World (VOTW) database records covering the past two millennia, nearly 90% of all eruptions are probably missing from that same interval.

Historical records of volcanic activity are heavily influenced by human factors. In our catalogue the breakpoint in the historical time series is the eruption of 1572 CE. The significant increase of eruption recording after that point perhaps reflects the increased capabilities for historical record after the Renaissance and the invention of the printing press by Johannes Gutenberg in the mid-15th century. In this sense we may support that the “increase” in volcanic activity is anthropogenic, due to better observation/record-keeping, rather than geological. The difference in the eruption record before and after 1572 CE could be attributed to reporting incompleteness rather than to a fundamental shift in magma supply. This is interpreted as a discovery bias, meaning that we simply haven’t found the records for the older events yet, if any. In the last years new documentary sources were revealed about the 1572 CE eruption [11] and the 1667 CE and 1773 CE episodes examined in this paper. The discovery bias, however, implies that it is less likely that new documentary sources would be found for the earlier period of the Santorini historical eruptions. If such a hypothesis is correct, the role of geological research becomes critical for completing our knowledge about the earlier historical eruption record. While magma supply is the physical driver of eruptions, it typically operates on different scales than short historical recording shifts as the ones examined in this paper. Local variability may include short-term increases in eruption rates, say on decadal to century scales, which can be caused by the recharge of magma reservoirs as the Santorini example indicates [47]. On the other hand, when geological data are used to augment historical records, e.g., [21,22], the “true” eruption rate may appear more stable than the written history suggests.

6. Conclusions

The compilation of a new catalogue of historical eruptions in Santorini has been possible after evaluating an enriched set of documentary sources and of geological observations when available. The catalogue covers the time interval from the second century BCE up to the present and lists 15 episodes including those of 1667 CE and 1773 CE that remained little-known so far.

Various uncertainties and inconsistencies involved in the historical and geological record of the eruptions necessitated the evaluation of the reliability and the re-evaluation of the eruption size. A reliability score, R, has been assigned for the first time to each episode based on a Reliability Assessment Matrix. The reliability scale ranges from 1 (lowest reliability) to 4 (highest reliability). Only the episode of 19 CE was assigned R = 1, while the episodes of 1457 CE and 1667 CE received R = 2. The rest episodes of 197/196 BCE, 46 CE, 725/726, 1572, 1650, 1707–1711, 1773, 1866–1870, 1925–1926, 1928, 1939–1941 and 1950 received a reliability score of either 3 or 4.

The size of an eruption has been estimated in terms of the 8-grade scale of the Volcanic Explosivity Index (VEI). The existence of many uncertainties in the documentary sources, VEI estimation in a bimodal form was selected for each episode, i.e., either “V” or “V + 1”. At all indications the largest eruptions of a VEI of 4–5 in the entire time interval examined were those of 725/726 CE in the Santorini caldera and of 1650 CE in the extra-caldera submarine Kolumbo volcano. The rest eruptions were of smaller size, of a VEI 2–3 or 3–4.

Although the new catalogue is the most complete published so far, the time series of the events indicates that the eruption of 1572 CE constitutes a breakpoint in the completeness of the historical eruption record. Considering eruptions receiving reliability score R ≥ 3, it was found that nine out of 12 eruptions occurred within the 450-years time interval after that point. The other three eruptions were recorded within the 1747-years time interval before that point. This 10-times difference, which is statistically highly significant, probably is a bias discovery effect of anthropogenic origin. Using the eruption record rate of the time interval of the last 450 years as a baseline for a steady-rate Poisson process, it was calculated that there are about 30 missing eruption events that escaped record in the older historical interval of 1747 years. However, this practically regards eruptions of size less than a VEI of 4–5. Further investigation is needed to support that at the same time this difference does not reflect a drastic change in magma plumbing rate.