Towards reconstruction of the lost Late Bronze Age intra-caldera island of Santorini, Greece

During the Late Bronze Age, the island of Santorini had a semi-closed caldera harbour inherited from the 22 ka Cape Riva Plinian eruption, and a central island referred to as ‘Pre-Kameni’ after the present-day Kameni Islands. Here, the size and age of the intracaldera island prior to the Late Bronze Age (Minoan) eruption are constrained using a photo-statistical method, complemented by granulometry and high-precision K-Ar dating. Furthermore, the topography of Late Bronze Age Santorini is reconstructed by creating a new digital elevation model (DEM). Pre-Kameni and other parts of Santorini were destroyed during the 3.6 ka Minoan eruption, and their fragments were incorporated as lithic clasts in the Minoan pyroclastic deposits. Photo-statistical analysis and granulometry of these lithics, differentiated by lithology, constrain the volume of Pre-Kameni to 2.2–2.5 km3. Applying the Cassignol-Gillot K-Ar dating technique to the most characteristic black glassy andesite lithics, we propose that the island started to grow at 20.2 ± 1.0 ka soon after the Cape Riva eruption. This implies a minimum long-term lava extrusion rate of ~0.13–0.14 km3/ky during the growth of Pre-Kameni.

On the other hand, the presence of abundant lithics, especially clasts of a chemically distinctive black glassy andesite, in the pyroclastic deposits of the Minoan eruption (see below) suggested that the central part of the caldera bay was occupied by an intracaldera island 10,[14][15][16] . We use the term 'Pre-Kameni' 14,17 for this pre-Minoan edifice, which may have been similar to the present-day, post-Minoan islands of Palaea and Nea Kameni, but completely destroyed during the Minoan eruption. The size and age of such a Pre-Kameni island have remained uncertain, although broad estimates of 3 to 5 km 3 were proposed 3,15 based on the poorly constrained total lithic ejecta 18 . For comparison, the volume of the present-day Kameni Islands is 3.2 km 3 derived from a merged LiDAR bathymetry grid 19 .
In this paper, by applying a photo-statistical analysis of outcrops complemented by granulometric analysis, we present quantitative results for the volume of the lithic clasts included in the Minoan deposits and, after proportioning their amount, determine the dimensions of the destroyed Pre-Kameni island. The timing of island growth is constrained using high-precision K-Ar dating of the black glassy andesite, which is regarded as the most significant island-forming lithic type 9,10,15 . The topography of Santorini just before the Minoan eruption is also shown by integrating previous and our own results using a high-resolution digitial elevation model (DEM) of Santorini 19 .
Late-stage evolution of Santorini. The pre-Minoan landscape of Santorini was dominated by a shallow, flooded caldera formed by the last major explosive eruption before the Minoan, the Cape Riva eruption. A synthesis of published dates for the Cape Riva eruption yields a mean age of 21.8 ± 0.4 ka 20 . This eruption is thought to have collapsed the pre-existing Skaros-Therasia lava shield 13,15 . For the subsequent interplinian period which lasted for ~18 ky only minor explosive activity is documented [20][21][22] . The 3.6 ka Minoan eruption consisted of four main phases, the deposits of which are denoted A to D by Druitt et al. 13 . The first phase was a plinian pumice fall 7,9,23 . A lower, crudely bedded and an upper, unbedded layer, up to 5.5 m thick in total, are collectively named unit A.
During the second phase, interaction with sea water and partial column collapse resulted in phreatomagmatic surges (unit B), ranging in thickness from some dm to up to 12 m 13 .
In the third phase, significant column collapse produced the most prominent unit of the eruption on land (unit C). This is a coarse-grained, massive, up to 55 m thick, phreatomagmatic ignimbrite up to 55 m thick 13 , still reflecting magma-water interaction and deposited at low temperatures 10,24 . The third eruptive phase may have created a tuff cone 3,11 , possibly a large pyroclastic construct (cf. 25 ) that filled the caldera bay. This phase is thought to coincide with the explosive disruption of the Pre-Kameni island (along with other parts of Santorini), given the occurrence of abundant, evenly distributed lithic clasts up to 10 m in size in the deposit.
In the fourth phase, finer grained, hot ignimbrite (unit D) was produced from multiple vents without phreatomagmatic character. This phase may have been coeval with major caldera collapse 9,10 . The unit thickens distally up to 80 m offshore and is thought to be the most voluminous Minoan deposit 4 . The ignimbrites contain thin lithic concentration zones and are capped (or cut) in some places by lithic-rich gravelly beds of fluvial origin up to 5-10 m thick 10,13 . The presence of the latter indicates considerable potential energy as well as transport distance required for fluvial redeposition of the lithics contained in unit C, B and A, implying a topography which may have been truncated by the caldera collapse only toward the end of phase 4.

Materials, Methods and Results
Petrography and geochemistry of lithic clasts in the Minoan deposits. The Minoan deposits contain a diverse range of juvenile components as well as lithic clasts derived from older parts of Santorini 10,13 . On the basis of Druitt 10 , the lithic clasts analysed here are grouped into black glassy andesite, flow-banded rhyolite, and miscellaneous lavas and tuffs ( Fig. 2A). Geochemically, the Minoan lithics range from basalt to rhyolite (50.8-71.0 wt.% SiO 2 ). In addition, a dominant low-Ba/Zr group of lithics is distinguished from a characteristic high-Ba/ Zr group of clasts of mainly intermediate (andesitic) composition, which are unique in the eruptive history of Santorini since 530 ka 10 . Among the high-Ba/Zr group, a conspicuous suite of fresh, black, glassy and porphyritic blocks of andesite lava (BGA hereafter) has been interpreted as fragments derived from a pre-Kameni edifice 7,9,10 . This high-Ba/Zr BGA is distinguished from similar blocks of dark-grey, porphyritic andesite of low-Ba/Zr nature interpreted as fragments of older lava sequences, such as the Therasia lavas 13,20 . Blocks characterised by a high Ba/Zr ratio also include a variety of slightly altered, light-grey and fine-grained andesite lava and an altered clast of dacitic tuff, grouped under 'Miscellaneous lavas and tuffs' ( Fig. 2A). A likely origin of these clasts is the early centres of the Akrotiri peninsula (650-550 ka 10,13 ). From the above, it can be concluded that the high-Ba/Zr BGA is the dominant lithology that constituted the Pre-Kameni island.
Volumetric analysis of the lithic clasts in the Minoan deposits. 76 sites (Fig. 1) have been selected for photo-statistics of the lithic clasts in order to infer the volume of Pre-Kameni. Unit C, with dispersed lithic clasts, is represented by 58 sites; the lithic-poor units of A and B by 4 and 5 sites, respectively, and unit D by 9 sites. The latter includes both the ash-rich, lithic-poor facies and the thin lithic concentration horizons in the ignimbrites. Details of the photo-statistical method are given in Fig. 2B and Appendix 1, and the numerical results are summarized in Table 1.
Although the clast proportion values of the 58 sites of unit C show a great dispersion, the average percentage of the lithic clasts (15.7 vol%) is statistically confident as the standard error for unit C is quite small (1.2%). As for the other units, they are undersampled, but the studied locations are considered to be representative. In units A and B, the percentage of lithic clasts is insignificant (1.8 and 0.8 vol%, respectively), so the standard error is also small (0.6% and 0.2%). In contrast, the average clast proportion increases to 13.8 vol% in unit D (with 5.7% standard error), and the data are even more scattered than in case of unit C. This fact is in accordance with the occasional occurrence of lithic-rich beds. All clast proportions were calculated for the >0.125 mm fraction (see Fig. 2B and Appendix 1).
Granulometric analysis of 4 samples, one from each unit, shows that the finest fraction of <0.125 mm diameter represents an important addition: 39-41 wt% of the bulk material (except the sample from unit A where it is only 6.5%). The proportion of BGA was also measured manually in each fraction of the sieved samples, and in the coarse fraction (>16 mm) of three other unit C locations. Results show that, for unit C, the proportion of BGA ranges between 18% and 31% in the coarsest fraction, whereas in the finer fractions (16-0.125 mm) it varies between 15.9 and 45.1%. These values support that the mean BGA proportion of unit C can be appropriately taken as 30% as proposed by Druitt (2014). As for the other units, we obtained 4.0-18.2% for the BGA proportion in unit D and 0-26.3% in unit A taken all fractions into account, and we did not find BGA clasts in the studied sample of unit B.
In general, both the proportion of lithics and the proportion of BGA within the lithics show a decreasing trend with clast diameter. Based on this trend, we propose that both the proportion of lithics and the proportion of BGA within the lithics in the finest (<0.125 mm) fraction is less than 10%. Accordingly, in our study we calculated with 10% as a maximum value.
K-Ar dating of the black glassy andesite lithic clasts. K-Ar dating of a single BGA sample was performed in the GEOPS laboratory (Orsay, France) using the unspiked Cassignol-Gillot technique, which is especially suitable for dating young samples with low radiogenic argon content (Gillot et al. 26 ; for details, see Appendix 2). The 40 Ar- 39 Ar dating technique was not preferred for the sample, because the irradiation may induce significant 39 Ar recoil from the vitrous groundmass and a high production of interferring 36 Ar from Ca. Five independent argon analyses of different aliquots of the sample yielded ages ranging between 18.7 ± 3.1 and 21.5 ± 2.0 ka leading to a weighted mean (using the inverse of the variance) 27 of 20.2 ± 1.0 ka (Table 2, Fig. 3).

Discussion
While the caldera morphology of Santorini prior to the Minoan eruption has been recently clarified 17 building on the earlier data of Druitt and Francaviglia 15 , the size and age of a Pre-Kameni island have remained tentative. The rationale for our analysis is that a significant fraction of the lithic clasts, predominantly the BGA included in the Minoan deposits, represents Pre-Kameni, whereas the rest is derived from other destroyed parts of Thera and Therasia. Moreover, dating the BGA can give insights into the timing of Pre-Kameni relative to the Cape Riva eruption.
In order to estimate the volume of the pre-Minoan intra-caldera volcanic island, we proportioned the lithic content between the four Minoan units. At first we considered the maximum estimate of 60 km 3 DRE of the on-land and offshore pyroclastic material 4 . Of this, 41 km 3 DRE submarine pyroclastic flow deposits (mostly unit D) corresponds to 54.5 km 3 bulk volume 4 , i.e. the ratio of volume to DRE is 1.33:1. Using this ratio, the on-land DRE values 4 correspond to the following bulk volumes: Plinian fall: 3 km 3 (2 km 3 DRE), on-land ignimbrite: 2 km 3 (1.5 km 3 DRE), and offshore co-ignimbrite ash: 22.5 km 3 (17 km 3 DRE), totalling 82 km 3 bulk volume.
Additional 18-26 km 3 DRE (31-41 km 3 bulk volume, i.e. 36 km 3 on average) of the intra-caldera infill was proposed by Johnston et al. 3 , representing units A, B and C (unit D is considered negligible within the caldera). As for their relative proportions, we used the isopach values published by Bond and Sparks 7 , and Heiken and McCoy 12 . Units A, B and C have average thicknesses of 2 m, 5 m, and 25 m, respectively, i.e. they represent 6%, 16%, and 78% of the intra-caldera infill. Dividing the bulk volume estimate of Johnston et al. 3 by these percentages gives 2.0-2.6 km 3 for unit A, 5.0-6.6 km 3 for unit B, and 25.0-32.8 km 3 for unit C, which are included in the volume values of the Minoan deposits given in Table 1.
This proportioning, along with the ~40% extra lithic content of the finest fraction from granulometric analysis, makes it possible to calculate the total lithics contained in each unit (see Table 1), as well as the BGA (i.e. Pre-Kameni island) volume within the units. An unsolved question is how to divide the 54.5 km 3 bulk volume of submarine pyroclastic flow deposits 4 between unit C and D. As unit C decreases rapidly in thickness away from the caldera, we propose a ratio of C:D = 1:2 or even 1:4 (both calculations are given). The final standard errors in Table 1 include uncertainty resulting from lithic clast proportions and BGA proportions, but do not take into account the uncertainty of deposit volumes and additional lithics proportion, because the latter values are estimates from the published literature.
As our results show, the total volume of the lithics in the Minoan deposits is ~16-17 km 3 , more than twice the value suggested by Pyle 18 . However, within this, the volume of Pre-Kameni, if we envisage that the island was   40 Ar ( 40 Ar*) × 10 10 in number of atoms per gram; age ± 1-sigma uncertainty of each measurement (in ka), and weighted mean age and uncertainty (see text) [40][41][42][43][44][45] . As for the calculated range, we consider that the smaller value of 2.2 km 3 (i.e. a lower C:D ratio) is more realistic due to the overwhelming presence of unit D in submarine settings. However, assuming that parts of Pre-Kameni may have collapsed into the caldera and were not incorporated into the Minoan deposits, this smaller value must be considered a minimum estimate. This may be offset against a possible, albeit unconstrained, decrease in lithic concentrations with distance from source and uncertainties about the precise volume relationships between units C and D. However, it is the unit C ignimbrite which contains the evenly distributed BGA clasts. Unit D, which is thought to be coeval with caldera collapse during the Minoan eruption, contains lithic clasts of BGA interpreted as predominantly remobilised from unit C. Therefore, it can be inferred that most of the Pre-Kameni island, which was located close to the main vents, was incorporated in unit C. Such a Pre-Kameni island, on the basis of a new K-Ar age obtained by the K-Ar Cassignol-Gillot technique, seems to have started to grow immediately after the Cape Riva eruption at 20.2 ± 1.0 ka, similar to the present-day Kameni Islands after the Minoan catastrophe 11,13 . However, the obtained age does not constrain the duration of island growth. The extrusive activity may have declined, and possibly the island was dormant by the Late Bronze Age. This would be consistent with the lack of clasts of BGA with radial jointing (indicative of hot emplacement) in the Minoan products 10 . Nevertheless, this age allows us to calculate a minimum long-term mean lava extrusion rate for the period from 20.2 to 3.6 ka assuming a constant activity: considering the 2.2-2.5 km 3 island volume, the rate is 0.13-0.14 km 3 /ky. However, because this long-term value is only one seventh of the average growth rate during the much shorter lifetime of the present-day Kameni Islands (~0.9 km 3 /ky since the Minoan eruption 19,28 ; Fig. 3), it is also possible that Pre-Kameni grew up in a short period at higher lava extrusion rates.
Finally, we visualize the proposed physiography Santorini prior to the Minoan eruption (Fig. 4) using a combined DEM with 10 m resolution offshore and 15 m resolution onshore 11 . Here, an unconstrained strait in the north, a closed caldera rim in the southwest, a smaller caldera bay (after filling a large part of the present-day southern caldera basin Druitt 10 ; Nomikou et al. 11 ; Athanassas et al. 17 ), and a 2.4 km 3 Pre-Kameni island (with ideal low dome shape) in the centre of the flooded bay, are depicted. The SW caldera rim was completed by adding some contours with interpolation, in order to connect Thera and Therasia. The presented DEM reconstruction, i.e. the proposed 3D view of Minoan Santorini, is considered a combination of the obtained quantitative results and qualitative information on the Late Bronze Age topography.