Quantitative morphology, recent evolution, and future activity of the Kameni Islands volcano, Santorini, Greece Open Access

A long-standing goal in volcanology is to develop ways of extracting quantitative information about past eruptions that will allow us to develop forecasts of the nature and timing of future activity. In the case of lava flows, a key area of investigation is understanding the relationship between the surface morphology of lava flows, and the bulk dynamics of the erupting material (e.g., Hulme, 1974; Fink 1980; Kilburn, 2004). Understanding this link is critical in general because of the importance of lava in planetary resurfacing, and, specifically, because it allows reconstruction of the detail of past eruptions from the morphology of the emplaced lavas, and underpins forecasts of how future lava flows will evolve.

Here, we present a new high-resolution digital elevation model (DEM) for the volcanic Kameni Islands (Santorini, Greece) based on a light detection and ranging laser radar (LiDAR) aerial survey carried out in April 2004. This is one of the first applications of an aerial LiDAR survey to a volcano, and the unprecedented spatial resolution that the technique offers opens up many new avenues for future work. In particular, the new data reveal details of the surface morphology (on 1–100 m length-scales) of young dacite lava flows, cones and domes, from which important rheological information can be extracted. In combination with a reanalysis of historical eruption accounts, we show how this information can be used to understand the emplacement processes of viscous silicic lavas.

Santorini is one of the active volcanoes of the South Aegean volcanic arc. It has a rich volcanic history, with more than 12 major explosive eruptions recognized over the past 250,000 yr (Druitt et al., 1989, 1999). Typically, episodes of caldera formation on Santorini have been followed by extended periods of lava effusion, leading to the intercalations of andesite to dacite lava piles and andesite to rhyodacite tephra formations that are exposed within the caldera cliffs at the present day (Druitt et al., 1999).

Currently, Santorini volcano is in an effusive phase, with the focus of intracaldera volcanism for the past 2200 yr being the dacitic Kameni Islands, which represent the emergent top of a 2.5 km³ volcano that has a basal area of ∼3.5 km² and that rises 500 m from the floor of the flooded caldera (Druitt et al., 1999). These islands, of which there are currently two (Palea and Nea Kameni, Fig. 1), have been the subjects of extensive petrological, geochemical, and textural investigations over the past thirty years because of their unusually uniform chemical compositions and their contrastingly heterogeneous population of exotic xenoliths and cognate enclaves (e.g., Nicholls, 1971; Barton and Huijsmans, 1986; Higgins 1996; Zellmer et al., 2000; Holness et al., 2005; Martin et al., 2006). There are considerable ongoing efforts to characterize and monitor the state of the Kameni Islands, in particular, the seismicity (Dimitriadis et al., 2005), hydrothermal and fumarolic activity, and ground deformation (Stiros and Chasapis, 2003; Vougioukalakis and Fytikas, 2005). Although the islands are currently in a state of intereruptive repose, there is no reason to suppose that there will not be future eruptions of a similar nature, perhaps within decades, and certainly within centuries.

Since the extensive work of Fouqué (1879), Kténas (1926, 1927), Georgalas and Liatsikas (1925a, 1925b, 1936a, 1936b), and Reck (1936a), which documented the course of many of the historic eruptions in great detail ⁰¹, little attention has been paid to the physical form, or posteruptive morphology and evolution, of the lavas and ash cones that make up the Kameni Island group. Indeed, many of these original works have been overlooked, and the wealth of relevant data they contain has been forgotten.

The historical activity of the Kameni Islands is well known from contemporary written records ⁰¹. Modern compilations (e.g., Georgalas, 1962; Fytikas et al., 1990) generally begin with the work of Fouqué (1879), who produced one of the classic modern scientific accounts of an active volcano and its evolution. Fouqué, in turn, drew on material collated by Pègues (1842), who published an extended account of the historical activity of the islands, and by Leycester (1851), who also published a detailed bathymetric map of the caldera. The 1866–1870 eruptions were closely documented by Fouqué, based in part on his own observations, as well as on the many books and pamphlets that were published within a year or two of the start of the 1866 activity (e.g., Virlet d'Aoust, 1866; von Seebach, 1867; Reiss and Stübel, 1868). The 1866 eruptions clearly drew considerable scientific interest at the time; for example, the eruption was responsible for the earliest medical work on the health effects of volcanic eruptions (da Cologna, 1867). Together, the many published works that describe the major eruptions of the eighteenth, nineteenth, and twentieth centuries present an excellent basis ⁰² from which to develop a quantitative analysis of the evolution of an intracaldera volcano.

The salient features of the Kameni Islands eruptions and their products are summarized in ⁰¹. The Kameni Islands make an excellent case study since the eruptions have exclusively involved dacite lava ⁰², in contrast to the many active lava-forming systems that have been studied in recent decades, which are predominantly basaltic (e.g., Etna, Hawaii; Walker, 1973; Pinkerton and Sparks, 1976). Eruptions typically include both dome-forming and mildly explosive (Vulcanian) phases, and the steady effusion of lava. During the most closely observed eruptions of the twentieth century, flows typically extended to more than 500–1000 m over durations of 30–200 d (e.g., Kténas, 1926; Georgalas and Papastamatiou, 1951, 1953; Georgalas, 1953), with flow-front advance rates ranging from 10⁻³ ms⁻¹ in the early stages to <10⁻⁴ ms⁻¹ after 3 mo, and averaged effusion rates on the order of 0.5–2 m³s⁻¹ (Fig. 2; ⁰³ and ⁰⁴). The erupting lavas are, therefore, classical examples of creeping viscous flows, with low Reynolds numbers (implying laminar flow) and moderate Peclet numbers, similar to those of slow-growing domes (Griffiths, 2000).

Airborne data were collected over the Kameni Islands, Santorini, in April 2004 during an overflight by the UK's Natural Environment Research Council and airborne remote-sensing facility (ARSF) Dornier 228 aircraft. This aircraft was equipped with a WILD RC-10 camera, and Cambridge University's Airborne Laser Terrain Mapper (ALTM) system (Model 3033, Optech Inc., Canada). The aircraft survey took place at an altitude of 650–780 m above sea level (asl) and at a ground speed of 60–75 ms⁻¹. The ALTM LiDAR was operated at 33.3 kHz and used a side-to-side scanning mechanism, which combined with the forward motion of the aircraft to provide swath coverage of the ground surface below. Data from an onboard global positioning system (GPS) and the local GPS base station (at Akrotiri) were used to derive the precise flight path. First pulse and last pulse data were recorded for each airborne laser measurement and converted to location (x, y, z) and intensity data by Gabriel Amable of the Cambridge University Unit for Landscape Modeling. During the overflight, 21 aerial photographs and 4.52 million light-imaging and detection-ranging (LiDAR) measurements were made. In addition, Airborne Thematic Mapper (ATM) data were collected across 11 spectral bands.

The processed LiDAR data comprised 4.52 million measurements over a surveyed area of ∼8 km², including a ground area of 3.86 km². Measured points have a nominal accuracy of 1.0 cm, 1.2 cm, and 5.9 cm in their x, y, and z directions, respectively. The data were imported into a geographic information system (GIS) application (ARCMAP), and anomalous data points (due to cloud cover) were removed. A large area of absorption due to low cloud cover appeared on one flight line (Fig. 3), so for this area there is no precise height information. Instead, this area was patched with a 15-m-resolution digital elevation model (DEM; from IntegralGIS). The LiDAR data were used to construct a high-resolution DEM by interpolating the LiDAR points to an ∼1 m post spacing using a cubic spline function and fitting a surface to these data.

A hill-shade rendition of the DEM is presented in Figure 4, and a contoured DEM is presented in Figure 5. In addition, maps of slope (not shown) were generated to help visualize surface morphological features. This digital elevation model may be referenced either to the UTM (Universal Transverse Mercator) grid, or to standard latitude and longitude coordinates (as shown in Figs. 3 and 4). It should prove a useful starting point for any future analysis of either pre-eruptive deformation of the Kameni Islands or of posteruptive morphological evolution. For the present time, and the purposes of this paper, the DEM allows us to proceed with a detailed quantitative investigation of the surface morphology of a number of key aspects of the Kameni Island lava flows and domes.

During the overflight, a set of 21 color analogue aerial photographs were captured using the onboard Wild RC-10 camera. Together, these images give a mosaic that covers the entirety of the Kameni Islands. Images were scanned digitally at high resolution, internally referenced using fiducial points, and linked to the x, y, z coordinates of identifiable ground control points on the digital elevation model. In turn, this allowed the photographs to be corrected for distortion and terrain effects. An orthorectified DEM-referenced aerial photograph mosaic is shown in Figure 6. This image is complete, save for the easternmost part of one lobe of the 1925 lava flow, which was obscured by a cloud. This mosaic image demonstrates clearly the ease with which the surface textures of the lava flows and ash cones can be identified, and with which flows of different age can be distinguished. We used the aerial photographs, combined with the DEM, in the following sections to develop an updated interpretation of the geological map of the Kameni Islands.

There have been several modern interpretations of the surface geology and eruption sequence of the Kameni Islands, based primarily on the observations and interpretations of Georgalas (1962). More recently, maps were published by Pichler and Kussmaul (1980), Huijsmans (1985), and Druitt et al. (1999). We present an interpreted map in Figure 7, which was prepared using a combination of feature mapping (e.g., using the slopes visualized in the DEM) and visual imagery. The new map is properly georeferenced to the UTM and latitude-longitude grids, and is offset by 280 m (S) and 225 m (W) from the UTM coordinates of the Kameni Island map of Druitt et al. (1999).

The main difference in the new map, compared to that of Druitt et al. (1999), is the interpretation of the lava-flow sequence from the 1939–1941 eruptions in the western portion of Nea Kameni. This is shown in the detail in Figure 8A, along with extracts from two aerial photographs of the same region: Figure 8B is a monochrome photograph taken in May 1944 by a British reconnaissance aircraft, and Figure 8C is the same view from the 2004 overflight. The new lava flows that erupted during 1939–1941 activity can be clearly seen in the two photographs, as can the minor changes in crater morphology associated with the 1950 eruption, and the emplacement of the very small Liatskias dome and lavas, as described by Georgalas (1953).

A second feature of the map is that, in contrast to the earlier mapping of Pichler and Kussmaul (1980), we see no evidence for exposed surficial faults on Nea Kameni. In our map, we do not explicitly identify the areas covered or partially covered by younger ash deposits (e.g., the portions of the 1866–1870 lava flows covered with ash from the 1925–1928, 1939–1941, and 1950 activity), but these areas can be identified from the surface texture and coloration on the aerial photograph mosaic.

The high-resolution DEM (Figs. 4 and 6) reveals a wealth of lava-flow surface textural information, most strikingly, the forms of lavaflow levées and fold patterns. Both of these sets of features, in principle, have scales that are thought to reflect the rheological properties of the underlying flow, and there are widely used models for both approaches (e.g., levées: Hulme, 1974; folds: Fink and Fletcher, 1978). The resolution of the DEM (with a point spacing on the order of 1 m) means that the meter-scale blocks that are often the most obvious surface feature of these and other silicic lavas in the field cannot be resolved.

One of the prominent features of the lavaflow morphology that is visible both from aerial photographs and the DEM is the abundance of gently arcuate ridges on the flow surfaces, which are convex downstream (away from the vent) and lie approximately perpendicular to the flow direction. These ridges are best developed on flows (such as the 1707–1711 flow) that also have well-defined levées. Ridges such as these have been described from a number of terrestrial and planetary volcanic settings (e.g., Fink 1980; Gregg et al., 1998). Analogue experiments, for example, using cooling wax flows, strongly suggest that ridge development happens early in the flow history as the flow surface develops a crust (whether by cooling or by crystallization; Griffiths et al., 2003). In the case of the Kameni flows, field observations show clearly that the ridges formed very close to the vent, essentially as the lava flow was extruded. In the case of the 1925–1926 eruptions, this much is also clear from the contemporary eruption reports (Kténas, 1926; Reck, 1936b). We quantify and interpret each of these morphological features in subsequent sections.

Dimensions of flow levées and surface folds were determined by taking transverse and longitudinal (flow-parallel) sections across flows (Table DR1¹). A typical cross-sectional profile of a flow is shown in Figure 9. This exhibits the classical form associated with a flow levée (e.g., Sparks et al., 1976). Levées are typically 12–30 m high, 30–60 m wide ⁰⁵, with outer slopes of 25–35°. While measurement errors associated with the DEM are small, more significant errors result from the subjectivity of identifying levée margins, from the additional complications where levée dimensions change downstream, or where flows bifurcate, and from uncertainty in the slope down which the flow was emplaced. The apparent viscosity and yield-strength estimates based on lava-flow morphology should only be regarded as order-of-magnitude estimates of flow parameters.

Estimates of yield strengths derived using the three complementary approaches are summarized in ⁰⁶. The yield-strength estimates derived from the gross-flow morphology appear to be the most internally consistent and suggest yield strengths on the order of 30 kPa (1925 flows) to ∼70 kPa (1940 flows). These values are consistent with yield-strength estimates for dacites elsewhere (Hulme and Fielder, 1977; Wadge and Lopes, 1991). Yield-strength estimates based on levée height are similar, in terms of order of magnitude, to those based on whole-flow morphology, while those based on levée width are substantially smaller. The poor agreement between the yield-strength estimates based on levée width and those based on grossflow morphology is consistent with observations elsewhere (e.g., Etna; Sparks et al., 1976) and most likely reflects the fact that the processes important in levée formation and evolution (including accretion, avalanching, and cooling, for example; Sparks et al., 1976) violate the assumptions required for Hulme's (1974) analysis to work.

In addition to the levées, another prominent surface morphological feature of the Kameni lava flows that can be seen clearly in the DEM and aerial photo images is folding. A typical longitudinal section along a flow, corrected for the general slope of the flow field, is shown in Figure 10. This shows folding with wavelengths on a 10–100 m scale, defined by a train of arcuate ridges that runs perpendicular to flow direction. Folds of these sorts have been recognized on a number of length scales on lava flows of all compositions, from basalt to rhyolite, and the usual interpretation is that they form as a result of buckling of a more rigid surface layer (e.g., lava crust) during flow of the underlying fluid. Detailed descriptions and analyses of examples of folding have been presented by Fink and Fletcher (1978), Fink (1980), and Gregg et al. (1998), among others; the motivation for this work, again, was the possibility of inferring flow rheology and compositional parameters for extraterrestrial examples (e.g., Warner and Gregg, 2003).

Nine lava flows were selected for surface folding analysis, based on the visible ridges seen in the DEM and aerial photographs. For each flow, five parallel profiles, running perpendicular to the axis of the folds, spaced 2 m apart, were taken along the middle sections of each flow. Fold scales were investigated by fast Fourier transform (FFT) analysis in MatLab using a custom series of scripts to investigate fold wavelength characteristics. Data were prepared for FFT analysis by interpolating the data to unit spacing, detrending the data stream, and applying a cosine taper (Kanasevich, 1975; Bracewell, 2000) to remove the baseline shift and sinc convolutions, which would otherwise have reduced the data quality. Typical examples of a contoured periodogram are shown in Figure 11, which shows the relative power (arbitrary scale) of fold wavelengths along the length of a 1925 lava flow. This gives a rapid visual assessment of the extent to which fold dimensions change, or otherwise, along a flow. In this case, the fold wavelength is ∼20 m close to the vent and ∼30–40 m further downstream.

Most of the flow profiles share a number of common features: folds form close to the original eruptive vent. Folding is most prominent close to the center-line of the flow, and in many cases there is a consistent pattern of flow generations that develops downstream, i.e., several fold generations of differing wavelength can be recognized. Results are summarized in ⁰⁷. It is clear from ⁰⁷ that most of the flows that show folding have one generation of folds with wavelengths (λ₁) on the order of 15–25 m. When multiple generations of folds exist, second-generation folds (λ₂) have wavelengths on the order of 25–65 m (or 1.5 ± 0.2λ₁), and third-generation folds have wavelengths of 30–110 m, or 1.8 ± 0.3λ₂. This pattern of low ratios (<2) between subsequent generations of folds is consistent with the few data that exist for other dacite (r = 2.1 ± 0.3) and rhyolite (1.8 ± 0.4) flows (Gregg et al., 1998), and contrasts strongly with the high ratios (5.1 ± 1.1) that characterize basaltic flows (Gregg et al., 1998). This confirms the potential of flow fold analysis for the identification of “silicic” flows by remote sensing.

The scale of the folds qualitatively constrains the minimum apparent viscosity of the underlying fluid (Fink, 1980; Gregg et al., 1998). Data from the Kameni Island flows are consistent with the view that in slow, effusive eruptions of silicic magmas, crust formation on the flow by cooling is much faster than crustal thickening by strain, and consequently the folds that form are of long wavelength. The slow effusion of the flows is consistent with the observation that the folds develop close to the vent.

One motivation for the study of the quantitative morphology of terrestrial lava flows of known composition is that it may reveal features that allow investigators to make inferences about the compositions of flows on distant planets, simply from the remote observations of morphology. One such approach that appears to have some promise was proposed by Bruno et al. (1992, 1994), which is to determine the fractal dimension, D, of the flow shape in planform. The technique relies on measuring the gross linear distance around an object as a function of the length scale of the measuring rod. On a log-log plot of distance versus length scale, data following a fractal distribution should define a trend with a linear slope of 1 – D. Bruno et al. (1992) showed that the parameter varies with lava type (a‘a versus pahoehoe) in basalts, but also showed (Bruno et al., 1994) that many more chemically evolved lavas do not show a consistent fractal behavior.

The Kameni lavas present an excellent opportunity to test whether this model is applicable to blocky dacite lavas.

Lava flows with well-defined margins were digitized from the orthorectified aerial photographic images to obtain an outline with a spatial resolution of better than 1 m. Spatial data were exported into a program (Fractals; http://www.nsac.ns.ca/envsci/staff/vnams/Fractal.htm; Nams, 1996) to determine the gross length as a function of length scale, for length scales of 1–100 m. Most of the analyzed lava flows had margins that could be traced for ∼1 km, which limited the upper range of length scale that could be used. A typical plot of gross distance against ruler length, confirming a fractal distribution, is presented in Figure 12. All of the lava flows that were analyzed showed clear evidence for a fractal distribution, and, surprisingly, there was no detectable difference between the fractal dimensions of younger flows (<200 yr old) that form the coastline and those with margins that are entirely exposed inland. The only example that was clearly nonfractal was the profile for the A.D. 726 flow on Palea Kameni, which presumably has been deeply eroded. The fractal dimensions for all of the lava flows that were analyzed are plotted in a histogram in Figure 13. It is clear from this graph that the blocky a‘a dacite lavas of the Kameni Islands have a fractal dimension, D (1.067 ± 0.006, n = 10), that is indistinguishable from that of Hawaiian a‘a, at least on the 1–100 m scale. Thus, fractal dimension alone may be a good discriminator for gross-flow type (a‘a versus pahoehoe), but is not necessarily a good discriminator for composition.

One typical feature of all of the historical eruptions of the Kameni Islands is the growth of lava domes. Field observations of the changing height of the summit of the 1866–1870 (Giorgios) dome and the 1939 (Fouqué) dome were recorded, respectively, by Fouqué (1879), and Georgalas and Papastamatiou (1953). Surprisingly, the importance (indeed, the existence) of these data seems to have been overlooked, and they are summarized for reference in ⁰⁸. The dome-growth data are plotted in Figure 14. Both dome-forming eruptions show closely similar growth rates in terms of dome height (H) with time (t), which parallel those of the well-known lava domes of the Soufrière volcano of St. Vincent (1979 eruption; Huppert et al., 1982), and Mount St. Helens (1980–1986; Swanson and Holcomb, 1990). The power-law dependence of dome height with time is close to the t^1/4 dependence predicted by models of domes that have growth controlled by a crustal yield strength (e.g., Fink and Griffiths, 1998; Griffiths 2000), and the best-fit curve with a t^1/4 dependence is H = 1.2t^1/4. Following Griffiths (2000), this suggests that the yield strength of the crust of the Kameni dacite domes is ∼4 × 10⁷ Pa, which is somewhat lower than for the domes of St. Vincent (1.5 × 10⁸ Pa), Mount St. Helens (1.3 × 10⁸ Pa), or Pinatubo (9 × 10⁷ Pa; Griffiths, 2000).

It is worth noting that while the dome-height data are consistent with a t^1/4 time dependence, alternative relationships cannot yet be ruled out. For example, measurement errors are realistically likely to have been of the order of 1 m, and the start time of the eruption (as opposed to the time when magma emerged above sea level) is not necessarily well known. In addition, Kameni domes often show a transition from endogenous growth early in the eruption, to modification by Vulcanian explosions at later stages (e.g., Kténas, 1926). For these reasons, the possibility of a different time dependence (e.g., t^1/3 for a fixed dome shape and a uniform eruption rate) cannot be ruled out; and in the event of a future eruption, careful quantitative observations of dome growth will be of great value.

The close similarity of the behavior of the 1866–1870 and 1939 domes allied with the uniformity of the composition of the Kameni Islands over the past 2200 yr suggests that we may use the heights of domes from earlier historic eruptions to infer the eruption durations. We summarize the results of this analysis in ⁰⁹. The predictions of both the t^1/4 model and a time model based simply on the empirical fit to the height data are consistent with reports of these early eruptions. Both models predict that the dome will grow to a height of ∼20 m within 1 d, and ∼30 m within 4–5 d. Thus, for a submarine eruption originating on the Kameni-Banco plateau (at ∼20 m below sea level), the dome should emerge above the surface within a day or two of the start of activity. This is consistent with descriptions of the start of the 1707 eruption (Goree, 1710; Tarillon, 1715a; Fouqué, 1879). For domes emerging from this submarine base level to the heights recorded in 1875 of ∼70 m (the 1573 dome), or 101 m (the 1707 dome), expected eruption durations are on the order of 0.3–1 yr and 1–3 yr, respectively. The latter is, again, consistent with the contemporary records of the 1707–1711 eruption.

Simply on the basis of observed height of the Mikra Kameni (1573) dome, it is plausible that the eruption did indeed last for only 1 yr or less. This supports our suggestion that the quoted 1570–1573 age range for this eruption derives simply from an uncertainty in the calendrical date of the eruption, and does not reflect the true duration of the event.

Apart from the anomalous event of 1950, all of the historic intracaldera eruptions for which there are at least adequate records (those since the eighteenth century) have shared a number of important common features: pre-eruptive uplift of parts of the submarine Kameni edifice, and the eruption of magma from at least two eruptive vents. All have also involved the early formation of lava domes, which later act as a focus for vigorous, intermittent explosive activity, as well as acting as the vent from which lavas emerge. There is no reason to suspect that a future eruption would not be preceded by the same general phenomena—including general uplift of the edifice, and discoloration of the sea—and so, given the current monitoring of the caldera, it is highly likely that the next eruption will be anticipated some days to weeks in advance. Some further inferences about a future eruption on the Kameni Islands can also be gained from a consideration of the vent distribution and intereruptive periods.

The distribution of volcanic vents across the Kameni Islands strongly suggests that there is an underlying tectonic control on the supply of magma toward the surface. As previous authors (e.g., Fytikas et al., 1990; Druitt et al., 1989, 1999) noted, the vent pattern defines a narrow NE-SW trend (Fig. 15) that extends northeast of Nea Kameni to include the submarine high (including the former Bancos bank), which divides the flooded caldera basin into two parts. Many of the prehistoric explosive eruptions also have tephra dispersal patterns that are consistent with vent locations along the same trend (Druitt et al., 1989, 1999), and the locations of previous (pre-1866) and current hydrothermal vents around the Kameni Islands also lie along this same trend. Purely on the basis of the activity of the past 1000 yr, one would anticipate that the next intracaldera eruption would start in the region between Palea Kameni (UTM zone 35N coordinates 40291, 3542) and northern Nea Kameni (40302, 3568), and, as long as the eruption lasted more than 2 mo, it would involve two or more eruptive vents.

The present state of the Kameni Islands can most simply be interpreted as a part of the intereruptive shield-building phase, which is a typical feature of the evoluti on of Santorini over the past 300,000 yr (e.g., Druitt et al., 1999). Despite the considerable work that has been completed on the dating of past eruptions of Santorini (Druitt et al., 1999), several of the major explosive eruptions still lack a good age control. Consequently, there are insufficient data from which to quantify the pattern of intereruptive shield-building periods, for example, using a rank-order approach (Pyle, 1998). Instead, we can use the known ages of a few key eruptions to infer the mean interval between all events. The last phases of seven explosive eruptions were separated, on average, by 28 k.y. of repose. Within this sequence, major lava shields were constructed between ca. 67 and 55 ka (the Skaros shield) and ca. 50 and 21 ka (the Therasia dome complex; Druitt et al., 1999). These crude constraints imply that the shield-building phase might last for tens of thousands of years. Assuming a mean interval between explosive eruptions of 20–30 k.y., then from Poisson statistics, there is a 50% probability that the post-Minoan shield-building phase will last between 14 and 23 k.y.

The relationship between eruption length and intereruptive interval for the last 5 eruptions of the Kameni Islands is shown in Figure 16. If we regard the 1950 eruption as anomalously short, and interpret this as a minor extrusion following the 1939–1942 activity, the notable feature is that the remaining four events lie along a line of best fit described by: eruption length (days) = 7.4 × (intereruptive interval, yr) + 578. This striking relationship is consistent with a model for the Kameni Islands of a constant time-averaged deep supply of magma, and with eruption lengths that are determined by available magma volume. Estimates of erupted volumes (Table DR1, see ¹) are consistent with this model; the largest recent eruption (1866–1870) followed the longest pre-eruptive repose period. If this empirical time-dependent relationship holds in the future, then the next Kameni Island eruption will last for more than 2.7 yr (in 2006) to more than 4 yr (in 2070).

A combination of high-resolution digital mapping and aerial photography with archived contemporary eruption reports reveals a wealth of detail relating to the emplacement of viscous, blocky dacite a‘a lava flows on the Kameni Islands, Santorini, Greece.

Lava flows from recent eruptions of the Kameni Islands exhibit the classic surface morphologies associated with viscous a‘a lavaflow emplacement: levées and compression folds. Levée structures, tens of meters wide and tens of meters high, develop close to the vent. Channelized flows within the levées show prominent ridges with ∼20–40 m wavelength and 1–4 m amplitude. Ridges show variable, but limited, evolution in terms of scale downstream, with secondary and tertiary fold wavelengths only 50–100% longer than the first-formed folds. Levée heights and flow widths are consistent with a Bingham rheology, and lava yield strengths are on the order of 3–7 × 10⁴ Pa.

Analysis of the shapes of flow edges confirms that the blocky a‘a dacite lava flows show a scale-invariant morphology, on length scales of 1–100 m, with a typical fractal D value of 1.067 ± 0.006, which is indistinguishable from Hawaiian a‘a. On these short length scales, at least, there may be limitations to the use of fractal dimension for inferring lava composition.

Dome-forming eruptions of the Kameni Islands in 1866–1870 and 1939–1940 showed similar patterns of behavior, with progressive increases in dome height with time (height = 1.2t^1/4) that are consistent with a model of slow inflation of a dome with a strong crust. Lava domes on the Kameni Islands have a crustal yield strength (4 × 10⁷ Pa) that is lower by a factor of 2–4 than the domes at Pinatubo (1991), Mount St. Helens (1981–1986), or St. Vincent (1979). The dome-height model enables us to infer the durations of early historic eruptions on the Kameni Islands. This, combined with the apparent time-predictable nature of volcanic eruptions of the Kameni Islands since 1573 A.D., allows us to suggest that the next eruption of the Kameni Islands will last for more than 2.7 yr (in 2006) or more than 4 yr (by 2070 A.D.), and may involve the formation of a dome of ∼115–125 m and 130–145 m height, respectively.

Data were collected during the Natural Environment Research Council (NERC) Airborne Remote Sensing Facility (ARSF) campaign to the eastern Mediterranean in April 2004. Light detection and ranging laser radar (LiDAR) data were expertly processed in Cambridge, Department of Geography Unit for Landscape Modelling, by Gabriel Amable. We thank David Cobby, Carl Joseph, Ivana Barisin, and the ARSF team, and George Vougioukalakis (IGME), Stathis Stiros, and Aris Chasapis for assistance with data collection; and Vikki Martin for discussion. Pyle thanks staff of the Earth Sciences and University Libraries, Cambridge; the British Library; and Keele University's ‘Evidence in Camera’ for their able assistance with archive material. Elliott thanks Shell (UK) for support toward field costs, and Zoë Rice for field assistance. Part of this work formed John Elliott's final-year undergraduate dissertation. We thank Ricky Herd and Steven Anderson for their detailed and helpful reviews.