Geochemistry and Structural Control of Hydrothermal Sediments and New Hot Springs in the Caldera of Santorini, Greece

Genetic models show that hydrothermal sources account for up to 90% of the iron in some caldera sediments, even far from the Kamenis. Manganese is selectively lost from the Kameni islands and is enriched in the caldera, while most hydrothermal Ba and Si disappear into the Aegean.

Anomalous iron concentrations in the sediments indicate that a separate submarine hot spring exists in the north-east part of the caldera on the same north-east structure that connects the springs on the Kamenis with the submarine volcano Kolombos 6 km north-east of Thera.

The relation with other caldera deposits along the Hellenic arc is obscure. The Santorini caldera sediments are highly Fe-dominant, and contain only a little Ba and Mn, whereas sediments in the Vani caldera on Milos show Fe-, Mn- and Ba-dominated sections.

INTRODUCTION

Much sedimentary iron and manganese in the caldera of Santorini is assumed to derive from hydrothermal springs of the Kameni islands (Butuzova 1969, 1978; Petersen and Müller 1978; Smith and Cronan 1978, 1983). In the following we use the term 'caldera sediments' for loose deposits that are located more than 300 m from the Kameni islands (Fig. 1). It is not known, though, how much of the iron, manganese and other elements derives from hot springs. The low contents of iron and manganese in most of the caldera sediments could equally well suggest a non-volcanic origin; even epicontinental seas without volcanism like the Baltic Sea have sediments with 10-40% Fe₂O₃, 3-7% MnO and 1000-3000 ppm Ba (Boström et al. 1978, 1983), while most caldera sediments contain less than 0.70% MnO and 400 ppm Ba (Butuzova 1969; Perersen and Müller 1978).

The exhalative-sedimentary nature of some caldera sediments thus needs further corroboration. This demands better data, particularly for Al₂O₃ and TiO₂ than has commonly been available, in order to identify contributions from pumice and hydrothermal sources. In this study we show that the relative role of hydrothermal and detrital inputs in the caldera sediments can be estimated by geochemical studies, but petrographic microscopy is an important auxiliary tool. These results, furthermore, suggest the existence of a new hydrothermal spring in the north-east part of the caldera.

BOTTOM PHYSIOGRAPHY AND SEDIMENT DISTRIBUTION

The caldera of Thera is located on a volcanic arc, which around Thera trends from north-west to south-east, as is shown by the 200 m isobath in Fig. 1. In the caldera the bottom topography has been partly controlled by volcanic processes since 1400 BC.

The old caldera-wall remnants, Thera, Therasia and Aspronisi, form a ring round a deep central basin exceeding 200 m in depth. The young islands of Palaea and Nea Kameni are located on a swell that divides this central basin into four sub-depressions, of which the north basin (NB) is the largest and locally reaches depths of 390 m. The east, south and west basins (EB, SB, WB) are small and exceed the 300 m depth only locally (Fig. 1). The sill depths between the west basin and adjacent depressions are about 270-280 m.

The central parts of the basins generally have almost flat bottoms judging from seismic profiling during recent cruises by the Institute of Geology and Mineral Exploration of Greece (IGME). These areas are underlain by soft sediments, which are about 100 msec in thickness in the north basin and up to 45 msec in the other basins. In the depth interval 0-75 m near the coasts, steep slopes and wave activities prevent major accumulation of fine sediments, except where narrow straits and sills break the waves; thus the two sediment-richest bays on the Kameni islands are facing each other at the strait between the islands, which act as wave breakers for each other. Sediment samplings and sampling attempts show that most soft deposits, rich in fines occur in the open part of the caldera at depths below 100 m as a consequence of these relations (Butuzova 1969, 1978; Petersen and Müller 1978; Smith and Cronan 1978, 1983; Perissoratis et al. 1989).

FIELD METHODS, SAMPLE PREPARATIONS AND LABORATORY TECHNIQUES

The samples were obtained from May 30-June 1, 1987, on board R/V Irini, operated by IGME (see Fig. 1), using a grab sampler. Depending on the circumstances, station positions were determined by means of Loran C and radar.

As a rule only soft top sediments were obtained. All samples were stored in plastic bags to prevent contamination before the laboratory studies. At IGME sample fractions were set aside for analysis at the geochemical laboratory of the Department of Geology and Geochemistry, Stockholm University, where the samples were further subdivided. One unseparated fraction was analysed, and other fractions were leached in dilute hydrochloric acid and hydroxylamin chloride at 50-70° for about 45 minutes to reduce ferric iron and dissolve other hydrothermal components. After leaching, the rock detritus was filtered off and rinsed with some dilute hydrochloric acid; these solutions were analysed by atomic emission spectrometry, using an inductively coupled plasma as excitation source (ICP-AES).

All untreated samples and leaching residues were ground in an agate Retch mikro schnell Mühle, and dissolved by means of lithium-metaborate and hydrofluoric-perchloric acid treatments and were analysed by ICP-AES (Burman et al. 1977, 1978, 1979). The results are given in Table 1; the standard errors are listed since we use t-tests extensively.

Good analytical values are needed for the data processing and the modelling, to avoid large propagation of error effects, particularly for Al and Ti, since these elements are used as conservative members. To check analytical accuracies and precision, reference rocks were analysed in parallel with these samples. Furthermore, duplicate samples were sent to the analytical laboratory of SGAB, the Swedish Geological Company, for independent accuracy-checks of their ICP-AES unit, utilizing analytical techniques identical to those at the Stockholm University.

The data sets (see Fig. 2) show excellent agreements, the mean relative error (RE) being only 0.7% for TiO₂ (or only 0.005% in absolute units) and 5% for Zr (or 10 ppm in absolute units). The plot in Fig. 2 reveals that the precisions apply at the individual sample level; older methods often had REs of almost 22% for Ti or more (Burman et al. 1978). Likewise, an inter-laboratory reference rock study yielded excellent results (Boström 1987), whereas many laboratories that partake in inter-laboratory comparisons of standard rocks report trace element data that may differ with a factor of 10 or more (Roubault et al. 1970; Roche and Govindaraju 1973, 1974; Govindaraju 1987).

RESULTS

Our data (Table 1) agree well with other results (Butuzova 1969; Smith and Cronan 1978). An abundance of pumice in the sediments was observed by Petersen and Müller (1978) and Smith and Cronan (1978). The results (Table 1, Fig. 3) corroborate these conclusions. However, the leaching residues show slightly less iron, manganese, phosphorus, calcium and vanadium than fresh pumice, suggesting that the detritus is also partly attacked by the acids.

We have estimated the non-detrital phase (or excess constituent) by three procedures: 1) analyses of untreated sediment were normalized to the same Al-level as the corresponding acid-leached sediment residues, after which the pumice phase was deduced; 2) the contents of the leaching solutions were re-calculated to solid-sediment equivalents; after this the relative acid-leach losses from the sediments could be assessed; 3) analyses of untreated sediments were Al-normalized to the same Al-level as fresh pumice from the quarries on Thera; pumice was thereafter removed. Procedure one is the poorest, because of several propagation of error effects.

The results of these analyses and processing procedures are given in Fig. 4-11 and in Tables 2 and 3, which include comparisons of the results attained by different methods.

DISCUSSION

       Total concentrations in the sediments:      The north-basin sediments show high contents of total iron, manganese and phosphorus (see Fig. 4a-c). The concentrations are somewhat lower in the east and south basins and lowest in the west basin. Our maximum values in the north basin are displaced towards the north-east for Fe and to the north for Mn; data published by Petersen and Müller (1978) show the same concentration patterns. The distributions of total SiO₂, Ba and V are different (see Fig. 5a-c), with a minimum in the north-east part of the north basin.

       Estimates of the hydrothermal phase:      The distribution of relative acid-leach losses differs considerably from the patterns for the total abundances (see Fig. 6-8). The largest percentages of Fe, Mn, P, Si, Na, Ba and Yb are lost from sediments in the north basin, but much is also lost from sediments in the west basin, whereas the sediments in the east and south basins have small losses. Analogous patterns for V and Cu (Fig. 8b, c) are less distinct, due to the propagation of error effects.

Geochemical maps often fail to show the significance of the change from one point to another. To improve the confidence in the results we arranged the data in 4 groups (Fig. 6-8) according to the level of admixed excess iron, presumably deriving from the Kameni islands: (However, not all excess matter is exhalative; most excess Ca, Mg and Sr derive from carbonates.)

   1) Sediments outside the caldera (SNT 75, 100 and 103; shown as OC in Fig. 9).

   2) Exhalative-poor sediments (SNT 127 and 129-133, shown as EP in Fig. 9).

   3) Exhalative-rich sediments (SNT 128 and 134-138, shown as ER in Fig. 9).

   4) Extremely exhalative-rich sediments (SNT 128 and 134-140, shown as EX in Fig. 9).

Groups 3 (ER) and 4 (EX) overlap considerably. However, EX is a very unhomogeneous group and the removal of SNT 139 and SNT 140 might leave a residual group, not very different from EP. Group 3 (= ER) was therefore established to broaden the test.

The mean excess abundances and the standard errors for these groups (Table 2, Fig. 9-10) show conclusively that procedures two and three give similar results for several constituents. Fig. 9 shows that the excess quantities of Fe, Mn, P, V, and Yb derived both by leaching and by pumice-deduction show remarkably similar trends, indicating that sediments outside the caldera (OC) are generally poor in excess components. The exhalative-poor group (EP) generally has significantly lower abundances of these constituents than the extreme group (EX) and in many cases is even significantly poorer than the exhalative-rich group (ER) in these excess components. Similar trends exist for Na, Si and Ba (Fig. 9c), but outside the caldera this trend is possibly broken for Ba.

Fig. 10 shows that ten constituents, namely Fe, Mn, P, V, Yb, Y, Zn, Mg, Ca and Sr show consistent results by these methods; most of these elements are easy to dissolve in acid and can be determined with good precision. Data for Na and Cu by the different methods co-vary, but still deviate much from the 1:1 line of best fit. The explanation is probably a poor reproducibility for Cu combined with propagation of error effect when the pumice phase is removed. Sodium is probably also selectively leached from pumice in the sediments, and propagation of error effects may occur when the pumice phase is deducted; the most likely excess values for Na possibly fall between the two sets of results.

Ti, Si, Ba, Zr and K were only partly dissolved, which is not surprising; Ti, Zr and Si are generally present in poorly soluble silicates and oxides, and Ba is possibly present to some extent in authigenic barite beside the common host mineral K-feldspar, both minerals being fairly insoluble. Furthermore, the excess Ti calculated by pumice-deduction is not significantly different from zero, as should be the case if Al and Ti co-vary in a pumice phase. The calculated excesses for Si, Ba, Zr and K appear significant, however.

       Geochemical significance of the results:      Fe, Mn, P.    The distributions of total iron and manganese oxides and associated phosphorus (Fig. 4) resemble those presented by Butuzova (1969, 1978) and Petersen and Müller (1978); all high values occur in the north and south basins, whereas the west basin is poor in these constituents. Furthermore, these authors report a local maximum in the strait between Nea and Palaea Kameni outside the most active hot-spring bays; our results suggest the same conclusion (Boström et al. 1990a, 1990b).

However, when the pumice contributions are removed (by leaching or data processing) the local maximum in the south basin vanishes, whereas the major excesses for iron, manganese and phosphorus occur along a broad band from the west basin to the north or north-east basin, covering the stations SNT 128 and 134-140 (Fig. 6-9).

SiO₂ Ba.    Total SiO₂ and Ba (Fig. 5a, b) reach the highest values around the Kameni island swell and the lowest values in the west basin and in the north part of the north basin. These patterns are probably related to the changing volcanism in Santorini; the silica content of the young dacites of the Kameni islands is 10% higher than corresponding rocks of the ring islands (Pichler and Kussmaul 1972). Such effects are to be expected in view of the total content of volcano clastics in the sediments, locally exceeding 90% (Petersen and Müller 1978; Smith and Cronan 1978; Perissoratis et al. 1990). Similarly, the trend in barium data suggests the existence of more silicic volcanoclastics near the Kameni islands; data published by Pichler and Kussmaul (1972) indicate higher Ba-abundances in the dacites than in andesites from Santorini. However, the excess Si and Ba (Fig. 7a, 7c) show distributions that are similar to those for Fe, Mn, and P, suggesting a genetic connection.

Nevertheless, only a few sites show excess silica and barium values exceeding 15% and 10% respectively. This suggests that most hydrothermally derived silica and Ba escape into the Aegean Sea. This conclusion is well based, since barium and silica are identified with great certainty in the hydrothermal solution at Nea Kameni (Boström et al. 1990b).

Na, Yb, Cu.    The distributions of excess Na (Fig. 7b) and excess Yb (Fig. 8a) resemble that for excess Fe, whereas the pattern for excess V differs slightly and the distribution of excess Cu bears no resemblance to that of Fe.

Ca, Sr, Mg.    Excess abundances for Ca, Sr and Mg do not co-vary with those for Fe, Mn, and P, but occur also in sediments outside the caldera; these excesses are probably present in carbonates.

       Statistical analysis and geochemical modelling:      A t-test (Crow et al. 1960) shows that many differences in excess element concentrations for OC, EP, ER and EX (Fig. 9) are significant. The inter-element relations, on the other hand, are not clarified by these graphs, although possible correlations are indicated. The present material forms too small a population (n = 17); in small data-sets freak values may even enhance erroneous correlations (Comrey 1973).

Another approach to study the problem of the provenance of the caldera sediments is to form model mixtures of the probable source components, and check how these mixtures explain the real sediments; this procedure has been used in genetic discussion of meteorites (Mason 1962), sediments (Boström 1976; Boström et al. 1978 a, b) and rain (Andren et al. 1977, Boström et al. 1990a).

One obvious component in such modelling is pumice (Table 3); the second is hydrothermal matter; studies of the extreme hydrothermal sediments at the hot springs on Nea Kameni suggest a realistic phase (NK) that can be used (Table 3).

Tests of a large number of models result in a best fit for SNT 139 with a mixture of 50% NK + 50% Pumice, and for ER with a mixture of 12.5% NK + 87.5% Pumice; see Fig. 11. Table 3 shows how the models explain the real sediments.

A satisfactory model should deviate by less than a factor of two from the real sediment, that is 50 to 200% of a given component. The best model for ER shows a correlation coefficient for the fit between the log-abundances for real and model sediment of 0.995; of 18 components only 4 (MnO, CaO, Cu and Sr) are present in too-small quantities, accounting for only 36-46% of the real abundances in question. Addition of small amounts of an MnO-phase, possibly with associated Cu, and a carbonate phase with CaO and Sr, might correct these shortcomings.

The model for the Fe-extreme sediment SNT 139 becomes poorer if the NK-input changes by ± 15%; the correlation between the model and real sediment is 0.981, using log-abundance data. Of 18 constituents 11 meet the quality criterion assumed above, the largest deficit being shown by MnO, which is only accounted for to 36%, and a surplus being shown by Na₂O (Table 3) and by the lithogenic components Al₂O₃, TiO₂ and Zr. This suggests that the hydrothermal input of iron, manganese and phosphorus has been underestimated in the model.

Two processes probably transfer extra Mn to the deep part of the caldera:

   1) Fe is mainly deposited in the Kameni bays, whereas Mn is selectively lost from the hydrothermal areas (Smith and Cronan 1983; Boström et al. 1990 a, b).

   2) Mn in the sediments is secondarily remobilized and migrates to the surface layers, an idea discussed and rejected by Petersen and Müller (1978). However, these authors and Butuzova (1969) noted the widely occurring reduced layers 10 cm below the surface in caldera sediments, which should mobilize manganese. The presence of Mn-enriched layers at depth in the caldera sediments (Petersen and Müller 1978) does not exclude upward migration of Mn; exceptionally reducing sediments retain Mn as a carbonate, but in intermediary Eh-ranges Mn may be quite mobile (Manheim 1965, 1982).

We conclude that the following processes explain the origin of most sediments in the Santorini caldera:

   a) Erosion on the large islands releases pumice, which is floated or blown over the surface of the caldera; even the Kameni islands have some small beaches of pumice. Much pumice settles on the caldera floor.

   b) Iron and several associated elements, such as P, V, and Zn partly escape from the hydrothermal bays at the Kamenis in a flocculate, which may drift for considerable distances with currents before settling.

   c) Much Mn is selectively lost from the hydrothermal bays, being transported in solution to a larger extent than iron.

   d) Some Mn may be diagenetically released in the reduced sediments.

   e) The spring systems at the Kamenis discharge large quantities of Si and Ba. Much of this escapes into the Aegean Sea.

   f) Biological processes may deposit some volcanogenic silica; the biosphere also deposits much carbonate in the caldera and on the schelves outside the caldera of Santorini.

One might well wonder, however, why the most extreme sediment outside the Kameni swell is located far to the north-east in the north basin, instead of near the Kameni islands. Microscopic studies of this sediment show that more than 50% of the material consists of limonitic, rounded concretions, with a diameter of about 0.5-2.0 mm, and broken fragments which appear to have been parts of crusts; only a minor fraction of the sediment consists of pumice and dacitic-like detritus. The material is most easily explained as an authigenic product (Perissoratis et al. 1990). This fact and the composition (Table 1) suggest that the sediment SNT 139 obtains its metal content from a local submarine hot spring. This suggestion has some support insofar as the site is located on the line connecting the Kameni islands with the submarine volcano Kolombos, about 6 km north-east of Santorini. The most active spring systems on Palaea and Nea Kameni occur along a NE-SW line, a weakness direction proposed already by Fouqué (1879); this zone has been further studied and corroborated by Heiken and McCoy (1984).

Caldera sedimentation varies considerably along the Hellenic arc. In Santorini the sediments are largely ferrugenous, but in the Vani caldera of Milos iron-rich, manganese-rich and barium-rich deposits alternate (Fig. 12). These formations occur as alternations of clastic and chemical sediments and show a maximum thickness of 40 m, overlying fine-grained to porphyritic dacitic lava flows and domes. The manganese oxides, e.g. pyrolusite, ramsdellite and cryptomelane occur as stratified horizons intercalated with barite layers. A late barite-manganese-silica phase cuts across the caldera sediments.

These distinct variations in the geochemistry of these caldera deposits may be due to differences in the chemical compositions of the volcanic rocks in the two islands. In Milos acid lavas and pyroclastics predominate, whereas Santorini is characterized by mafic to intermediate volcanites (dacites). Furthermore, the degree of hydrothermal alteration is very pervasive in Milos (Fytikas 1977; Liakopoulos 1987). These problems are further discussed in Galanopoulos et al. 1990.

CONCLUSIONS

Sediments in the caldera of Santorini show increased concentrations of iron, manganese, phosphorus, vanadium and possibly zinc, that derive from hot springs on the Kameni islands. Most excess values explain some 30-60% of the concentrations found for these elements, suggesting that the volcanic input is considerable. In areas north-east of the Kameni islands these excesses reach very high values, accounting for 80-90% of the iron concentrations, suggesting the occurrence of a local submarine hot spring. Some elements that are distinctly volcanic, such as Si and Ba, deposit only to a small extent in the sediments; most surfacing silica and barium appears to escape from the caldera region. This new hot spring is located near the NE-SW trending tie line between the hot springs on the Kameni islands and the submarine volcano Kolombos, north-east of Thera.

------------------------------------------------

------------------------------------------------