Statistical Study on Santorini Pumice-falls

The samples originate from four different pyroclastic levels (pumice falls): 131 samples from the Upper Pumice Series or Bo; 24 samples from the Middle Pumice Series or Bm II; 63 samples from the Middle Pumice Series or Bm I; 117 samples from the Lower Pumice Series or Bu. A statistical study carried out on the analytical results shows considerable compositional differences and four different, easily distinguishable distribution areas.

From a magmatological point of view, two evolutive trends (sub-alkaline trends) are recognizable: both evolve from andesitic towards (rhyo)dacitic terms. The Bu evolutive trend is richer in K₂O and something poorer in SiO₂, probably due to interaction with the metamorphic basement. The Bo evolutive trend, although parallel to the Bu one, is poorer in K₂O and richer in free SiO₂.

The two pyroclastic Bm I and Bm II horizons are chemically similar but different from an evolutive point of view. A rhythmical ascent of more basic melts and some interaction processes with SiO₂-rich surrounding rocks are to be taken into consideration. As a first application of this study to the solution of archaeological problems, it is now possible to clearly characterize and distinguish between pumices from Bo and pumices from Bu.

INTRODUCTION

The stratigraphy, volcanology, petrography and geochemistry of Santorini volcano have been the subject of extensive works by Fouqué (1879), Reck (1936), Neumann van Padang (1936), Nicholls (1971), Puchelt and Schock (1972), Pichler and Friedrich (1980), Pichler and Kussmaul (1980) and Druitt et al. (1989).

The aim of this paper is to add some statistical data obtained from the compositional analysis of more than 330 pumices for 25 elements each.

The pumices were collected from the Upper Pumice Series (Bo), the ignimbrites horizon (or Bm II, as it is called in this paper), the Middle Pumice Series (Bm I) and the Upper Pumice Series. Many samples - especially those from lower topographical levels - have been rejected due to high contents of marine-derived NaCl.

The Upper Pumice Series was sampled in the following localities (Fig. 1, total of 131 samples): the pumice quarries of the Gulf of Ammudi (west of Oia); at Phirostefani (the northern quarter of Phira); the Phira quarries; at the upper stretch of the road leading to Athinios; at the upper stretch of the footpath leading to Balos; on the beach and on the Akrotiri excavations; on the northern part of Therasia.

The Middle Pumice Series (Bm I + Bm II) was sampled in the following localities (total of 87 samples): along the steps leading from Oia to the Ammudi Gulf (Bm II); in the Phira quarries (Bm I); in the Cape Therma quarries (Bm I); along the road leading to Athinios (Bm I).

The Lower Pumice Series was sampled in the following localities (total of 117 samples): in the small harbour of Armeni, below the village of Oia; in the Phira quarries; along the road leading to Athinios; at Balos.

FIRST CONSIDERATIONS

Before discussing the results and their implications, we first address some preliminary considerations. The plinian deposits of the Bu and Bo series contain a small percentage (2%) of basic pumices, which can be easily distinguished on the basis of their greyish colour. These pumices have been previously recognized by Bond and Sparks (1976) and explained as an example of 'magma mixing' by Sparks et al. (1977). All the plinian deposits considered in this paper contain variable amounts (the Bu is richer) of grey cauliform scoriae. Whether these scoriae represent (an end-member) phase of a more basic juvenile magma, a hybrid magma (in this case, the grey pumices should represent the mixed phase), or whether they represent an immiscible basic magma intruded from beneath is a matter of controversy. The chemistry of these scoriae will be the subject of a forthcoming paper. Pollution of pumices by marine aerosol is common, as is evident from the standard deviation value for Na₂O and Cl (see analyses). Moreover, this pollution causes considerable scattering in those plots in which Na is taken into consideration.

The two plinian phases of the Lower Pumice Series are hardly distinguishable from each other, chemically. Therefore, they have not been distinguished in this paper.

DIFFERENTIATION PROCESSES

Pichler and Kussmaul (1972) have outlined the existence - among the Santorinian volcanics - of three differentiation trends: the first shown by volcanics from Akrotiri I and the Kameni islands, but this is not addressed in this paper. The second (which consists ofthe Upper Lavas and the Middle Pumice Series) and the third (represented by the Lower Pumice Series) are addressed in this work. It should be noted that Pichler and Kussmaul's third trend is envisaged by means of two analyses only, whereas we used 117 analyses, showing a clear evolutive trend. In our opinion, a certain problem exists in the case of the second trend. We agree that the Upper Pumice Series shows a separate trend, which, however, includes the Upper ignimbrites (Bm II) but not the Middle Pumice Series (Bm I), as Pichler and Kussmaul maintain. Even in this case, the two above-mentioned authors made use of a limited number of analyses.

The Na₂O-SiO₂ and K₂O-SiO₂ (Fig. 2a) show a wide compositional spreading of Santorinian volcanic products, ranging from basalt-andesites up to rhyolites. The Upper Pumice Series consists, on average, of dacites, but some of its pyroclastics are of rhyolitic composition. On the other hand, the Bo products richer in Fe and Mg cluster in the field of andesites, very close to the boundary of basalt-andesites. The Bo differentiation trend can be defined as sub-alkaline.

The sub-parallel series composed by Bm I and Bm II, although somewhat richer in alkalis, can be also defined as sub-alkaline. The products of the Bm II series are more differentiated in comparison with those of the Bm I series and cluster in the field of dacites. Moreover, the older Bm I series presents a more basic composition which implies ascent of more basic magmas, probably due to volcano-tectonic activity between the eruptions of Bm I and Bm II series.

The Lower Pumice Series (Bu), at least with regard to the less NaCl-polluted samples, shows a sub-alkaline evolutive trend, very similar to those of Bm I and Bm II.

An AFM plot (Fig. 3) shows a continuous compositional variation in which only a weak initial role of the M-components can be noticed. The F-components are evident in the initial stages of differentiation and become prevalent in the intermediate and final ones. This plot does not allow a clear distinction between evolutive trends. More information can be obtained from the MgO-CaO-Fe₂O₃ (Fig. 4), AFM (Fi.g 5) and K₂O-Na₂O-SiO₂  (Fig. 2b) plots. An AFC diagram allows a clear distinction between the Bo and Bu series; the Bo series seems poorer in F at A/C constant ratio. The CaO, MgO and Fe₂O₃ versus SiO₂ plots show a progressive impoverishment in these elements as evolution advances. A decrease in the importance of Mg-rich minerals in the phenocryst assemblage with increasing SiO₂ is noticeable in the MgO-SiO₂ plot, but the different trends are indistinguishable. In the case of the CaO-SiO₂ (Fig. 6) a control of the Ca-rich minerals is evident, but the Ca-rich Bu series is distinguishable. The Bm I and Bm II series are more basic, but a greater scatter of SiO₂content at the same CaO content is recognizable. Similar conclusions can be drawn from the Fe₂O₃-SiO₂ (Fig. 8) and TiO₂ (Fig. 9) diagrams; in both cases, they show a SiO₂ variability under the same Ti and Fe contents. These trends result from differentiation of different phenocryst assemblages under different P, T and P-H₂O

The Bu series shows, in any case, a wide Na dispersion which is not entirely imputable to marine aerosol but also to other factors, such as alteration or other secondary processes. The CaO-SiO₂ and MgO-SiO₂ plots (Fig. 6 and 7) display a molecular CaO/MgO ratio clearly favourable to the alkali-poorer series; in the former a marked CaO decrease is noticeable in the first differentiation. A considerable Al₂O₃ decrease with evolution is also noticeable in the Al₂O₃-SiO₂ plot (Fig. 11), notwithstanding increase in the importance of feldspars with progressive crystallization. The above-mentioned factors, together with the CaO-MgO plot (Fig. 12), lead to the conclusion that Santorinian magma differentiation is mainly controlled by plagioclase + rhombic pyroxene, while the role of monoclinic pyroxene is very limited. Analyses carried out by microprobe display plagioclase (By-An) and rhombic pyroxene crystals concurrence. From an evolutive point of view, the most significant plots are K₂O-SiO₂ (Fig. 13), Rb-SiO₂ (Fig. 14), Zr-SiO₂ (Fig. 15) and the An-Ab-Or system (Fig. 17). The K₂O-SiO₂ plot clearly shows that all the pumice-falls taken into consideration present clear indications of differentiation processes. This is more meaningful given the fact that all the samples analysed originate from plinian horizons whose thickness rarely goes beyond 5 m. This speaks for a considerable compositional gradient (zonation) in magma chambers. The K₂O-SiO₂ (Fig. 13), as well as Rb-SiO₂ (Fig. 14) and Zr-SiO₂ (Fig. 15), allow a clear distinction of two parallel series. The first one (poorer in K₂O) is represented by the pyroclastic products of the Bm II and Bo. The second one, represented by the Bu and Bm I products, is richer in K₂O. The same trends are outlined in the Rb-SiO₂ (Fig. 14) and Zr-SiO₂ (Fig. 15) plots. 

On the contrary, the CaO-Na₂O-K₂O (Fig. 16) and An-Ab-Or normative (Fig. 17) plots show a more complicated evolution. Besides trends which can be explained by a plagioclase fractionation becoming more and more acidic (Bm II to Bo), there are perhaps mixing processes (Bm I to Bo), whereas the transition Bm I to Bm II takes place by means of An-enrichment at Or-constant rate.

CONCLUSIONS

The petrochemical evolution of the Santorinian pumices analysed shows complex volcanological sequences in which crystal fractionation, magma mixing and assimilation play a major role. Three main series have been identified, all of sub-alkaline affinity, ranging from basalt-andesites up to rhyolites. The first one, with low K₂O-content, consists of the medium differentiated Bm II pyroclastics and those - more evolved - of the Bo Pumices. The second one, richer in K₂O, lines up the Bm I pumices and those of the Bu sequence. The third one is represented by the Bm I to Bm II alignment, intersecting the others.

From a chronological point of view, it is possible to draw some conclusions about the evolution of Santorinian pyroclastics, taking into account that some intermediate pyroclastic products are missing, notwithstanding the conspicuous amount of analytical data at our disposal.

The Bu (Lower Pumice Series) pyroclastics show a distribution consistent with a crystal fractionation mainly controlled by By-An-plagioclase, rhombic pyroxene and spinels.

The transition Bu to Bm I and the chemical shifting towards more basic compositions attest to the ascent of deeper melts and probable mixing in shallow magma chambers containing differentiated Bm I melts. Evidence for this mixing includes the absence of parental magmas, the chemical shifting towards more basic compositions and the increasing in CaO, MgO, Fe₂O₃ and TiO₂. Most likely, ascent of hotter melts and re-opening of ancient fractures have been associated with some assimilation of crustal rocks.

The Bm II (Upper ignimbrites) pyroclastic sequence continues the same evolutive trend of the Bm I (Middle Pumice Series), due to SiO₂ assimilation (see SiO₂ chemical scattering) and - to some extent - to crystal fractionation. More synthetically, the Bm I 'horizontally' evolves towards Bm II by assimilation of crystalline basement. It is also true that some isotope investigations are required to distinguish mantle versus crustal causes (Huijsmans). However, assimilation of metamorphic basement was previously envisaged by Pichler and Kussmaul.

The Bo (Upper Pumice Series) pyroclastic sequence contains a wide compositional range of pumice types. The most basic inclusions (but non-parental), those intermediate and the most evolved products belong to the K-poor evolutive sequence. Basic inclusions attest to a relatively long lapse of time between the Bm II deposition and the Bo eruption, enough to allow emptying and cooling of previous magma chambers. Nevertheless, between the more basic products and the more evolved pumices, there exists a significant compositional gap. According to the An-Ab-Or plot magma evolution does not seem to be related to crystal fractionation only. The low phenocryst content of rhyodacitic pumices, the high H₂O-content, the rapid evolution of pyroclastic products, the existence of calderas, are evidence for more complicated differentiation processes, including those of assimilation and re-melting of the crystalline basement.

Finally, as a first application of this study to the solution of those archaeological problems in which sea-borne pumices are to be identified, it is now possible clearly to characterize and distinguish between pumices from Bo (Minoan) and pumices from Bu (60,000-100,000 years old).

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

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