Controls of magma chamber zonation on eruption dynamics and deposits stratigraphy: The case of El Palomar fallout succession (Tenerife, Canary Islands)

Abstract Anticipating volcanic eruptions at central volcanoes require knowing how magma chambers prepare for new eruptions. Pre-eruptive processes that occur in such magma chambers are recorded in the products of these eruptions, so their characterisation in terms of magma composition and physics offers the clues to understand past eruptions and predict future ones. Here, we study the very well preserved pyroclastic succession of El Palomar Member (712 ± 41 ka), in Las Canadas caldera, Teide, Canary Islands. This deposit resulted from a single explosive eruption of a phonolitic magma that started with a sustained eruption column (sub-plinian or plinian) that formed a massive, 40 m thick non-welded fallout deposit, progressively changing into a lower intensity fire fountain that deposited a 25 m thick fallout succession of non-welded to strongly welded pumices. Stratigraphic and petrological data suggest that this eruption was related to a thermally-compositionally zoned and relatively shallow magma chamber in which the arrival of a hotter and more mafic magma rapidly triggered the eruption. The studied deposit shows how this zoned structure was maintained during the whole process, which allows one to reconstruct what happened during the eruption. Comparison of this eruption with the current situation at Teide volcano alerts on the potential rapid preparation for new eruptions in the case that sufficient phonolitic magma was available in the shallow plumbing system of this active volcano if new inputs of deeper magma take place.


Introduction
A good knowledge of the formation and dynamics of magma chambers is crucial to understanding how magmas evolve and how they subsequently erupt at the Earth's surface. The dynamics of magma chambers prior to and during volcanic eruptions has been widely studied through experimental and theoretical models (e.g., Spera, 1984;Sparks and Huppert, 1984;Sparks et al., 1984;Turner and Campbell, 1984;Huppert et al., 1984;Spera et al., 1986;Blake, 1981a, b;Blake andIvey, 1986a, 1986b;Oldenburg et al., 1989;Rice, 1993;Trial et al., 1992;Tait et al., 1988;Bower and Woods, 1998;Folch et al., 1998Folch et al., , 2001Mourtada-Bonnefoi et al., 1999;Bea, 2010;Dufek et al., 2013, among many others), which have investigated how a magma chamber structures itself chemically and physically and how this structure may remain unchanged or can be partially or totally perturbed during an eruption.
Chamber zonation models have been fuelled by field observations (e.g., Hildreth 1979Hildreth , 1981Cioni et al., 1992;Wallace et al., 1999;Coombs et al., 2000;Bryan et al., 2000;Hildreth and Fierstein, 2012;Marcaida et al., 2014;Weidendorfer et al., 2014) showing that pre-eruptive, broadly continuous zoning is indeed a common characteristic of eruption products of many small to intermediate volume eruptions (e.g., Cioni et al., 1992;Bryan et al., 2000), but is also present in some of the largest caldera-forming eruptions (e.g., Wolff  scale, it is observed that there is an upward increase in the content of lithic fragments in each set, which marks a first order alternation in U1. Its top is eroded, with numerous small size channels and reworking of the pyroclastic material just below this erosion surface (Fig. 4a).
The lower zone (Z1) of U2 (El Palomar Member) covers the U1 paleosurface and has a total thickness of 40 m. It is a massive, grain supported, pumice deposit, composed of angular pumice fragments of several centimetres to a few decimetres in size, immersed in more abundant pumices of a few centimetres in diameter. The largest pumice fragments show a reverse grading throughout Z1. Lithic fragments up to a few centimetres in diameter are scarce, while minor grey pumice fragments with borders of obsidian up to a few decimetres across can be recognised throughout the whole Z1 part of El Palomar deposit. The lowermost portion of Z1 (< 1.5 m) is poorly sorted and comprises dense, poorly-vesicular pumice blocks, 25-30 cm in diameter, with dense, poorly-vesicular pumice fragments (2-6 cm). Rare lithic fragments (< 10 cm) include phonolitic lava and/or welded rock fragments, basaltic lava and obsidian fragments. Toward the top of Z1, welding of the pumices, marked by the collapse of the vesicles in some pumices and the obsidian-like aspect of their inter-porous walls, starts to appear (Fig. 4b), progressively increasing upward ( Fig. 4c) and grading into Z2. In this transition towards Z2, three different kinds of pumice fragments occur: i) white pumices; ii) brown pumices, and; iii) streaky white and brown pumices ( Fig. 4b and 4c).
Z2 and Z3 have an exposed thickness of 8 and 12 meters, respectively. The main macroscopic difference between zones Z2 and Z3 of the El Palomar deposit is the vitrification versus devitrification of the pumices, which gives their black versus white color (Fig. 4d), respectively. Z2 corresponds to a moderately welded pumice deposit, showing a rapid upward increase in the stretching of pumices and degree of welding. It is mostly formed by dense, nonvesicular, slightly sintered, brown pumice fragments (10-15 cm). The brown pumice fragments are increasingly flattened upwards, but the clast-supported fabric is also visible throughout this part of J o u r n a l P r e -p r o o f El Palomar Martí et al 9 the deposit. Lithic fragments are scarce to absent. The contact between Z2 and Z3 is sharper (Fig.   4d). , but still gradual. Z3 is even more strongly welded, (Fig 4e and 4f) with mesoscopic flow structure from ductile deformation. Extensive devitrification has affected Z3. The devitrified deposit (light color) shows an abrupt color change compared with Z2 (brown color). Despite the intense welding, Z3 contains unwelded, dense pumices, around which the welded pumices (fiammae) are deformed by rheomorphism (Fig 4e). Upwards, Z3 acquires a densely welded,.
rheomorphic aspect, the clastic nature of which is still revealed by the occurrence of rare lithic clasts included in the grey pumice spatter and few unwelded pumices (Fig. 4f). As in Z1, Z2 and Z3 deposits are characterised by the absence of any inter-clasts matrix.

Grain-size distribution, bulk density and porosity.
Vertical variations in the grain-size distribution and modal variations were analysed by selecting a total of 12 representative samples along the non-welded part (Z1) of El Palomar Member (Fig. 5). Results are reported in Table 1. This part of the deposit includes certain vertical variations in the size of pumice clasts (Fig. 3) mostly in the middle part, with the largest pumices up to 13-14 cm (Fig. 6). Median diameters (MdΦ) show values between -1.1 Φ and -3.5 Φ. In general, most of the Z1 deposit is poorly sorted (between 2.1 Φ and 2.7 Φ) although in some parts it appears better sorted with values between 1.  processes (Noble, 1965). The main expected effect of such processes is a decrease in total alkali content, and a correlative enrichment in other components, the largest of which will be for the major oxides SiO 2 and Al 2 O 3 . However, other major, trace, and rare earth elements should be essentially unaffected by alteration (Cox et al., 1979).

Intensive parameters
The crystal-poor character of the phonolitic products (≤ 5.5 vol.%) prevented finding enough magnetite-ilmenite pairs to apply the Fe-Ti oxide geothermometer in order to estimate the pre-eruptive temperature (T) and oxygen fugacity (ƒO 2 ). Although eleven pairs were found within the different thin sections of the products, none of them passed the equilibrium test of Bacon and Hirschmann (1988), suggesting either crystal inheritance upon mixing or a thermal perturbation shortly prior to eruption that affected FeTi oxides primarily owing to their propensity to rapidly record new conditions, although not at the same rate for ilmenite and magnetite (e.g., Venezsky and Rutherford, 1999;Scaillet and Evans, 1999). The determination of intensive parameters can still be made by comparing the phase assemblage and compositions of the rocks with available phase equilibrium data from Tenerife phonolites (Andújar et al., 2008(Andújar et al., , 2010(Andújar et al., , 2013Andújar and Scaillet, J o u r n a l P r e -p r o o f 2012). To assess the pre-eruptive temperature of the Palomar phonolitic magmas, we have linearly regressed the variation of CaO, TiO 2 , and FeO* melt contents (calculated as raw Fe 2 O 3 *0.899, Table 2) with the experimental temperature using the phase equilibrium experiments performed by Andújar et al. (2008Andújar et al. ( , 2010Andújar et al. ( , 2013 and Andújar and Scaillet (2012)  the in situ portable XRF raw data are also shown, as derived from eq. 1 (Fig. 8, Table 2). Apart from two extreme T values of 1100-1200°C that were retrieved at the uppermost part of the stratigraphical log, there is a general good correspondence between both sets of temperatures. The two outliers in T come from the CaO contents obtained from the portable apparatus, which form spikes on some of the horizons (Fig. 7a).
Although we could not determine the temperature of the5welded layers (Z2-Z3) by Fe-Ti oxide geothermometry, the similarity in composition of these welded horizons with other tephriphonolitic magmas of the same volcanic system, for which a T of ∼ 1050°C was determined ), which is characteristic of Tenerife phonolitic magmas (Andújar et al. 2008(Andújar et al. , 2013, we can constrain a pressure range of 100 ± 50 MPa for the production of the El Palomar mineral assemblage. These shallow conditions resemble those previously determined for other Tenerife phonolites, and are essentially constrained by the presence of sodalite, a mineral that, along with haüyne, set an uppermost pressure value of 150 MPa for magma crystallisation (Andújar et al.

Discussion
On the light of the data presented above, we can now discuss on whether the physico-  , 1984, , Edgar et al., 2017Granadilla, Bryan et al., 2000), also suggesting that injection of more mafic magmas into a compositionally zoned phonolitic magma reservoir was a common mechanism to trigger Las Cañadas explosive eruptions. This situation has been described at Teide and Pico Viejo (Ablay et al., 1995(Ablay et al., , 1998, so understanding what happened during the construction of Las Cañadas complex may provide the clues to understand and anticipate the future behaviour of Teide-Pico Viejo Zonation of magma chambers is a common process that has been described for many volcanic centres, including phonolitic ones (Vesuvius, Laacher See, Tambora) (Tait 1988;Cioni et al., 1995;Cioni, 2000). Identifying the causes for the preservation of zonation in erupted products is crucial to understanding eruption dynamics and the behaviour of the reservoir prior and during the eruption.
Magma zoning can correspond to the juxtaposition of magmas of contrasted compositions and physical properties in the same reservoir, or to the development of compositional and density gradients in an originally homogeneous magma due to its internal differentiation (i.e., by fractional crystallisation through crystal settling, or by double diffusive convection). The existence of either sharp or smooth chemical gradients has obviously different implications in terms of magma dynamics. While the first case suggests a relatively short time scale for magma coexistence that allows macroscopic identification of blended magmas from hand samples, the second suggests a sufficiently long lasting process to allow magma differentiation. In a similar way, the fact that any of these zonations are preserved or not in the erupted products also implies different eruption dynamics, particularly concerning the way in which the magma chamber withdraws. In this sense, contrasted models implying either a progressive sequential emptying of the magma chamber that allows the compositional stratigraphy of the chamber to be inverted by the eruptive process (e.g., Blake, 1981b;Spera, 1984), or a chaotic withdrawal implying possible overturning and syn-eruptive mixing between diverse magma compositions (e.g., Folch et al., 1998) have been proposed, both being supported by natural examples.
In the case of El Palomar Member, lithological and compositional vertical variations along the whole deposit offer insights into the structure of the magma chamber just before the onset of the eruption and how magma was withdrawn. The thickness of the deposit, the relative large size of the pumice fragments, and the position of the studied outcrop on the caldera border, indicate its proximal character. The whole stratigraphic succession shows a continuous deposition of pumice fragments, without any minor (layering, bedding) or major (erosional surfaces, paleosoils) stratigraphic discontinuity. This suggests that the succession was deposited during a single, sustained eruption, rather than in different pulses or phases. The grain supported character of the whole deposit and the total absence of any inter-clasts matrix, as well as the grain size homogeneity shown at horizontal level, confirms its fallout character. However, variations in grain size, density, degree of welding of the pumice fragments along the studied vertical profile, reveal that some changes in eruption dynamics occurred during the deposition of El Palomar Member.
The almost continuous vertical compositional and mineralogical variations through the main plinian fall unit (Z1 and Z2 in U2) (Fig. 3) argues against any major compositional gaps or abrupt compositional interfaces, and implies a slight pre-eruptive zonation of the phonolite in terms of composition (minor and trace element abundances, mineral chemistry, volatiles) and temperature.
Z1 deposit, which represents the main part of the eruption, is rather homogeneous in terms of lithology and grain-size distribution, which suggests that significant changes did not occur in the chamber and conduit during that phase of the eruption. However, the progressive increase in the degree of welding that characterises the uppermost part of this deposit and Z2 and Z3, is consistent with the withdrawal of progressively hotter and less volatile-rich magma during that part of the eruption, rather than changes occurring in the conduit, as no evidence (variation in lithics content, sudden variation in grain size, etc) exists to support the latter hypothesis.
The lithological and depositional characteristics of the Palomar fallout deposits observed in the field, as well as its grain-size analysis, indicate that these deposits derived from a quasi-steady subplinian to plinian (?) eruption column during Z1, which then waned and changed into a lower intensity fire fountain in Z2 and Z3, which was associated with a progressive lower degree of fragmentation, thus giving rise to the gradual deposition of larger pumice fragments. As already stated, the grain size and density variations of pumice fragments along the stratigraphic profile suggest that this eruption was characterised by a single magmatic episode, as indicated by the presence of highly vesicular juvenile fragments, and a low proportion of fine particles (F1 values between 8% and 23% and F2 values between 2% and 9%; Table 1 and Fig. 5). The lack of internal bedding and having either normal or inverse grading both indicate steady growth of the eruption column without major fluctuations until its climax. This was probably facilitated by the gradual stabilisation of the conduit walls associated with the increasing vent diameter and magma discharge rate (see Houghton and Carey, 2015). The unimodal trend of Z1 samples together with the coarse grain size of the deposits (medium diameter values that are more or less consistent with values between -1.1 Φ and -3.5 Φ; Table 1  show values between 1.3 Φ and 1.8 Φ, typical of dry fallout deposits-although some of them are characterised by sorting > 2 Φ, which might be related in this case to the proximity of the vent (see Houghton and Carey, 2015). Up to the top of Z1 and for the Z2 and Z3, the size of pumice fragments increases progressively. Nevertheless, this apparent decrease in the degree of magma fragmentation supports a variation in the internal conditions of the magma chamber (e.g., a progressive decrease in volatile content, a decrease in viscosity, increase in temperature, etc.) rather than being due to changes in the conduit or other parts of the eruptive system. Moreover, bulk density increases upwards, inversely proportional to porosity, as a consequence of the degree of welding, and this makes difficult to correlate this with a primary density gradient. However, the existence of a density gradient in the magma chamber is supported by the change in composition along the sequence.
Taken altogether, mineralogical and geochemical evidence indicate that the physical gradients observed are controlled by compositional changes in the pre-eruptive magma. The phenocryst content changes upwards in the pyroclastic succession, being less than 1% in the nonwelded part and increasing upwards to 5.5% in the welded part. Moreover, vertical (in stratigraphic order) trace element plots (Fig. 7) also show the existence of gradual vertical chemical variations, which cannot be explained only considering the upward increase in phenocrysts: for instance, by adding ~5.5% crystals in the observed proportions to the composition of the lower part of the deposit, we do not obtain compositions similar to those from the uppermost part (Z3). Finally, T-P constraints support a thermally-compositionally zoned and relatively shallow magma reservoir operating prior to the El Palomar eruption.
The onset of El Palomar eruption is marked by the deposition of a poorly sorted, coarse grained pumice horizon with coarse lithic fragments, which rests on the erosion surface formed at The occurrence of the welded and non-welded pumice in the same layers is a remarkable feature of the El Palomar deposit, which indicates that there was not sufficient time during eruption for the different magma compositions to thermally equilibrate. There is not a limit to the size range for which pumices weld or not, as in the intermediate zone (Z2) of the deposit welded and nonwelded pumices have similar sizes, but the size of deformed (flattened) pumices increases progressively towards the top of the deposit (Z3). However, there are compositional differences between non-welded and welded pumices, being the first similar to those from Z1, while the welded ones correspond to the slightly more mafic compositions of Z2 and Z3 (Zafrilla, 2001). This suggests that fragments from different parts of the magma chamber were erupted simultaneously.
The fact that a broadly constant temperature is expressed by Z1 suggests that a volume of and t is the time of contact in seconds. This allows assessing the depth over which Z1 could have been heated following the Z3 intrusion. Taking the distance between the last measured point in Z2 and the first measured one in Z3 (0.5 m) yields a time duration of about 25 days, which should be taken as a maximum contact time before eruption (given the possibility that the distance could be shorter than the measured one). Alternatively, though injection of tephri-phonolite is our preferred model, it could be considered that the coexistence of two contrasted magma layers reflects the operation of double diffusive convection in the reservoir (Huppert and Turner, 1981). Physical This eruption occurred in a similar way to that predicted by the theoretical model of Blake

Conclusions
Field observations, T-P constraints and compositional analysis point towards a thermallycompositionally zoned and relatively shallow magma chamber structure operating prior to the El Palomar eruption. The eruptive mechanism and magma withdrawal that generated the El Palomar sequence seems to have been related to the arrival of a hot and more mafic magma that triggered the eruption, as documented for many other volcanic centres during the construction of the Las Cañadas central complex and currently at Teide-Pico Viejo stratovolcanoes, and elsewhere. The data reported here suggest a relatively short contact time between resident and invading magmas, on the order of several weeks to a few months, although such an estimate needs to be refined. By comparison, this short time implies that anticipating the next eruptive events produced by magma mixing at Teide-Pico Viejo will require careful and continuous data monitoring in order to be able to detect magma intrusions into the shallow reservoir, such as geodetic or gravimetric changes, before the magma chamber ruptures.      Methodology section). The degree of correspondence between the two measurements depends on the chemical element and on the detection limit and accuracy of each method for each case. It is obvious that there is no correspondence in some elements, but there is good agreement in others. c) The eruption that occurs before mixing and/or overturning in the chamber can affect its internal structure. Grey circles indicate the withdrawal isochrones (see text for more explanation).  s d Si 5 1 6 2 6 1 6 2 5 5 5 7. 6 6 0 6 2 6 4 6 0 6 0 5 0 6 2 5 0 6 6 1 6 4 6 1 6 7 7 1 7 5 7 8 7 3    0 8 N n n n n n n n n n n n n n n n n n n n n n n n n n n n n n n n n n n n n n n n n n n n n n n