The Cerro Blanco sequence (CB) consists of at least seven sub–units, which we have grouped into three packages defined largely by observable unconformities (Fernandez-Turiel et al., 2013, 2014, 2015) (Table 1).

The lowest unit CB₁ is exposed around the outer northern margin of the Cerro Blanco Caldera (Fig. 2), and is made up of angular decimetric rhyolitic blocks immersed in a rhyolitic lapilli and ash matrix with diffuse metre–scale stratification showing reverse grading. The deposits are clast–supported and some blocks exceed 1 m in size. There are not pumice clasts.

The CB₂ unit comprises three sub–units (Table 1). The lower CB₂1 sub–unit mantled palaeotopography and is preserved as patches on hill slopes and above alluvial terrace deposits incised by active erosional channels (Fig. 3d). The thickness of these deposits ranges from 26 to 5 cm at distances of 11 to 200 km from the CBVC (Fig. 6). It consists in alternating, parallel, very thin beds of lapilli and ash (Fig. 7). The CB₂2 sub–unit directly overlies CB₂1. The contact is sharp and not erosional. It consists of an ash deposit without grading, with poorly defined stratification. CB₂2 is locally overlain by alluvium or colluvium and elsewhere the surface has incipient soil developed on top (Figs. 7b, 7c, 7e, 7f and 7i).

The deposit thins from several meters near the vent to ~30 cm thickness at 170 km, but then thickens to > 300 cm thickness at ~200 km (Tafí del Valle area) before thinning once more at greater distances, e.g., ~20 cm thickness at 370 km from the CBVC near Santiago del Estero (Figs. 6 and 7). As a first approach, an isopach map was constructed taking into account the field measurement of the fallout deposit thickness of CB₂2 (Fig. 8). To validate these isopachs, the calculated areas were examined in a semilog plot of the square root of the contour area versus isopach thickness (Fierstein & Hildreth, 1992; Fierstein & Nathenson, 1992; Hildreth & Fierstein, 2012a, 2012b; Pyle, 1989), testing three different models for tephra deposition (exponential, power–law, and Weibull) using the AshCalc code (Daggitt et al., 2014). The lower mean relative squared error was obtained with the Weibull model (Fig. 8) (Fernandez-Turiel et al., 2019; Table S4).

These numerical approaches allow us to have a rough estimation of the volume of the Plinian deposits, despite the uncertainties owing to different features, e.g., thickness reduction by erosion of the upper part of deposits or error associated with techniques of estimating volume of fallout deposits by extrapolating thickness versus distance plots and assuming continuous decrease in thickness regardless of secondary thickening. The Weibull model provides a volume estimate of 172 km³, higher than the power–law model (164 km³) and lower than the exponential model (194 km³) (Fernandez-Turiel et al., 2019; Table S2).

The particle size distributions of CB₂1 and CB₂2 are unimodal in proximal deposits and bimodal in distal deposits (>100 km from source) (Fig. 5b). In bimodal distributions, the lower size mode (~20 µm) does not change with distance (Fig. 5c). Instead, the upper size mode shows a thinning from 600-700 to 100 µm particle size with distance up to ~200 km (Fig. 5c). Figure 5d shows the abundance of coarse and fine populations against distance for CB₂1 and CB₂2 sub–units. The coarse mode is less abundant with distance, but we cannot properly distinguish changes in volume of particles of finest mode (~20 µm) with distance.

The CB₂3 sub–unit filled the valleys north, northwest, and south of the CBVC and can be followed for more than 35 km. It directly overlies sub–unit CB₂1 in outcrops of the southern valleys with a sharp or, occasionally, erosive contact. The thickness reaches 30 m. Although strongly eroded, these deposits are well exposed in the valley of Las Papas (Figs. 1b, 9) where they are largely unstratified, homogeneous deposits with only locally developed, impersistent and diffuse layering of pumice and concentrated pumice lenses. The deposits are whitish, non–welded, and consist of poorly–sorted coarse pumice lapilli in a coarse ash matrix. Sub–rounded clasts of pumice are very abundant reaching ~20 cm in size. Sometimes the clasts show a subtle imbrication within the deposit. Sub–rounded to sub–angular exotic lithic clasts of local metamorphic and volcanic rocks, 1–20 cm in diameter, are abundant at the basis of the unit in many places. Parallel and cross bedded, 0.2 to 2 m thick beds occurs particularly on palaeotopographic highs.

The unit CB₃ consists of four sub–units (Table 1), all of them exposed close or within the caldera of Cerro Blanco. The exceptional preservation of these sub–units is mostly due to the arid environment of the region (Aulinas et al., 2015). The CB₃1 sub–unit comprises crystal poor, very vesicular, rhyolite lava domes. These domes were described in detail by Báez et al. (2016). The largest dome appears in a rectangular shape of 2.7 × 1.4 km, and cut the SW caldera margin. In addition, three other smaller, pinkish, extra–caldera domes are aligned along a trend N100°E that includes the above–mentioned larger dome. The largest dome cut the caldera margin, indicating their post–caldera emplacement and a change in the eruptive style from highly explosive to effusive–moderately explosive.

The CB₃2 sub-unit is made up of 3-30 cm thick, white lapilli and ash deposits organized in poorly to well–defined layers (Fig. 2a), forming the highest peaks of the CBVC, surpassing 4700 m a.s.l.; lava blocks are common within the deposit and, in some cases, show a bread crust structure. The thickness reaches 150 m. These deposits are around the main lava dome, except on the side oriented towards the caldera centre where the lava dome is directly in contact with CB₃3. These deposits are interpreted as fallout from eruptions associated with these structures (Báez et al., 2016). Within the unit also phreatomagmatic deposits with lapilli and ash can be observed, characterized by low–angle cross stratification.

The deposits of CB₃3 form several overlapping lobes reaching ~3.5 km from the main lava dome to the opposite wall of the caldera (Fig. 2), almost completely covering the bottom of the Cerro Blanco Caldera. They consist of 10-100 cm thick beds, diffusely stratified, and made up of decimetric angular blocks of pumice immersed in a coarse lapilli matrix. The total thickness reaches up to 50 m.

The Cerro Blanco sequence ends with 1–5 m thick deposits of siliceous sinter in the middle of the caldera (CB₃4).

A third sequence is defined on basis of the geomorphological, stratigraphical, mineralogical, and chemical characteristics of numerous outcrops occurring in the northwest of Bolsón de Fiambalá (BdF), 40–60 km to the south of CBVC (Fig. 1). We combine these deposits in only one lithostratigraphic unit called BdF₁ (Table 1). It mantles the palaeotopography with a series of unstratified and diffusely parallel medium bedded layers which together make a total thickness between 0.8 and 1.6 m. Some layers are pumice lapilli and others are largely ash. Occasionally, layering is highlighted by alignments of pumice of ~1 cm in size. BdF₁ is locally overlain by alluvium or hillslope colluvium, but also by aeolian deposits of lapilli and ash.

The ash of the CdP₁ unit in the Cueros de Purulla sequence is made up of glass shards and micropumices, with angular to sub–angular blocky shapes containing tubular vesicles (Fernandez-Turiel et al., 2019; Fig. S1). The dominant phenocrysts are biotite, which is the distinguishing feature of these deposits in the field, feldspars (plagioclase >> K–feldspar), quartz, titanomagnetite, and ilmenite. Amphiboles and clinopyroxenes are minor contents. Apatite, allanite–epidote, muscovite, titanite and zircon can be found in trace amounts.

The ash-dominant matrix in the CdP₂ unit is fine to coarse in size and consists of pumiceous glass shards, and very abundant feldspar (plagioclase >> K–feldspar), biotite, and quartz crystal fragments (Fernandez-Turiel et al., 2019; Fig. S1).

The glass compositions of the CdP units are rhyolitic (74.8–79.2 wt% SiO₂; 5.8–10.8 wt% Na₂O+K₂O) (Fig. 10). Feldspars are dominated by plagioclase (Fig. 11). In the different plagioclases, the anorthite (An) decreases from 41.7 to 23.4 mol %, with an increase in orthoclase (Or) from 2.8 to 6.6 mol % (n = 80). The Or content of K–feldspar ranges from 69.5 to 71.3 mol %, with an An < 1.2 mol % (n = 11).

Biotite typically contains 36.9 ±0.8 wt% SiO₂ (mean ±1 standard deviation), 13.7 ±0.3 wt% Al₂O₃, and has a Mg number Mg# = (100 × [mol Mg] / [mol Mg + Fe]) of 59 ±1. The TiO₂ content is 5.0 ±0.2 wt%, and the MnO content is 0.35 ±0.03 wt% (Fig. 12).

Titanomagnetite is more common than ilmenite. The TiO₂ content ranges 5–7 wt% and MnO is ~1 wt% in titanomagnetite, while ilmenite contains ~35 wt% TiO₂ and ~1 wt% MnO (Fig. 13).

The rare clinopyroxene found in the CdP₁ unit (n = 5) are diopsides (Fig. 14). Amphibole crystals are also rare in the Cueros de Purulla sequence and were only observed in CdP₁ unit (Fig. 14). Compositions include magnesio-hornblende and pargasite, according the IMA 2012 recommendations (Locock, 2014). These samples showed values of Al# exceeding 0.21, the threshold at which the amphiboles are inferred to represent xenocrysts of crustal or mantle materials (Ridolfi et al., 2010).

The products of Cerro Blanco sequence showed a large homogeneity in composition, mineralogically and chemically. For example, both CB₂1 and CB₂2 ash sub–units are mainly composed of cuspate and blocky glass shards (~90%), but also contain feldspars (~5%), quartz (~3%), and biotite (< 2%), often as free crystals (Fernandez-Turiel et al., 2019; Figs. S1 and S2). Minor contents include magnetite, ilmenite, clinopyroxene, orthopyroxene, amphiboles, allanite–epidote, muscovite, titanite, and zircon. Lithic fragments are very rare.

The lava domes and the deposits of CB₃2 and CB₃3 are porphyritic and very vesicular; their glassy matrices include quartz and alkali feldspars as the main phenocrysts, and biotite and Fe–Ti oxides as minor phases. There are also rare apatite, allanite–epidote, and zircon phenocrysts. A detailed description of lava domes was perfomed by Báez et al., 2016.

The glass composition of all CB units is rhyolitic (74.6–80.2 wt% SiO₂, normalized to anhydrous contents) with alkali concentrations of 3.1–5.4 wt% K₂O (Fig. 10), 2.1–6.2 wt% Na₂O, and CaO < 0.9 wt%.

In the CB sequence, K–feldspar is more abundant than plagioclase and, while two groups of plagioclases are observed, only one group of K–feldspars can be distinguished (Fig. 11). Both groups of plagioclases coexist in the same sample. In one group of plagioclases, the An content in different crystals decreases from 44.3 to 28.5 mol %, with an increase in Or content from 2.3 to 8.1 mol % (n = 16). On the other hand, the second group shows An content in different crystals decreasing from 19.5 to 10.2 mol %, with an increase in Or content from 4.5 to 7.2 mol % (n = 138). The Or content in different crystals of K–feldspar ranges from 61.6 to 72.5 mol %, with an An content of < 1.3 mol % (n = 259).

The TiO₂–MnO diagram of biotite provides a good chemical discrimination between two groups of biotites of the CB sequence: a high–MnO (1.11 ±0.06 wt%) and a low–TiO₂ (3.5 ±0.2 wt%) group and a low–MnO (0.21 ±0.07 wt%) and high–TiO₂ (4.7 ±0.7 wt%) group (Fig. 12). Both groups can be observed in the same sample. The rest of the major elements show similar concentrations in both groups. The high–MnO group is relatively compact and includes samples from all sub–units of the CB sequence. However, the low–MnO group of biotites occurs only in CB₂.

The CB sequence contains two distinct titanomagnetite and ilmenite populations. Examples of both groups coexist in the same sample and exsolution features are not observed within the grains. One group is compositionally compact, showing minimal variability among CB sub–units with a TiO₂ content 2.6–3.7 wt% and MnO 2.4–3.2 wt% in titanomagnetite, and 41.3–45.4 wt% TiO₂, and 4.7–7.9%, m/m MnO in ilmenite (Fig. 13). The other group was found only in the three sub–units of the CB₂ unit, i.e., those related to the paroxysmal phase, and it contains lower concentrations of MnO in both titanomagnetite (0.3–1.9 wt% MnO) and ilmenite (0.3–0.7 wt% MnO), and wide ranges of TiO₂ in titanomagnetite (1.5–8.2 wt%) and in ilmenite (0.3–0.7 wt%). The abundance ratio of titanomagnetite/ilmenite is ~50/50 in the high–Mn group, while the titanomagnetite dominates ilmenite in the low–Mn group.

The rare clinopyroxene and orthopyroxene in the CB sequence (n = 9) were found in sub–units CB₂1, CB₂2, and CB₂3. They have a greater compositional variability than the pyroxenes of the CdP₁ unit (Fig. 14). They were not in equilibrium with melt, being probably xenocrystic. Amphibole crystals are found only in the CB₂ unit. Compositions include magnesio–ferri–hornblende, magnesio–hastingsite, and pargasite, following IMA 2012 recommendations (Locock, 2014) (Fig. 14).

The tephra deposits of BdF sequence are crystal–poor and lithic–poor made up of pumice, glass (subangular blocky shards) (~50%), amphiboles (~5%), plagioclase (~35%), biotite (~3%), and quartz (~5%); magnetite, ilmenite, apatite, and titanite are scarce (2-3%). We did not find pyroxene in the BdF sequence.

The glass compositions of BdF₁, with values normalized to anhydrous compositions, are rhyolitic (75.6–79.0 wt% SiO₂, with alkali concentrations of 3.5-4.6 wt% K₂O, 3.1-4.3 wt% Na₂O, and 1.0-2.1 wt% CaO) (Fig. 10). In different crystals, the An content of plagioclase of the BdF sequence decreases from 61.5 to 24.5 mol %, with an increase in Or content from 1.1 to 6.6 mol % (n = 65) (Fig. 11). The biotite of the BdF sequence typically contains 37.4 ±0.7 wt% SiO₂ (mean ±1 standard deviation), 13.7 ±0.2 wt% Al₂O₃, and have a Mg number (100 × [mol Mg] / [mol Mg + Fe]) of 59 ±2. TiO₂ content is 4.1 ±0.2 wt% and MnO is 0.24 ±0.05 wt% (Fig. 12). Unfortunately, Fe–oxides were not analysed in this sequence.

Compositions of amphibole phenocrysts of the BdF sequence include magnesio–ferri–hornblende, magnesio–hastingsite, Ti–rich magnesio–hastingsite, and magnesio–hastingsite, according the IMA 2012 recommendations (Locock, 2014). Their A–C variation is close to the CB sequence and quite different from the CdP sequence (Fig. 14).

Feldspar data were used to estimate the temperature. Three pairs of alkali feldspar–plagioclase crystals from CdP₁, using the two–feldspar thermometer (eqn. 27b, Putirka 2008), yield temperatures between 780 and 800 °C for 5 kb.

Coexisting titanomagnetite–ilmenite pairs were evidenced in three samples, one each from CdP₁, CdP₂1, and CdP₂2. Fe–Ti exchange temperatures, calculated using the Fe–Ti two oxide geothermometer and oxygen–-barometer (Ghiorso & Evans, 2008), fall in the range of 750 to 800 °C. Results are also indicative of oxidizing conditions (log fO₂ relative to NNO between 1.2 and 1.4) and elevated TiO₂ activity (a_TiO2 ~0.8) in the magma just prior to eruptive quenching.

The diopsides of CdP₁ unit allowed the use of the equilibrium pairs of clinopyroxenes and melt to determine both temperature and pressure conditions (eqn. 35 for T and eqn. 30 for P, Putirka, 2008). For melt composition, we used average glass composition of samples with clinopyroxene. Mineral–melt pairs were checked to ensure they satisfy the equilibrium condition K_D(Mg–Fe)^cpx–liq = 0.28 ±0.08 (eqn. 35, Putirka, 2008) (Gilbert et al., 2014; Schindlbeck et al., 2014). Clinopyroxene–liquid pairs satisfying the before–mentioned equilibrium condition (n = 5) yielded a temperature of 795–828 °C and a pressure of 3.7–8.3 kb with 0.21 < K_D(Mg–Fe)^cpx–liq < 0.23.

Pre–eruption conditions were constrained using the chemistry of feldspars (Putirka, 2008), Fe–Ti oxides (Ghiorso & Evans, 2008), and amphibole (Ridolfi et al., 2010). Twenty pairs of alkali feldspar–plagioclase from CB₁, CB₂, and CB₃ provided temperatures between 610 and 700 °C (modelled at 1 kb) for plagioclase with An_10–20, and four pairs of CB₂1 and CB₂2 sub–units resulted in 750 to 810 °C (modelled at 1 kb) for plagioclase with An_28–44.

Fe–Ti oxides in CB₁ (n=1 pair) yielded an equilibration temperature of ~590 °C and a log fO₂ value of 1.2 (relative to NNO). CB₂ pairs (n = 10) provided from 580 to 725 °C and log fO₂ of 0.4–0.9. Only one pair of low–MnO titanomagnetite–ilmenite was found and it showed 794 °C and log fO₂ of 1.3. Finally, CB₃3 pairs (n = 3) indicated ~610–630 °C and log fO₂ of 0.7–0.8.

Physical–chemical conditions during amphibole formation were estimated by applying the thermobarometric method based on Al contents resulting in temperatures from 801 ±22 to 1009 ±22 °C, pressures from 108 ±12 to 554 ±61 MPa, water contents of the melt from 3.8 ±0.4 to 7.6 ±0.4 wt%, and log f (O₂) from -8.8 ±0.4 to -12.8 ±0.4. The reliability of this thermobarometer was discussed (Erdmann et al., 2014; Ridolfi & Renzulli, 2012), showing incorrect pressure values but reasonable estimations of temperature and oxygen fugacity (Pignatelli et al., 2016). These higher temperatures in comparison with feldspars and Fe-Ti oxides possibly indicate that these crystals are xenocrysts.

Applying the thermobarometric method for amphibole phenocrysts of the BdF sequence (Ridolfi et al., 2010) yielded temperatures from 793 ±22 to 950 ±22 °C, pressures from 94 ±10 to 317 ±35 MPa, water contents of the melt from 4.1 ±0.4 to 5.3 ±0.8 wt%, and log f (O₂) from –10.4 ±0.4 to –12.4 ±0.4. The pressure values could be incorrect according to discussions on this thermobarometer (Erdmann et al., 2014; Pignatelli et al., 2016; Ridolfi & Renzulli, 2012).

The CdP₁ unit is interpreted as air-fallout deposits from the buoyant eruption plume of a highly explosive eruption (Cas & Wright, 1988). The mineralogical and chemical correlation of proximal and distal deposits showed that ash-fall reached more than 180 km eastward Cueros de Purulla volcano (CdP). Reworking is discarded as the origin of grain–size bimodality, because this feature is observed in primary deposits. This bimodality is interpreted as evidence of aggregation of particles in the eruption cloud (Bagheri et al., 2016; Brazier et al., 1983; Brown et al., 2012; Sorem, 1982). The CdP₂ deposits are interpreted as ignimbrites emplaced in successive eruptive pulses.

The Cerro Blanco sequence (CB) was interpreted largely by observable unconformities that developed in response to caldera collapse during the evolution of the Cerro Blanco Volcanic Complex (CBVC) (Báez et al., 2015, 2016; Fernandez-Turiel et al., 2013, 2014, 2015): the pre–caldera CB₁ unit, the syn–caldera CB₂ unit, and the post–caldera CB₃ unit (Table 1).

Deposits of the lowest unit CB₁ were interpreted as produced by block–and–ash flows from pre–caldera dome structures (Báez et al., 2015, 2016). The mantled topography, the poorly defined stratification or parallel thin-bedding as well as the particle size (lapilli and mostly ash) of CB₂1 and CB₂2 deposits are consistent with air-fallout deposits from the buoyant eruption plume of a highly explosive eruption (Bagheri et al., 2016; Cas & Wright, 1988). The lack of unconformities between CB₂1 and CB₂2 indicates no time gap between the emplacements of the deposits. The angular to sub–angular morphology of the pumice clasts and glass shards indicates a low degree of abrasion during transportation in the eruptive plume. The grain–size bimodality observed in primary distal CB₂1 and CB₂2 deposits (>100 km from source) is interpreted as evidence of particle aggregation and sedimentation of cored ash clusters. They are mostly sub–spherical fragile aggregates, observed with high–speed, high–resolution videos, that have never been observed in the deposits as they typically break at impact with the ground. They consist of a core particle (200–500 µm) fully covered by a thick shell of particles < 90 µm (Bagheri et al., 2016; Brown et al., 2012). The breaking of cored ash clusters and sedimentation of the integrating particles explain the bimodality in the distal CB₂1 and CB₂2 deposits, where the core particles are 100–400 µm in diameter and the shell particles are ~20 µm in diameter (Fig. 5c).

The CB₂3 sub–unit is interpreted as an ignimbrite massive lapilli–ash (mLT) lithofacies with inner transitions to diffuse–stratified lapilli–ash (dsLT) lithofacies, and changes to parallel and cross bedded lapilli–ash (//sLT and xsLT) lithofacies towards the top and on topographic highs (Branney & Kokelaar, 2002) as well as in distal regions. Lateral and vertical variations are related to changes in the pyroclastic flow dynamics and their interactions with topography. The poorly defined stratification indicates pyroclastic flows under high depositional rates, transporting high concentrations of pumice and ash (Branney & Kokelaar, 2002; Cas et al., 2011). The pumice imbrication shows the shear sense within the flow at that location. The interaction of these flows with the valley bottoms produced the ingestion of metamorphic and volcanic clasts from alluvial beds and terraces and incorporation into the bottom of ignimbrite. The change from a poorly defined stratification to parallel and cross–bedded structures evidences the action of traction–dominated flows, especially onto palaeohills. The last deposits were interpreted as stratified topographic veneers (Báez, 2014; Báez et al., 2015).

Although there are no geochronological dates, the stratigraphy indicates that the paroxysmal phase of the eruption (CB₂ unit deposits) was followed by the emplacement of various lava domes (CB₃1 sub-unit) with associated fallout and phreatomagmatic deposits (CB₃2 sub-unit) (Báez et al., 2016). The dispersion direction of the lobes, the structure of deposits, the absence of pumice and exotic lithic clasts, and the same rhyolitic composition shared by all components, all indicate an origin of CB₃3 sub-unit related to successive block–and–ash flows generated by gravitational or explosive lava–dome collapses (Charbonnier & Gertisser, 2008).

Finally, the siliceous sinters in the middle of the caldera (CB₃4) are interpreted as result of amorphous silica deposited around hot springs.

The BdF₁ unit is interpreted as fallout deposits from the buoyant eruption plume of a highly explosive eruption (Cas & Wright, 1988). The overlying lapilli and ash deposits with heterogeneous clasts of regional rocks indicate reworking by water or wind (Figs. 3e and 4). Aeolian pumice–rich dunes are very common in the central and western areas of Bolsón de Fiambalá. Between BdF₁ unit and these overlying sediments there is evidence of a time gap, recorded by an erosional contact and occasional alluvial deposits.

The Nevado Tres Cruces Volcanic Complex, 90 km W of the Bolsón de Fiambalá basin, is the strongest candidate for the source of the BdF₁ deposits. Although proximal samples are not available to test this hypothesis, the geographical distribution, the geomorphological features observed in satellite images, and the petrological information from previous works (Baker et al., 1987; Gardeweg et al., 2000) describing the main trends of the stratigraphy and mineralogy (abundance of hornblende and biotite, and absence of K-feldspars) in the younger proximal dacitic ash-fall deposits, ignimbrites, and lava–domes of the Nevado Tres Cruces complex, favours this edifice as the possible source.

The post-depositional dissection of the block–and–ash flow deposits of CB₁ by the caldera edge allows the interpretation that it is a pre–caldera unit from an extrusive dome that is compositionally linked to CB. Therefore, this unit is included in the CB sequence assisted by its similar petrography and geochemistry to the other CB units that differentiate them from the older products of the CBVC (Báez et al., 2015; Montero-López et al., 2010b).

The lack of geochronological data does not allow estimation of the time gap between CB₁ and CB₂, i.e., the caldera collapse. The absence of lithic breccias or pyroclastic products around the caldera edges, as well as the low lithic content in pyroclastic products of CB₂, suggest that a single vent in the SW of the caldera rim acted as emission conduit during the eruption, favouring optimal conditions for a sustained Plinian column and fallout deposition.

The fall deposit CB₂1 of alternating layers, 1-3 cm thick, some of lapilli and some of ash are the first record of the Plinian column. The lower thickness of this deposit among CB₂ sub–units seems to be consistent with a relatively brief episode in comparison with those generating the overlying sub–units. Thin bedding suggests instability because of changing column heights and therefore variable vertical wind patterns (Cioni et al., 2000). The change to the thicker, unstratified rhyolitic ash (CB₂2) indicates that the feeding system reached and maintained steady conditions during a single climactic column phase (Cioni et al., 2000; Pardo et al., 2012). The eruption column height for this phase reached 27 km, according the results of Tephra2 modelling, with an ESE dispersal on a large surface of the Chacopampean Plain (Fernandez-Turiel et al., 2019; Fig. S3).

Bimodal particle size distributions, as observed in CB₂1 and CB₂2, are a common feature in many distal ash deposits, such as in the eruptions of Campanian Ignimbrite (CI) 39 ka BP (Engwell et al., 2014), Pinatubo in 1991 (Wiesner et al., 2004), and Chaitén in 2008 (Durant et al., 2012; Ruggieri et al., 2012; Watt et al., 2009). The occurrence of these particle size subpopulations in fallout deposits was explained by the propensity of aggregation in historic and ancient eruptions, independently of their style (Brown et al., 2012; Durant et al., 2009, 2012), but also by other processes without the need for aggregation, e.g., particle differences in texture and componentry, convective instabilities in eruptive plume, and lower atmospheric circulation patterns (Durant et al., 2009; Watt et al., 2015).

An interesting new approach to explain the particle size bimodality is through particulate cored clusters. They are sub–spherical fragile aggregates consisting of a core particle of hundreds of microns in size covered by a thick shell of particles of tens of microns in diameter (Bagheri et al., 2016). These aggregates are formed in the buoyant column and plume of the eruption. These ash clusters tend to disaggregate on impact with the ground and, as consequence of their very low preservation potential, they are difficult to find and document in the historic and ancient ash-fall deposits, although their presence can be inferred by polymodal particle size populations. In this way, the bimodality of the Plinian CB₂1 and CB₂2 sub–units from 100 km away from the Cerro Blanco through more than 350 km (Fig. 5c) could be due to the clustering phenomena of the ashes described in the literature.

A distal mass accumulation maximum or secondary thickening was observed in many historic ash-fall deposits. For example, in the eruptions of Quizapú, Chile in 1932 (Hildreth & Drake, 1992), Mt. St. Helens, Washington, USA in 1980 (Carey & Sigurdsson, 1982; Sarna-Wojcicki et al., 1982), Unzen, Japan in 1991 (Watanabe et al., 1999), Mt. Hudson, Chile in 1991 (Scasso et al., 1994), Pinatubo, Philippines in 1991 (Wiesner et al., 2004), Crater Peak, Alaska, USA in 1992 (McGimsey et al., 2001), and Chaitén, Chile in 2008 (Watt et al., 2009). This phenomenon has been related to ash aggregation in the eruptive plume (Brown et al., 2012; Carey & Sigurdsson, 1982). Aggregation may be enhanced by the freezing of water in the high troposphere and the melting of the ice in aggregates is postulated as the mechanism that triggers their fall en masse (Brown et al., 2012; Mastin et al., 2016).

The distal mass accumulation maximum of the Plinian CB₂2 sub–unit reached several meters thick ~210 km from the volcano, in the eastern ranges that surrounds the southern Puna (Figs. 6, 7e, 7f, and Fernandez-Turiel et al., 2019; Fig. S4). Semi–arid conditions prevail west of this area, while warm, moist, and rainy conditions prevail over the easternmost slopes of these ranges and adjacent lowlands. The climate of Central Andes is characterized by a complex interplay of large–scale atmospheric circulation with local orographic effects (Garreaud et al., 2003; Garreaud, 2009; Neukom et al., 2015), and the main moisture source for the region is the easterly influx from the Amazon Basin. In this setting, the eastern upslope moist warm flow could accentuate the ice melting in aggregates, collapsing their structure and triggering ash-fall en masse.

The secondary thickening related to upslope moisture transport by easterly atmospheric flow, should also elucidate the presumable secondary thickenings of the Cueros de Purulla eruption near Cafayate, where there are several metre-thick fallout ash deposits of CdP₁ (Fig. 3b), and the BdF₁ ashes in the northwest of Bolsón de Fiambalá (Fernandez-Turiel et al., 2019; Fig. S4). However, the warm and moist easterlies hardly reach these areas implying that these deposits have not necessarily affected by wet disaggregation and that other process may be highly significant in driving the fallout of particles. If we observe the location of the secondary thickenings for CdP, CB and BdF fallout units in topographic profiles (Fernandez-Turiel et al., 2019; Fig. S4), the affected areas are in the eastern slopes of important ranges. This fact allows us to consider in the conceptual model the effect of the topographically induced turbulences in the disaggregation, e.g., the breaking of lee waves, generated by winds passing over elevated topography beneath the eruption plume (Watt et al., 2015). As a consequence, this spatial association of the secondary thickening and topography could help to locate affected areas in similar cases. It may also help to prevent hazards associated with the increased loading of ash on infrastructures and buildings (Wilson et al., 2014), because it allows us to constrain more precisely the most hazardous areas during large explosive eruptions around the world and, particularly, in the Central Volcanic Zone of Andes, which are the eastern slopes of the mountain ranges bordering the Altiplano–Puna plateau.

The CB₂3 PDC’s maybe were partially synchronous with the CB₂2 ash-fall. Many Plinian columns generate regional fall deposits and large-volume ignimbrites at the same time. Marginal collapse forms PDC’s while central column sends up regional fall (e.g., Bishop Tuff, 1912 Novarupta) (Hildreth & Fierstein, 2012a). The transition to non–optimal conditions for a buoyant column, possibly with a significant increase in discharge rates, may also produce large PDCs preceding caldera collapse (Legros, 2000).

Thus, on the whole, the CB₂ unit follows the idealised sequence of a caldera–forming eruption, displaying a Plinian fall deposit overlain by an ignimbrite (Druitt & Sparks, 1984; Legros et al., 2000). However, in the Cerro Blanco caldera this model is more complex, as is revealed by the two different compositions of plagioclases, biotite, and Fe–Ti oxides observed in the CB₂ syn–caldera sub–units which seem to indicate mixing of two magma batches during the eruption. The compositional homogeneity of CB₁ and CB₃ deposits suggests a single magma body. Within the paroxysm of the eruption, however, it is likely that another significant magma input occurred, recorded in the compositional bimodality of CB₂ deposits. The most frequent compositions in CB₂ are also the same as the most common observed in CB₁ and CB₃ (Figs. 10–14). A possible explanation for the minority compositions could be the mobilization of hotter magma, evidenced by geothermobarometry data, from a different part of the magmatic chamber. Similar mixing–derived bimodality was observed in titanomagnetite compositions of magmatic products from, for example, Mono Basin in California (Marcaida et al., 2014) and Usu volcano in Japan (Tomiya & Takahashi, 2005).

There are no geochronological data to determine the time gap between the syn–caldera CB₂ and the post–caldera CB₃ units. The CB₃ unit represents a significant change in the eruption style of the CBVC, from an explosive phase (CB₂) to an essentially effusive magmatic phase characterised by the extrusion of the CB₃1 lava domes in the SW edge of the Cerro Blanco caldera, and the occurrence of different fallout episodes of pyroclasts and block–and–ash flows related to them. Block–and–ash flows were topographically confined towards the caldera’s interior, filling the bottom of this 5 km wide circular structure almost completely (Fig. 2a). These flows involved both single and multiple–collapse events such as those documented in the eruptions of Unzen in Japan, Colima in Mexico, Merapi in Indonesia, Arenal in Costa Rica, and Montserrat in Lesser Antilles (Charbonnier & Gertisser, 2008; Freundt et al., 2000). A post–eruption geothermal field produced white sinter deposits within the Cerro Blanco caldera. The associated hot springs, located in the central part of the caldera, are active in the present day (Viramonte et al., 2005). Finally, the Cerro Blanco caldera area showed a deformation pattern subsiding about 1-3 cm/year at least since 1992 (Pritchard & Simons, 2004, Brunori et al., 2013, Henderson & Pritchard, 2013, Yazdanparast et al., 2017).

Only proximal and medial fallout deposit thickness data, reaching 400 km eastward, are available for the Cerro Blanco eruption, which is common in many prehistoric deposits (Bonadonna & Houghton, 2005). Consequently, large uncertainties in volume estimates are expected. Despite this, the order of magnitude obtained for bulk volume estimates always exceeds 100 km³, i.e., the threshold value for a volcanic explosivity index or VEI of 7 (Fernandez-Turiel et al., 2013, 2014, 2015). Taking as a point of reference the density of the 2008 Chaitén rhyolitic ash deposits, that ranged from 1,000 to 1,250 kg·m^–3 (Alfano et al., 2011; Watt et al., 2009), we have assumed a deposit density of 1,000 kg·m^–3 excluding lithics. With this value and a dense–rock density of 2,300 kg·m^–3, the 172 km³ of bulk volume amounts to a dense–rock equivalent (DRE) volume of 75 km³. Based on a model of the pre–eruptive topography, the volume of ignimbrite associated with Cerro Blanco accumulated 15 km³ without lithics (Báez et al., 2015), corresponding to 8.5 km³ DRE, assuming a bulk density of 1,300 kg·m^–3, and a dense–rock density of 2,300 kg·m^–3. Based on these results, the Cerro Blanco eruption is among the largest volcanic eruptions of the Holocene globally, similar to Kuril Lake (Kamchatka, Russia), Crater Lake (Oregon, USA), Tambora (Indonesia), Samalas (Indonesia), Aniakchak (Alaska, USA), Santorini (Greece), or Changbaishan (China–North Corea) (Brown et al., 2014; Crosweller et al., 2012; Lavigne et al., 2013; Vidal et al., 2015). It clearly exceeds the magnitude of the 1600 Huaynaputina eruption, the largest historical eruption known in the Central Volcanic Zone of the Andes (Global Volcanism Program, 2013b; Stern, 2004; Tilling, 2009).