Syn- and post-eruptive origins of leachate spatial features
We identified two distinct deposit features: (i) high mean soluble Ca, Cl, Na and S concentrations per unit mass (Fig. 5) and per unit tephra particle surface area (Fig. 8), in fine-grained samples from blast zone deposits to the north of the volcano, and (ii) a region of low S/Cl, Na/Cl and SSA-normalised soluble S concentrations in the RSSM region (Fig. 9). The effect of leaching time does not compromise the validity of these spatial features; only one location of five in the proximal region includes data from 25 h leaching experiments, and the exclusion of 25 h data in the distal field only serves to magnify the disparity between the two regions. Similarly, the variation in S/Cl ratios observed is slight and cannot account for the observed discrepancy between low S/Cl ratios in the RSSM region and higher values in the surrounding area. We therefore interpret these identified spatial features by reference to our current understanding of gas-tephra interactions.
Proximal enrichments in soluble S and Cl
The proximal enrichments in soluble S and Cl were previously noted by Stoiber et al. (1981) who suggested that this was indicative of interactions with a ‘large magmatic gas component in the directed blast and early in the eruption which decreased with time’. This gas component was considered to be SO2-rich and HCl-poor, in contrast to that erupted in the later stages of the eruption. However, the proximal enrichments in soluble S and Cl can also be attributed to a prolonged period of pre-eruptive gas-rock interactions within the cryptodome. Rock samples taken from the dome formed at MSH in 2004 showed extensive cubic and bleb-like surface deposits on internal surfaces (Fig. 11), morphologically identical to CaSO4 and NaCl deposits formed by high-temperature adsorption on volcanic glass surfaces (Ayris et al. 2013, 2014). Similar deposits were observed on internal surfaces of blast zone deposits from the MSH eruption (Fig. 11). As the cryptodome was emplaced over a period of several weeks (Cashman 1992), the timescale of pre-eruptive gas adsorption may be up to five orders of magnitude longer than during the eruption itself (e.g. Mastin 2007). Crucially, this model can account for the observed leachate feature without necessitating a change in magmatic gas composition during the eruption.
Distal depletions in elemental ratios and SSA-normalised S
The lowest S/Cl, Na/Cl and SSA-normalised soluble S concentrations in tephra leachates are identified in the RSSM region, although the validity of the region of low SSA-normalised soluble S, which may have its onset as far west as Moses Lake, WA, is uncertain. These features coincide with increasingly Si-rich, Ca-poor tephra deposits (Fig. 7b, c), with an apparent locus in the RSSM area. They also coincide with the distal maximum in the deposit mass accumulation, where the heaviest deposition of weakly bound tephra cluster aggregates occurred (Durant et al. 2009; Sorem 1982).
Compositional dependences on leachate chemistry across the tephra deposit have been noted by Hinkley et al. (1987); soluble S and Ca were found to be positively correlated with tephra deposit CaO content and negatively correlated with tephra deposit SiO2 content. Such correlations may be indicative of high-temperature adsorption of SO2 by tephra surfaces in the first seconds after tephra emission (Ayris et al. 2013, 2014). In short duration experiments on silicate glasses, Ayris et al. (2013) observed that high-temperature SO2 adsorption, forming CaSO4, increased with glass Ca content. That study assumed crystal phases to be unreactive to SO2, but Henley et al. (2015) stated that, albeit over longer timescales, crystalline and amorphous anorthite (CaAl2Si2O8) exhibited comparable reactivity with SO2. Under the assumption of the same comparable reactivity, high-temperature SO2 uptake, and accordingly CaSO4 formation, may be broadly correlated with bulk tephra CaO content, irrespective of mineralogy.
The potential for HCl adsorption within the hot eruption plume may be significantly less than that of SO2. Ayris et al. (2014) observed that high-temperature adsorption of HCl was negligible in dacite and rhyolite glasses, attributing this to limited reactivity of Na+ coordinated with tetrahedral [AlO4]− and [FeO4]− groups within those materials. Limited high-temperature adsorption may instead imply that HCl uptake is dominated by scavenging mechanisms acting in the cold volcanic cloud. Based on numerical simulations using the Active Tracer High Resolution Atmospheric Model (ATHAM), Textor et al. (2003) predicted that the high solubility of HCl would result in its rapid dissolution into liquid water and ice, whether as individual hydrometeors or as coatings on tephra surfaces (hydrometeor-tephra aggregates). In their simulations of a large stratospheric eruption, approximately half of all erupted HCl was sequestered into hydrometeors and hydrometeor-tephra aggregates within 60 min of the eruption onset. In contrast, virtually all SO2, being poorly soluble in either water or ice, remained within the volcanic cloud. As Sarna‐Wojcicki et al. (1981a) report that within 1 h of the start of the eruption, tephra deposition was confined to areas west of Yakima, it would be expected that hydrometeor-tephra aggregate scavenging of HCl would be a predominantly proximal phenomenon.
In combination with tephra dispersal, aggregation and sedimentation processes, the SO2 and HCl uptake models proposed can explain the observed spatial features within MSH leachate data. In the high-temperature eruption plume, SO2 would be most efficiently scavenged by the most Ca-rich particles; at MSH, these were dense crystal-rich tephra, more extensively produced during the early eruption of highly evolved cryptodome material than in the later eruption of juvenile magma (Sarna‐Wojcicki et al. 1981a; Scheidegger et al. 1982). In either case, these tephra were preferentially deposited in proximal regions. In contrast, Ca-poor silicic tephra with limited reactivity to SO2 was deposited in the RSSM region. If HCl scavenging occurs in proximal regions and is thus dictated by solubility in water and ice coatings on tephra surfaces, then there should be no spatial trend, other than a dependence on tephra SSA, and hence at least partially on granulometry, in soluble Cl concentrations across the tephra deposit. In this scenario, S/Cl ratios across the deposit would be driven by the variable reactivity of tephra to SO2, mediated by temporal changes within the eruption and by tephra sedimentation patterns. Such an uptake-dependent model explains the low SSA-normalised soluble S and S/Cl regions noted here and also offers an alternative explanation for the varying S/Cl ratios in ‘early’ and ‘late’ tephra noted by Stoiber et al. (1981).
The abundance of Ca-poor tephra in the RSSM region, and hence the particular spatial location of low S/Cl ratios, can be attributed to the formation and fallout of tephra aggregates. Durant et al. (2009) proposed that observations of weakly turbulent mammatus lobes at the base of the volcanic cloud over a wide area (including Ephrata, WA; Moses Lake, WA; and Vantage, WA) implicated bulk settling of the cloud layer driven by ice crystal formation and sublimation at the cloud base. Preferential aggregation of ice-laden, and hence Cl-rich, ultra-fine tephra (p
d 8–31 μm), principally comprised of silicic pumice and glass shards (Carey and Sigurdsson 1982), occurred in this region. Subsequent passage through the 0 °C isotherm caused ice to melt and form a liquid phase, which increased the rate of particle aggregation (Durant et al. 2009). Although Textor et al. (2003) predicted that HCl would be degassed during ice melting and/or sublimation, their models exclude the chemical interaction of HCl with the tephra surface. It may be possible that both prior to freezing and after thawing, acidic liquid films leach alkali and alkaline-earth cations from tephra surfaces. However, in the RSSM region, the limited capacity for leaching of Na+ by HCl in highly silicic glass shards (Ayris et al. 2014) may promote the additional leaching of other cations. These would be ultimately deposited as assorted chloride salts on tephra aggregate surfaces during evaporation of the newly thawed liquid film, resulting in a low Na/Cl ratio in leachate compositions, as is evident in Fig. 9.
Implications for leachate analysis
Standardised analytical techniques
The 2013 IVHHN working group report ‘Protocol for analysis of volcanic ash samples for assessment of hazards from leachable elements’ (Stewart et al. 2013) offers a revised protocol of recommended practices for sample collection, storage, preparation and leaching, to promote acquisition of high-quality leachate compositions which can be more easily compared to that of other studies. Our analysis illustrates the utility of such protocols, as their use would have precluded any assumptions regarding leachate composition comparability. However, we emphasise that standardised leachate protocols do not guarantee a dataset free from analytical artefacts, and thus, should be complemented by secondary supporting analyses. In our interrogation of the MSH data, we noted that the short leaching times used in some studies, comparable to those recommended in the IVHHN guidelines, only achieved partial dissolution of soluble salts. Confidence in the representativeness of these leachates was only acquired via comparison with data derived from longer duration leaching experiments (e.g. Taylor and Lichte 1980; Smith et al. 1983; Jones and Gislason 2008). Our analysis also highlighted the possibility of systematic analytical error in the data of Stoiber et al. (1981), whereby soluble S and Cl concentrations were consistently higher than those of other studies at the same location. As systematic analytical error is difficult to detect, future leachate studies would benefit from a universal reference material, i.e. a well-characterised tephra sample with known leachate composition, verified by independent laboratories, or a synthetic tephra material which can be consistently reproduced in large quantities.
Spatial and temporal variability
The spatial features identified in our analysis demonstrate that small leachate datasets from large tephra deposits can fail to represent the complexities of the wider deposit. Whilst Stoiber et al. (1981) examined seven samples from Yakima, Spokane and Missoula and noted that S/Cl ratios increased with increasing distance from the volcano, our analysis identified a region of low S/Cl ratios near Spokane, east of the WA-ID border. Thus, the inferred trend of Stoiber et al. (1981) is an artefact of undersampling. However, even our collated dataset is subject to sampling density limitations; the deposit margins and most distal deposits were poorly sampled, notably in the heavily forested regions of northern Idaho (Fig. 1). This undersampling may mask unidentified spatial trends, or alter the extent, and hence interpretation, of those already identified. A more extensive leachate dataset with a homogeneous distribution of samples across the deposit would have better resolved the observed, or additional, spatial features.
Additionally, although the uptake-dependent model proposed in the section “Syn- and post-eruptive origins of leachate spatial features” offers an explanation for features noted in time-series leachate compositions (e.g. Stoiber et al. 1981; Hinkley et al. 1987), we note that such data are scarce. It is possible that with a greater quantity of similar time-series leaching, if coupled with sampling of other deposit properties (i.e. chemical composition, mineralogy) that further evidence in support of, or perhaps contrary to, the proposed model, could have been obtained. However, it is crucial to emphasise that in a time-dependent analysis, leachate datasets must still be (a) spatially representative and (b) coupled with detailed analysis of deposit stratigraphy and tephra physical and chemical properties.
In any spatio-temporal interrogation of leachate data, it is vital to consider the influence of local-scale intra-deposit variability. Field duplicate variability of total soluble Ca, Cl, S and Na concentrations per unit mass of tephra was low and may be most strongly influenced by varying SSA and the influence of different leaching times. However, the large variability of S/Cl and Na/Cl ratios (38 %, Table 2), presumed to be independent of these variables, may therefore be indicative of natural deposit variation. This possibility highlights the risk that single tephra samples may poorly represent local deposit variability in leachate compositions, requiring more extensive sampling of each location. In the current study, such data could have validated samples previously considered to be outliers, or identified additional data as anomalous. Current IVHHN guidelines recommend acquiring and compositing multiple samples from an area where deposits appear heterogeneous. For any spatial analysis of leachate compositions, greater sampling may be necessary in all cases, as there is no visible indicator of leachate heterogeneity. Similarly, whilst appropriate for impact assessment, pre-analysis compositing of samples to create a ‘blind’ mean would be undesirable for spatio-temporal or mechanistic interrogations, as it prevents any measure of local-scale variability. This could lead to over-interpretation of small variations in leachate compositions and the conflation of local and regional-scale variability.
It is well established that soluble salts can be dissolved by rainfall, and we accordingly excluded 49 leachate compositions from our analysis. However, the extent to which leachate compositions can be compromised by rainfall bears emphasis. At MSH, Hinkley et al. (1987) reported that tephra recovered from the Ritzville area, which received 45 mm of rain between May 18 and June 18, had lost in excess of 75 % of soluble S and Cl. We additionally compared the concentrations of soluble Ca, Cl, Na and S and the S/Cl ratios, of all 49 unpristine samples to the pristine dataset (Fig. 12). For all elements in unpristine tephra, all concentrations below the 75th percentile of their respective datasets are lower than even the lowest concentrations in the pristine dataset. Furthermore, the S/Cl ratios of unpristine samples are dissimilar to those of pristine tephra, perhaps reflecting the dissolution of S- and Cl-bearing compounds at different rates or in response to varying quantities of rainfall. Neither of these observations can be attributed to the spatial distribution of tephra samples, as the majority are recovered from areas either previously sampled, or in proximity to those areas. Thus, we emphasise that for any quantitative analysis of leachate compositions, the collection of pristine samples is of absolute importance, and echo the recommendations of the 2013 IVHHN guidelines in that researchers must ‘try to collect tephra in a pristine (dry, not rained on) condition’.