I welcome the opportunity to continue the discussion of my mass analysis methodology, and I would like to take this opportunity to acknowledge the workers who have contributed to the extensive literature on the mapping and geochemistry of Columbia River Basalt Group (CRBG) lava flows. Notably, the junior authors of the comment by Baker et al. (2019) have produced much of the formative work on CRBG chemostratigraphy over the past several decades. To each of them I express my appreciation for providing the foundation on which I have built my own research.
In their comment, Baker et al. state that careful sampling by previous workers, and sample selection through examination of numerous thin sections, have resulted in a geochemical database for samples from the Sentinel Bluffs Member of the Grande Ronde Basalt that are mostly unaffected by secondary alteration (e.g., Reidel and Valenta, 2000). Moreover, they state that their geochemical methods are “rigorous,” “well established,” and have a “long history” of application in the study of CRBG lavas. They claim, however, that my mass analysis method (see Section 3) is “flawed,” “invalid,” and “arbitrary,” and that “it seems improbable that samples with fresh outcrop appearance and commonly glassy and unaltered mineralogy in thin section could simultaneously show chemical trends resulting from considerable mineral dissolution.”
In reply, I will show that both anoxic alteration and oxic surface weathering of CRBG lavas are prevalent both east and west of the Cascade Range, that my methodology is based on quantitative techniques that have been used in the analysis of weathered rocks and soils for decades (Nesbitt, 1979; Brimhall et al., 1992; Anderson et al., 2002), and that my own careful field and laboratory sampling, as documented in Sawlan (2018), can result in an unweathered and relatively unaltered sample collection, whereas those samples previously collected in SB lavas (e.g., Reidel and Valenta, 2000) have not only experienced significant anoxic water-rock interaction with groundwater but have subsequently undergone appreciable oxic weathering at the ground surface.
2. COMPARISON OF MASS ANALYSIS AND ROCK WEATHERING EXPRESSIONS
The mass analysis methodology introduced by Sawlan (2018) applies a fundamental term, the parent:sample immobile element ratio, which is incorporated in all prior expressions used to quantify low-temperature chemical alteration from water-rock interaction (e.g., Nesbitt, 1979; Brimhall et al., 1992; Anderson et al., 2002). This ratio equals the sample mass fraction remaining after alteration (if any), and its application as a multiplier to element abundances is termed mass-normalization (Sawlan, 2018). Baker et al. claim that these expressions are flawed and invalid, but they do not substantiate such claims. To dispel any doubt about the validity of the mass analysis methodology, here I show how these expressions relate to the expression widely applied in evaluating chemical effects of weathering.
Baker et al. contend that Equation (1) accounts for bulk density and volume changes and thereby assert that Sawlan’s (2018) “data correction methodology,” apparently referring to mass-normalization, is invalid because this expression does not include density and volume terms. This claim by Baker et al. is plainly incorrect: a cursory examination of Equation (1) shows that there are no terms for either volume or density.
The two left terms, , correspond to the parent:sample immobile element ratio, which was defined by Sawlan (2018) as the sample mass fraction (remaining), Ms, or, expressed as a percentage, the sample mass index, MIs. Because immobile elements (IE) are not transported in low-temperature water-rock interaction, their abundances in normalized analyses (to 100% volatile-free) of altered rock change only from depletion (or addition) of mobile elements (ME), which causes a change in sample mass. IE do not “accumulate,” as described by Baker et al., but their increase in normalized abundances is simply an artifact of the analysis process whereby mass-depleted samples are analyzed as a unit mass. The multiplication of normalized element abundances by the sample mass fraction, Ms, reverses this inherent normalization and gives element abundances for a sample’s mass-depleted state. As explained in Sawlan (2018), for normalized analyses of mass-altered samples IE abundances “change” as the reciprocal of the remaining sample mass fraction, 1/Ms, and such changes are predominantly positive due to sample mass loss from mineral dissolution.
For IE, the product of the sample mass fraction, Ms, and sample abundances, , removes the analytical process artifact and yields their parent abundances because these elements are conserved. For ME, this product also removes the analytical process artifact but gives their abundances as depleted (or enriched) without the normalization in proportion to depleted abundances.
The two terms in Equation (5), in addition to mass-normalized abundance, convert the mass-normalized abundance to a fraction of the parent abundance, and reference element-abundance changes to zero instead of 1. Thus, depleted and enriched ME have negative and positive tau values, respectively, and IE have a tau value of 0. Although unessential, these same two terms facilitate visual comparison of alteration-generated changes for multiple elements in relation to a unit parent abundance, such as in a vertical weathering profile. Excluding these terms allows direct examination of ME abundances (i.e., in wt% or ppm) remaining after mass changes from alteration and enables the determination of important relationships between mass-normalized abundances and remaining sample mass. These relationships will be discussed below along with the differences between alteration in reducing anoxic groundwater alteration and oxic surficial weathering, which can be used to identify the environment(s) in which the rock samples were altered.
3. PARENT COMPOSITIONS AND ALTERATION INDEXES
For mass-normalized IE abundances of Sentinel Bluffs (SB) and other CRBG lavas to correspond to parent abundances, and be applicable to making precise chemostratigraphic distinctions, parent compositions must be determined from independent criteria. Sawlan (2018) quantified SB parent IE abundances from a negative Al2O3n-TiO2n magmatic trend derived from Cpx+Pl fractionation in subequal amounts. In addition, given that the dominant effect of alteration is mineral dissolution resulting in mass loss, which leads to increased IE-normalized abundances, those samples having the lowest IE abundances along the fractionation trends represent parent magmatic compositions.
Baker et al. do not acknowledge the fundamentally different influences from magmatic and alteration processes on sample chemistry, documented by Sawlan (2018), but instead evaluate Sawlan’s SB samples using mafic indexes of alteration for reducing (MIAr) and oxic conditions (MIAo) (Babechuk et al., 2014). They assert that unaltered CRBG lavas have “typical” values for alteration indexes (MIAr = 25–26; MIAo = 39–40), but they do not explain how they determined that samples with these values correspond to parent compositions. They also fail to account for the differences in magmatic compositions that cause variations in parent alteration index values.
Figure 1 shows a plot of the mafic index of alteration for reducing conditions, MIAr versus sample mass index, MIs, for the SB analyses reported in Sawlan (2018). In this plot, samples are assigned symbols according to Sawlan’s (2018) SB chemical series identified using IE ratios and mass-normalized abundances, and the physical stratigraphy. These data show a negative correlation between MIAr and MIs values, having a slope of approximately −0.4 over the 9 wt% range of MIs; each unit change in MIAr therefore is equivalent to 2.5 wt% mass loss. In their table 1, Baker et al. report MIAr values rounded to the nearest integer, which implies that by their criteria, “unaltered” samples can be distinguished only within an uncertainty of ∼2.5 wt% mass loss.
SB parent magmatic compositions are indicated for the samples in Figure 1 between the vertical dotted lines at MIs values of 99.75 and 100.25. These sample values are used to define the MI100 magmatic trends (Sawlan, 2018) and span a range in MIAr of 24.6–25.9. Differences in MIAr values for samples from different SB series having MIs values of ∼100 correspond to differences in magmatic compositions that are not accounted for by the MIAr alteration index. For example, SB Series V samples generally have higher MIAr values compared to the earlier SB lavas (Series I–IV). This relative offset is maintained with decreasing MIs values indicating that the reactive minerals were similar in composition and dissolved in similar proportions among all SB lavas (Fig. 1).
The dashed line with positive slope at MIs ∼98 in Figure 1 divides those samples with lower MIs, identified by Baker et al. as altered, from those with higher MIs values that they considered unaltered. The anti-correlation of MIAr and MIs values is maintained among samples within SB series having MIs values of 100–97 indicating that many samples deemed “unaltered” by Baker et al. have actually lost as much as 2 wt% of their original mass from dissolution. Baker et al. interpret these samples as unaltered because of their inability to identify parent compositions in reference to magmatic processes.
A negative correlation between MIs and MIAr values is not surprising given that both values account for the loss of mobile elements from alteration, although in different ways. The sample mass index, MIs, accounts for ME loss through the increase in IE abundances relative to those of unaltered magmatic compositions. The mafic indexes of alteration (Babechuk et al., 2014) are ratios of one or two IE (Al2O3 ± Fe2O3) divided by the sum of the same IE and selected ME, which can be summarized by the ratio IE/IE+ME (usually expressed as a percentage). As ME are depleted by alteration, the denominator value (IE+ME) decreases relative to the numerator, and the ratio value increases. MIs values are referenced to a magmatic value of 100% and decrease due to ME loss from alteration. On the other hand, MIA values are referenced to a value of 100 that corresponds to complete ME removal by alteration but magmatic values are undefined. Thus, with sample MIA values, there is uncertainty in parent MIA values because these are not referenced to compositions that are demonstrably magmatic in origin.
The use of MIA values is suited to characterizing relative differences in the degree of alteration for samples from an individual lava that were variably altered under the same conditions (e.g., Babechuk et al., 2014). When applied to samples whose parent magmatic compositions differ, they can give misleading results. Furthermore, MIs values are independent of alteration conditions, but additional information is needed to determine which index (MIAr or MIAo) is appropriate. As will be shown, this determination can be complicated by samples that have experienced both oxic and anoxic alteration.
4. Ti-Fe VARIATIONS
Baker et al. plotted Ti-Fe variations of the Sawlan (2018) samples in which all samples were given the same symbol and noted that a negative trend expected by them between the immobile Ti and mobile Fe (for anoxic conditions) was not apparent. Figure 2 shows Ti-Fe variations for SB sample analyses given in Sawlan (2018), with samples identified by SB chemical groups. These groups include compositionally distinct lavas or possibly two flows differing only slightly in composition. The positive Ti-Fe trend at right is defined by ∼70% of samples that have high MIs values (≥99) and corresponds to a magmatic trend resulting from Cpx+Pl fractionation (Sawlan, 2018). Baker et al. postulate that Ti and Fe should be anticorrelated if the variations were due to alteration (under reducing conditions), and that Ti and Fe should be enriched and depleted, respectively. Such an inference assumes, however, a single parent composition, ignores the chemostratigraphic units (SB chemical series and groups) identified by Sawlan (2018), and presumes that alteration-generated Ti-Fe variations exceed magmatic variations.
Given the large range of magmatic Ti-Fe compositions, compared to the limited range of compositional changes caused by relatively minor sample mass loss (<9 wt%), an overall negative Ti-Fe trend among all SB samples should not be expected as predicated by Baker et al. With the identification of samples by SB chemical groups in Figure 2, negative Ti-Fe trends are readily apparent among samples from individual SB groups that span a substantial range of mass loss (generally 5 wt% or more; e.g., Groups 7, 11, and 15). With the identification of magmatic compositions, the influences of alteration on lava compositions are clear. It is also apparent from the Ti-Fe relationships between parent and variably altered samples of SB groups, as well as from figure 10 of Sawlan (2018), that Fe loss is a major contributor to sample mass loss under reducing anoxic conditions.
Sawlan (2018) applied a rigorous sampling protocol that limited samples to unoxidized lava internal to alteration rinds, which yielded 70% of samples showing less than 1 wt% mass loss, with 30% of all samples showing no detectable mass loss (MIs = 100 ± 0.25). The previously analyzed SB samples, mostly compiled by Reidel and Valenta (2000), however, exhibit a continuum in MIs values between MI100 and MI96 (Sawlan, 2018; figure 13 therein), but there is no record of the protocol used in collecting these samples. An attempt to assess whether samples are altered or not from Al-Ti outliers in this older SB data set using an outlier identification approach would incorrectly designate samples that have lost as much as 4 wt% of their mass as unaltered.
In disregarding Sawlan’s (2018) SB chemostratigraphic framework, Baker et al. present a confusing interpretation of the Ti-Fe variations shown in their figure 3. They attribute the positive Ti-Fe correlation among high-MIs samples to “some variation in the abundance of Fe-Ti oxides in these lavas,” and they speculate that I (Sawlan, 2018) would have interpreted the positive Ti-Fe trend as being caused by “leaching of Fe in these samples under reducing conditions.” This, however, is incorrect in both respects. As discussed above, the positive Ti-Fe correlation does not result from Fe-Ti oxide dissolution (under reducing conditions), which instead generates negative Ti-Fe trends. The positive Ti-Fe trend among SB lavas overall is a magmatic characteristic resulting from fractional crystallization of an assemblage lacking Fe-Ti oxide, such that both Ti and Fe are concentrated in the fractionated magmas.
Moreover, Baker et al.’s interpretation that the positive Ti-Fe trend reflects differences in the abundance of Fe-Ti oxide among SB lavas implies that Fe-Ti oxide was redistributed in SB magmas prior to eruption. This interpretation is clearly incorrect. Fe-Ti oxide has been reported only as a groundmass phase in SB lavas and was not observed as a phenocryst phase in the sequential fractional crystallization (FC) experiments on tholeiitic magma having compositions similar to those of SB lavas (Villiger et al., 2007). Both Fe and Ti are concentrated in residual liquids from fractional crystallization of magma having sufficiently low oxygen fugacity to suppress Fe-Ti oxide crystallization. As discussed by Sawlan (2018), the low Fe-Mg distribution coefficient in clinopyroxene and other ferromagnesian silicates (∼0.3 ± 0.03) results in Fe enrichment in the residual magmatic liquids from which such minerals have been removed by crystallization.
5. Ti-Mg VARIATIONS
Sawlan (2018; figure 12 therein) showed that low-MIs samples from SB chemical groups have higher TiO2n and lower MgOn abundances for each group than high-MIs samples (parent abundances), and, as for the Ti-Fe variations, attributed these differences to alteration under reducing anoxic conditions. Baker et al. interpret these same sample analyses using alteration index values and a AF–CNK–M (AF–Al2O3 + Fe2O3; CNK–CaO + Na2O + K2O; M–MgO) ternary plot (their figure 2A). They conclude that these samples “do not display any sign of having undergone significant alteration” (except for the one oxic-weathered sample) and that they “show no evidence of Mg loss.” Baker et al. therefore claim that mass-normalization (though not identified by name) consists of unjustified “arbitrary adjustments” to the analyses.
The ternary AF–CNK–M diagram shown in Baker et al.’s figure 2A is intended to illustrate the effects of oxic weathering but is inappropriate for the included SB samples of Sawlan (2018), which were altered (if at all) under anoxic conditions in which Fe (as Fe2+) was mobile. With all three apexes including mobile elements, the SB samples plot in a small cluster because losses of mobile elements cancel each other and not because these samples are unaltered as claimed by Baker et al. As shown by Sawlan (2018; figure 12 therein), MgO is clearly depleted by alteration under reducing anoxic conditions.
Baker et al. subsequently present a much different interpretation based on negative tau values calculated for Fe and Mg (from unidentified parent compositions), outliers identified in the Ti-Fe sample distribution, and a trend away from the FM apex in a A–CNK–FM ternary plot, in which samples underwent Fe and Mg depletion during alteration under reducing conditions. This interpretation directly contradicts their earlier erroneous claims noted above.
6. FORENSIC CHEMICAL ASSESSMENT OF ALTERATION ENVIRONMENT AND SAMPLING PROTOCOL
Given Baker et al.’s assertion of having applied diligent sampling procedures, I have examined the SiO2–MIs–FeO relationships for sample data mainly from Reidel and Valenta (2000), and also from Reidel (2005) and Derkey et al. (1999, 2004). The overall data set includes analyses of samples from Basalt Waste Isolation Project (BWIP) studies, which were discussed by Landon and Long (1989) but not included in their paper.
Figure 3 presents plots of SiO2 versus MIs for samples from this earlier SB data set assigned to Series II–IV (Sawlan, 2018; supplemental file 1 therein), with data points color-coded by FeO abundance. MIs values for these samples were determined using parent abundances established from a magmatic Al-Ti “baseline” (Sawlan, 2018; see figure 13 therein). This data set includes many samples showing alteration-generated mass loss, mostly <4 wt% (MIs > 96), but as much as 12 wt% (MIs ∼88). Because sample mass changes from mineral dissolution could occur from either alteration under reducing anoxic conditions in groundwater or from the oxic conditions of surficial weathering, this discussion will focus first on identifying the conditions of alteration.
The plots shown in Figure 3 leverage the limited variations in parent magmatic SiO2 and FeO abundances among SB Series II–IV lavas, as well as the large differences in alteration-generated depletions for SiO2 and FeO between reducing anoxic and oxic conditions. Parent SB Series II–IV lavas (MIs = 100 ± 0.25) span only 0.6 wt% SiO2n and 0.8 wt% FeOn (Sawlan, 2018; supplemental file 2 therein). The limited range in magmatic SiO2n results from Cpx+Pl fractionation in subequal amounts, which generates only minor Si enrichment and only slightly greater Fe enrichment (Villiger et al., 2007).
Figure 3A shows normalized SiO2 abundances (SiO2n) from the older SB data set as a function of sample mass index (MIs), with samples color-coded by FeOn abundance. Overall, these samples define a distinctive pattern, a fanning array from a narrow range in SiO2n at MIs ∼100 to both higher and lower SiO2n with decreasing MIs, such that SiO2n abundances span nearly 4 wt% at MIs ∼93. FeOn abundances decrease sharply with decreasing MIs and increasing SiO2n along the upper bound of this array. FeOn abundances, however, remain near parent FeOn abundances and SiO2n decreases with decreasing MIs along the lower bound of this array.
To understand the origin of this sample distribution, variations in normalized and mass-normalized SiO2 and FeO abundances as a function of MIs are examined for samples from individual lavas altered exclusively under either reducing anoxic or oxic conditions. Samples from the higher-Ti SB Group 7 lava, which were altered (if at all) under anoxic conditions, are plotted in Figures 4A and 4C. Samples from a basalt of Dodge member lava, Wanapum Basalt (herein, “Dodge”) (Thomson et al., 2014), which were altered under oxic weathering conditions, are plotted in Figures 4B and 4D. The Dodge parent composition differs from that of SB compositions in having ∼2 wt% lower SiO2 and FeO (with total Fe as FeO) and may not be precisely applicable to SB compositions; however, the Dodge samples illustrate the fundamental differences in chemical changes resulting from oxic alteration as compared to anoxic alteration.
Figure 4A shows that normalized SiO2 abundances (open symbols) for the mass-depleted higher-TiO2 SB Group 7 samples (altered under anoxic conditions) increase with decreasing sample mass. This increase in SiO2n abundances from mineral dissolution under anoxic conditions may seem counterintuitive, but the increase is an artifact of the analysis process which imparts an inherent normalization of analyses for mass-depleted rock to 100% in proportion to depleted abundances. When the aggregate SiO2 abundance of the dissolved minerals is less than that of the parent (MI100) abundance, the SiO2 depletion from alteration is exceeded by the increase from the inherent normalization discussed earlier, resulting in an increase in normalized abundances. In Figure 4B, which shows the Dodge samples altered under oxic conditions, SiO2n decreases slightly with decreasing MIs, which indicates that the SiO2 abundance of the dissolved minerals exceeds that of the parent.
The mass-normalized SiO2 abundances, SiO2mn, shown in Figure 4A (solid symbols), clearly show that SiO2 is depleted, not enriched, relative to the parent abundance for alteration under anoxic conditions. The mass-normalized SiO2 abundances for the oxic-weathered Dodge samples (Fig. 4B, solid symbols), however, show that SiO2 is more strongly depleted under oxic conditions than under anoxic conditions.
For both anoxic and oxic alteration, the changes in SiO2n and SiO2mn with MIs are linear, which indicates that steady-state mineral (±glass) dissolution is the predominant influence on the chemistry of altered, mass-depleted samples. The slope of the variation of mass-normalized abundances of any element as a function of MIs gives the fraction of the total mass removed by mineral dissolution. For instance, a slope of 1 would indicate that the element accounts for all mass loss, and a slope of zero would indicate that the element accounts for none of the mass loss (i.e., for an immobile element). As such, SiO2 accounts for 29 wt% of the material transported under anoxic conditions but accounts for nearly double this amount, 55 wt%, under oxic conditions (Figs. 4A and 4B). The linear ME depletions with MIs in both environments not only enable the identification of those ME that have been depleted but enable the weight percentage of the total material removed to be quantified.
Figures 4C and 4D show equivalent variations for FeOn and FeOmn with MIs for the same samples as in Figures 4A and 4B, respectively, and illustrate the difference in Fe mobility between anoxic and oxic alteration. FeOn is strongly depleted with decreasing MIs in the reducing anoxic environment (Fig. 4C, open symbols), and the slope of FeOmn–MIs variations, 0.50, indicates that FeO accounts for half of the total mass loss (Fig. 4C, solid symbols). For oxic alteration, FeOn increases with decreasing MIs (Fig. 4D, open symbols), but the approximately zero slope of FeOmn–MIs variations (Fig. 4D, solid symbols) shows that Fe effectively behaved as an IE.
Baker et al. generally attributed Fe and Mg depletion in the SB samples of Sawlan (2018) to ferromagnesian mineral dissolution. The high proportion of FeO of the total mass loss for anoxic alteration, ∼50 wt%, far exceeds that of any silicate in SB lavas and clearly indicates that the Fe depletion results primarily from Fe-Ti oxide alteration in which Fe is transported and Ti is retained. While Baker et al. emphasize thin-section examination to identify samples lacking alteration, the alteration of opaque oxides would only be apparent with the examination of polished sections using reflected-light microscopy.
These differences in Si and Fe behavior between the anoxic and oxic alteration environments relate to differences in phase reactivity in these two environments, and, for Fe, differences in the solubility of Fe2+ and Fe3+. The difference in Fe solubility exerts a major influence on the composition of the material transported from mineral dissolution, and, because this difference results from an integral difference in its oxidation state, there is not a continuum of element depletion amounts between anoxic and oxic alteration.
Returning to the SiO2-MIs-FeO variations for the earlier SB data set (Fig. 3), the divergent SiO2n limiting trends of the fanning array shown in Figure 3A are readily explained by alteration exclusively under either reducing anoxic or oxic conditions. This explains the paired decrease in FeOn and increase in SiO2n with decreasing sample mass, MIs, along the anoxic alteration trend (e.g., Figs. 4A and 4C), and the persistence of high FeOn and the decrease in SiO2n with decreasing MIs along the oxic alteration trend (e.g., Figs. 4B and 4D). The mass-normalized SiO2n abundances, SiO2mn, demonstrate, however, that SiO2 is depleted in both alteration environments but at a much higher rate under oxic conditions (Fig. 3B).
Most samples from the older SB data set plot between the limiting trends of anoxic and oxic alteration (Figs. 3A and 3B), however, and their analyses cannot be explained by alteration solely under either condition. The FeOmn abundances of such samples (Fig. 3B) provide clear insights into their alteration history. The samples plotting between the limiting trends form bands of samples having subequal FeOmn abundances subparallel to the oxic alteration trend indicating that these samples were “weathered” under oxic conditions under which Fe is immobile. This is more explicitly illustrated in Figure 3C, which shows regressions of sample FeOmn in 0.5 wt% intervals, the same intervals used to color-code the data points in Figure 3B. For samples with FeOmn < 11 wt%, the regressions intersect the anoxic limiting trend at progressively lower MIs values. This indicates that samples between the limiting trends require a two-stage alteration process in which the samples were initially altered under reducing anoxic conditions and subsequently altered under oxic conditions.
7. GLASS REACTIVITY IN ANOXIC AND OXIC ALTERATION ENVIRONMENTS
Baker et al. also emphasize that the presence of “fresh, unaltered” glass in samples, as identified from petrographic examination, indicates a lack of chemical alteration. In addition, they suggest it is “improbable” that “commonly glassy” samples could show chemical trends resulting from mineral dissolution.
Figure 5 compares the variations in K2Omn abundances as a function of remaining sample mass (MIs) for samples altered under reducing anoxic conditions (in groundwater) (Fig. 5A) and under the oxic conditions of surficial weathering (Fig. 5B). Samples plotted in Figure 5A are the higher-TiO2 SB Group 7 samples, and those plotted in Figure 5B are the Dodge samples (Thomson et al., 2014) identified earlier. The SB samples altered under reducing anoxic conditions generally have higher K2Omn abundances than high-MIs samples, which indicate non-systematic K enrichment on the order of several tenths of a weight percent compared to high-MIs samples. This shows that glass reactivity under reducing conditions is minimal, such that it is not dissolved to any appreciable extent. It is also consistent with the low depletion of SiO2, which comprises only ∼29 wt% of the material transported for anoxic alteration (Fig. 4A), whereas glass compositions in SB lavas are dacitic to marginally rhyolitic (∼65%–73%) (Hoover and Murphy, 1989).
In contrast to the K enrichment shown for reducing anoxic conditions, the K2Omn abundances for the Dodge samples are strongly depleted with decreasing MIs (Fig. 5B), which, given that the K is concentrated in groundmass glass during lava solidification, clearly indicates glass dissolution during oxic surficial weathering. Glass dissolution in this environment is also consistent with higher SiO2 depletion associated with oxic weathering, which, as shown earlier, is about twice the rate for alteration under anoxic conditions.
The K2Omn-MIs variations (Fig. 5A), together with the SiO2mn-MIs variations (Fig. 4A) show that glass is minimally reactive and is not dissolved to an appreciable extent under reducing anoxic conditions. The use of petrographic identification of “fresh” glass, therefore, is not a diagnostic criterion for establishing whether or not a sample has undergone chemical alteration. Samples altered under oxic weathering conditions, however, undoubtedly show evidence of glass dissolution and, likely, partial replacement by secondary minerals. Such samples can be readily identified in hand specimen from the obvious visual cues (color, texture, and secondary minerals) associated with oxic surficial weathering. Furthermore, Baker et al. cite only 766 thin sections of SB samples compared to 2268 SB samples analyzed. This disparity indicates that thin-section examination was not consistently applied in evaluating SB samples for chemical analysis.
8. INTRAFLOW CHEMICAL VARIATIONS AND LAVA FLOW MIXING
In their figure 1, Baker et al. present a vertical profile showing TiO2 and P2O5 abundances for samples from a borehole in the Pasco Basin; this figure was presented first in Reidel (2005) and has been presented in multiple papers since. Despite having ample opportunity to present the underlying data and appropriate documentation (e.g., stratigraphic position, accurate location coordinates, analysis date, sampling protocol, sample context, and sampling geologist), Reidel (±coauthors), and now Baker et al., still have not presented the analyses and documentation that would allow independent evaluation of these data.
Sawlan (2018) showed that 16 compositionally distinct SB groups within five chemical series can be distinguished by IE ratios and abundances of samples that have retained their original mass, or nearly so. Moreover, mass-depleted samples can be related to specific chemical groups from their mass-normalized IE abundances as well as to the broader SB series from their IE ratios. Following Reidel’s (2005) lead, Baker et al. base their interpretations of lava flow chemical heterogeneity entirely on variations in the normalized abundances of Ti and P, both of which can be modified by alteration. Furthermore, they hypothesize that four separately and simultaneously erupted lavas flowed long distances and eventually comingled and mixed in the Pasco Basin area, by “multiple-flow injection” according to Reidel (2005). They claim that this process resulted in flow intervals or “layers” (Reidel, 2005) distinguished by their Ti and P abundances, within the SB Group 11 flow (Sawlan, 2018), which was informally named the Cohassett flow during the BWIP studies. Notably, Baker et al. do not present an alteration assessment of the analyses used in their figure 1 as was done for the Sawlan (2018) SB samples.
Baker et al.’s figure 1 shows that high Ti and P abundances in samples from the Cohassett flow’s top and bottom are confined within the flow contacts. Other profiles given in figure 6 of Reidel (2005) (RRL-6, White Salmon), however, show that his “compositional types” traverse flow boundaries. This indicates that such compositional differences are postemplacement chemical characteristics that may occur from preferential alteration of the more porous flow tops and bottoms of successively emplaced flows.
Furthermore, Baker et al.’s claim that four compositionally distinct lava flows erupted simultaneously is not supported by the physical stratigraphy. They infer that the high-Mg lavas assigned to their Airway Heights and California Creek compositional types, which are found only in the northeastern Columbia Basin, erupted at the same time as the SB lavas assigned to the Stember Creek and Spokane Falls types. The Airway Heights and California Creek lavas, however, could be significantly older than SB lavas because their lower possible stratigraphic position is poorly constrained. Other high-Mg lavas (Slack Canyon and Indian Ridge GRB members) having limited extents within the Columbia Basin are known to have erupted during the GRB N2 magnetozone prior to SB eruptions (Landon and Long, 1989; Reidel et al., 1989; Reidel and Tolan, 2013). In addition, neither the Airway Heights nor California Creek lavas are found in a stratigraphic succession with the SB Group 11 (Stember Creek) lava. The intraflow chemical differences therefore are more plausibly explained by differential alteration of the dense and porous parts of the lava flows.
9. GEOGRAPHIC DISTRIBUTION OF ANOXIC GROUNDWATER ALTERATION AND OXIC SURFICIAL WEATHERING
From the common occurrence of thick paleosols formed by surficial weathering west of the Cascade Range, Baker et al. contend that CRBG lavas in the Columbia Basin to the east of the Cascades have remained unaffected by any alteration. The type of alteration identified by Sawlan (2018), however, is not visually apparent and is caused by interaction with anoxic groundwater. Alteration under anoxic conditions can occur anywhere within the flood basalt province given that these lavas currently host aquifers on both sides of the Cascades and have undoubtedly done so for the past ∼16 m.y. Baker et al.’s claim of no anoxic alteration east of the Cascades also contradicts statements elsewhere in their comment, in which they discuss “early” leaching of Mg and Fe under reducing conditions from altered CRBG lava in southeastern Washington and nearby Idaho (e.g., Thomson et al., 2014).
The assertion by Baker et al. that CRBG lavas east of the Cascades in the Columbia Basin are mostly unaffected by surficial weathering, even in recent road cuts, is simply not true. Figure 6 shows examples of alteration rinds and weathered rock in road cuts and natural exposures in various parts of the Columbia Basin (northeastern, southeastern, south-central, and southwestern) not scoured by the Missoula floods (Sawlan, 2018). Areas where the floods were erosive tend to have thin alteration rinds from the removal of weathered rock, but such areas are not representative of the flood basalt province as a whole. Thick concentric weathering rinds are commonly encountered in the Columbia Basin and can be up to ∼15 cm thick in some outcrops. In some instances, oxic surficial weathering is pervasive, or nearly so, such that many fracture-bounded blocks lack a “corestone” of less altered or possibly unaltered inter-rind rock. In each Columbia Basin location where small corestones were sampled from within thicker weathering rinds, multiple fracture-bounded blocks that were broken open contained only weathered rock.
The mass analysis expressions used to determine the remaining sample mass and mass-normalized abundances (Sawlan, 2018) are incorporated within those expressions that have been widely used in rock and soil weathering studies (e.g., Anderson et al., 2002). They differ from those applied in conventional weathering studies, however, in two important respects: (1) mass-normalized abundances are not converted to a dimensionless ratio of the parent abundance from which a value of 1 is subtracted, but are presented directly in wt% or ppm. This allows direct comparison of mass-normalized abundances for altered and unaltered samples; and (2) parent abundances are referenced to a magmatic trend between immobile elements (Al and Ti), thereby establishing a unit parent mass to which altered samples are referenced. Parent compositions of altered samples can be determined from their Al/Ti ratios, and, as in conventional expressions, the parent:sample ratio of an immobile element (e.g., Ti) is used to quantify the remaining sample mass. It appears that Baker et al. did not compare the simple expressions for sample mass index and mass-normalization to the terms in the expression for tau given by Anderson et al. (2002). As shown above, the mass analysis (Sawlan, 2018) and weathering (Anderson et al., 2002) calculations are in complete agreement. Therefore, if the mass analysis formulas are incorrect, then those given in Nesbitt (1979), Brimhall et al. (1992), and Anderson et al. (2002), as well as Baker’s own “mobility ratio” (Thomson et al., 2014), are also incorrect.
Baker et al. state that “good” samples can be obtained with care in “collection and preparation,” and that “excellent samples” can be obtained “if one takes care and collects unaltered rock from the core of altered columns.” Workers with decades of experience are said to have established a “standard procedure” to avoid sampling altered rock, but descriptions of such a procedure cannot be found in the methods sections of prior reports. Baker et al. cite Swanson and Wright (1981) as an example of the standard sampling procedure for CRBG lavas, but Swanson and Wright did not discuss their sampling protocol.
Baker et al. also claim that identification of (apparently) unaltered glass using petrographic examination is sufficient to characterize a sample as unaltered. This criterion does not apply, however, to samples altered under anoxic reducing conditions due to the low reactivity of glass relative to that of Fe-Ti oxide and pyroxene. Although glass is altered in surficially weathered samples, thin-section examination is not necessary to identify oxic weathering because there are other visual cues indicating this type of alteration. Furthermore, excluding one strongly oxic-weathered sample intentionally collected, all SB samples collected by Sawlan (2018), without benefit of thin-section examination, showed no detectable chemical evidence of oxic alteration. In addition, 70% of Sawlan’s (2018) analyses indicated less than 1% mass loss, and, among the samples exhibiting mass loss, progressive Fe depletion with decreasing sample mass showed they were altered exclusively under reducing anoxic conditions.
Despite the general agreement of Baker et al. with Sawlan (2018) that alteration of a subset of samples reported in Sawlan (2018) occurred under reducing conditions, there are important differences stemming from the different approaches. Baker et al. cannot identify specific parent compositions of lavas such that they can only broadly characterize the chemical effects of alteration. As shown in Sawlan (2018), together with the analysis given in this reply, the application of mass analysis can (1) quantify alteration in terms of a physical property, (2) identify parent compositions, (3) quantify a threshold of alteration, and (4) determine the composition of the material removed by dissolution.
Sawlan (2018) showed that the earlier SB data set includes many samples showing mass loss, mainly <4%, as well as some samples showing higher amounts of mass loss, up to 12%. Most of these samples experienced alteration under oxic surficial weathering conditions, such that these samples undoubtedly exhibited characteristic color changes associated with Fe oxidation. Although Baker et al. claim they exercised diligence in applying a sampling protocol similar to that described in Sawlan (2018), supplemented by petrographic examinations, the chemical evidence indicates otherwise.
A complete data set of the earlier SB analyses, accompanied by the appropriate documentation to allow independent evaluation of these data, must be published. Following Reidel (2005), Baker et al. have interpreted sample analyses from the earlier SB database to represent magmatic compositions, but most of these samples were variably altered under both anoxic and oxic conditions. Intra-flow chemical differences, therefore, are most likely due to alteration of chemically homogeneous lava flows fed by well-mixed magma reservoirs rather than to “comingling and mixing” at the surface of four presumably coeval and compositionally distinct lava flows. In addition, previously proposed SB chemostratigraphies are inexact because magmatic compositions could not be distinguished from those modified by alteration.
Constructive reviews of the manuscript by Jon Hagstrum significantly improved the manuscript and are much appreciated. Shan de Silva is to be commended for his editorial handling as well as for his review comments.