We tested the associations of TOC and secondary Fe oxide with total Mo, V, and Cr concentrations and concentrations within secondary Fe oxide phase using linear regressions for each CZO. All regressions were performed using Matlab (Matlab, 2016, Mathworks, Natick, MA, United States). For elemental concentrations, error bars were propagated based upon the coefficient of variation for each sample measurement and the standard error measured through duplicate and triplicate samples.

Mass transfer coefficients (τ) of Mo, V, and Cr were calculated to assess the degree of chemical weathering for specific elements using Ti as an index element (Brantley and Lebedeva, 2011). Mass transfer coefficients are defined for the element of interest (Mo, V, and Cr) (j) in the soil or regolith sample normalized to an immobile index element Ti (i) in the parent material:

where C_j,w and C_i,w are total concentrations of the element of interest (Mo, V, and Cr) and Ti in the soil and regolith samples, respectively, and C_j,p and C_i,p are the element of interest and index element concentrations in the least weathered regolith material (e.g., Brimhall and Dietrich, 1987; Kraepiel et al., 2015). Negative τ values indicate a net loss of an element from a specific soil depth due to leaching. Positive τ values indicate a net enrichment of an element. Although Ti can be mobilized at strongly acidic, heavily weathered tropical soils (Kurtz et al., 2000), application of Ti-based mass transfer coefficients is effective for estimating relative losses and gains in the CZOs of our study.

Stream water concentrations were grouped across collection dates and compared across the four CZOs using the Kruskal–Wallis non-parametric tests. Using measured watershed area and mean annual stream discharge, we calculated the export of Mo, V, and Cr at each of the four CZOs. To accomplish this, we obtained the total watershed area, MAP, and mean annual discharge for each of the four CZOs from existing online data ¹ (accessed May 2018). Stream water Mo, V, and Cr concentrations were multiplied by total discharge to estimate export rates for each element.

Regolith profiles exhibited a range of TOC and secondary oxide Fe with depth across the four CZOs. Overall, %TOC decreased with depth from approximately 1% down to 0.1% with subsurface peaks present at all four CZOs (Figure 2). Calhoun CZO had <0.1 %TOC below 2 m depth, which was lower than the other three CZOs (Figure 2). Total Fe was relatively constant with depth at Boulder Creek and Southern Sierra CZOs. Subtropical Calhoun CZO had low total Fe concentrations in the surface soils while tropical Luquillo CZO had higher total Fe concentrations above 2 m depth (Figure 2). Secondary oxide Fe concentrations were highest throughout the regolith profile at Luquillo CZO by nearly two orders of magnitude compared to Boulder Creek and Southern Sierra CZOs (Figure 2). Calhoun CZO had a subsurface peak of secondary oxide Fe concentration between 1 and 1.5 m depth (Figure 2).

Total Mo profile concentrations ranged from 0.4 to 2.5 mg kg^-1 across the four CZOs (Figure 3 and Supplementary Table 1). The highest total Mo concentrations at the Calhoun and Luquillo CZOs were in the surface soils, at 0.6 m and 0.4 m in depth, respectively. The colder, higher elevation Boulder Creek and Southern Sierra CZOs had their highest total Mo concentrations deeper in the regolith, at 2.4 m and 4 m in depth, respectively. Total V profiles exhibited two orders of magnitude variation in concentrations across regolith profiles. Total V concentrations were highest in the top meter for Boulder Creek, Luquillo, and Calhoun CZOs, and were generally consistent with depth at the Southern Sierra CZO. Total Cr concentrations at the Calhoun CZO had subsurface peaks of total Cr between 2.0 and 4.0 m. Boulder Creek exhibited a significant total Cr peak between 1.0 and 2.0 m, which decreased with depth. Southern Sierra and Luquillo CZO exhibited nearly homogeneous (±3 mg kg^-1) total Cr profiles with depth.

In the regolith, organic matter and amorphous oxide bound Mo, V, and Cr were substantially lower than their respective secondary oxide and total concentrations. Across the four CZOs, organic matter and amorphous oxide bound Mo ranged between 0.01 and 0.13 mg kg^-1 (Supplementary Table 2), which on average was 0.5–0.9% of the total Mo concentration (Supplementary Table 4). The organic matter and amorphous oxide bound fraction of Mo were greatest near the surface (<1 m depth) for Calhoun and Southern Sierra CZOs while Boulder Creek and Luquillo CZOs had subsurface peaks (>2 m depth). Vanadium varied between 0.11 and 3.1 mg kg^-1 (Supplementary Table 2), which on average was 0.3–2.1% of the total V concentration (Supplementary Table 4). The organic matter and amorphous oxide bound fraction of V were greatest near the surface (<1 m depth) for Luquillo and Southern Sierra CZOs while Boulder Creek and Calhoun CZOs exhibited subsurface peaks (Figure 3). Organic bound Cr concentrations ranged between 0.01 and 1.07 mg kg^-1 (Supplementary Table 2), which on average was 0.9–3.5% of the total Cr concentration (Supplementary Table 4). Organic matter and amorphous oxide bound Cr was greatest in the top 2 m of regolith at Luquillo and Southern Sierra CZOs but were greater below 2 m in depth at Boulder Creek and Calhoun CZOs (Figure 3).

Secondary Fe oxide Mo concentrations ranged between 0.02 and 0.68 mg kg^-1 (Supplementary Table 3), which on average was 6.6–40.1% of the total Mo concentration (Supplementary Table 5). Vanadium concentrations in the secondary Fe oxide ranged between 1 and 52 mg kg^-1 (Supplementary Table 3), which on average was 6.0–39.7% of the total V concentration (Supplementary Table 5). Secondary Fe oxide Cr concentrations varied between 0.2 and 8.7 mg kg^-1, which on average was 13.0–40.5% of the total Cr concentrations (Supplementary Table 5).

Using Pearson correlations, we observed that total Mo, V, and Cr were correlated with secondary oxide Fe concentrations [a proxy for long-range ordered to crystalline Fe oxy(hydr)oxides] (Table 2). At Calhoun CZO, secondary oxide and total Mo, V, and Cr concentrations were positively correlated with secondary oxide Fe concentrations. For Southern Sierra CZO, secondary oxide V and Cr were positively correlated with secondary oxide Fe concentrations. At Boulder Creek CZO, secondary oxide and total concentrations of V and Cr were negatively correlated with both secondary oxide Fe and TOC. At Luquillo CZO, secondary oxide Mo and V were positively correlated with %TOC.

Calculated tau values for Mo, V, and Cr show both net enrichment and depletion across the CZOs (Figure 4). At the montane coniferous Boulder Creek CZO, regolith down to 2.0 m depth were depleted in Mo and V, but Cr was enriched below 1.0 m depth. At the subtropical Calhoun CZO, V was depleted while Mo and Cr were enriched throughout the regolith profile. The tropical Luquillo CZO showed a wide variation but overall Cr was depleted throughout the entire profile while Mo and V were enriched at certain depths. The Southern Sierra CZO regolith profile was depleted in V and Cr but enriched in the surface and subsurface between 2.0 and 5.0 m depth.

Analysis of the μ-XRF spectra for the weathered regolith from Luquillo CZO showed Fe, V, and Cr coinciding together in hornblende and biotite primary minerals (Buss et al., 2005, 2017). This is particularly observable in the upper left-hand corner of the images and the outlines of the hornblende minerals are distinguishable in all four images. Moreover, Fe, V, and Cr also correspond with each other in the Fe oxide weathering rind on the right side of the image (Figure 5). More information on the weathering rinds of quartz diorite from Luquillo CZO is available in Buss et al. (2008).

Overall, stream water concentrations of oxyanions were generally comparable across CZOs (Table 3 and Figure 6), but there were a few significant differences. Stream water V concentrations were similar across CZOs and ranged between 0.04 and 0.70 μg L^-1 (Figure 6). Stream water Cr concentrations at Calhoun and Boulder Creek CZOs were significantly than at the Southern Sierra and Luquillo CZOs (Table 3 and Figure 6) using the non-parametric Kruskal–Wallis Test (p < 0.05). Stream water Mo concentrations were significantly higher at Boulder Creek and Calhoun CZOs compared to Southern Sierra CZO (p < 0.05, Table 3 and Figure 6).

We compared the concentration of oxyanions in the stream water to the deepest, least weathered regolith to determine if variations in lithology explained the differences in stream water concentrations across CZOs and among oxyanions. In Figure 7, Mo stream water to regolith ratios were significantly greater at subtropical Calhoun and tropical Luquillo CZOs than montane Southern Sierra and Boulder Creek CZOs. In addition, Mo stream water to regolith ratios at Calhoun, Luquillo, and Southern Sierra CZOs were significantly greater than V and Cr ratios. Stream water to regolith ratios for V were not significantly different for Boulder Creek, Calhoun, and Luquillo CZOs but were greater than at Southern Sierra CZO. In addition, Stream water to regolith ratios for V were generally lower than Mo and Cr across CZOs (Figure 7). Stream water to regolith ratios for Cr were significantly higher at Calhoun CZO than the other three CZOs. Similarly, stream water to regolith ratios for Cr were significantly higher at Boulder Creek CZO than Luquillo and Southern Sierra CZOs.

Our results support the hypothesis that deep regolith play a substantial role in sorption of oxyanions. Overall, both surface soils (<0.3 m depth) and deeper regolith served as important sources and retainers of oxyanions. We observed many peaks in Mo, V, and Cr in both the organic matter and amorphous oxide and the secondary Fe oxide extractions at depths are far deeper than typical 0.3–1 m depth investigated for oxyanion sorption and weathering in soils (e.g., Shiller and Mao, 2000; Wichard et al., 2009; Wisawapipat and Kretzschmar, 2017). For example, at the montane regolith of Southern Sierra CZO had >39% of Mo and Cr were in the secondary Fe oxide fraction from 3 to 4 m in depth and the tropical forest Luquillo CZO, 52% of Cr and 23% of V were in the secondary Fe oxide fraction at 3 m in depth.

In addition, enrichment and depletion of oxyanions >2 m in depth indicated that regolith played both a significant role in the sorption and retention of oxyanions (Figure 4). Except for Cr, Boulder Creek CZO demonstrated the most depletion of oxyanions, extending down 2 m in depth (Figure 4). This is particularly important as the mountain-steppe Boulder Creek CZO has the least intensive weathering environment with the lowest MAP and MAT, but due to its long exposure, has been moderate to extensively weathered. The more tropical regolith profiles at Calhoun and Luquillo CZOs show much more complicated patterns of accumulation and depletion (Figure 4), likely due to variations in retention as described later in this section.

These results support other weathering observations by Dethier and Bove (2011) at Boulder Creek CZO, Bacon et al. (2012) at Calhoun CZO, and Buss et al. (2005) at Luquillo CZO. Subtropical Calhoun CZO and tropical Luquillo CZO were the most weathered regolith profiles and had the highest accumulations of secondary Fe oxides (Figure 2), consisting of hematite and goethite as observed by Buss et al. (2005); Fimmen et al. (2008), and Dethier and Bove (2011). In addition, neoformed kaolinite was present at the previously mentioned CZOs at >4% by mass (Fimmen et al., 2008 at Calhoun CZO; Dethier and Bove, 2011 at Boulder Creek CZO; Buss et al., 2017 at Luquillo CZO) demonstrating their extensive weathering history. The extent of weathering corresponded with the sorption of oxyanions throughout the regolith profiles. Southern Sierra had lower secondary Fe oxide fractions of Mo, Cr, and V compared to the other three CZOs. Moreover, secondary oxide Mo, Cr, and V were greater in regolith with higher secondary Fe oxide concentrations at much warmer and wetter Calhoun and Luquillo CZOs when compared to the colder and drier Boulder Creek CZO. Thus, we conclude that weathering has influenced the formation and retention of oxyanions with secondary Fe oxides.

A generalized trends across the three oxyanions were present, demonstrating some linkages between weathering intensity and climate with the sorption of oxyanions to secondary Fe oxides. Total Mo, secondary Fe-bound Mo and organic and amorphous oxide bound Mo concentrations were comparable across the four CZOs, suggesting weathering intensity nor lithology did not strongly influence Mo in the soil and regolith profiles. Secondary oxides did not appear to play a dominant role for Mo retention at Luquillo CZO as total Mo was not correlated with secondary Fe oxides. Instead secondary oxide Mo was correlated with %TOC despite the fact that organic matter and amorphous oxide bound Mo was not a substantial fraction of the total Mo concentrations at any of the CZOs, generally ranging below 5%. This largely agrees with previous studies that the sorption and retention of Mo is dominated by both, secondary Fe oxides and organic compounds (e.g., Wichard et al., 2009; King et al., 2016, 2018). Due to the relatively small range in %TOC, there may not have been enough variability to capture the relationship between Mo and TOC.

Total V concentrations varied significantly across the CZOs (Supplementary Table 1), suggesting lithologic controls on the sourcing of V to the soil and regolith profiles. Secondary Fe oxide fraction of V, however, coincided with weathering intensity. Average secondary Fe oxide concentrations were highest for the more tropical Calhoun (14 mg kg^-1) and Luquillo CZO (26 mg kg^-1) compared to mountain-steppe Southern Sierra (9 mg kg^-1) and Boulder Creek CZOs (3 mg kg^-1). Vanadium exhibited significant linear relationships with secondary Fe oxide at Calhoun, Boulder Creek, and Southern Sierra CZOs. The substitution of V(III) for Fe(III) in minerals has been widely observed (Schwertmann and Pfab, 1996; Shiller and Mao, 2000; Wisawapipat and Kretzschmar, 2017) and V(II) and V(III) for Fe(II) and Fe(III) in primary Fe-silicate minerals during magmatic crystallization (e.g., Goldschmidt, 1954). Even though this relationship was not clearly observed at Luquillo CZO as with the other three CZOs using linear regressions with secondary Fe oxide, we did observe this relationship in μ-XRF maps of regolith samples. Our μ-XRF imaging of V in an intact weathering rind sample from Luquillo demonstrates its spatial distribution with Fe, both in primary minerals like hornblende and biotite and in secondary Fe oxides within regolith at Luquillo CZO.

Chromium concentrations exhibited a pattern similar to V. Total Cr concentrations were varied strongly with lithology but Cr concentrations in secondary Fe oxides appeared to be climate influenced. The more tropical Calhoun and Luquillo CZOs had higher % Cr in the secondary Fe oxide fraction than the montane-steppe Southern Sierra and Boulder Creek CZOs, suggesting intensive weathering promotes Cr sorption by secondary Fe oxides. The sorption of Cr by secondary Fe oxides and organic matter has been well observed by previous studies (e.g., Gustafsson et al., 2014; Pérez et al., 2014; Saputro et al., 2014).

Some potential variables that may explain the variation of oxyanions not explored here are the effects of erosion, atmospheric inputs, hydrological properties, and other source of sorption such as clay minerals and other oxides. Luquillo CZO differs from the other three CZOs due to its substantially greater annual precipitation (Table 1), which could not only flush and dissolve solutes at rates that may prevent sorption (see Dosseto et al., 2012) but also lead to increased atmospheric inputs of oxyanions in the surface soils (Marks et al., 2015; King et al., 2016). This may also influence τ values by enriching oxyanions at the surface and a net depletion due to leaching losses with depth. Nevertheless, the effect of atmospheric inputs on the deep regolith is likely minimal. In addition, Luquillo CZO undergoes more erosion in the form of landslides and surface soil erosion than the other three CZOs (Larsen, 2012). Lastly, Luquillo CZO also had much greater clay content in its soils and upper regolith than Calhoun, Boulder Creek, and Southern Sierra CZOs. Clays can provide an unaccounted source of oxyanion sorption capacity (Goldberg et al., 1996).

Our study observed low fractions of oxyanions in the organic bound and amorphous oxide phase and poor correlations between %TOC and oxyanion concentrations in the deep regolith. However, this conclusion should not be extrapolated to surface soils. Many previous studies have observed a significant relationship in surface soils and oxyanion concentrations (see Grove and Ellis, 1980; Wichard et al., 2009; King et al., 2014, 2016, 2018; Pérez et al., 2014; Marks et al., 2015; Siebert et al., 2015; King, 2017; Larsson et al., 2017; Wisawapipat and Kretzschmar, 2017). Therefore, sorption of oxyanions with organic matter may be most important spatially but not with depth into deep regolith (>2 m).

Oxyanion concentrations and export from the Critical Zone via stream water varied substantially across CZOs. The most intensively weathered regolith had higher average Mo concentrations in stream water (Table 3 and Figure 6). However, Mo concentrations were generally comparable across CZOs when removing the outlier samples at Calhoun and Luquillo CZOs. Calhoun and Luquillo CZOs had higher Mo stream water to regolith ratios compared to V and Cr (Figure 7). This implies that Mo had lower relative retention in regolith, increasing its export by stream water compared to V and Cr at Calhoun and Luquillo CZOs. Alternatively, the stream water to regolith ratios increase with increasing MAP across the CZO (Table 1 and Figure 7). This suggests that atmospheric inputs of Mo via precipitation (Marks et al., 2015; Siebert et al., 2015; King et al., 2016) may lead to increased export of Mo from the regolith. There may be additional transport processes influencing Mo transport such as colloidal Al, Fe, and Mn oxides and dissolved organic matter (Pearce et al., 2010; Trostle et al., 2016; Prouty et al., 2017), which can enhance its transport through soil. Furthermore, fine colloids <0.45 μm in solution may also be responsible for the two elevated Mo concentrations at Calhoun and Luquillo CZOs. When these high concentrations are not considered, Mo concentrations were similar. Thus, there is not a strong trend in Mo concentrations with weathering intensity but potential variability due to colloids was greatest at the most intensively weathered CZOs.

Few studies have examined the controls on V concentrations in stream water (Shiller and Boyle, 1987; Shiller, 1997; Shiller and Mao, 2000; Strady et al., 2009). Stream water V concentrations were not significantly different across the four CZOs (Table 3). The similarity in stream water to regolith ratio for V suggests that regolith at Boulder Creek, Calhoun, and Luquillo CZOs are weathering and exporting V at consistent rates. Further, the stream water to regolith V ratios were lower than Mo and Cr, suggesting that V is the least mobile of the three oxyanions during weathering (Figure 7). The elevated stream water V concentration measured in one sample at Luquillo CZO may be due to colloids, particularly DOC and Al and Fe particulates. Stream water to regolith ratios for V was not significantly different than Fe for each CZO, suggesting V export is directly controlled with Fe movement. This may be as either dissolved or fine colloids <0.45 μm, the size of the filter pores used in this study (e.g., Pokrovsky and Schott, 2002; Gardner et al., 2017). Stream water Cr concentrations exhibited the greatest variation among the four CZOs. Boulder Creek CZO had the highest stream water Cr concentrations (Table 3). Moreover, Boulder Creek CZO had significantly higher stream water to regolith ratio, despite having the lowest MAT and lowest MAP compared with the other four CZOs in this study (Figure 7).

Using measured watershed area and mean annual stream discharge, we calculated the export of Mo, V, and Cr at each of the four CZOs. Luquillo CZO had Mo, V, and Cr fluxes (10⁵ to 10⁶mg year^-1) up to four orders of magnitude greater than the other CZOs (10² to 10⁴mg per year) due to 10² to 10³ times greater stream discharge than the other three CZOs. We observed that Boulder Creek CZO had greater Mo, V, and Cr fluxes than Southern Sierra CZOs, despite having lower MAP. Even when scaled for different watersheds size, Boulder Creek had >10× greater export of oxyanions per km². This highlights the importance that more weathered systems (based upon interpretation of mass transfer coefficients) export more oxyanions at greater rates than fresher regolith. It must be noted these approximations are based upon oxyanion concentrations in baseflow conditions. While this assumption might not hold for each of the oxyanions at the four CZOs, previous studies such as Shiller and Mao (2000), Aguirre et al. (2017) have found oxyanions can be isostatic with respect to concentration-discharge (CQ) dynamics. However, this provides an estimate in the variation of Mo, V, and Cr export via stream waters across different watershed with contrasting climates.