A dual-domain column experiment was designed to test the ability of the detection method to map the differences in redox reactivity between core and interface of anoxic microsite under oxygenated hydrologic flow simulated in the laboratory.

Columns (30 cm long, 7 cm diameter) were filled with natural sediments, collected ∼2 weeks prior from the Wind River-Little Wind River floodplain outside of Riverton, WY (42°59′19.1″N, 108°23′58.6″W). Columns were set up by embedding one fine-grained, organic-rich, reducing lens (∼3 cm diameter, 17–20 g dry weight made from sulfidic sediments collected underneath an oxbow lake on the floodplain) into aquifer sand (1.3–1.4 kg dry sand in the column). Columns were fed artificial groundwater (in upflow mode using a peristaltic pump at a flow rate of 90 mL/day, equivalent to ∼1 pore volume exchange per 48 h; Figure 2) made to have a chemical composition similar in concentrations of the major solutes, except organic C, measured in Riverton floodplain groundwater at the site over the past 2 years (Kumar et al., 2020). The column experiment was run for a total of 70 days. More information of the column design and description of sediment are detailed in Kumar et al. (2020). At the end of column experiment, the core was preserved at −20°C until prepared for X-ray analyses.

Cores with Fe^II and Fe^III microsites were artificially created in sandy matrices with and without a natural Fe background to develop and calibrate data processing methods for detecting Fe^II and Fe^III microsites in Fe-“poor” and Fe-“rich” environments.

To prepare synthetic cores with either Fe-“poor” or Fe-“rich” background matrices, a 50 mL falcon tube was cut lengthwise down the middle and the halves were separately capped. One-half was then filled with 350 g dry, Fe^III-free sand, to represent an Fe-“poor” background and the other half was filled with a mix of 60 weight-% Fe^III-smectite (210 g) and 40 weight-% Fe^III-free sand (140 g) to represent a 1.6% Fe-“rich” background. Fe^III-free sand consisted of HM-106 Density Test Sand (Sand Cone Test Sand) and Fe^III-smectite consisted of SWy-2 (Na-rich Montmorillonite), a source clays repository from the Clay Minerals Society composed of ≈3.35% Fe₂O₃ and ≈0.35% FeO (Crook County, Wyoming, USA; https://www.clays.org/sourceclays_data/). After filling the half-cores, they were smoothed to ensure a plane surface and then covered in epoxy (EPO-TEK^® 301-2FL) to stabilize the sand. Before the epoxy fully hardened (∼12 h), ∼2–10 mm diameter and ∼1–2 mm deep holes were carved along the surface of each synthetic core, using a simple needle. After fully hardening (>72 h), the holes were then filled with solid fraction (dried and powdered) of either Fe^II-S (i.e., Fe^II-sulfide), Fe^IICO₃ (i.e., siderite), or Fe^III-(hydroxy)oxide mixed with the background material to generate microsites with 5 weight-% Fe of varying redox properties. Thus, Fe^II-S and Fe^IICO₃ microsites in the Fe^III-smectite sand synthetic core would represent a depletion of Fe^III compared to the background, whereas Fe^III-(hydroxy)oxide microsites would represent a gain of Fe^III compared to the background. Fe^II-S and Fe^IICO₃ consisted of natural Fe^II-sulfide and siderite, respectively, from the Stanford University mineral collection, while Fe^III-(hydroxy)oxides consisted of synthesized goethite (α-Fe^IIIO(OH)) nanoparticles characterized in Kumar et al. (2018). Phase purity of the natural Fe^II-sulfide, siderite, and synthetic goethite were confirmed by X-ray diffraction (XRD) analysis (XRD method and diffractograms are provided in Supplementary Material). Additionally, Fe K-edge XANES spectra of natural Fe^II-sulfide, siderite, and synthetic goethite composing Fe^II-S, Fe^IICO₃, Fe^III-(hydroxy)oxide microsites, respectively (Supplementary Figure S2), compared well with spectra of previously reported references (Supplementary Figure S3).

Soil cores were collected from different field sites to detect and quantify anoxic microsites within natural samples.

Soil aggregates were also collected to map spatial distribution of Fe reducing zones in natural aggregates to complement other analyses targeting anoxic microsites from soil cores and aggregate samples collected at the same locations (Lacroix et al., 2022).

For X-ray chemical imaging, all samples (i.e., the dual domain column, natural soil cores, and natural soil aggregates) were sectioned in half to map Fe oxidation spatial distribution in 2D at the center of each sample. For this purpose, it is crucial to preserve the chemical and physical state of the samples during sectioning (including limited oxidation and physical smearing). Therefore, all samples were transferred (while frozen and sealed) into an anoxic chamber (∼4% H₂, 96% N₂) equipped with an O₂ detector and Pd catalysts (Coy Laboratories) and all sample preparations were conducted inside this chamber. Importantly, the synthetic cores (§ 2.2), having been prepared directly for X-ray analyses, were not treated as described below, but were kept in anoxic containers until X-ray analyses.

The dual domain column and natural soil cores were cut in half (lengthwise) in a two-step process: (i) the PVC “shell” was cut with a rotary cutter and (ii) the frozen column/soil cores were cut with a sharp knife – this two-step process prevents PVC contamination of the core and minimizes smearing of the cut surfaces. For each sample, one half of the cut core was selected for X-ray imaging and a 2-mm section of the surface was removed delicately using a knife (to further remove any smearing or contamination from the cutting process). Thereafter, the soil core halves were left to thaw and dry partially, inside the anoxic chamber at ambient temperature under suction (vacuum) for 24-48 hours, prior to being fixed with epoxy (EPO-TEK® 301-2FL). It is necessary to allow the cores to partially dry to enable penetration of the epoxy while preserving physical structure; applying suction accelerates the drying and, hence, minimizes microbial activity and associated redox changes during thawing/drying. The suction also enhances the penetration of epoxy into the sample, which with 1-2 cm depth penetration from the exposed surface serves to: (i) prevent movement of core sections and/or particles during manipulations and X-ray analyses, and (ii) minimize redox alterations during transfer for X-ray analyses.

Because soil aggregates were too fragile to be cut in their untreated state, they were first epoxied before cutting in following steps; 1) a 50mL capped plastic centrifuge tube was cut longitudinal to the axis of the tube and one-half was filled with epoxy; 2) soil aggregates were partially submerged so that roughly half of each aggregate remained above the surface of the epoxy in the cut centrifuge tube; 3) after allowing the epoxy to harden for 72 hours, the non-fixated half of soil aggregates were cut along the surface of the epoxy to generate a smooth surface for X-ray mapping. Once prepared, all samples were kept separately in anoxic containers until X-ray analyses.

The natural and synthetic samples were transferred to the beamline in anoxic containers where they were directly inserted into an He flow-box (completely sealed and continuously saturated with He) to avoid oxidation. The entire surface area of the experimental (dual-domain columns, §2.1), synthetic (§2.2), and prepared natural (cores and aggregates, §2.3 and §2.4) samples were analyzed by synchrotron-based multiple energy μ-XRF mapping across the Fe K-edge to determine the spatial distribution of Fe^II vs. Fe^III species. Multiple-energy Fe maps were collected at Beamline 7-2 at the Stanford Synchrotron Radiation Lightsource (SSRL, SLAC National Accelerator Laboratory, Menlo Park, CA), a 20-pole 2-Telsa wiggler device equipped with a double crystal monochromator (Si [111],φ = 0°). μ-XRF maps were collected at 7,116, 7,123, 7,127, 7,132, 7,138, and 7,145 eV at ambient temperature using a 4-channel Vortex Silicon Drift detector (Hitachi) in fluorescence mode. These energies were selected based on the reference compound spectra of the most likely Fe species to be found in both oxic and anoxic Fe containing soils to give the highest contrast between oxidation/chemical states (see Supplementary Material for detailed explanations, Supplementary Figure S3 and Supplementary Table S1). An alternative to analyzing less maps would be to collect μ-XRF maps at 7,134 eV instead of 7,132 eV and 7,138 eV, which is sensitive to Fe^III-O species (Supplementary Figures S3B, C). The map step size was 30–75 μm with a dwell time from 25 to 50 ms. The angle subtended by the incident beam, sample, and detector was 90°, with the sample normally bisecting this arrangement.

The μ-XRF maps were processed at the end of each mapping and several spots (i.e., putative microsites, see § 2.7) were further selected to collect full μ-XANES spectra across the Fe K-edge (6,862–7,512 eV) that were used to reconstitute Fe^II and Fe^III species maps using XANES fitting (see § 2.7). Scans were background subtracted and normalized using ATHENA (Demeter v0.9.24) (Ravel and Newville, 2005).

In a fluorescence image map, each pixel registers the respective Fe fluorescence peak intensity at the incident energy. Thus, in the multi-energy X-ray fluorescence maps, each pixel registers a set of Fe intensities for each incident energy, resulting in a shortened spectrum of Fe fluorescence intensity as a function of energy, similar to a XANES spectrum. Because, the μ-XRF maps were collected at six different energies (7,116, 7,123, 7,127, 7,132, 7,138, and 7,145 eV), each pixel registers the equivalent of a six-point XANES spectrum. Thus, data processing of the Fe K-edge multiple energy μ-XRF maps consist of fitting the six-point XANES spectrum of each pixel. Fe K-edge multiple energy μ-XRF maps were analyzed first via Simplex Volume Maximization (SiVM) and second via XANES fitting, using the MicroAnalysis Toolkit (Webb, 2011), to assess spatial relationships of similarities and differences in oxidation states/chemical forms and, thereby, identify spatial heterogeneities composed of contrasted redox with the background, corresponding to putative anoxic microsites. This method provides a reliable estimation of the spatial distribution of Fe redox for each pixel, where measuring the full μ-XANES spectrum of each pixel (i.e., 30–75 μm) across the analyzed sample (up to 100 cm long) is not possible.

Fe K-edge multiple energy μ-XRF maps of a synthetic column of coarse aquifer sand containing reducing organic-rich lenses (i.e., microsites) from Riverton, WY (Kumar et al., 2020), and under oxygenated hydrologic upflow simulated in the laboratory, are displayed in Figure 2. At 7,123 and 7,127 eV, X-ray energy is preferentially absorbed by Fe^II species, and possibly Fe^II−III-clays (Supplementary Figures S3A, B). Thus, the higher intensity at 7,123 eV in the column microsite compared to the microsite-sand interface, and the inverse pattern observed at 7,127 eV, clearly suggests the occurrence of different Fe^II-species composition in the microsite core vs. the interface. Fe^II-S species (i.e., Fe^IIS₂, Fe^IIS (Fe^IIFe^III ₂S⁴) are the only Fe^II species whose absorption intensity does not increase between 7,123 and 7,127 eV (Supplementary Figure S3A). Other Fe^II-species (such as Fe^II-carbonate, Fe^II-phosphate, Fe^II-C, Fe^II-hydroxide, Fe^II-sulfate) and Fe^II−III-clay species preferentially absorb X-rays at 7,127 eV rather than 7,123 eV (Supplementary Figure S3A). Thus, although the core of the microsite primarily comprised Fe^II-S species, another Fe^II-species was preferentially accumulated at the interface. The accumulation of Fe^II-species at the interface upstream of the microsite, that was continuously exposed to influent oxygenated water, suggests the presence of Fe^II-species that are relatively resistant to oxidation, such as Fe^II-C or Fe^II−III-clay species. In this context, the spatial distribution of different Fe^II-species constitutes forensic evidence of the presence of anoxic microsites and the spatial distribution of their internal redox chemistry.

At 7,134 eV, X-ray energy is preferentially absorbed by Fe^III-O species (Supplementary Figures S3B, C). Thus, the preferential decrease of X-ray energy absorption intensity at 7,134 eV at the microsite-sand interface downstream (compared to upstream) of the microsite suggests a depletion of Fe^III-O species. This is not surprising as oxygen has likely been consumed due to anoxic conditions in and downstream the microsite. Indeed, Babey et al. (2022) demonstrated that the aquifer sediments located downstream the anoxic microsite could develop secondary reducing zones. The depletion of Fe^III-O species offers another piece of forensic evidence to help identify and characterize anoxic microsites, and in this case, indicates the direction of flow through the reduced microenvironment.

Visual contrasting of μ-XRF maps from different Fe K-edge energies is rarely conclusive and requires fitting all multiple energy maps to reconstruct the spatial distribution of Fe to detect microsites containing different Fe^II and Fe^III species. Contrasting the Fe^II and Fe^III microsites is further challenged by the likely presence of Fe in the surrounding matrix (e.g., silicate clays), that may obscure signatures of anoxic microsites. In this part, the Fe K-edge multiple energy μ-XRF mapping of Fe^II and Fe^III microsites in synthetic cores was analyzed to test our methodology and data processing approach to detect redox microsites in Fe-“poor” and Fe-“rich” environments. As previously underlined in §3.1, a shift in absorbed X-ray energy between Fe^II-S species and other Fe^II-species (such as siderite, Fe^IICO₃) enables identification of the spatial distribution of microsites containing different Fe^II-species. Thus, our initial approach compared maps of each Fe component via different fitting methods (XANES and SiVM fitting) with the actual and known locations of Fe^II-S, Fe^IICO₃, and Fe^III–(hydroxy)oxide microsites to validate and improve the data processing approach for detecting different redox microsites.

Eliminating false positive redox microsites based on comparison of the same Fe component obtained from XANES and SiVM fitting. To compare maps of the spatial distribution of Fe^II-S, Fe^IICO₃, and Fe^III–(hydroxy)oxide components of XANES and SiVM fitting without potential interferences from Fe-background, maps collected in the Fe^III-free sand synthetic core were compared first (Figure 3 and Supplementary Figure S5). Overall, the XANES and SiVM maps of each Fe component compared well, and the spatial distribution of Fe^II and Fe^III microsites were accurately identified. However, some low-intensity inconsistencies were also observed. The SiVM fitting identified the presence of low-intensity Fe^II-S microsites, that in fact were Fe^IICO₃ (Supplementary Figure S8), and the presence of low-intensity Fe^IICO₃ (Supplementary Figure S9) and Fe^III–(hydroxy)oxide (Supplementary Figure S10) microsites that were in fact Fe^II-S. Similarly, XANES fitting identified the presence of low-intensity Fe^IICO₃ microsites that were in fact Fe^III–(hydroxy)oxide (Supplementary Figure S9). These inconsistencies are referred to as false positive microsites. Theoretically, if the contribution of each Fe component was properly detected from both fitting methods, i.e., SiVM and XANES, the correlation plot between both methods should show y = x, i.e., a slope of 1. To suppress the low-intensity false positives, only the cloud of points (i.e., pixels) near the y = x line was selected for re-mapping. The re-mapping of spatial distribution of the contribution of each Fe component using exclusively the correlated points between XANES and SiVM fitting successively removed the false positive microsites and identified accurately the real locations of Fe^II-S, Fe^IICO₃, and Fe^III–(hydroxy)oxide microsites (Supplementary Figure S8, Supplementary Figure S9, and Supplementary Figure S10, respectively).

Eliminating false positive redox microsites based on comparison of distinct Fe component. Conversely to the Fe^II-S, Fe^IICO₃, and Fe^III–(hydroxy)oxide component maps of XANES fitting in the Fe^III-free sand synthetic core (proxy of Fe-“poor” background), the spatial distribution of Fe^II-S, Fe^IICO₃, and Fe^III–(hydroxy)oxide microsites of XANES fitting in the Fe^III-smectite sand synthetic core (1.6% Fe background; proxy of higher Fe background) showed significant false positives (Figure 3). These false positives exhibited a similar intensity range compared to the real microsites, which prevents the distinction between real Fe^II-S, Fe^IICO₃, and Fe^III–(hydroxy)oxide microsite and their false positives based on simple thresholding of XRF intensity. SiVM fitting in Fe^III-“rich” smectite sand environment also showed significant false positives (Figure 3). However, the nature, spatial distribution, and intensity of the false positives differed between XANES and SiVM fitting. Theoretically, if the contribution of each Fe component was properly distinguished in each pixel, no correlation should be observed while comparing with the contribution of a distinct Fe component, regardless of the fitting method. The map of the spatial distribution of the contribution of a specific Fe component (from both fitting methods, i.e., SiVM and XANES) is thus compared with the map of the spatial distribution of the contribution of each distinct Fe component via correlation plot from both SiVM and XANES fitting (Figure SM-11, SM-12, and SM-13). For example, Fe^IICO₃ component map of XANES fitting is compared separately to Fe^II-S and Fe^III-(hydroxy)oxide component map of XANES and SiVM fitting. Using Fe component maps generated from both fitting methodologies increases the statistical comparison. Comparisons that showed correlation curve(s) were then selected for re-mapping of each Fe component (see detailed explanations in Supplementary Material for further information; Figure SM-12 and SM-13). Then, the spatial distribution of the contribution of the same Fe component re-mapped from different paths of re-mapping was compared and their correlation curve selected for removing the last false positives (Figure SM-14, SM-15, and SM-16 for Fe^II-S Fe^IICO₃, Fe^III–(hydroxy)oxide component). Finally, Fe^II-S, Fe^IICO₃, and Fe^III–(hydroxy)oxide microsite maps were reconstructed following this new refined data processing approach and evidently all false positive redox microsites were successfully removed (Figure 4), validating this refinement in data processing of µXRF maps for detection and quantification of redox microsites using Fe.

The new refined data processing approach described above for synthetic cores was systematically applied to successfully detect anoxic microsites from natural cores. The choice of reference compounds of Fe species for the XANES fitting is systematically based on analysis of full μ-XANES spectra across the analyzed sample from multiple putative redox anomalies previously highlighted by SiVM fitting at the end of each mapping. The results below highlight the ability of our X-ray mapping approach to detect and quantify the anoxic microsites in natural systems. Given this approach has been successful, our future investigations will focus on characterizing the internal biogeochemical mechanisms of anoxic microsites spatially localized and the impact of these microsites on the biogeochemical processes at the scale of ecosystems probed below.

Detection and quantification of anoxic microsites from toeslope to floodplain in a Colorado mountain watershed. The spatial distribution of Fe species in cores collected from the toeslope and down-gradient floodplain in a mountainous watershed, East River (CO), were mapped and reconstructed (Figure 5). Fe in these cores mainly occurred as Fe^III-(hydroxy)oxides, Fe^II/III-silicates, and Fe^II species. The full μ-XANES spectra of Fe^II microsites shown a mixture of Fe^II and Fe^III, which make the characterization of the nature of Fe^II species complicated. While the Linear Combination-Least Squares fitting of the full μ-XANES spectra of Fe^II microsites do not provide clear differentiation on the characterization of Fe^II-species, the best XANES fitting of Fe K-edge multiple energy μ-XRF mapping was clearly obtained by including Fe^II-S component. In opposite, including other Fe^II-species component into XANES fitting of Fe K-edge multiple energy μ-XRF mapping did not reconstruct the putative redox anomalies previously highlighted by SiVM fitting that we confirmed to be Fe^II-microsites by collecting the full μ-XANES spectra.

In the toeslope, Fe^III-(hydroxy)oxides dominated, whereas, soil Fe in the floodplain mainly occurred as Fe^II/III-silicates. Floodplains typically remain saturated in water for certain period of the year, limiting oxygen supply, which could explain the lower abundance of Fe^III-(hydroxy)oxides that would be susceptible to reduction under anoxic conditions. In contrast, the drier conditions on the toeslope would not traditionally be associated with soil oxygen limitations. However, Fe^II-S microsites were detected and accounted for 2% of the surface area in cores from both sites (Figure 5). Although, it is necessary to analyze multiple cores to obtain statistical representative of (in this case) Fe^II-S microsites, these preliminary results indicate a high frequency of anoxic microsites, even in the drier toeslope soil that typically would be expected to be fully oxic. Also, these investigations suggest a change in the size distribution of Fe^II-S microsites in soil cores with larger anoxic (i.e., Fe^II-S) microsites being more frequently distributed in floodplains.

Spatial distribution of reduced Fe in aggregates in well-aerated California oak savanna soils. The reconstruction maps of Fe species distribution inside aggregates from the toeslope and ridge soils of an oak savanna, Stanford (CA) are displayed in Figure 6. Although, oak savanna is not traditionally associated with soil oxygen limitations, soil aggregates show significant proportions of reduced Fe, suggesting current or past development of anoxic microsites inside the soil aggregates.

Chemically specific X-ray imaging can record “forensic” evidence of Fe and S reducing microsites, ranging from μm to cm in size (length/width), from different soils and environments (Figure 5; 6). The location, along with the size, ultimately determines the susceptibility of the anoxic microsites for disturbance and the absolute volume over which anaerobic processes may occur (Lacroix et al., 2023). In our study, for example, the contrast of microsite size between toeslope and floodplain could influence the impact of anoxic microsites might have on biogeochemical cycling for each soil type, and consequently, constitutes a fundamental parameter to include in models. Although, direct evidence of anoxic microsite sizes and distribution is crucially missing to support model parametrization (Lacroix et al., 2023), there are many approaches for including their functional contributions. For example, certain models based on spatially discretizing soil into distinct biogeochemical environments (Currie, 1961; Parkin, 1987) are dependent on aggregate size (Wang et al., 2019). Similarly, the 3-D angular pore network model was expanded to include interactions of aggregates of different sizes and orientations (Ebrahimi and Or, 2016) and upscaled to demonstrate, how aggregate size distributions alters anoxic microsite distribution and, thus, biogeochemical gas fluxes from soil profiles (Ebrahimi and Or, 2018a). Revealing the distribution and size of Fe and S reducing microsites could thus drastically improve the accuracy of biogeochemical models.

Our X-ray mapping revealed also the spatial distribution of Fe-reducing microsites inside aggregates from an oak savanna soil. Aggregates typically predominantly contain clay particles and organic matter adhering together and expanding/contracting through actions of plant roots, soil organisms, and/or hydration and thermal cycles (Brady and Weil, 2008; Totsche et al., 2018). As a result of their composition (i.e., small pores, enhanced water retention), aggregates typically experience limited O₂ supply (Currie, 1961; Sexstone et al., 1985; Ebrahimi and Or, 2016; Ebrahimi and Or, 2018b; Keiluweit et al., 2018), and favor the formation of anoxic microsites in the interior of these aggregates (Currie, 1961; Sexstone et al., 1985; Ebrahimi and Or, 2018b). It is commonly postulated that due to diffusion kinetics, anoxic conditions develop preferentially in the core of aggregates (Currie, 1961; Figure 6). Microelectrode measurements confirmed localized anoxia at the interior of both saturated (Greenwood and Goodman, 1967) and unsaturated (Sexstone et al., 1985) aggregates. Our results, however, challenge this classical view of the anaerobicity inside aggregates, suggesting preferential spatial distribution of Fe-reducing conditions that are not specifically concentrated in the center of the aggregate (Figure 6). Instead, we hypothesize that Fe reducing conditions depend on heterogeneous pore structure (preferential water transport paths and/or compaction zones), as well as heterogeneous organic matter and root spatial distribution. Thus, the complementary approaches are necessary to interpret the insights gained from X-ray 2D imaging. For example, we recommend future research to overlap the 2D imaging techniques detecting anoxic microsites with those revealing spatial distribution of physicochemical parameters (such as organic matter and pore structure; Vongsvivut et al., 2019; Noël et al., 2022), in order to identify the key “drivers” that control the spatial distribution of the redox heterogeneity. Similarly, the omic methods such as the meta-transcriptomics and fluorescence methods could identify active microbial communities (Djemiel et al., 2022; Rosado-Porto et al., 2022). Coupling these microbial measurements at the time of soil core collection could help determine if the biogeochemical products composing anoxic microsites are from active anaerobic microbial processes in soils, providing a critical linkage of mapped element species with the timeframe of redox processes.

Lacroix et al. (2023) highlighted three domains of variability in how anoxic microsites are manifested: anoxic microsite distribution, redox gradient magnitude, and temporality. As discussed in §4.1, Fe K-edge X-ray mapping largely covers the detection of anoxic microsite distribution, including the size and location of anoxic microsites. However, Fe reduction represents only one set of anoxic processes among many. Anoxic microsites would be expected to host gradients of redox potentials below the anoxic threshold with multiple chemical transformations at various distances from each other (Lacroix et al., 2023). Driven by electron donor/acceptor concentrations and their ability to move through the soil pore space (i.e., exhibiting diffusion vs. advective transport; Keiluweit et al., 2017), redox gradients (i.e., the magnitude of redox decline over a given distance) influence the products generated inside anoxic microsites and the potential for those products to diffuse to the oxic surrounding environment (Pallud et al., 2010a; Harris et al., 2021; Naughton et al., 2021). For example, a relatively rapid increase in concentration of electron donors within a model aggregate system would promote a sharp redox gradient (i.e., a large redox decline over a relatively small distance), favoring Fe^IICO₃ neoformation, whereas Fe^IICO₃ is unlikely to precipitate in a system with a more gradual transition of redox potentials (Pallud et al., 2010b). Similarly, the interaction of hydrogen sulfide with Fe^III-oxides is dependent on gradient magnitude, with sharper gradients (i.e., shorter distances between sulfidic species and Fe(III) species) that are potentially more likely to support coprecipitation of metastable Fe^II-S (Lacroix et al., 2023). Thus, the occurrence and distribution of biogeochemical products, such as Fe^IICO₃ (Figure 4) and Fe^II-S (Figure 5), could record indirect “forensic” evidence of redox gradient magnitude. Furthermore, the development of the new capabilities combining the High Energy Resolution Fluorescence Detection (HERFD) XANES with µ-XRF map could improve spectral energy resolution as compared to conventional XANES spectroscopy (Edwards et al., 2022), improving sensitivity to local- and intermediate range structure/composition, and thus the detection of different Fe species (as an example). However, Fe K-edge X-ray mapping alone is not sufficient to capture the large set of different redox potentials and their associated redox gradient. We envision that spatial correlation of multiple energy X-ray maps of multiple redox-sensitive elements with varying redox potentials (e.g., Mn, Fe, S and trace metal (loid)s; Jones et al., 2018; Noël et al., 2017a) would extend our ability to quantify the 2D distribution of anoxic microsites with different redox potentials (Figure 7), redox gradients and, thereby, biogeochemical functions. Additionally, overlapping multiple energy μ-XRF 2D maps of different redox-sensitive elements could rule out false positives and complement the information gained individually by each element. As an example, while Fe multiple energy µ-XRF map cannot distinguish Fe-monosulfide from Fe-disulfide, the µ-XRF map at different S K-edge energies could confirm the presence of Fe-sulfides detected by Fe µ-XRF map and determine their nature (i.e., FeS vs Fe₂S; Noël et al., 2017a).

In our study, we have clearly shown that Fe K-edge X-ray mapping records “forensic” evidence of anoxic microsites ranging from a few hundred μm to cm scales (Figure 4). We cannot exclude that smaller anoxic microsites may also exist and were not captured with this approach. Thus, we cannot assess the potential relevance or distribution of anoxic microsites <100 μm, but given the size of plant root hairs, microbial cells, organic compounds, and mineral particles it is likely that the abundance and biogeochemical relevance of anoxic microsites is potentially even larger than what is currently suspected and indicated by results in this study. Expanding the range of anoxic microsite sizes (especially smaller) detectable using X-ray mapping requires refinement of the coring methodologies and sample preparation approaches. Coring techniques should be designed to limit physical (e.g., compaction) and chemical (e.g., change in oxygen availability) alterations of cores. Preparation approaches for sectioning samples should avoid physical smearing, which will particularly help with preserving smaller microsites. We note that maintaining the sample physical integrity while producing a plane surface is extremely challenging; a difference in surface topography along a sliced core may directly impact the spatial resolution for detecting anoxic microsites, as well quantifying the concentration of elements within these microsites.