Mineral element stocks in the Yedoma domain: a first assessment in ice-rich permafrost regions

With permafrost thaw, significant amounts of organic carbon (OC) previously stored in frozen deposits are unlocked and become potentially available for microbial mineralization. This is particularly the case in ice-rich regions such as the Yedoma 15 domain. Excess ground ice degradation exposes deep sediments and their OC stocks, but also mineral elements, to biogeochemical processes. Interactions of mineral elements and OC play a crucial role for OC stabilization and the fate of OC upon thaw, and thus regulate carbon dioxide and methane emissions. In addition, some mineral elements are limiting nutrients for plant growth or microbial metabolic activity. A large ongoing effort is to quantify OC stocks and their lability in permafrost regions, but the influence of mineral elements on the fate of OC or on biogeochemical nutrient cycles has received less 20 attention. The reason is that there is a gap of knowledge on the mineral element content in permafrost regions. Here, we use a portable X-ray fluorescence device (pXRF) to provide (i) the first large-scale Yedoma domain Mineral Concentrations Assessment (YMCA) dataset (doi:10.1594/PANGAEA.922724; Monhonval et al., in review), and (ii) estimates of mineral element stocks in never thawed (since deposition) ice-rich Yedoma permafrost and previously thawed and partly refrozen Alas deposits. The pXRF method for mineral element quantification is non-destructive and offers a complement to the classical 25 dissolution and measurement by optical emission spectrometry (ICP-OES) in solution. This allowed a mineral element concentration (Si, Al, Fe, Ca, K, Ti, Mn, Zn, Sr and Zr) assessment on 1292 sediment samples from the Yedoma domain with lower analytical effort and affordable costs relative to the classical ICP-OES method. pXRF measured concentrations were calibrated using standard alkaline fusion and ICP-OES measurements on a subset of 144 samples (R2 from 0.725 to 0.996). The results highlight that (i) the most abundant mineral element in the Yedoma domain is Si (2739 ± 986 Gt) followed by Al, 30 Fe, K, Ca, Ti, Mn, Zr, Sr, and Zn, and that (ii) Al and Fe (598 ± 213 and 288 ± 104 Gt) are present in the same order of magnitude than OC (327-466 Gt). https://doi.org/10.5194/essd-2020-359 O pe n A cc es s Earth System Science Data D icu ssio n s Preprint. Discussion started: 8 December 2020 c © Author(s) 2020. CC BY 4.0 License.


Introduction
To improve our understanding on the potential effects mineral elements can have on OC from permafrost regions, it is essential to have better knowledge on mineral element stocks in these regions, and on the transformation of these mineral element stocks following thawing. Ice-rich permafrost regions are particularly well suited areas to study the impact of thawing processes on 70 the mineral elemental stocks in deposits. Indeed, 'never' (since deposition) thawed deposits can be compared with previously thawed deposits resulting from thermokarst processes of the warmer Pleistocene/Holocene transition period (13 to 9 ka BP; Morgenstern et al., 2013;Walter et al., 2007). Assessing the evolution of the mineral element stocks in these deposits will contribute to better predict the impact of the ongoing climate change on mineral element content in ice-rich permafrost regions, with implications for the fate of OC in these deposits (Strauss et al., 2017). This is particularly relevant given that thermokarst 75 processes are projected to spread across the Arctic and will potentially unlock additional OC stocks (Abbott and Jones, 2015; Lacelle et al., 2010;Nitzbon et al., 2020;Schneider von Deimling et al., 2015).
The aim of this study is to provide a first large-scale assessment of a climate sensitive mineral stock. We choose the Yedoma domain and provide the Yedoma domain Mineral Concentrations Assessment (YMCA) dataset including mineral element 80 concentrations (Si, Al, Fe, Ca, K, Ti, Mn, Zn, Sr, and Zr) and the stocks for these mineral elements in deposits from ice-rich permafrost regions. This comprises 'never' thawed since syngenetic freezing Yedoma Ice Complex deposits (in the following referred as Yedoma) and in at least once previously thawed and then refrozen drained thermokarst lake deposits (in the following referred as Alas). 85 2 Environmental settings

The Yedoma domain
The Yedoma domain is part of the permafrost region characterized by organic-and ice-rich deposits as well as thermokarst features. Today, the Yedoma domain covers areas in Siberia and Alaska ( Figure 1) which were not covered with ice sheets during last glacial period (110 ka -10 ka BP; Schirrmeister et al., 2013). 90 https://doi.org/10.5194/essd-2020-359 from the Circum-Arctic map of permafrost and ground-ice conditions shapefile (Brown et al., 1997). Yedoma domain coverage from Strauss et al. (2016) (Database of Ice-Rich Yedoma Permafrost (IRYP) Pangaea shapefiles).
Recent estimates indicate that the current Yedoma domain is ~1.4 million km² in extent and contains between 327 -466 Gt OC (Strauss et al., 2017). Frozen deposits from the Yedoma domain alone store probably at least as much carbon as the tropical 100 forest biomass (Lai, 2004). These deposits were formed by long-term continuous sedimentation and syngenetic freezing.
Depositional processes are polygenetic including alluvial and aeolian deposition and re-deposition, as well as in situ weathering during the late Pleistocene cold stages . Despite evidence of homogeneity of Yedoma deposits aggradation, grain-size analyses show the large diversity of Yedoma deposits which result from multiple origins and transport of sediments, as well as (post)depositional sedimentary processes Strauss et al., 2012). For 105 millennia, continuous sedimentation lead to the accumulation of several tens of meters thick permafrost deposits with characteristic ice-wedge formation. Harsh cold late Pleistocene climate triggered the formation of freezing cracks within deposits in which water filling and refreezing created, with time, wide and up to 40 m high ice wedges. Those large volumes of ice within sediments together with the rise in temperature during the Holocene Thermal Maximum (between 11-5 ka BP depending on the region; thermal maximum reached at 7.6 -6.6 ka BP in North Alaska and variable with time in Siberia 110 regions; Porter and Opel, 2020) lead to a vast reshaping of the landscape with formation of thermokarst lakes and Alas basins, resulting from the drainage of former lakes Kaufman et al., 2004;Velichko et al., 2002). Alas deposits in drained thermokarst lake basins are composed of reworked Yedoma deposits as well as Holocene accumulation during sub-https://doi.org/10.5194/essd-2020-359 aquatic and sub-aerial phases (Strauss et al., 2017;Windirsch et al., 2020). The Yedoma domain area includes 30% of unthawed Yedoma deposits composed of homogeneous silty deposits with polygenetic origins (eolian, alluvial or colluvial deposition) 115 between large ice-wedges. Alas deposits have experienced thawing processes and drainage during early Holocene until more recently and account for 56% of the Yedoma domain area. Deltaic deposits (4%) as well as lakes and rivers (10%) complete the Yedoma domain deposits area distribution ( Figure 2).

Sampling sites
This large-scale mineral element concentration assessment from Yedoma and Alas deposits is the result of more than 20 years of sampling in remote locations such as Siberia and Alaska. This dataset includes 22 locations from north and Interior Alaska to Laptev Sea coastal regions, Kolyma region and New Siberian Islands (Figure 1). For each location, Yedoma and Alas deposits profiles were sampled if both types of deposit were present. In total, 1292 different samples were analyzed for their 135 mineral element concentration. Sampling strategies were specific for each sampled features. During the frozen season, samples were collected by drilling from the surface down with drilling rigs whereas during the summer season, drilling was performed below the active layer. Frozen samples from cliffs or headwall exposures were cleaned and sampled with hammers and axes.
For headwall or cliff sampling, sub-profiles from different vertical exposures were included when needed to reconstruct a complete composite profile. Additional information on specific site sampling techniques can be found in reference papers cited 140 in Table 1. Recovered samples were air-or freeze-dried before being stored and archived.

'Classical' method for total elemental analysis: ICP-OES measurement after alkaline fusion 150
Inductively coupled plasma optical-emission spectrometry (ICP-OES) is a classical method to assess mineral element concentrations in solutions from the environment accurately. Solid phases such as soil samples should be first digested prior to ICP-OES analysis. Here, soils were digested by alkaline fusion. Briefly, air-or freeze-dried soil samples are carefully milled for homogenization (agate mill). We mixed a portion of the milled sample (80 mg) with Lithium metaborate and Lithium tetraborate and heated it up to 1000 °C for 10 minutes. Then we dissolved the fusion bead in HNO3 2.2N at 80°C and stirred 155 until complete dissolution (Chao and Sanzolone, 1992). We measured the mineral element concentrations in that solution by ICP-OES (iCAP 6500 Thermo Fisher Scientific). We assessed the loss on ignition at 1000°C, and the total element content in soils is expressed in reference to the soil dry weight at 105 °C. The analytical measurement is validated by repeated measurements on the USGS basalt reference material BHVO-2 (Wilson, 1997). To assess the precision of the method, we conducted three repetitions on three individual samples from Yedoma and Alas deposits. For each mineral element relative 160 standard deviations on the repetitions, expressed in % to the mean, are available in Table 2.
Out of the 1292 deposit samples retrieved from the Yedoma domain (Sect. 2.2), a subset of 144 samples has been analyzed by ICP-OES after alkaline fusion to determine their mineral element concentrations. We used this subset of analysis to calibrate pXRF-measured element concentrations (Sect. 3.2) for accurate determination of concentration values. For this first assessment, we measured the following elements on the subset of samples (except for Zn, which was measured on 119 out of 165 the 144 samples): Si, Al, Fe, Ca, Mg, Na, Cr, Ba, K, Ti, P, Cu, Mn, Ni, Zn, Sr, and Zr.

3.2
'Alternative' method for total elemental analysis: ex situ direct measurement by portable XRF X-ray fluorescence (XRF) spectrometry is an elemental analysis technique with broad environmental and geologic applications, from pollution assessments to mining industries (Ravansari et al., 2020;Rouillon and Taylor, 2016;Weindorf et 170 al., 2014a;Young et al., 2016). In addition, there is a growing use of portable XRF for soil science (McLaren et al., 2012;Ravansari and Lemke, 2018). Portable XRF is used primarily for solid elemental analysis (soils, sediments, rocks or even plastics) but can also deal with oil chemical characterization (Weindorf et al., 2014a). XRF is based on the principle that individual atoms emit photons of a characteristic energy or wavelength upon excitation by an external X-ray energy source.
By counting the number of photons of each energy emitted from a sample, the elements present may be identified and 175 quantified (Anzelmo and Lindsay, 1987; Appendix A). XRF-scanning results represent elements intensities in "counts per second" (cps) which are proportional to chemical concentrations in the sample but depend also on sample properties (Röhl and Abrams, 2000), ice and water content (Tjallingii et al., 2007;Weindorf et al., 2014b) and interactions between elements called "matrix effect" (Fritz et al., 2018;Weltje and Tjallingii, 2008). In situ pXRF measurement often involves variability from uncontrolled environmental factors, such as water content, organic 180 matter content or sample heterogeneity (Ravansari et al., 2020;Shand and Wendler, 2014;Weindorf et al., 2014b). To avoid such variability in water content, measurements were performed on dried samples in laboratory (ex situ) conditions with the handheld device in the laboratory. The particle size distribution of these deposits allows considering sample homogeneity within the fraction inferior to 2 mm (described in reference paper from Table 1). For the pXRF measurement, the dried sample is placed on a circular plastic cap (2.5 cm diameter) provided at its base with a transparent thin film (prolene 4µm). To avoid 185 the underestimation of the detected intensities sample thickness in the cap needs to be higher than 5 mm to 2 cm, depending on the element of interest (Ravansari et al., 2020). For a precise measurement, the sample thickness in the cap is set to >2 cm.
Above 2 cm, the width is considered as "infinitely thick" for all elements. Measurements on the 1292 samples from the Yedoma domain were performed using the pXRF device Niton xl3t Goldd+ (Thermo Fisher Scientific), which has two specific internal calibration modes called mining and soil. Each internal calibration is dealing with different energy range and filters to scan the 190 complete energetic spectrum from low to high-energetic fluorescence values. Both modes were used on each sample and depending on the element, the calibration with the best correlation with the classical method (Sect. 3.1) was kept for further calculations (Sect. 3.3). To standardize the analysis, total time of measurement was set to 90 seconds. We conducted the analysis in laboratory conditions, using a lead stand to protect the operator from X-rays.
In theory, the pXRF device used to generate this dataset can measure simultaneously elements of atomic mass from Mg to U. 195 Because ambient air annihilates fluorescence photons that do not have enough energy, low atomic mass elements from Na and lighter cannot be quantified by pXRF. Note that Na quantification would be possible in controlled void conditions during analysis (Weindorf et al., 2014a). In this study, we focussed on the concentrations in 16 elements (Si, Al, Fe, Ca, Mg, Cr, Ba, K, Ti, P, Cu, Mn, Ni, Zn, Sr, Zr) by pXRF and by the classical ICP-OES method (Sect. 3.1) to allow for quality check, calibration and correction. Some elements are at the limit of detection (LOD) for pXRF device (e.g., Cu, Ni). The LOD is 200 reached when the sample is lacking a specific mineral element. LOD concentrations are set to 0.7 times the minimal concentration measured for this element, which is an arbitrary number but conventionally used for data statistical treatment (Reimann et al., 2008). Depending on the considered element, pXRF measured concentrations are highly precise but not always accurate (far from the true value). Trueness is achieved after correction using a regression with concentration values measured by the classical method to avoid systematic bias (Sect. 4.1). Using a well-defined regression to correct pXRF measurements 205 for trueness allows using the pXRF method to measure mineral element concentrations on a large number of Yedoma and Alas sample (n=1292), a valuable alternative method to assess the mineral element content on a large sample set ( Figure 4).  To assess the precision of the pXRF method, three to five repetitions were conducted on 20 individual samples from different locations from Yedoma and Alas deposits. Between each repetition, instrument/sample repositioning is used to mitigate the 215 "nugget effect" due to heterogeneity in the sample (Ravansari and Lemke, 2018). Relative pooled standard deviations (SDpooled, weighted average of standard deviations divided by weighted average of the means), expressed in %, of the repetitions for the ten elements used for stock calculation (Sect. 4.1) are available in Table 2.  (Strauss et al., 230 2013) and was improved by Jongejans and Strauss (2020). Because Yedoma domain deposits are composed of massive ice volume, stock calculations need to take into account not only the bulk density (BD) of the samples but also total thickness and total wedge-ice volume (WIV) of the deposits. In order to avoid overestimation of the mineral stocks, WIV was subtracted from the total Yedoma domain deposit volume, since the proportion of mineral elements locked inside ice wedges is negligible.
Assumptions and calculations for BD determination, WIV estimations, and deposits thickness are fully explained in Strauss et 235 al. (2013) and are summarized here. The bootstrapping statistical method use resampled (10 000 times) observed values (mineral element concentrations, BD, WIV and deposits thickness (Appendix B)) and derive the mean afterward. The BD determination for each sample was obtained by using an inverse relationship with porosity ( ) Eq. (1) (Strauss et al. 2013): (1), whereas is the solid fraction density. We assume that pore volume, in saturated ice samples, is directly measured with segregated ice volume. For samples where no BD determination is available, BD is inferred from the 240 total organic carbon (TOC) content using equation Eq. (2): = 1.126 −0.061× (2) (Strauss et al., 2013). If neither BD or TOC is available, BD is fixed to 0.88 10 3 kg.m -3 and 0.93 10 3 kg.m -3 for Yedoma and Alas deposits, respectively, i.e., the mean BD measured in such deposits (Strauss et al., 2013) and comparable with other studies on Yedoma deposits (0.98 10 3 kg.m -3 ; Dutta et al., 2006). The WIV estimations were performed using ice-wedge width from field measurements (Strauss et al., 2013), ice-wedge polygon size determinations from high-resolution satellite, and additional geometrical tools (assuming ice-245 wedges have an isosceles triangle or rectangle shape depending on the type of ice-wedge; Strauss et al., 2013;Ulrich et al., 2014). To obtain the best approximate deposits thickness, the mean profile depths of the sampled Yedoma and Alas deposits were used following Strauss et al. (2013). Using these parameters, the overall mineral element stock is determined with Eq. (3) included into the bootstrapping calculation: Since the Yedoma domain is composed of ice-rich Yedoma deposits (30% of the area coverage) or Alas deposits (56% of the total area coverage; Figure 2) and because Yedoma and Alas deposits have different properties, from BD to ice volume content and thickness, stocks were estimated for each deposit feature individually. Calculation for Yedoma deposits stocks are 255 conducted as follows: the mineral element concentrations from Yedoma deposits samples (n=814) are selected (Alas deposits are not taken into account), Yedoma deposit thicknesses (n=19), and WIV properties from Yedoma deposits (n=18) are selected from Strauss et al. (2013). Total coverage is set to 410 000 km², following the last Yedoma deposits coverage estimation (Strauss et al., 2013). With all these input parameters ( paired (as they are not independent), a specific stock is estimated based on one plausible value for WIV and thickness. The stock is then multiplied by total coverage for total mineral stock estimation to the Yedoma domain scale. From these input 265 parameters, multiple mineral element stocks are computed, one for each bootstrapping step. Eventually, from those multiple steps (n=10 000), a mineral element stock distribution is estimated from which the mean represents the best stock estimation of the considered mineral element. The error estimates in this study represent 2 standard deviations (mean ±2σ), with the assumption of a normal distribution. Computations were performed using R software (boot package; R Core Team, 2018).
Supplementary information on input parameters (deposits thickness, WIV) are available (Appendix B). 270

Linear regression for accurate mineral elements concentration measurements
Mineral  ; Table 3). Linear regression plots for these elements are presented in Figure 5. The other six elements (Mg, Ba, Cr, Cu, Ni and P) present weak (R² < 0.5) or no correlations between the two methods. According to these correlations, the mineral element stock quantification will be performed on the 10 elements reliably measured by pXRF (Si, Al, Fe, Ca, K, Ti, Mn, Zn, Sr and Zr), and the other six elements will not be discussed 280 further in this study. The uncertainty on the ICP-OES measurement after alkaline fusion is always lower than pXRF measurements (Table 2).
Nonetheless, the advantages of a non-destructive, rapid and cheaper method predominate for a large-scale mineral concentration assessment given that the uncertainty (5.5% to 17%) on the pXRF measurement is satisfying to differentiate 290 between high and low concentrations values measured in these deposits (i.e., comparing the horizontal error bar with the total range of values represented on the X-axis in Figure 5). In this dataset, the risk of overestimating the element concentration by pXRF in organic-rich samples (Shand and Wendler, 2014) is limited. Among the 1292 Yedoma and Alas samples used in this study to build the YMCA dataset (Sect. 4.2), less than 0.5% of the samples are characterized by TOC content similar to or higher than 40 wt% (Table 4). Overestimation of the 295 concentration was only observed for a single sample for Fe ( Figure 5c) and for three samples for Ca (Figure 5d) out of the 144 samples from the subset (for samples with TOC content > 40 wt%), and not observed for Si, Al, K, Ti, Mn, Zn, Sr and Zr.
Since the points corresponding to these organic-rich samples are excluded from the robust linear regressions (Figure 5), it can be considered that the matrix effect is not a source of significant bias to correct pXRF measurements of element concentrations using these regressions. Thus, the pXRF method can be applied for this first large-scale mineral element assessment of Yedoma 300 domain deposits.

Mineral element stocks in the Yedoma domain
The quantification of mineral element stocks is a major step to assess the distribution of mineral elements in the Yedoma domain. We performed mineral elements stock estimations with the bootstrapping technique (Sect. 3.3).

Advantages and limitations of the ICP-OES and pXRF method for mineral element concentration measurements
Although considered as a classical method for total element analysis, the main drawbacks of the ICP-OES technique after alkaline fusion are to be a destructive, multiple steps and time consuming (involving more risks of error), resulting in high costs analytical protocol when dealing with thousands of samples. Indeed, this protocol includes: (i) the homogenization of samples before weighting, a critical step to avoid heterogeneities or nugget effect, (ii) the determination of dry weight for each 390 sample to correct for sample's moisture, (iii) the need to assess loss on ignition to close the budget of the sample oxide content ( Figure 4). For these reasons, we have investigated for a suitable alternative method. obtained with the classical ICP-OES method after alkaline fusion (n=144). From this YMCA dataset, mineral element stocks were estimated using a mean-bootstrapping technique previously applied to estimate the OC budget in the Yedoma domain 400 (Strauss et al., 2013). However, some important mineral elements could not be assessed because of poor correlation with classical ICP-OES method (Mg, P, Cu, Ni, Ba) or because of their low atomic masses which make XRF-measurement impossible (N, Na, S).

Implications for mineral element mobility upon thermokarst processes 405
The YMCA dataset allows investigating the mineral element behavior upon thaw. Mineral element concentrations of some soluble elements, such as Ca, could be influenced by thermokarst processes. Indeed, Alas formation history includes lake formation and drainage, and the dynamism of such formation over the past thousands of years may lead to leaching processes of soluble elements, a commonly observed process in soils (Stumm and Morgan, 1995) and cryogenic soils (Ping et al., 2005).
As shown on a density plot (Figure 8), sediments from Alas deposits are characterized by lower Ca concentrations compared 410 to Yedoma deposits. We made this observation between Alas and Yedoma in Alaska, and between Alas and Yedoma in Siberia ( Figure 8). The higher Ca concentrations in Yedoma and Alas deposits from Alaska relative to the deposits from Siberia can be explained by the local lithology, i.e., carbonate rocks from the northern Brooks Range contributing to the deposits in Alaska (Till et al., 2008). The lithology of the underlying bedrock inferred from the GLiM map is similar between Yedoma and Alas deposits (Appendix C). This can likely be explained by the fact that the deposits from the Yedoma domain mainly originates 415 from Quaternary deposits overlying the bedrock (Grosse et al., 2007;Strauss et al., 2017). Therefore, mineral element concentrations are more influenced by the mixing of the unconsolidated sediments contributing to the Quaternary deposits rather than by the lithology of the underlying bedrock. The lithological similarity between Yedoma and Alas deposits can be also explained by the fact that Alas deposits are dominated by reworked sediments from former Yedoma deposits . Therefore, Ca depletion in Alas deposits relative to Yedoma deposits from one region is not 420 lithology dependent but potentially results from leaching processes of soluble elements such as Ca during former thawing periods.  This observation is based on the YMCA dataset comprising more than a thousand samples supporting that Ca solubility and leaching mechanisms following Holocene thermokarst processes is a suitable hypothesis to explain lower Ca concentration in Alas deposits relative to Yedoma deposits. This highlights the potential of our study and the YMCA dataset to investigate 430 dynamic processes controlling mineral element concentrations in thawing environments.

5.3
Implications of mineral elements release upon permafrost thaw

Implications for organic carbon stabilization/degradation 440
The evolution of interactions between mineral elements and OC upon thaw is of major concern to predict OC stabilization/degradation patterns in the context of permafrost thaw. Changing conditions for mineral protection of OC upon thawing is likely to influence OM degradation (Herndon et al., 2017;Kögel-Knabner et al., 2010;Opfergelt, 2020). Indeed, mineral elements play a key role for OC microbial degradation. This was illustrated by incubation experiments highlighting the importance of organo-mineral interactions for soil OC sequestration (Gentsch et al., 2015), or by the inhibition effect of 445 some mineral elements (e.g., Fe, Mn) on methane production (Beal et al., 2009;Herndon et al., 2015;Lovley and Phillips, 1987;Sowers et al., 2018). The OC can be stabilized via i) spatial inaccessibility of OM against decomposers organisms due to occlusion in aggregates; ii) organo-mineral associations with Fe-, Al-, Mn-oxides or clay minerals using polyvalent cations bridging between OC and mineral surfaces (e.g., Ca 2+ , Sr 2+ in neutral and alkaline soils or Fe 3+ and Al 3+ in acid soils); or iii) organo-metallic complexes involving Fe 3+ , Fe 2+ and Al 3+ ions (Lutzow et al., 2006). Up to 80% of the total soil C can be stored 450 within mineral horizons in the first meter of tundra soils of Arctic Siberia (Gundelwein et al., 2007). Providing a first assessment of the mineral element content in the Yedoma domain deposits (the YMCA dataset) identifies the mineral elements (e.g. Al, Fe, Ca, Mn and Sr) potentially available to interact with OC in permafrost landscapes. Mineral elements from permafrost could contribute to modulate the permafrost carbon feedback through the evolution of mineral-OC interactions upon thaw. 455

Implications for the supply of nutrients
Other processes through which mineral elements from the permafrost can affect the carbon cycle and the permafrost carbon feedback is by indirect pathways such as nutrient supply for terrestrial or aquatic Arctic ecosystems. Macro-(e.g., K, Ca) and micro-nutrients (e.g., Fe, Mn, Zn) are required for plant nutrition and also regulates other vital processes for plants and 460 microorganisms growth and metabolic activity (DalCorso et al., 2014). Briefly, K regulates vital processes, such as photosynthesis, water and nutrient transportation, or protein synthesis (Marschner, 2012). Ca is a major second messenger in plant signal transduction, mediating stress-and developmental processes (Liese and Romeis, 2013). Micro-elements are required as cofactor for some essential proteins (Fe; Morgan and Connolly, 2013), play an essential role as a photosynthetic function or in metallo-protein conformation (Mn; Yang et al., 2008) or can have enzymatic functions (Zn; Lindsay, 1972). 465 Some non-essential elements, such as Si and Al, can stimulate plant growth by playing with abiotic-biotic stress resistance and symbiosis (DalCorso et al., 2014;Richmond and Sussman, 2003). In aquatic ecosystems, silicon is a limiting nutrient for diatoms and other siliceous organisms, thereby controlling diatom abundance and community structure in the ocean, and as a result, food web and CO2 uptake by photosynthesis (Smetacek, 1999;Yool and Tyrrell, 2003). Assessing the concentration of elements considered as immobile such as Ti or Zr can be useful to evaluate the advance of weathering in a soil relative to its 470 https://doi.org/10.5194/essd-2020-359 parent material, and thereby the soil mineral reserve (Hodson, 2002;Jiang et al., 2018;Kurtz et al., 2000): the leaching of more mobile cations such as Ca or K can be estimated by comparing ratios between mobile and immobile elements. Other important nutrients (e.g., P, N, S, Mg) could not be estimated in this study (Sect. 5.1). Therefore, the YMCA dataset is far from a comprehensive dataset but remains a first step needed for ice-rich permafrost mineral assessment.

Conclusions
This study provides the first mineral element inventory of permafrost deposits focusing on the ice-rich Yedoma region, i.e., never thawed Yedoma deposits and previously thawed Alas deposits. Based on the YMCA dataset, the stocks of 10 mineral 480 elements in Yedoma domain deposits have been quantified. Mineral elements stocks are shown to be in the same order of magnitude for Al and Fe than for OC, and to decrease from Si, Al, Fe, K, Ca, Ti, Mn, Zr, Sr, to Zn. Among the 10 investigated elements, Si has the highest stock with about 2700 Gt. This dataset allows tracking dynamic processes controlling mineral element concentrations in thawing environments, as illustrated by lower Ca concentration in Alas deposits relative to Yedoma deposits highlighting potential Ca leaching upon thawing. This dataset also provides a vertical distribution of 10 important 485 mineral elements in 75 different profiles from the Yedoma domain. In permafrost soils, between 30 and 80% of OC is considered to be mineral-protected, i.e., involving interactions between OC and mineral constituents or mineral elements.
Providing the YMCA dataset is contributing to improve the knowledge on the mineral side of permafrost, i.e., a necessary step to better understand the evolution of mineral-OC interactions upon permafrost thaw. The YMCA dataset is particularly relevant given the increasing occurrence of abrupt thaw in ice-rich permafrost regions. Abrupt thaw exposes deep mineral horizons, 490 and thereby a stock of mineral elements potentially available to interact, directly or indirectly, with OC and influence the fate of OC upon thaw.

Appendices
Appendix A 500 Fig. A1: Principles of the X-ray fluorescence methodology. An X-ray source emitted by the portable XRF device (here, Niton xl3t Goldd+, Thermo Fisher Scientific) excites and ejects an electron from a specific atom. An outer electron fills vacancy to stabilize the overall atom and the electron translocation process implies an energy loss (fluorescence). This energy, specific to each atom, is detected, amplified and processed by the pXRF device. The processing converts photon energy of a specific wavelength to counts per seconds to concentration of a specific element. Yedoma (Y). The bootstrapping statistical method use resampled (10 000 times) observed values (circled in red; i.e., mineral element concentration, bulk density, deposits thickness and ice-wedge volume (WIV)) and derive the mean afterward. This bootstrapping technique is used due to the non-normal distribution of the parameters. We used sampling with replacement, which means that after each step of the random draw from the original sample, we put the observation back before the following step. The process is done with 10 000 steps from which a distribution is obtained. To estimate mineral elements stocks, we use the arithmetic mean and standard deviation assuming normal 515 distribution (Strauss et al., 2013).

Authors contribution
AM, SO, and JS conceived the project. JS, GG, LS, MF and PK retrieved samples from Alaska and Siberia from many different field expeditions. AM did the pXRF measurements with the help of EM. AM, SO, EM, BP and AV contributed to set up the 535 YMCA database. AM analyzed the data and calculated to the stocks based on the code developed by JS for carbon stock estimation using mean-bootstrapping. AM prepared the manuscript with contributions from all co-authors.

Competing interests
The authors declare that they have no conflict of interest.

Acknowledgments 540
The authors acknowledge Hélène Dailly and Anne Iserentant from the MOCA analytical platform at UCLouvain for mineral elemental analysis. We thank Waldemar Schneider (AWI logistics) and Dmitry Melnichenko (Hydrobase Tiksi) for decadal logistical support, and Catherine Hirst, Maxime Thomas, and Pierre Delmelle for fruitful scientific discussions. This project received funding from the European Union's Horizon 2020 research and innovation program under grant agreement No. 714617 to SO (WeThaw). SO also acknowledges funding from the National Funds for Scientific Research FNRS (FC69480). 545 This work was embedded into the Action Group "The Yedoma Region" funded by the International Permafrost Association.