Agricultural acceleration of soil carbonate weathering

Soil carbonates (i.e., soil inorganic carbon or SIC) represent more than a quarter of the terrestrial carbon pool and are often considered to be relatively stable, with fluxes significant only on geologic timescales. However, given the importance of climatic water balance on SIC accumulation, we tested the hypothesis that increased soil water storage and transport resulting from cultivation may enhance dissolution of SIC, altering their local stock at decadal timescales. We compared SIC storage to 7.3 m depth in eight sites, each having paired plots of native vegetation and rain‐fed croplands, and half the sites having additional irrigated cropland plots. Rain‐fed and irrigated croplands had 328 and 730 Mg C/ha less SIC storage, respectively, compared to their native vegetation (grassland or woodland) pairs, and irrigated croplands had 402 Mg C/ha less than their rain‐fed pairs (p < .0001). SIC contents were negatively correlated with estimated groundwater recharge, suggesting that dissolution and leaching may be responsible for SIC losses observed. Under croplands, the remaining SIC had more modern radiocarbon and a δ13C composition that was closer to crop inputs than under native vegetation, suggesting that cultivation has led to faster turnover and incorporation of recent crop carbon into the SIC pool (p < .0001). The losses occurred just 30–100 years after land‐use changes, indicating SIC stocks that were stable for millennia can rapidly adjust to increased soil water flows. Large SIC losses (194–242 Mg C/ha) also occurred below 4.9 m deep under irrigated croplands, with SIC losses lagging behind the downward‐advancing wetting front by ~30 years, suggesting that even deep SIC were affected. These observations suggest that the vertical distribution of SIC in dry ecosystems is dynamic on decadal timescales, highlighting its potential role as a carbon sink or source to be examined in the context of land use and climate change.


| INTRODUC TI ON
Carbon sequestration by storage of organic material in plants and soils is an active policy mechanism to mitigate rising atmospheric CO 2 , the key driver of climate change (Intergovernmental Panel on Climate Change [IPCC], 2014). Although the global pool of soil inorganic carbonate (SIC) is comparable in size to the carbon pool in the terrestrial biomass, and carbonate weathering is one of the dominant drivers of atmospheric CO 2 levels on geologic timescales, SIC has received substantially less attention by scientists and policymakers (Jobbágy & Jackson, 2000;Schimel et al., 1995;Schlesinger, 1982). This lack of attention may partly stem from the relatively slow formation rates of SIC (Berner, 2003;Goddéris & Brantley, 2013;Liu & Dreybrodt, 2014), even though the long residence time can be a desirable quality for carbon sequestration (Berner, Lasaga, & Garrels, 1983;Lal & Kimble, 2000;Williams, Walter, Ku, Kling, & Zak, 2003).
Pedogenic carbonates form in chemical weathering reactions involving water, CO 2 , and a cation, expressed as (Birkeland, 1999): Dissolution of CO 2 in water produces bicarbonate ions (HCO − 3 ), which can react with Ca 2+ or Mg 2+ to precipitate into solid carbonates, releasing water and CO 2 . Additions of CO 2 , water, and/or acids thus can push the equilibrium toward dissolution of carbonates. For example, depth to caliche is often positively correlated with depth of rainwater infiltration (Arkley, 1963;Jenny & Leonard, 1934). Once dissolved, bicarbonate may be susceptible to downward leaching under high soil water fluxes or to CO 2 degassing in the presence of acids. For these reasons, pedogenic carbonate formation is more stable under arid and semi-arid climates, with Ca supply limiting precipitation rates and the Ca source (e.g., calcareous or siliceous dust or parent material) determining whether carbonate formation results in carbon sequestration (Schlesinger, 2017). Some recent studies have hinted at a more dynamic SIC pool in dry ecosystems than previously acknowledged (Liu, Li, & Wang, 2011;Tamir et al., 2011). A study estimated that the carbon sink associated with the carbonate weathering process is comparable to that of the global forest biomass (Liu & Dreybrodt, 2014), and other studies have reported SIC sequestration to be in excess of 0.02 Mg C ha −1 year −1 under natural vegetation and up to 0.4 Mg C ha −1 year −1 in managed lands (Bughio et al., 2015;Landi, Mermut, & Anderson, 2003).
The current literature offers more than one hundred comparisons of SIC under natural vegetation and croplands with nearly an even split of comparisons reporting higher SIC with agricultural conversions and vice versa (60 vs. 54 cases, respectively; Appendix S1: Literature synthesis). The majority of these studies focused on the surface soil (sampling depth of <2.5 m with median depth of 0.75 m). Studies that observed decreases in SIC with agricultural conversion have hypothesized erosion and dissolution as responsible mechanisms (e.g., Kalbitz et al., 2013;Papiernik et al., 2007). Studies reporting increases in SIC have generally attributed them to carbonate precipitation triggered by the addition of Ca-rich irrigation waters or by enhanced freeze-thaw cycles (e.g., Bughio et al., 2015;Yu et al., 2014). Analyses of national soil inventories and a recent meta-analysis have similarly found littleto-mixed evidence of land-use effect on SIC while also noting shallow sampling depths of many studies and inventories (An, Wu, Zhang, & Tang, 2019;Berhongaray, Alvarez, De Paepe, Caride, & Cantet, 2013;Wu, Guo, Gao, & Peng, 2009). Lack of consensus and limited sampling depths seem to impede conclusions on the net effects of cultivation on SIC ( Figure S1).
Our goal was to compare deep SIC storage and dynamics based on carbon isotopes under paired land uses to see whether this supposedly stable pool of carbon may respond to land-use-induced changes in soil water fluxes at decadal timescales, potentially forcing soils across the threshold between net carbonate accumulation and loss (Chadwick & Chorover, 2001). In this framework, anthropogenic activities such as farming act as a sixth state factor dictating soil formation (Kuzyakov & Zamanian, 2019;Richter Jr., 2007). Water flux and acid-base reactions are primary determinants of whether weathering products (in our case, dissolved carbonate) are retained or flushed (Equation 1). We hypothesized that increased deep drainage and acidification accompanying agricultural conversion of natural vegetation dissolve carbonates (Kim & Jackson, 2012;Moody & Aitken, 1997). If deep drainage is the only important change, transport of dissolved carbonate to groundwater would prevail, whereas if acidification is the dominant change, release as CO 2 would prevail. Although a negative relationship between rainfall and SIC storage across space is well established (Kelly, Amundson, Marino, & DeNiro, 1991;St. Arnaud, 1979), to our knowledge no studies have tested the temporal effect of increased soil water storage and transport on SIC based on the conversion of natural vegetation to croplands.
To examine whether decades-old (30-100 years) cultivation of natural dry ecosystems results in SIC losses, we collected soil samples to 8.5 m depth in eight sites with adjacent or nearby natural vegetation and rain-fed cropland plots, with four of the sites having in addition irrigated cropland plots. We hypothesized that (a) SIC present under natural vegetation will experience net losses under cultivation that will be proportional to increases in storage and transport of soil water and these losses would be larger under irrigated conditions, (b) given accelerated weathering and equilibration in wetter conditions (Amundson et al., 1994;Gocke, Pustovoytov, & Kuzyakov, 2010;Van Breemen & Protz, 1988), SIC under croplands will reflect modern crop inputs compared to the more stable SIC pools under natural vegetation.
Specifically, we expected SIC under croplands to have higher percent modern carbon (pMC) and to reflect exchange with crop-respired CO 2 in its δ 13 C compared to those of SIC under natural vegetation.

| Site description and soil sampling
We selected five sites in the southern Great Plains of the United States and three sites in the Pampas grasslands of Argentina along a (1) precipitation gradient (400-900 mm/year) to examine potential interactions between land-use change and climate for altering ecosystem storage of SIC (Kim & Jackson, 2012;Kim, Jobbágy, & Jackson, 2016; see Figure S2 map of our sites). Arid and semi-arid regions, including our study areas, contain virtually all of the terrestrial SIC (Lal & Kimble, 2000) and have also experienced widespread conversion to crop cultivation (Ramankutty, Evan, Monfreda, & Foley, 2008), providing a relevant and representative setting to test the effects of land-use change on SIC storage. Rain-fed and, where available, irrigated cropland plots were paired with an adjacent or nearby natural vegetation plot (<1 km away; Table 1). All plots were greater than 2.5 ha in size, and the cropland plots had been under the current land uses for 30-100 years. The sites were located on loess or calcareous deposits (Table 1). Soil maps were consulted prior to site selection to ensure that the land-use plots were on the same soils at our sites (United States Department of Agriculture (2 mm mesh) and analyzed separately for SIC content (see below).
Samples were collected before crop plantings occurred (February-March in the USA and November-December in Argentina).
One limitation of the space-for-time substitution approach is the unknown initial conditions prior to the land-use changes. Our adjacent or nearby paired land-use plots were located on similar topographic setting and soil (Table 1). To further test whether the plots shared similar geomorphic history, we examined stable carbon isotope ratios (δ 13 C) in nodular concretions of SIC. Nodular SIC data provide a good test of our plot selection as they represent records of paleoclimate and geomorphic processes but are less vulnerable to shortterm land-use changes than diffuse SIC (Zamanian, Pustovoytov, & Kuzyakov, 2016). δ 13 C of nodular SIC were not significantly different to p > .1 between land uses (Student's t test; Figure S3), indicating that processes governing the formation of these segregated SIC nodules such as soil formation and deposition, carbon inputs, climate, and drainage on geologic timescale were likely consistent between our land-use plots. In addition, radiocarbon values show overall more modern SIC in croplands compared to natural vegetation pairs (see Section 3), a trend opposite to that expected from agricultural topsoil erosion or liming. This evidence suggests our data are likely capturing SIC dynamics associated with enhanced agricultural drainage and acidification rather than other processes associated with cultivation practices or artefacts of site selection.

| Laboratory analysis
Soil samples were analyzed for soil moisture, bulk density, SIC content, carbonate nodule mass, carbon isotopes for SIC, and anions. Intact soil core sections were oven-dried at 104°C for estimating gravimetric moisture content and at 60°C for homogenized samples for carbon measurement with a Carlo Erba Elemental Analyzer using the two-temperature combustion method (Chichester & Chaison, 1992). Soil texture was determined at 0.15, 0.9, 3.4, and 6.4 m depths for the US sites and at 0.4, 1, 2.5, 4, 6, 8 m depths for the Argentina sites by the pipette method (Klute, 1986) and ranged from loam to clay for the different sites (Table 1).

| 13 C analyses of SIC
Stable carbon isotope ratios of SIC (δ 13 C SIC ) were analyzed on an isotope ratio mass spectrometer (IRMS) coupled to a Gasbench II (Finnigan MAT DeltaPlus XL) and a CTC PAL-80 autosampler. Soil samples were transferred to 12 ml screw capped Labco exetainer vials with butyl rubber septa, which were pre-flushed with N 2 in a glove box (Assayag, Rivé, Ader, Jézéquel, & Agrinier, 2006). We determined sample weights (0.1-250 mg) from SIC contents to produce approximately 1,500 ppm of CO 2 from acidification of SIC. 1.5 ml of water and 0.5 ml of 85% H 3 PO 4 were added to acidify the samples.
Vials were shaken and left to equilibrate for 24 hr. Stable carbon isotope ratios are reported as 13 C/ 12 C ratios in per mil (‰) relative to the international reference material Vienna Pee Dee Belemnite: To verify whether agricultural conversion resulted in incorporation of CO 2 from crops into SIC, we examined whether δ 13 C SIC values shifted toward δ 13 C of carbonates expected from precipitation of crop-respired CO 2 . This comparison was made possible by shifts between C3-and C4-dominated communities in transitions from natural vegetation to croplands (Table 1). To estimate δ 13 C of carbonates formed with crop-respired CO 2 , we used the following equation: where δ 13 C SIC,crop is δ 13 C of carbonates precipitated exclusively from crop-respired CO 2 , δ 13 C POM,crop is the δ 13 C of crop residue or light-fraction (2) δ 13 C = 13 C∕ 12 C sample ∕ 13 C∕ 12 C reference × 1, 000.

| 14 C analyses of SIC
Radiocarbon concentrations of SIC were measured by accelerator mass spectrometry at the Jena 14 C facilities (Steinhof, 2014). Because Samples were prepared in a similar fashion to that of the aforementioned δ 13 C analysis, except for using larger sample quantities (7-800 mg) to yield about 0.8 mg of C as CO 2 and caps with Teflon/silicone septa and additionally underlain Black Viton septa (Sigma Aldrich; Gao et al., 2014). The sample was shaken gently and left to equilibrate at room temperature for at least 24 hr.
After preparation, CO 2 was directly extracted cryogenically from the vial headspace into a customized high vacuum extraction line using a 1:1 ethanol-dry ice mix as a water trap and liquid nitrogen for freezing out the CO 2 . All radiocarbon values are reported in pMC, which is defined as the fractionation corrected 14 C/ 12 C ratio (R) between the 14 C activity of the sample compared to the new oxalic acid standard (C OX ; NBS SRM 4990C) according to Steinhof (2013) and Steinhof, Altenburg, and Machts (2017):

| Groundwater recharge estimations
To examine the potential role of deep drainage on SIC storage, we looked for relationships between average SIC contents from individual boreholes and groundwater recharge rates previously reported in Kim and Jackson (2012) calculated from chloride mass balance and tracer displacement methods (Allison & Hughes, 1983;Phillips, 1994;Walker, Jolly, & Cook, 1991). In these methods, Cl − is taken to be a conservative tracer for soil water movement. For the mass balance approach, atmospheric and anthropogenic inputs are equated to Cl − output from the root zone into the groundwater table. Assuming steady-state conditions, the recharge rate is expressed as follows: where Q out is the water flux exiting the root zone per ground area per year or recharge rate (m 3 m −2 year −1 or m/year), Q in is the average rain and irrigation water flux entering the root zone per ground area per year (m/year), Cl in is the average atmospheric and anthropogenic Cl − input expressed as the concentration in precipitation (g/m 3 ), and Cl out is the concentration of Cl − in the soil water exiting the root zone (g/m 3 ). The approximate root zone was taken to be the top 2.4 m, below which we found a linear relationship between cumulative Cl − and cumulative soil moisture content, except for cropland plots where we assumed the root zone to be the top 1 m (Phillips, 1994). At the Tribune, Vernon, and Riesel cropland plots, where we did not observe complete leaching of the Cl − peak formed under natural vegetation, we also used the Cl − displacement method to estimate recharge rates based on the migration of the original natural vegetation Cl − and changes in water profiles (Walker et al., 1991).
Calculations for the Cl − tracer displacement method are as follows: where Q is the recharge rate (m/year), z 1 and z 2 are the depths (m) of the Cl − fronts corresponding to land uses at years t 1 (new, rain-fed cropland) and t 2 (old, natural vegetation), and Δθ is the average change in soil moisture content (g water/g soil) from natural vegetation to cropland of this depth interval (i.e., θ 1 -θ 2 ). Recharge rates were estimated for all sites except two Argentinean sites with saline soils with naturally high Cl − levels (General Levalle and Rio Bamba).

| Dissolution of SIC under the new land uses
Soluble salts that are flushed down the soil profile with the wetting front under agricultural conversion (e.g., Cl − ) may take years to reach the water table in semi-arid regions with thick unsaturated zones (Scanlon, Reedy, Stonestrom, Prudic, & Dennehy, 2005). We hypothesized that for the less soluble soil carbonates, there may be an additional lag in their dissolution and mobilization after the initial infiltration of the wetting front. We estimated the time it takes for the observable SIC loss to occur after infiltration of the wetting front under a new land use (i.e., the number of years for the increase in soil moisture at a given layer to produce statistically significant changes in SIC under new land uses) with the following moisture balance: where L is the lag in years between the onset of higher drainage and detectable loss of SIC at a given depth, D wf is the depth of the wetting front (m), defined as the maximum depth at which statistically significant differences in soil moisture between cropland and natural vegetation plots are observed based on soil moisture content from our replicate cores (p < .05; see Section 2.5 below), D SIC is the maximum depth (m) of statistically significant SIC loss (i.e., SIC loss front), Δθ is the average change in gravimetric soil moisture (g water/g soil) from natural vegetation to cropland within the two fronts (D wf and D SIC ), ρ BD is soil bulk density (g soil/cm 3 ), and Q is recharge rate (m/year) under the cropland plots estimated by Equations (5) and (6) as reported in Kim and Jackson (2012 Table 3).
We also estimated DIC flux to the ground water underneath croplands using bicarbonate (HCO − 3 ) concentrations below the root zone: where Q DIC is the downward DIC flux out of the root zone into the groundwater table (Mg C ha −1 year −1 ), Q is the recharge rate (m 3 ha −1 year −1 ), and DIC is the average concentration of bicarbonate-C in soil water extracts (Mg C/m 3 soil water) below the root zone (>2.4 m for natural vegetation and >1 m for cropland).

| Statistical analysis
To examine whether land-use changes reduced SIC storage by dissolution with increasing deep drainage (i.e., SIC in natural vegetation > rain-fed > irrigated), we used a multinomial test to compare SIC stocks between the rain-fed and irrigated croplands and native vegetation pairs and triplets. Additionally, we used regression analyses between SIC contents and recharge rates from individual boreholes at each site, with log-transformed recharge rates to satisfy homoscedasticity. To test whether irrigation alone can also result in further SIC losses in croplands, we compared the SIC stocks under irrigated versus rain-fed croplands with paired t test.
To determine at which depths SIC losses under cropland plots were significant compared to natural vegetation, we used Student's t tests on data from individual soil samples. To correct for multiple comparisons, we used Benjamini-Hochberg procedure to control the false discovery rate (Benjamini & Hochberg, 1995). Because data from different depths on same cores are not independent, we corrected for our multi-site comparison only. We used the same procedure to compare soil moisture under the different land uses to determine the wetting fronts under croplands. SIC loss fronts (D SIC ) and wetting fronts (D wf ) for Equation (7) were determined from these tests. In addition, SIC storage at three depth increments (the top 2.4 m, 2.4-4.9 m, and 4.9-7.3 m) were compared between land uses using Student's t test.
To examine how environmental factors affect soil carbonate storage, we analyzed SIC content from individual soil samples in our field data (n = 1,573) with a linear mixed model using the following environmental variables: land use (natural vegetation, rain-fed cropland, irrigated cropland), potential water excess (PWE = mean annual precipitation − potential evapotranspiration [PET]), mean annual temperature (MAT), depth, and sand content as fixed effects and site as a random effect. PET was calculated by the Penman-Monteith equation using high-resolution climate datasets (Allen, Pereira, Raes, & Smith, 1998;New, Lister, Hulme, & Makin, 2002). Because effective moisture (precipitation − evapotranspiration) is a critical determinant of pedogenesis and soil hydrology (Birkeland, 1999;Kim & Jackson, 2012), we used PWE as an aggregated climate variable to represent aridity index. The same mixed model was used for bicarbonate concentrations in soil water extracts to examine whether net acidification occurred under crop cultivation at our sites. The same mixed model analyses were also performed for percent modern carbon (pMC SIC ) and δ 13 C of SIC (δ 13 C SIC ) to examine whether land-use changes resulted in incorporation of modern and crop-respired carbon into SIC and how this was affected by the environmental variables. Because we analyzed a subset of our samples for 14 C and 13 C measurements, land-use designations for these mixed models were natural vegetation versus croplands. For pMC SIC data, we used a generalized linear mixed model as the data were lognormally distributed. We expected δ 13 C SIC to be depleted or enriched depending on whether agricultural conversion resulted in shifts to more C3-or C4-dominated communities, respectively, for each site. Most agricultural conversions resulted in shifts to C3 crops, but irrigated cultivation in Tribune and rain-fed cultivation in Parera resulted in more C4-dominated communities and hence more enriched δ 13 C carbon inputs. Because of this, we shifted 13 C data for these two conversions around the means of δ 13 C SIC under the natural vegetation for our statistical analysis. This allowed us to represent the change in δ 13 C SIC in the direction toward the shift expected based on new crop carbon inputs, rather than negative and positive for depletion and enrichment, respectively. The modified δ 13 C (δ 13 C′ SIC,Aij ) for the two croplands was calculated as follows: where δ 13 C SIC,Ni is the average δ 13 C of SIC at depth i under natural vegetation, δ 13 C SIC,Aij is the δ 13 C of SIC at depth i under cropland core j, and sng(x) is the sign function that returns −1, 0, or 1 depending on the sign of x. δ 13 C SIC,crop is δ 13 C of carbonates precipitated exclusively from crop-respired CO 2 from Equation (3).

| RE SULTS
We observed significantly less SIC storage to 7.3 m depth under cropland plots compared to their native vegetation in all 12 pairs, indicating that a shift from natural vegetation to cropland may result δ 13 C � SIC,Aij = δ 13 C SIC,Ni + δ 13 C SIC,Aij −δ 13 C SIC,Ni ⋅ sgn δ 13 C SIC,Ni −δ 13 C SIC,crop , in loss of soil carbonates (multinomial test p < .0001; and 20 Mg C/ha for rain-fed and irrigated cultivation, respectively; Table 2). SIC losses were also statistically significant down to 7.

R-f
Irr. of native soil carbonates formed or maintained under the drier natural vegetation (.0001 < p < .037; average R 2 = .51; Figure 2). This possibility is further shown by overall deeper SIC losses (5.3 vs.

San
2.5 m on average;  (Table 2). At the same sites, SIC content also peaked at or below 2 m depth regardless of land use, with half of SIC stored below 4.9 m depth on average (Figure 1). We also observed the largest changes in SIC with land-use conversions below 2.4 m, with half of SIC loss occurring below 4.3 m depth on average (Figure 1; Table 2). The deep distribution and large potential for both storage and change in SIC indicate that SIC at even deeper depths may potentially be vulnerable to the downward-advancing wetting front under the cultivated land uses.
Radiocarbon δ 13 C data for SIC also support equilibration of SIC with more modern carbon sources and point to the crops and dissolved SIC percolating from above soil layers as likely sources of the newly incorporated modern carbon. δ 13 C SIC values under cropland moved toward expected δ 13 C of carbonate precipitated from crop-respired CO 2 , accounting for fractionation during CO 2 dissolution and carbonate precipitation (mixed model p < .0001; Table S1, Equation 3; Figure 3). However, in some deeper depths of our sites Tribune and Parera, δ 13 C SIC of croplands changed in the opposite direction, reflecting instead the δ 13 C of carbonate in shallower layers under natural vegetation and suggesting some equilibration of agricultural SIC with bicarbonate (HCO − 3 ) leaching from dissolved SIC of the above soil layers (Figure 3). F I G U R E 2 (a) Average soil inorganic carbon (SIC) content (%C by weight) to 7.3 m depth by borehole versus estimated groundwater recharge rates from the corresponding boreholes. SIC decreased significantly with higher recharge rates at all sites except at Tribune. Averaging SIC content to 4.9 m depth significantly improves the regression for Goodwell and Tribune, two sites with the most recent conversion to rain-fed cultivation (R 2 of .79 and .41, respectively), indicating that SIC at deeper depths may not have equilibrated with the new recharge rates at these sites. Error bars represent standard error; bars for SIC (y-axis) are based on propagation of analytical errors for the weighted averaging (Skoog, Holler, & Crouch, 2017), and bars for recharge rates (x-axis) are based on measurements at different depths within a borehole. (b) Change in soil water storage versus change in SIC stock with agricultural conversion of natural vegetation (see also Figure S4). Soil water storage was calculated using gravimetric water content and soil bulk density. All comparisons are to 7.3 m depth except San Angelo (4.9 m), General Levalle (2.4 m), and Rio Bamba (2 m) sites. Error bars represent standard error from replicate cores In our mixed model, bicarbonate-C in soil water extracts was lower on average by 3.9 and 5.7 mg C/kg soil in rain-fed and irrigated cropland plots, respectively, compared to their paired natural vegetation plots, indicating that agricultural acidification may also have contributed to SIC loss by reducing the buffer capacity of the soils and degassing of the carbonates ( Note: Depth to SIC loss front: the maximum depth (m) of statistically significant SIC loss based on moisture balance (Figure 1; Equation 7), Depth to wetting front: the maximum depth at which statistically significant differences in soil moisture between cropland and natural vegetation plots are observed based on soil moisture content from our replicate cores (for more details, see Section 2.4 and Equation 7 in the methods).
a Years since the infiltrating wetting front under cultivated plots drained past the SIC loss front. Average difference in the SIC loss front and wetting front are 2 m for rain-fed and 4 m for irrigated. Average of upper bounds for the lag is ~30 years for carbonate loss to occur after the initial wetting. b The ">" sign indicates that the value is likely underestimated as our sampling depth could not capture the wetting front under some plots.  Table S1). Most depths also move toward the dotted vertical lines with agricultural conversion, suggesting incorporation of crop-respired carbon into SIC (p < .0001, Table S1). However, some notable opposite trends at Tribune and Parera sites also suggest some of the new precipitates under cultivation may be from bicarbonate of dissolved SIC leaching from above layers p < .0007 and .0001 for rain-fed and irrigated plots, respectively).
Significantly lower bicarbonate concentrations were observed for the entire sampled profile at some cropland plots, suggesting agricultural soil acidification at these sites may have permeated through >8 m depth ( Figure S5; Goodwell and Quanah sites).

| D ISCUSS I ON
Our results point to a potentially dynamic soil carbonate pool that is sensitive to contemporary changes in vegetation and hydrology.  Table 3). If the dissolved carbonates (DIC) do not reprecipitate during transport along the soil-, ground-, and surface-water flow path, our results would add to the growing body of evidence that SIC may interact with atmospheric CO 2 in decadal timescales, as discussed in more detail below.
Our data suggest a strong coupling between the SIC cycle and hydrology as mediated by the vegetation. Greater deep drainage with agricultural conversion (often by three orders of magnitude; Figure 2) may represent a sudden pedogenic forcing that elicits a rapid response (Chadwick & Chorover, 2001;Chen, Zhang, & Effland, 2011). Decreasing SIC with increasing soil moisture and deep drainage, both between land uses and across our sites along the humidity gradient, suggests greater leaching of weathered products under wetter conditions (Figure 2; Figure S6). Greater loss of SIC with irrigation relative to rain-fed cultivation further highlights the influence of water, rather than differences in land-use management such as tillage or fertilization, on SIC losses. In addition to being a medium for dissolution and transport, greater water content may allow SIC to quickly re-equilibrate with pore space CO 2 (Amundson et al., 1994;Gocke et al., 2010;Van Breemen & Protz, 1988), as indicated by more modern C signatures with increasing humidity in our mixed model analysis (i.e., increasing PWE; Table S1). Carbonate chemistry in variably saturated media, such as the vadose zone we sampled, is sensitive to the hydrologically accessible reactive surface area (Xu, Sonnenthal, Spycher, & Pruess, 2006). Relative to the previously dry and inert soil matrix under natural vegetation, increased reactive area of dampened finer laminar coating of disseminated carbonates under croplands may be responsible for the greater SIC dissolution (Stumm & Wollast, 1990;Zamanian et al., 2016).
Deep distributions of both SIC storage and losses from agricultural acceleration of weathering (50% below 4.9 and 4.3 m on average, respectively) highlight the vulnerability of the large SIC stocks held in the deeper soil to land surface changes, in our case the downward-advancing wetting front generated under cultivated land uses (Marion, Schlesinger, & Fonteyn, 1985). We expect SIC losses to continue as these wetting fronts infiltrate deeper horizons where larger SIC stocks are held (Table 3). The SIC losses were also more than an order of magnitude larger than the changes at the same sites in the organic carbon stocks (462 vs. 29 Mg C/ha on average), a pool often considered for carbon sequestration in climate change mitigation strategies, and larger even than biomass in some of the most carbon-dense ecosystems in the world (~400 Mg C/ha,  (Zamanian, Zarebanadkouki, & Kuzyakov, 2018). The soil acidification under cropland plots throughout our sampling depth indicates some degassing may have occurred at our sites (Table 2; Figure S5).
We also note that greater fluxes of water and strong acids such as nitric acid from agricultural activities may increase CO 2 consumption via silicate weathering, but this may be offset by increased CO 2 degassing (Guo, Wang, Vogt, Zhang, & Liu, 2015;Pacheco, Landim, & Szocs, 2013;West, Galy, & Bickle, 2005).
The fate of the carbon (i.e., DIC) mobilized by the leaching pathway and its impact on the carbon cycle hinge on the hydrologic conditions and connectivity of the soil-, ground-, and surface waters. Barnes and Raymond (2009) found ~4 times greater export of DIC from agricultural watersheds compared to forested ones. 14 C analyses revealed that DIC in ground-and river-water in agricultural watersheds was also older than that of natural vegetation, supporting its SIC origins (Grimm, Spero, Harding, & Guilderson, 2017;Li, Wang, Houghton, & Tang, 2015). As carbonate dissolution can consume CO 2 (Equation 1), the leached DIC entering groundwaters with long residence times may be sequestered long-term (i.e., carbon sink). Bicarbonate leaching to ground water could account for up to 0.067 Mg C ha −1 year −1 using bicarbonate concentrations below the root zone in soil water extracts and recharge rates from our field data (Table 2), which is much lower than the range of 0.14-3.6 Mg C ha −1 year −1 from other studies (Bughio et al., 2015;Kalbitz et al., 2013;Monger et al., 2015;Yu et al., 2014). In contrast, rapid connections to surface waters may create more uncertainty in the fate of the mobilized carbon, as DIC in surface waters may be exposed to conditions more conducive to degassing during transport to the ocean, where the composition of the mixing seawater will ultimately affect further carbonate reactions or uptake and sequestration by the biota (e.g., Berner, Westrich, Graber, Smith, & Martens, 1978;Liu & Dreybrodt, 2014;Liu et al., 2010;Mucci, 1986).
The estimates of the two pathways discussed above are modest relative to SIC losses we observed, implying either reprecipitation of dissolved SIC past our sampling depth (no net carbon flux), storage of bicarbonate in slow-moving aquifers, or greater interactive effects of agricultural practices on SIC (Ahmad et al., 2015;Dong, Duan, Wang, & Hu, 2016). δ 13 C SIC data support some reprecipitation of SIC with crop-respired CO 2 and/or DIC leaching from the above layers; however, given the loss of alkalinity under croplands despite regular liming (Table 2; Figure S5), the extent of this pathway may also be limited (Figure 3).
Though we could not assess the fate of the accelerated weathering of carbonates, dissolution of solid-phase SIC to more reactive mobile ions would likely increase exposure to different environmental conditions during transport and probability of further transformations. In the context of enhanced weathering as a means to achieve negative emissions (Fuss et al., 2018;Lawrence et al., 2018), our evidence for a sensitive SIC pool reacting to hydrological changes, along with the quantity, rate, and extent of the SIC losses, highlights the need for further evaluation of net sink or source potential of the SIC pool (i.e., fate of DIC) and associated time lags under changing land uses.