Higher Long-Term Soil Moisture Increases Organic Carbon Accrual Through Microbial Conversion of Organic Inputs

High long-term soil moisture may either stimulate or inhibit soil organic carbon (SOC) losses through changes to mineral and chemical composition, and resultant organo-mineral interactions. Yet, the trade-off between mineralization and accrual of SOC under long-term variation in unsaturated soil moisture remains an uncertainty. In this study, we tested the underexplored relationships between long-term soil moisture and organo-mineral chemical composition, and its implications for SOC persistence. The results provide new insights into SOC accrual mechanisms under different long-term moisture levels commonly observed in well-drained soils. Differences in long-term mean volumetric water content ranging from 0.4 - 0.63 (v/v) on fallow plots in an experimental field in New York, USA, were positively correlated with SOC contents (R2 = 0.228; P = 0.019, n = 20), mineral-associated organic matter (MAOM) (R2 = 0.442; P = 0.001; n = 20) and occluded particulate organic matter (oPOM) contents (R2 = 0.178; P = 0.033; n = 20). Higher long-term soil moisture decreased the relative content of sodium pyrophosphate extractable Fe (R2 = 0.33; P < 0.005; n = 20), increased that of sodium dithionite extractable Fe (R2 = 0.443; P < 0.001; n = 20), and increased the overall importance of non-crystalline Al pools (extracted with sodium pyrophosphate and hydroxylamine extractable) for SOC retention. Higher long-term soil moisture supported up to a four-fold increase in microbial biomass (per unit SOC), and lower C:N ratios in MAOM fractions of high-moisture soils (from C:N 9.5 to 9, R2 = 0.267, P = 0.011, n =20). This was reflected by a 15% and 10% greater proportion of oxidized carboxylic-C to aromatic-C and O-alkyl C, respectively, as measured with 13C-NMR, and a more pronounced FTIR signature of N-containing proteinaceous compounds in high-moisture MAOM fractions, reflective of microbial metabolites. SOC accrual increased with increasing soil moisture (P = 0.019), exchangeable Ca2+ (P = 0.013), and pyrophosphate-extractable Al content (P = 0.0001) and Al/Fe ratio (P = 0.017). Taken together, our results show that high long-term soil moisture resulted in SOC accrual by enhancing microbial conversion of plant inputs to metabolites that interact with reactive minerals.


Introduction 39
Soils comprise the largest terrestrial store of organic carbon (OC) (Friedlingstein et al., 2019) and 40 play a substantial role in the global C cycle (Scharlemann et al., 2014). As a result, increasing soil organic 41 carbon (SOC) storage by decreasing losses to CO2 may be a feasible strategy to withdraw atmospheric CO2 42 and partially offset anthropogenic emissions driving climate change (Minasny et al., 2017). As climate 43 change is expected to drastically alter soil moisture conditions globally (Seneviratne et al., 2010;Grillakis, 44 2019), a better understanding of the trade-off between SOC mineralization and stabilization under long-45 term changes to soil moisture is needed to manage SOC stocks (Falloon et al., 2011). Importantly, unlike 46 soil temperature which is expected to rise and increase SOC mineralization (Soong et al., 2021), soil 47 moisture is a parameter that can be managed -through irrigation, controlled drainage, and wetland 48 restoration -and can indirectly curb soil warming through evapotranspiration (Seneviratne et al., 2010). 49 Soil moisture controls SOC turnover and storage by regulating fundamental processes such as soil 50 biotic activity (Moyano et al., 2013), solute transport, gaseous exchange, and mineral weathering 51 persistence of SOC by determining mineralizability in laboratory incubations. We hypothesized that (1) 89 soils experiencing higher long-term moisture that is more conducive to microbial activity will have more 90 oxidized SOC functional groups and higher contents of non-crystalline oxide phases due to higher 91 decomposition and weathering, respectively, and (2) that organo-mineral interactions in high-moisture soils 92 will support SOC accrual.  (Table S1). 114 The driest subplots (denoted as Q1) had the lowest water contents relative to the field mean, and the wettest 115 subplots (denoted as Q5) had the highest water contents relative to the field mean. 116 The plant cover at the time of sampling consisted of fallow (unmowed for circa 10 years) grasses 117 dominated by legacy reed canarygrass (Phalaris arundinaceae L.) interspersed with numerous other 118 grasses and broadleaf forbs. The plots were undisturbed aside from small (1 m 2 ) hand harvests at subplots 119 after dormancy to characterize yields for comparison with other cropping treatments. Approximately two kg of soil were dug from each of the two locations, composited in a bucket, mixed, and 124 transported to the lab in a cooler. Half of each sample was air dried and passed through a 2 mm sieve, and 125 half was passed through a 4 mm sieve and stored at -20 °C for microbial biomass and soil respiration 126 analyses. Visible rocks and plant material were removed from each sample. Composited samples, prepared 127 by combining equal amounts of the soils from quintiles Q1, Q3, and Q5 (Table S1), were used for 13 C 128 NMR and C NEXAFS analyses. These composited samples are referred to as Low moisture, Mid moisture, 129 and High moisture, respectively. Collection of above-and below-ground biomass was described previously 130 extracted using 1 N ammonium acetate at pH 7 and measured on an inductively coupled plasma 138 spectrometer (Thermo iCAP 6000 series). 139

Soil fractionation 140
We used a combination of size and density fractionation to isolate operationally defined -but 141 ecologically relevant -fractions. Soil samples (10 g) were gently agitated for 10 s with 35 mL of sodium 142 polytungstate (SPT) adjusted to a density of 1.65 Mg m -3 . The samples were left to settle overnight, 143 centrifuged (3000 RCF, 30 min), filtered (GF/F, 0.7 µm glass fiber filter), and washed with 500 mL of 144 deionized water. The obtained material is referred to as the free particulate organic matter (fPOM). Fresh 145 SPT solution (35 mL, 1.65 Mg m -3 ) was added to the samples and a vortex was used to re-disperse the soil. 146 The samples were sonicated (XL 2020, QSonica, Newtown, CT, USA) at 350 J mL -1 of energy (operated at 147 75 J s -1 ), left overnight to allow the particles to settle, and centrifuged (3000 RCF, 45 min). The floating 148 material, referred to as the occluded particulate organic matter fraction (oPOM), was isolated as described 149 above. The remaining pellet was washed with deionized water and centrifuged (3000 RCF, 30 min) three 150 times to remove the SPT (supernatant density was 1±0.02 Mg m -3 ). Next, the soils were shaken end-to-end 151 with sodium hexametaphosphate (35 mL, 0.5% w/v) for 16 hours and wet sieved (53 µm) to separate the 152 sand sized fraction (material remaining on the sieve) from the silt and clay size fractions (material passing 153 the sieve). The material passing the sieve was referred to as the mineral associated organic matter 154 (MAOM) fraction. The MAOM and sand fractions were transferred to a pre-tared aluminum tin and dried 155 at 60 °C. The four obtained fractions -fPOM, oPOM, MAOM, and sand -were weighed, ball milled 156 (except the sand), and stored. Total C, N, and isotope ratios of the fractions (except the sand) and bulk soil 106%. Sand was assumed to contain zero C and N. The average C concentrations in fPOM and oPOM were 162 35% and 25%, respectively, indicating that some minerals were present in these isolated fractions. 163

Oxide extraction 164
We performed a sequential oxide extraction following a modified protocol (Heckman et  only approximately correspond to specific phases present in the soil as these extractions are not perfectly 172 selective (Parfitt and Childs, 1988;Kaiser and Zech, 1996) inductively coupled plasma spectrometer (Thermo iCAP 6000 series). The concentration of OC in each 182 extract, considered the C associated with Al and Fe, was measured by combustion catalytic oxidation

Microbial biomass carbon 185
Microbial biomass C was measured following the chloroform-fumigation-extraction method (Witt et 186 al., 2000). Briefly, ethanol-free chloroform (4 mL) was added to 10 g of field moist soil (<4 mm) in 187 stoppered 250 mL Erlenmeyer flasks. The samples were incubated for 24 hours, after which the flasks were 188 vented in a fume hood until the chloroform had fully evaporated.

13 C Nuclear Magnetic Resonance (NMR) 206
The molecular structure of organic matter in the bulk soil and soil fractions was analyzed using 13 C 207 time of 0.001 sec with a pulse delay of 0.4 sec for bulk soil and silt and clay fractions and 1 sec for fPOM 209 and oPOM fractions. At least 100,000 accumulated scans were performed. The spectra were integrated 210 using four major chemical shift regions: 0 to 45 ppm (alkyl-C), 45 to 110 ppm (O/N-alkyl-C), 110 to 160 211 (aryl-C), and 160 to 220 ppm (carboxyl-C) (Knicker and Lüdemann, 1995). Sample pre-treatment with 212 hydrofluoric acid was not necessary to obtain a well-resolved spectrum. We applied a molecular mixing

231
For the incubation experiment, soil samples were thawed and air dried. 5 g from each soil sample 20 mL glass vial containing a CO2 trap (15 mL KOH, 0.18 M) made with CO2-free deionized water. CO2-234 free deionized water (5 mL) was added to the bottom of the jar to maintain a humid atmosphere. To 235 account for the small amount of CO2 present in the jar, measurements from blank jars with no soil were 236 used. Samples were hydrated to a moisture level equivalent to 50% of water filled pore space with 237 deionized water and incubated for 53 days at 20 °C in the dark. On days 2, 7, 18, 33, and 53 the jars were 238 opened, and the electrical conductivity of the KOH solutions was determined. After each measurement, the 239 CO2 traps were replaced with fresh KOH solutions in new vials, and fresh deionized water was added to 240 the bottom of the jar. At each sampling event, the average (n = 3) electrical conductivity value of the KOH 241 solution from the blank jars was subtracted from each jar's KOH electrical conductivity value. This 242 corrected value was then converted into volume of CO2 released by the sample using a standard curve, and 243 further converted to mass C-CO2 by applying the ideal gas law. The standard calibration curve was made 244 by injecting known of volumes of 99.99% CO2 (Airgas, Inc, Elmira, NY) into septa-lidded Mason jars 245 containing empty Qorpak vials, CO2 traps, and 5 mL CO2-free water on the bottom. The electrical 246 conductivity of the KOH solution was measured 24 hours after injection. Cumulative respiration was 247 reported per unit soil (termed mineralization; mg CO2-C/g soil), and per unit SOC (termed mineralizability; 248 mg CO2-C/g SOC). 249

250
All statistical analyses were done in R (Version 4.04). Linear regression of response variables to 251 normalized moisture was performed using the lm function. The effects of normalized moisture values and 252 soil fraction on response variables were analyzed using ANOVA, followed by Tukey's HSD post hoc test 253 to determine significant differences between treatments. Pearson's correlation matrix was plotted using the 254 corrplot package. Partial least squares (PLS) regression models were constructed using the mdatools 255 package, to identify the variables that best explain SOC content, mineralization, and mineralizability. PLS 256 regression models reduce dimensionality and create components that maximize the covariance between similar number of predictor variables and observations, and/or when predictors are highly correlated. 259 Selectivity Ratio and Variable Importance in Projection tests were used to inform variable selection. 260 Variable Importance for Projection provided the best goodness of fit for all models, which was evaluated 261 based on the highest cross-validated R 2 (R 2 cv), and lowest root mean of square error (RMSE). 262 3. Results increasing normalized soil moisture (R 2 = 0.33, P = 0.005) and the relative Fe concentration in crystalline 301 phases (FeDITH) increased with increasing normalized soil moisture (R 2 = 0.443, P = 0.001) ( Figure 3B). In 302 total, the ratio of Al to Fe in the PY and HH extracts increased with increasing normalized moisture (R 2 = 303 0.287, P = 0.009 for PY; R 2 = 0.363, P = 0.003 for HH), but remained similar in the DITH extract (Figure 304 S6). In addition, the concentration of extracted DOC increased as the ratio of Al/Fe in the PY and HH 305 extracts increased, (R 2 = 0.26, P = 0.013 for PY; R 2 = 0.18, P = 0.035 for HH) ( Figure 3C), indicating that 306 the increase in Al/Fe ratio in these extracts with increasing moisture was consistent with increasing DOC. 307 While the distribution of C across oxide pools varied with normalized moisture, total oxide-associated C 308 normalized to SOC content (which constituted on average 44% of the SOC) did not differ across moisture 309 values (P = 0.79, Figure S7). Linear regression analyses show that extractable Al was better correlated with 310 respective extract (R 2 = 0.513; P = 0.0002, and R 2 = 0.411; P = 0.0014 respectively) ( Figure 3E)  1250-1800 cm -1 consisted of bands tentatively assigned to stretching of C=O at 1645 cm -1 , COOand C=C 332 at 1600 cm -1 , C=N and bending N-H vibrations at 1545 cm -1 , and COOat 1420 cm -1 . We found higher 333 absorbance in bands assigned to asymmetric and symmetric carboxylate COOstretch and amide/ketone 334 C=O stretch (1420, 1600, and 1645 cm -1 , respectively), symmetric O-H stretching (3340 cm -1 ) and 335 asymmetric and symmetric CH2 stretching (2930 and 2855 cm -1 ) with increasing moisture level. 336 Furthermore, there was a relative increase in the absorbance at 1545 cm -1 , assigned to stretching of 337 aromatic C=N and bending of N-H. 338

339
In the MAOM fraction, the ratio of carboxyl-C/aromatic-C increased with increasing moisture level 340 (R 2 = 0.731; P = 0.239, Table S3), as measured by 13 C NMR (Figure 4 and Figure S9). In addition, the ratio 341 of carboxyl-C/O-alkyl-C and the ratio of O-alkyl-C/alkyl-C in the MAOM fraction increased with 342 increasing moisture level (Table S3). Comparing fractions, carboxyl-C and alkyl-C forms increased, and

357
Confirming the 13 C-NMR results, we found that the ratio carboxyl C / (aromatic + substituted 358 aromatic C), measured with C NEXAFS (Figure 4D), increased with increasing long term moisture level. 359 We did not evaluate changes in alkyl-C and O-alkyl-C forms since NEXAFS spectra do not have a strong 360 and well-defined feature corresponding to these bonding environments (Heckman et al., 2017). Overall, 361 carboxyl-C forms relatively increased, while aromatic-C forms relatively decreased in the MAOM fraction 362 relative to the fPOM and oPOM fractions (R 2 = 0.866; P = 0.166, Table S3). The correlations between all variables are reported in Figure S12.  increasing moisture, indicating that they consisted of more processed and oxidized microbial products. 432 Although microbial biomass was measured only once, we assume that this finding reflects a more general Further insights into the chemical composition of MAOM fractions across the moisture gradient was 435 provided by spectroscopic techniques. MAOM fractions from high-moisture soils were enriched in 436 proteinaceous compounds, oxidized carboxylic-C and O-alkyl-C functional groups, and depleted in alkyl-C 437 and aromatic-C functional groups (Figure 4, Table S3, Figure S9, Figure S10). These results indicated that 438 MAOM of high-moisture soils contained more products of microbial decomposition of aromatic and 439 aliphatic plant residues, with concomitant enrichment in microbial proteinaceous components ( Figure S10). 440 The possible role of mineral composition in shaping SOC composition is discussed below. Taken together, 441 our results clearly show that high long-term moisture led to greater oxidation of SOC that explains its 442 accumulation. 443 Although SOC content has previously been shown to correlate with the amounts of non-crystalline Fe 455 oxides (Mikutta et al., 2006), surprisingly, DOC extracted from different oxide pools did not significantly 456 correlate with Fe content in each pool ( Figure 3D). This suggests that OC contents associated with oxide 457 minerals did not consistently follow the shift in Fe phases along the moisture gradient. However, our 458 results clearly show the increasing importance of AlPY and AlHH phases for SOC retention with increasing long-term moisture ( Figure 3C, 3D, and 3E). Several recent studies have similarly observed a greater 460 contribution of non-crystalline Al pools for SOC retention, albeit with greater saturation extent and 461 frequency (Possinger et al., 2020)  America showed that oxalate extractable non-crystalline Al was a better predictor of SOC storage than 467 oxalate extractable Fe , suggesting that this effect was more common than previously 468 thought. Taken together, our results indicate that SOC stabilization by AlPY and AlHH was directly 469 influenced by long-term moisture levels that are not commonly associated with a loss of FePY. 470

Long-term moisture shaped mineral composition and C interactions with minerals
There is also a possibility (which we did not test) that the chemical composition and mode of 471 interaction of SOC associated with Al differs from that of Fe, and that it may vary across soil moisture. 472 Considering the mounting evidence of the critical role of Al in SOC stabilization at a wide range of 473 moisture conditions, we suggest that future research should seek to understand the composition of SOC 474 associated with non-crystalline Al, and the processes that influence the persistence of this important SOC 475 pool. 476 Higher long-term moisture was also positively correlated with CaEX and pH (Figure S12 expected to prevail in the higher pH values found in the high-moisture soils which, since high pH promotes correlation between MAOM and SOC contents indicate that specific effects of CaEX, rather than greater 510 contents of silt-and clay-sized particles, played a role (Table 1, Figure 6, and Figure S12). As highlighted 511 recently, mineralogical composition is a better predictor than particle size distribution for SOC content 512 (Khomo et  In the trade-off between mineralization and accrual, high long-term moisture increased SOC accrual, 524 even at a range of soil moisture not expected to constrain decomposition or alter soil oxide mineral 525 composition, indicating that such changes are likely to occur even along moderate soil moisture gradients. 526 The direction and magnitude of changes to SOC stocks under changing long-term soil moisture are likely 527 to be dependent on additional soil properties that affect microbial conversion of organic inputs and mineral 528 interaction of these products. For example, soils which do not have the mineral capacity to interact with 529 oxidized organic compounds may not benefit from increases to moisture. Climate change is not only 530 changing soil moisture, but critically, also altering the magnitude and temporal patterns of moisture 531 variability through extreme floods and drought. The relative roles of mean long-term moisture vs. moisture 532 temporal variability on stabilization mechanisms are still not clear. Currently, conventional C models 533 consider soil moisture as a physical variable only, and models that account for microbial traits do not include the potential effect of moisture on carbon use efficiency. These processes will have to be clarified 535 to maintain SOC stocks by driving forward stabilization of organic inputs. 536 6. Acknowledgements