Elevated moisture stimulates carbon loss from mineral soils by releasing protected organic matter

Moisture response functions for soil microbial carbon (C) mineralization remain a critical uncertainty for predicting ecosystem-climate feedbacks. Theory and models posit that C mineralization declines under elevated moisture and associated anaerobic conditions, leading to soil C accumulation. Yet, iron (Fe) reduction potentially releases protected C, providing an under-appreciated mechanism for C destabilization under elevated moisture. Here we incubate Mollisols from ecosystems under C3/C4 plant rotations at moisture levels at and above field capacity over 5 months. Increased moisture and anaerobiosis initially suppress soil C mineralization, consistent with theory. However, after 25 days, elevated moisture stimulates cumulative gaseous C-loss as CO2 and CH4 to >150% of the control. Stable C isotopes show that mineralization of older C3-derived C released following Fe reduction dominates C losses. Counter to theory, elevated moisture may significantly accelerate C losses from mineral soils over weeks to months—a critical mechanistic deficiency of current Earth system models.

Hi there I enjoyed reading your manuscript. It is well written, framed by well-developed hypotheses and backed by a solid experimental design.
Indeed your data does add new knowledge to a growing pool of evidence that soil moisture dynamics may moderate the soil C loss through anaerobic processes co-occuring in otherwise upland or drained minerogenic soils. Especially, addition of CH4 dynamics as an integral component of C loss under rewetting is in my opinion a major strength of your manuscript and linking it to the biogeochemistry of Fe dynamics points a way forward for field testing and development of models.
I have added my review as an annotated file.
I have two minor issues though I believe you should address in more detail in the introduction and discussion 1. Period of redox change/flooding You should discuss in more detail what the effect on C mineral zation is of shorter periods of redox changes than the 82 days you are using. What is the argument for 82 days? In my opinion the moisture manipulation period is a major factor and an important dynamics to consider when scaling these ideas to a potential field study. In a natural setting, yes 82 days, may be realistic for sustained high moisture content in depression and footslope soils as you simulate here, but is this the case for artificially drained soils or even naturally in the ridge soils? Thus, how representative is 82 days compared to field observations of soil moisture in this landscape -considering also artificially drained soils? Looking at Fig 1 the effect of soil moisture become less and less important a short intervals which is not surprising. So to maximize impact of your study I do believe a more thorough discussion of the flooding period is needed.

Soil types and hydrological regime
Convincingly, your results does indicate that added C loss is mainly driven by hydrology and less by soil type. However, this may be because the soils are relatively similar. They are rich in clay, which promotes protection and complexing of organic C with Fe-oxides and allows to sustain high levels of WFPS. How would your results turn out if the same study was carried out in soils with a coarser texture, e.g. more sand and less clay? This would also alter the hydrological regime with shorter periods of flooding under field conditions (see comment above). Translating your findings in to an operational routine for a model would need to consider the role of soil texture as well as this may constrain rates. In my opinion this is also important to discuss if you want your results to connect to a broader audience.
That being said, your study in isolation makes up for a very good study as is, but you can increase the value by including the abovementioned points.
Reviewer #2 (Remarks to the Author): The manuscript by Huang & Hall challenges our current understanding that elevated soil moisture and associated anaerobic conditions suppress soil organic C mineralization. Results indicate that under reducing conditions there is even a greater C mineralization than under aerobic ones. As the lost C is old and very likely result from a dissolution of stable mineral associated organic matter, the identified processes may have a great impact on soil's C balance. The relation of soil organic matter and moisture in moister soils is an emerging and so far broadly overlooked topic and the manuscript makes a novel and interesting contribution to it. It may stimulate further research on the duration and susceptibility of SOM stabilization by Fe-oxides.
While soil C models currently apply the concept that C mineralization decreases with water saturation, the relative high C release observed by huang & Hall could also be deduced from various other studies. For instance, there are a number of studies showing an enhanced DOC release under reducing conditions. Also, Hanke et al (2013, European Journal of Soil Science) found that paddy soils lost more C and substantially older C by soil respiratory processes when they became anaerobic. However, although the findings and the ideas are not entirely novel, Huang & Hall present the first thorough assessment including CO2 and CH4 production, stable isotopes, redox potentials, released Fe (II), which will receive a high attention by modeler and experimentalists. Overall, the paper is also nicely written and well set in a theoretical framework.
My main concerns are: 1.
The sampling and the set-up of the incubation experiment are somewhat unclear (or sloppy; see more detailed comments below). For instance, it is unclear how large the field was, where the samples had been taken from, if all cores sampled in the field had been mixed to one composite sample and then split for the incubation experiment. This would definitely affect the statistical analysis as there might be no real (field) replicates, but it doesn't affect the story and the mechanisms identified but of course, the quality of the experiment.

1.
The sampling and the processing of the samples in the lab is rather unclear. How large was the field? How many soil cores were taken to have 'spatially representative samples'? Have all cores sampled in the field mixed to one composite sample and then split for the incubation experiment or have the soil from the cores been used directly for the incubation? To me it seems that the set-up seems to be pseudo-replicated as the samples were taken from one field per site (soil type) only and lab replicates have been incubated and analyzed. This does not change the mechanism identified, but at least site (here called soil) effects cannot be analyzed and interpreted in a statistical sense… 2.
Line 105 'CO2 production from the footslope soil was equivalent among the three moisture treatments'. This is not clear from Figure1a.
Line 110 Explain why the intermediate soil samples were not allowed to slowly drain as the saturated ones. This clarification needs to be included in the paper.
Line 110: It is not clear whether the saturated soils were drained to field capacity? Were the saturated/drained soils aerobic after day 82? No information on changes in oxygen concentration or Eh is provided.
Line 118 Which is the reason of the secondary peak in CO2 measured at day 30 for both the intermediate treatment and the control? May this increase be associated with the manual addition of water?
Line 125. 'Methane emissions....were negligible (< 0.2 % of total C mineralization) in the control' This result is expected due to the aerobic conditions of the control.
Line 155 'However, after 82 days, the cumulative C4-derived C mineralization was significantly higher in the saturated treatment (p < 0.05) and the intermediate treatment (p < 0.01) than in the control' The control of the footslope soil in Figure 3a seems to show values of mineralized C4-C that are comparable to ones the saturated treatment.
Line 178 The drop observed in Eh for both the intermediate and saturated treatments is in line with the previous comment on Figure1a.
Line 175: Why did the authors limit the Figure 1 of the Supplementary information to 75 days and did not report the changes in Eh following the drainage of the saturated soils?
Line 185 Why did the authors conduct the companion experiment only on the footslope soils?
Line 176 Why not measure directly oxygen? Line 176 Which were the oxygen levels in the saturated soils after day 82 (when the soils were allowed to drain)? There are no changes after day 82 in Eh values for the saturated soil, please explain why the redox values did not increase during the drainage phase.
Line 182: Is the lower concentration of Fe(II) for the saturated treatment a consequence of the drainage?
Line 191 Is the DOC in Figure 4b due to C release from C-Fe associations? Discussion: Line 201-203 This work is not challenging the relationship between soil moisture and C mineralization found by previous authors (e.g., reference [5]), rather it extends their finding to include long-term incubations for which the redox conditions promote the mineralization of protected C.
Line 250. For the saturated soils, there would be also an advective flux of oxygen due to the drainage of the samples. Please add this information for the sake of physical rigor.
Line 252 Is Figure3 solely accounting for the CO2 produced by heterotrophic respiration or also for the oxidation of methane? Did the authors quantify the methane oxidation rates for the saturated/drained samples after day 82? Line 273. Please quantify the lag at which the C release due to Fe-mediated processes starts. Is the lag associated with a specific range of redox or oxygen conditions? Methods: Line 301. Were the ground corn leaves added to analyze the consumption of readily C source and compare it to the mineralization of the occluded C source? Please clarify this point  In Figure 4 the lag time seems of 12 days. Is this lag consistent with the CO2 production observed    As stated in the new text, we chose 82 days because it is representative of seasonal 31 periods of saturation (spring -early summer) that commonly occur in the most poorly 32 drained soils in our region, and which occurred to an even greater extent prior to 33 European settlement and drainage infrastructure installation (van der Valk 2005).

35
The extended period of saturation and drainage provides a useful end member for 36 challenging the traditional moisture response relationship of soil C mineralization, 37 which was the focus of this particular study. We agree that shorter-term fluctuations 38 are also of interest, and these are the focus of ongoing work in our research group. We  Furthermore, we wanted to use a period of flooding that was long enough to achieve 47 pseudo-steady state Eh and trace gas emissions to use as an end-member for 48 comparison with the other treatments. We note, however, that our high-frequency gas 49 flux measurements allowed us to explore the temporal dynamics of microbial 50 responses to saturation over the entire period of 82 days. (2) Soil types and hydrological regime 59  where Fe-associated organic matter may respond dynamically to redox cycling. We water. This will make it easier to follow your argumentation below.

116
Response: Agreed. We have clearly stated that this is C substrate diffusion through 117 soil water in the sentence. The phase "oxygen availability and substrate diffusion" has 118 been revised to "oxygen (O 2 ) supply from the atmosphere and C substrate diffusion 119 through soil water" (Lines 35-36).    Response: We note that these values were from rapidly biodegradable DOC, not total 243 DOC. The enrichment could be due to transient dynamics of C substrate accumulation 244 and respiratory kinetic fractionation following flooding. We are presently exploring 245 this phenomenon in greater detail, but as these data may distract from the primary 246 message of the paper, we have elected not to present them in the revised manuscript.

248
We therefore used another more conclusive method in this revised manuscript to 249 measure the C isotope ratio of total DOC (Lines 489-502 of Methods). The δ 13 C 250 values of total DOC ranged from -20‰ to -13‰ for the saturated treatment and from 251 -23 ‰ to -16‰ for the control, as would be expected from a mixture of C 3 -and   Response: Agreed. We have revised "C" to "organic C". (Line 358) Response: Agreed. We have added "and including CH 4 as well as CO 2 production" in Response: Agreed. We have clarified that the mineralized C included CH 4 and CO 2 in 340 the y-axis title. The "Mineralized C4-C" and "Mineralized C3-C" have been revised 341 to "C 4 -derived CO 2 + CH 4 " and "C 3 -derived as CO 2 + CH 4 ", respectively. Please see   suggestions. In the first manuscript, we cited many of the earlier studies that showed   (2) The δ13C value in DOC seems unrealistic and greater than any value reported so 405 far (see specific comments). Also the method how DOC was obtained is unclear. The δ 13 C values of total DOC ranged from -20‰ to -13‰ for the saturated treatment  to CO 2 . The CO 2 produced from blanks containing the oxidizing agent + deionized 435 water was also analyzed to correct the sample δ 13 C values."

437
We have now specified the method for how DOC was obtained: "For DOC rather use the term 'turn over' because at the same time, there will be a C input to 471 mineral associated C and thus a 'replacement'.

473
Response: Agreed. We have changed "decompose" to "turn over" in the sentence.   "Soils at our study site supported mixed C 4 -C 3 prairie and wetland vegetation over 534 the last 10,000 years 33 , and have been cultivated under C 4 -C 3 crop rotations for at 535 least the past 50 years. For this study, soils were collected following corn harvest and 536 amended with corn residues, such that the most recent C inputs had a C 4 isotope 537 signature and C 3 -derived C was older by at least one year." "Isotope mixing models suggested that the DOC accumulated under saturated 545 conditions in the footslope soil was primarily of C 4 origin, but that C 3 -derived DOC 546 also increased relative to the field capacity control (Fig. 5). These data help reconcile  and C 4 soil organic C have different ages and turnover times. I.e., we would expect C 4 637 C losses to be greater in our study system because of the fresh C 4 C that was added at 638 the beginning of our study.   Supplementary Fig. 11. The new text has been added (Lines 251-253): "The 669 mean percentage of readily bioavailable DOC was similar between the control (33 ± 670 5%) and the saturated treatments (31 ± 5%) ( Supplementary Fig. 11).    We have now clarified that the additional month-long experiment with the footslope 788 soil was conducted to explore the relationships between Fe reduction, Eh, pH, and 789 DOC at higher resolution with high replication (total n = 96), which required   Fig. 7)."

797
Finally, this work is not strictly challenging the relationship between soil moisture   Response: Agreed; we have the rephrased this sentence to "Counter to theory, 894 elevated moisture may significantly accelerate C losses from mineral soils over weeks 895 to months-a critical mechanistic deficiency of current Earth system models".

897
Introduction: 898 (7) Line 27: Please add a reference to support this statement, such as reference [1] of 899 the bibliography provided below (which is already included in the reference list).

901
Response: Agreed. We have added the reference (Skopp et al. 1990) to support the 902 statement, as well as Linn and Doran 1984.   Response: Agreed. The decadal turnover times are averages that include some 927 Fe-associated C and some C associated with other mineral phases (see also response 928 to reviewer 2 on Lines 546-574 above). In our conceptual model (Supplemental 929 Figure 1), Fe-complexed C may persist for some time but may rapidly be released and  Response: As also suggested by the second reviewer, we have used thinner lines and 965 smaller symbols in the figure to make this point clearer. We have clarified that the 966 CO 2 production "became statistically equivalent" during this period (Line 128).

968
Response: Agreed. We have quantified the lag time at which the C release due to 1145 Fe-mediated processes starts. We observe an increase an Fe(II) after as little as two 1146 days, accompanied by increased DOC, as indicated in Fig. 5. Please see new text on 1147 361-363: "When Fe reduction occurs (e.g., after as little as two days of elevated 1148 moisture), it can accelerate C loss in mineral soils by facilitating microbial access to 1149 previously protected labile C."

1151
We caution to ascribe a specific Eh range to Fe reduction as others have done in the 1152 past, given that Eh is a mixed potential and therefore Eh values for Fe reduction may 1153 vary among systems.