Soil organic carbon decomposition rates in river systems: effect of experimental conditions

Rivers receive large amounts of terrestrial soil organic carbon (SOC) due to the action of different erosion processes. Mounting evidence indicates that a significant fraction of this SOC, which is often very old, is rapidly decomposed after entering the river system. The mechanisms explaining this rapid decomposition of previously stable SOC still remain unclear. In this study, we investigated the relative importance of two mechanisms that possibly control SOC decomposition rates in aquatic systems: (i) in the river water SOC is exposed to the aquatic microbial community which is able to metabolise SOC much more quickly than 10 the soil microbial community and (ii) SOC decomposition in rivers is facilitated due to the hydrodynamic disturbance of suspended sediment particles. We performed different series of short-term (168 h) incubations quantifying the rates of SOC decomposition in an aquatic system under controlled conditions. Organic carbon decomposition was measured continuously through monitoring dissolved O2 (DO) concentration using a fiber-optic meter (FirestingO2, PyroScience). Under both shaking and standing conditions, we found a significant difference between SOC with aquatic microbial organisms (SOC+AMO) and without aquatic microbial 15 organisms (SOC-AMO). The presence of an aquatic microbial community enhanced the SOC decomposition process by 70 %– 128 % depending on the soil type and shaking/standing conditions. While some recent studies suggested that aquatic respiration rates may have been substantially underestimated by performing measurement under stationary conditions, our results indicate that this effect is relatively minor, at least under the temperature conditions, the soil type and for the suspended matter concentration range used in our experiments. We propose a simple conceptual model explaining these contrasting results. 20

ecosystems (Kling, 1995). A key question is then which factors control the decomposition of terrestrial SOC, which has been stable 35 in soil for decades to centuries, when it enters into a river system.
In recent years, the mechanisms controlling this rapid riverine mineralization process have gained increasing attention (Aufdenkampe et al. 2011;Guenet et al. 2014;Ward et al. 2017). While in transit, SOC can be degraded by microbial degradation (Ward et al., 2013) and photochemical oxidation (Spencer et al., 2009). As can be expected, these processes are closely associated with a suite of factors such as temperature (Lapierre et al., 2013;Gudasz et al., 2015), the availability of oxygen (Koehler et al., 40 2012), and the presence and composition of microbial communities (Ward et al., 2019), along with physical river properties, such as river velocity and hydrodynamic conditions (Ward et al., 2018). For example, Ward et al. (2018) incubated river water and sediment under three rotation regimes to mimic river flow and found respiration rates were 1.4 (under 0.22 m·s −1 ) and 2.4 times (under 0.66 m· s −1 ) higher compared with stationary conditions. The physical breakup of large particles induced by river water disturbance may increase the accessibility of microbial enzymes to SOC, thus enhance SOC mineralization (Lal 2003; Richardson of 1 mole C consumed 1 mol O2. Using a respiration ratio of 1.0 might result in an overestimate of the total amount of C mineralized, but since all treatments were calculated with the same respiration ratio and the same soil samples were used throughout the experiments, the relative variations will not be affected by this choice. After the incubation period, the entire volume of ~320 ml water was filtered for later determination of total POC, particulate nitrogen (PN) and the stable C isotope ratios (δ 13 C) of POC on pre-combusted 25 mm Whatman GF/F filters (pore size: 0.7 μm).

115
Due to the high concentrations of soil materials in each bottle, each water sample was filtered on 2-5 filter papers. For arable land soil incubations, water samples were filtered on 5 filter papers, we measured all filter papers for 2 repetitions, and 2 of 5 filters papers for the other 4 repetitions. Then we used the suspended sediment weight and POC concentrations measured on this subsample to calculate the POC and PN content as well as the δ 13 C signature of the entire sample; for forest soil incubations, water samples were filtered on 2-5 filter paper, all of the filter papers were oven dried at 50 ℃ and preserved for POC, PN and δ 13 C 120 measurements. Inorganic C was removed from the filters by exposing them to HCl fumes overnight in a desiccator. Subsequently, the dried filters were packed in Ag cups for analysis on an elemental analyser-isotope ratio mass spectrometer (EA-IRMS, ThermoFinnigan Flash HT and Delta V Advantage). Certified (IAEA-600, caffeine) and in-house laboratory standards (leucine and tuna tissue) were analysed throughout each run.
To determine the DOC concentration and its stable isotope composition, 40 ml filtered water samples (0.2 μm) were collected and 125 stored in glass vials with Teflon-coated screw caps and 100 μL of H3PO4 was added for preservation. Analysis of DOC and δ 13 CDOC was performed on a wet oxidation TOC analyzer (IO Analytical Aurora 1030W) coupled with an isotope ratio mass spectrometer (ThermoFinnigan Delta V Advantage). Quantification and calibration were performed with IAEA-C6 (δ 13 C = -10.4 ‰) and an internal sucrose standard (δ 13 C = -26.99 ± 0.04 ‰).

130
Statistical tests were performed in R. The normality of data was tested with the Shapiro-Wilk test. Paired sample t test was applied to test for differences in DO consumption rates, total amount of C mineralized and POC loss between treatments under rotation and stationary conditions, and between treatments with and without the presence of AMO. Average values are given ± the standard deviation.

Effect of water disturbance and AMO on DO consumption rates
In all treatments, DO concentrations followed a decreasing trend, and the DO consumption rates were relatively constant over time ( Fig. 3). For incubations where soil was present, cumulative DO consumption ranged between 0.7 and 2.6 mg L -1 during the 168h incubation period. The highest DO consumption occurred in the SOC+AMO treatment with an average decrease rate at 0.015 mg O2 L -1 h -1 (arable soil) and 0.010 mg O2 L -1 h -1 (forest soil). DO consumption was lowest for treatments where soil was present 140 without AMO. We found that keeping soil particles in suspension resulted in a relatively small acceleration of the OC decomposition process. With the presence of aquatic microbial organisms (SOC+AMO), DO consumption rate was increased by 13 % (p < 0.05) for the arable soil while no significant effect was found for forest soil (p > 0.05).

Total amount of C mineralized
We calculated the total amount of C mineralized in each treatment, using an O2: CO2 ratio of 1.0 and expressed mineralisation 150 rates on a per carbon basis. Obviously, trends in total C mineralisation are similar to those in oxygen consumption. For SOC+AMO incubations with soil from arable land, keeping particles in suspension increased total amount of C mineralized by 13 % (p < 0.05, Fig. 4). Also for forest soil a small increase was noted (11 %) but this was not statistically significant (p > 0.05). In addition, when comparing the mineralisation of oven-dried and air-dried soil without AMO addition, oven-dried soil incubation showed a 40 % lower C loss in comparison to air dried soil incubation at the end of the incubation (Fig. 5). However, the addition of AMO resulted in higher C loss with both treatments (air drying: 70 %; oven drying:165 %).

Particulate and dissolved organic carbon concentrations and δ 13 CPOC, δ 13 CDOC values
Measurements of final POC concentrations showed a reduction of POC at the end of the incubations, where 5 %-13 % of POC was mineralized for incubations with soil from arable land, and 1 %-11 % for incubations with forest soil (Fig. 6, Table 3). Keeping particles in suspension resulted in ca. 4 % more POC loss in SOC+AMO incubation series for both soil types (arable land: 13 % vs. 9 %, p < 0.05; forest: 11 % vs. 7 %, p > 0.05). Conversely, keeping particles in suspension in SOC-AMO incubation series 165 showed negative effects with 1 % (arable land soil)-4 % (forest soil) less POC loss, but this was not statistically significant (p > 0.05). The presence of AMO led to ca. 8-10 % (with rotation) and 2-3 % (under stationary conditions) more POC loss for both soil types. These trends are generally consistent with the patterns derived from the oxygen consumption measurements, i.e. a larger reduction in POC when AMO is present and a limited effect of hydrodynamic disturbance. However, it is also clear that variations in residual POC are less consistent than those observed from oxygen consumption rates. In 7 of 12 incubation series, the combined 170 POC and DOC losses exceeded the total amount of C mineralized calculated from DO consumption (Table 3). δ 13 CPOC values showed similar variation patterns for both soil types ( Fig. 6): firstly, an increase occurred during the first 24 h (for forest soil) or 48 h (arable land soil). The increase in δ 13 CPOC was more important for forest soil (0.4 ‰-0.6 ‰) in comparison to arable land soil (0.2 ‰-0.4 ‰). After this initial period, δ 13 CPOC values stabilised.
DOC concentrations were relatively stable during the 168 h incubation ( Fig. 7; Table 3). Initial δ 13 CDOC was lower in arable land 175 soil experiments where no AMO was present. When AMO was present the initially higher δ 13 CDOC values declined to similar values as those observed in experiments without AMO in the first 48 h. In the experiments with forest soils initial δ 13 CDOC were ca. 0.6 ‰ higher than the initial δ 13 CDOC values for arable land soils without AMO. δ 13 CDOC were stable throughout the incubation period for all forest soil experiments.
For both soil types the equilibrium δ 13 CDOC values are somewhat lower than the equilibrium δ 13 CPOC values: this difference is more 180 pronounced for the forest soil (ca. 1 ‰ in comparison to ca. 0.5 ‰ on average) (p > 0.05).

Discussion
Our results revealed that terrestrial SOC can indeed be mineralized relatively quickly when introduced in an aquatic environment.
With the presence of aquatic microbial organisms (SOC+AMO), up to 0.58 (forest soil) or 0.97 (arable land soil) mg C L −1 was mineralized within the 168 h incubation period, equivalent to 83-139 μg C −1 d −1 . Comparable respiration rates of 20-80 μg C L −1 185 d −1 were reported under similar temperature conditions by Berggren et al. (2010). Similar incubation experiments with water samples collected from northern temperate lakes and streams reported respiration rates from 16 to 54 μg C L −1 d −1 (Mccallister and Paul, 2012). We found that C mineralization rates were quite significant: in the presence of an AMO (which is always the case in natural conditions), 4 to 6 % of the OC present at the initiation of the experiment was mineralized during the experiment (Fig. 4). These mineralization rates are generally much lower than those observed in our study, where 6-9 μg C is mineralized per mg C 195 per day. This suggests that SOC indeed decomposes more rapidly in aquatic systems than in the terrestrial environment because SOC is not as recalcitrant as preciously thought in aquatic systems (Mayorga et al., 2005;Mccallister & Paul, 2012). Several studies already suggested that the transition from terrestrial to aquatic conditions likely facilitated SOC decomposition rates because of potential shifts in environmental conditions (Gurwick et al., 2008;Butman & Raymond, 2011;Mccallister & Paul, 2012). In soils, sorption of OC to mineral surfaces and encapsulation of C within soil aggregates may protect SOC from complete 200 mineralization (Bianchi et al., 2011;Schmidt, et al., 2011). This results in the accumulation of older SOC in pools that are less accessible to decomposers and their extracellular enzymes (Marí n-Spiotta et al., 2014). When SOC enters aquatic systems, a disruption of the mechanisms protecting C from mineralization, such as a physical disturbance due to the physical action of transport in water but also due to aggregate slaking (Le Bissonnais, 1996), may lead to the exposure of these protected pools to decomposers and therefore to an increase of the SOC decomposition rate. Alternatively, SOC decomposition may be accelerated microbial community is present, there is also a significant degree of SOC decomposition. Again, this suggests that the simple https://doi.org/10.5194/bg-2020-267 Preprint. Discussion started: 5 August 2020 c Author(s) 2020. CC BY 4.0 License.
immersion into water results in the breakdown of the physical protection of SOC, so that a similar microbial community becomes much more effective in decomposing SOC in an aqueous rather than a soil environment.
The presence of an aquatic microbial community caused a much more rapid mineralization of SOC (Fig. 3, Fig. 4). In all experimental runs that we performed (n = 12), there was a significant difference observed between treatments with and without 225 aquatic microbial organisms (p < 0.05). Given that the soil samples were oven-dried at ~55 ℃ before the incubation, the effect of inoculation with AMO may at least partly be explained by the fact that the soil microbial community was killed by the drying process. Comparing the results of oven-dried and air-dried soil without AMO addition, it is clear that oven drying indeed led to a 40 % decrease in the total amount of C mineralized in comparison to air dried soil, indicating that oven drying indeed eliminated an important fraction of OC-consuming soil microbial organisms (Fig. 5). However, the addition of AMO resulted in clearly higher 230 C losses with both soil treatments (air drying: ca.70 %; oven drying: ca.165 %), indicating that aquatic microorganisms indeed have the capacity to rapidly consume SOC that is not readily mineralized by soil microorganisms. This is not surprising. It is well known that a significant fraction of the SOC is very old (Trumbore, 2000). The presence of such old fractions in the soil is only possible if the soil ecosystem does contain no or only a very small number of consumers capable of mineralizing this POC fraction.
Clearly, some consumers in aquatic ecosystems do have this capability so that significantly more POC can be consumed. The 235 higher POC consumption rates observed when an AMO is present may, of course, partly be due to the fact that more microbes were present in those experiments where AMO were introduced. We compared the initial population of bacteria with and without the addition of AMO. The addition of AMO led to 20-30 % more bacteria present in the water at the beginning of the experiment (arable land: 4.30 × 10 5 vs. 3.52 × 10 5 cells ml −1 ; forest soil: 3.67 × 10 5 vs. 2.86 × 10 5 cells ml −1 ). This larger initial population could partly explain the higher SOC decomposition rates with the addition of AMO, but the fact that the addition of AMO increases 240 SOC decomposition rates by 70-165 % rather than 20-30 % does suggest that the aquatic microbial community is indeed capable of attacking old, stable SOC more effectively than the soil microbial community. Although the microbial community is considered to play a central role in shaping OC reactivity in both terrestrial and aquatic systems (Schmidt et al., 2011), such strong stimulation effect of the addition of AMO on SOC has rarely been reported.
The evolution of POC and DOC concentrations during the experiments is generally in agreement with the patterns derived from 245 the oxygen consumption measurements. However, the variations in residual POC are less consistent than those observed from oxygen consumption. Several reasons might explain the discrepancy between the total C mineralisation as calculated from oxygen consumption in comparison to direct measurements of POC. Firstly, POC and DOC samples were collected from 4 incubation bottles for one treatment. Even though we controlled the initial conditions of each bottle as closely as possible, there might be heterogeneity between different bottles with respect to the OC content of the soil sample that was placed into the bottle. Secondly, 250 we compared POC measurements from 2-5 filter papers for soil from arable land and 5 filter papers for forest soil. For each individual measurement filter weight has to be subtracted from the gross weight of the filter plus the sediment. Given the small quantities of sediment present on the filters, small weighing errors will result in relatively large errors in the calculation of the amount of POC that is remaining. Oxygen measurements are non-intrusive and are not subject to measurement errors related to the weighing of small quantities. We therefore believe that the oxygen consumption measurements provide us with more robust 255 measurements of OC decomposition in comparison to direct measurements of the OC content of the remaining sample. The direct measurements suggest that, overall, POC mineralisation was more important than DOC mineralisation when POC was present.
Indeed, DOC concentrations showed little variation during the experiments, despite significant oxygen consumption rates.
The increase of δ 13 CPOC values during the first 24-48 hours suggests that during this period an isotopically lighter POC fraction was preferentially mineralised. This resulted in the POC in the aquatic environment becoming enriched in 13 C by 0.2-0.6 ‰ adjustment is indeed due to the preferential consumption of a somewhat lighter, less recalcitrant POC-fraction rather than continued preferential consumption of lighter POC. The fact that, over the whole course of the experiments, the δ 13 CDOC values are lower than the corresponding δ 13 CPOC values, is on the other hand, best explained by a continuous leaching/release of DOC with a somewhat lower δ 13 C signature from the soil POC, replacing the mineralised DOC as it is unlikely that the DOC fraction would 265 be entirely stable while POC is continuously mineralised. Mineralisation of the original DOC and its replacement with soil-derived DOC could also explain the drop in δ 13 CDOC during the initial phase of the experiments with arable soil and AMO, because the lighter DOC that was originally present in the river water is replaced by soil-derived DOC. However, this drop was not observed in the experiments with forest soil. This may be partly explained by the higher δ 13 C value of the forest soil (-28.6 ‰) in comparison to the δ 13 C signature of the arable soil (-29.4 ‰) causing the DOC released from the forest soil to have an isotopic signature close 270 to that of the river water.
While the patterns described above would be consistent with the preferential decomposition of isotopically lighter POC, we did not observe an increase in δ 13 CPOC during our experiments as might be expected by the selective mobilisation of an isotopically lighter soil fraction: this can be explained by the relatively small differences in δ 13 C values between POC and DOC in combination with the fact that only a small fraction of the POC is ultimately mineralised, whereby most of this mineralised fraction may have 275 been directly transformed to CO2. If this mineralization does not selectively affect specific fractions of the POC pool, the δ 13 CPOC values can be expected to remain more or less constant throughout the incubation period.
Based on our findings, we propose that the relative importance of physical disturbance vs. exposure to a novel microbial community is likely to depend on (i) the level to which the SOC is indeed physically protected and (ii) the extent to which this protection is destroyed when soil particles are introduced in river water. When the protection level is relatively important but, at the same time, 280 sensitive to water immersion, further physical disturbance is unlikely to strongly increase SOC breakdown in aquatic conditions. If, on the other hand, physical protection is strong and resistant to immersion, physical disturbance may be necessary to break down soil aggregates to the extent that is needed to expose a significant fraction of SOC to the microbial community present in water (Fig. 8). The effect of the presence of an aquatic microbial community, on the other hand, will depend on its composition and its vigour. The addition of SOC may shift aquatic microbial metabolisms and make it more prone to SOC decomposition 285 (Lennon and Pfaff, 2005). Priming effects, whereby the exposure of old SOC to a different pool of microbes leads to increased SOC decomposition have also been documented in soil environments (Fontaine et al., 2007). External drivers such as water temperatures may play a role here. However, it is fair to state that our current understanding of the interaction between old SOC and different microbial communities is too limited to develop general principles describing which factors may stimulate or slow the decomposition of SOC exposed to a new microbial community.

Conclusions
We investigated the relative importance of physical disturbance vs. exposure to a novel microbial community on SOC decomposition rates in aquatic environments. While some recent studies found that the impact of mechanical disturbance on SOC decomposition rates was very important, we found only a very modest increase in SOC decomposition when soil particles were mechanically disturbed and kept in suspension. A simple conceptual model, whereby the effect of mechanical disturbance is 295 assumed to depend on the initial structural stability of soil aggregates can explain this difference in findings: mechanical disturbance is only important when soil aggregates are strong enough to withstand the disruptive forces imposed by immersion in water.
Our study also highlights the role of aquatic microbial organisms in SOC decomposition in river systems. Aquatic microbial organisms are capable of attacking old, stable SOC, leading to rapid SOC decomposition in river systems. Given the variability of 300 aquatic microbial community composition along different aquatic systems, understanding the linkage between aquatic microbial community composition and abundance on the one hand and the resultant SOC mineralization rates on the other hand would be important to better understand CO2 outgassing in aquatic systems.

Data availability
All data used and produced through this study are available upon request.

305
Author contribution GG, SB, LJ and MZ involved in the design of the experiments, and MZ carried out the experiments. SB and MZ conducted the laboratory analysis. All authors offered advice on the data analysis and contributed to the paper preparation.

Declaration of Competing Interest
The authors declared that there is no conflict of interest.     https://doi.org/10.5194/bg-2020-267 Preprint. Discussion started: 5 August 2020 c Author(s) 2020. CC BY 4.0 License.