36Cl, a new tool to assess soil carbon dynamics

Soil organic carbon is one of the largest surface pools of carbon that humans can manage in order to partially mitigate annual anthropogenic CO2 emissions. A significant element to assess soil sequestration potential is the carbon age, which is evaluated by modelling or experimentally using carbon isotopes. Results, however, are not consistent. The 14C derived approach seems to overestimate by a factor of 6–10 the average carbon age in soils estimated by modeling and 13C approaches and thus the sequestration potential. A fully independent method is needed. The cosmogenic chlorine nuclide, 36Cl, is a potential alternative. 36Cl is a naturally occurring cosmogenic radionuclide with a production that increased by three orders of magnitude during nuclear bomb tests. Part of this production is retained by soil organic matter in organochloride form and hence acts as a tracer of the fate of soil organic carbon. We here quantify the fraction and the duration of 36Cl retained in the soil and we show that retention time increases with depth from 20 to 322 years, in agreement with both modelling and 13C-derived estimates. This work demonstrates that 36Cl retention duration can be a proxy for the age of soil organic carbon.

bonds as organic chloride compounds, of various high molecular weights 11 , that are then fragmented into various low molecular weight compounds.Laboratory experiments conducted on different soil samples evidenced a link between Cl chlorination and soil organic matter [13][14][15][16][17] .Although based on short-term laboratory experiments, these results, together with the fact that Cl is covalently bonded to organic compounds, suggest that organochlorine follows the same dynamics as soil organic matter, moving from one molecule to another through bioassimilation or being released by mineralization into the soil solution, when carbon returns to the atmosphere as CO 2 . 36Cl is a naturally occurring radionuclide (half-life: 301,000 year) formed in the atmosphere by spallation of 40 Ar.Its production increased by three orders of magnitude above its natural level during the marine nuclear bomb tests that started in 1950, reached a peak in the early sixties, and lasted until the late 1970s [18][19][20][21][22][23] .Although stable chlorine isotopes ( 35 Cl and 37 Cl) have different sources, mainly produced by marine sprays, 36 Cl follows the same biogeochemical cycle without isotopic fractionation 24 .Since organochlorine mimics the fate of the global soil organic matter, we propose that the fraction of 36 Cl from the nuclear bomb tests retained in soils can be used to trace the SOC dynamics.This requires being able to quantify the fraction of 36 Cl retained and the duration of 36 Cl retention in soils.

Results and discussion
The fraction of 36 Cl retained in soils is a function of the SOC content.Batsviken et al. 15 assessed experimentally, by incubating soil samples with a fixed amount of 36 Cl, that 20% of 36 Cl was retained in soils a few hours after injection but that only 4% remained after a few weeks.Gustavsson et al. 16 suggested, also experimentally, that the amount of 36 Cl retained in soils could vary 3-to 4-fold with the vegetation cover.However, the amount of 36 Cl retained in soils under natural conditions has not yet been assessed.To fill this gap, we measured and modelled both Cl and 36 Cl concentrations in a soil profile from a mature beech forest in Northern France.Figure 1 shows (1) the measured Cl and 36 Cl inputs to the soil (by rainfall 25 , throughfall-portion of the rainfall that reaches the soil after penetrating the canopy-and litterfall) and outputs from the soil by soil water drainage in 2012-2013; (2) the Cl and 36 Cl stocks in the different soil layers measured in 2010 and the corresponding 36 Cl/ Cl ratios (see "Methods").
The 36 Cl/Cl ratios measured in the annual rainfall and litterfall are not significantly different (Fig. 1a), inducing no isotopic fractionation against 36 Cl by the vegetation.Cl is mainly in soluble anionic form in the leaves 24 , probably accumulated in the vacuole, and thus passively absorbed with the water flux by the plant, explaining the absence of isotopic fractionation.Since the 36 Cl/Cl ratios of the input and output fluxes are not significantly different (Fig. 1a), we can also assume the absence of isotopic fractionation associated to the transfer of 36 Cl through the soil and, as a corollary, a chlorination process also devoid of significant isotopic fractionation.
On the contrary, the 36 Cl/Cl ratio measured in soil organic matter is two to six times higher than the ratio measured in the rainfall and litterfall and in drainage (Fig. 1a).The soil Cl and 36 Cl stocks result from multiannual accumulation and this result suggests that some of the retained 36 Cl comes from the high 36 Cl input due to nuclear tests.This also suggests that the retention duration of 36 Cl at our site is higher than 40-60 years (i.e., duration estimated between the nuclear bomb testing period and the measurement dates).
The contribution of litterfall to soil 36 Cl and Cl represents 11% of that due to rainfall (Fig. 1b,c).Throughfall Cl and 36 Cl inputs equal rainfall Cl and 36 Cl inputs.The 36 Cl and Cl inputs to the soil over time can therefore be estimated on the sole basis of the 36 Cl and Cl rainfall flux reconstruction.
We developed a model based on a mass balance calculation of the forest stocks and fluxes for each of the five soil layers (Fig. 2).A Monte Carlo approach was used to determine the fraction of the Cl and 36 Cl input retained in the soil (X k ) by adjusting output 36 Cl stocks to the measured ones in 2010 (Fig. 2).Simulations were run from the pre-bomb period (1910 AD), when both Cl and 36 Cl can be considered at steady state, up to 2020 AD. 36 Cl anthropogenic production by nuclear bomb tests between 1952 and 1972 18 was used as model input.Once the X k had been determined, the Cl retention duration in the soil organic matter was calculated (for more details see "Methods").
Figure 3 displays the depth-evolution of the X k values, showing that they decreased exponentially, with a ninefold higher value at the surface than at depth.The associated uncertainties also decreased with depth.This result agrees with a high chlorination of organic matter, which is indeed classically observed in the upper soil layer 26,27 .The fraction of 4.5% yielded for the [0-5 cm] layer (X 1 ) is in good agreement with the one obtained experimentally by Batskviken et al. 15 (e.g., 4% after 4 months of incubation of topsoil samples).A more recent study 17 reported X k of the same order of magnitude for a forest soil with the same type of humus (mull).Lower X k were however derived for forest soil associated with different types of humus layers (thicker: moder or mor).As in our results, the authors also observed a decrease in X k with depth.Some much higher X k were also observed in other studies (15-35%) with variable experimental conditions, and no clear drivers were found to explain these differences 16,28 .
The exponential decrease of X k with depth (Fig. 3) is in good correspondence with the exponential decrease of SOC concentration with depth (Supplementary data Table 1) and the 36 Cl fractions retained in the soil for the different soil layers are linearly correlated to SOC concentrations of these layers (Fig. 4a).This confirms that the dynamics of the 36 Cl in organic form follows that of the soil organic matter.
A 36 Cl retention duration in soils equivalent to the age of the SOC.The retention duration of 36 Cl in the SOC increased with depth from 20 ± 15 to 305 ± 110 years (Fig. 3b).The layer-by-layer 36 Cl retention durations and the median SOC ages calculated by Balesdent et al. 9 were of the same order of magnitude and the two datasets show a very nice correlation for all depths (Fig. 4b).While both methods obtained a SOC age of about 50 years for the first 0-30 cm of the soil, the SOC ages obtained from 14 C measurements for the same depth were 1390 ± 310 8 .This discrepancy could be due to the presence in the soil of a very old fraction of SOC, even  www.nature.com/scientificreports/ in topsoil, that no longer contributes to the active soil carbon cycle but contributes to the 14 C age estimates 29 .Indeed, even in small proportions, old organic carbon leads to older average ages especially in the surface layers (0-30 cm) that are not representative of the real organic carbon dynamics of these layers 30 .The median SOC age obtained by Balesdent et al. 9 provides more precise information on the dynamics of carbon in the different soil layers, while the 14 C method makes it possible to estimate the proportion of very old organic carbon, part of which can be considered as inert.The ages obtained for 36 Cl are very similar to the median ages obtained by the stable carbon isotope method 9 and are therefore a good indicator of the active fraction of the SOC.
In conclusion, the 36 Cl retention duration can be used as a proxy for the median age of active SOC, i.e., the reactive part of soil organic carbon.This age is also provided by the δ 13 C analysis, in the special cases of sites that have experienced a change in vegetation photosynthesis, at a known date 9 .Because it is not conditioned by this change, the 36 Cl-based method may have greater potential for application.However, it has to be kept in mind that the present study is a proof of concept done on a single soil.It will have to be reproduced on other soil types: saline soil that requires adaptation of the model and other forest soils whose humus types could induce a variation in Cl and 36 Cl cycles.Nevertheless, this method potentially provides strong constraints on the average age of reactive soil carbon that can be used by ESM models.In this light, the 36 Cl method opens up a more accurate constraint on soil sequestration potential and a better assessment of anthropogenic carbon compensation by soils in the future.

Methods
Study site, flux and stock sampling.The experimental site consists in the 73 ha mature beech forest of the Perennial Environmental Observatory (PEO) located at Montiers-sur-Saulx, in the north-east of the Paris Basin, France (48° 31′ 54″ N, 5° 16′ 08″ E).The forest is a mature beech forest with a mean tree age of 50 yrs.It is part of a national forest, already recorded on the Cassini maps (XVIIth century), the Napoleonic land registry (1820-1866) and on old aerial photographs (1950-1965, Supplementary data Fig. 2).The soil is a Cambisol developed on Tithonian limestone.The humus of this forest soil was a mull.
A one-hectare station was established to monitor the following fluxes: rainfall, throughfall, drainage water at 60 cm depth and litterfall.Rainfall was sampled by open rainfall collectors located over the canopy as described by Pupier et al. 25 .Throughfall was collected monthly from March 2012 to February 2013 by four rain gutters placed 1 m above the forest floor and spatially distributed in order to capture the canopy variability and representing an equivalent surface of 0.39 m 2 .The water drainage at 60 cm depth was collected with the same frequency and time period by three lysimetric plates measuring 0.12 m 2 .This device only collects soil gravity water.An average sample of 1 L was made on site from these three lysimetric plates.All the devices were of high density polyethylene (HDPE).In order to avoid possible contamination, all the material was washed with ultra-pure water.
The litterfall was collected over the year 2012 thanks to 6 L trays of 1 m 2 to capture the canopy variability.The litter was weighed, dried at 65 °C in a ventilated oven for 1 week and ground in a ring mill.An annual composite sample was created.
The soil was sampled by coring in 2010.The following depth intervals were sub-sampled: 0-5 cm, 5-15 cm, 15-30 cm, 30-45 cm and 45-60 cm.Three cores were sampled and mixed to form an average sample.The soil samples were dried at 35 °C and sieved at 2 mm. 36Cl extraction and analysis.The Cl and the 36 Cl in the solid matrices were extracted by hydropyrolysis and trapped in ultrapure water after a protocol adapted from Cornett et al. 31 and Herod et al. 32 .A selective extraction of soluble Cl and 36 Cl was also performed by equilibrating 5 g of dry soil with 40 mL of ultra-pure water for an hour.

Cl and
The Cl and 36 Cl in the liquid samples were precipitated as AgCl by adding nitric acid and silver nitrate following a standard procedure as described in Pupier et al. 25 and Bouchez et al. 33 . 36Cl and Cl concentrations were measured by isotope dilution accelerator mass spectrometry at ASTER-CEREGE 34,35 (Accelerator for Earth Sciences, Environment, Risks).Both the 36 Cl/ 35 Cl and the 35 Cl/ 37 Cl ratios were measured and normalized to the in-house standard SM-CL-12 ( 36 Cl/ 35 Cl value of (1.428 ± 0.015) × 10 -12 ), assuming a natural 35 Cl/ 37 Cl ratio of 3.127.
Cl and 36 Cl flux and stock calculation.While both soluble and total 36 Cl and Cl concentrations were measured, we decided to calculate total stocks that were considered as organic as: 1. soluble fractions were very small, especially for 36 Cl (one order of magnitude lower), compared to the total fraction; 2. the considered soil does not contain salts 3. therefore, the soluble Cl and 36 Cl can either come from the Cl and 36 Cl contained in the soil water at the sampling time, or from Cl and 36 Cl released in solution from the organic matter at the extraction time.The concentrations recorded in the soil water were too low to explain the Cl and 36 Cl extracted experimentally 36 .We therefore considered that the soluble Cl and 36 Cl were organically bound Cl and 36 Cl The 36 Cl and Cl stocks (S k,2010 ; at m −2 ) were calculated using the 36 Cl and Cl concentrations (C k,2010 ; at g −1 ), the soil layer bulk density (ρ k ; g cm −3 ) and the soil layer thickness (d k ; cm) according to Eq. ( 1): The calculated stocks are reported in Supplementary data Table 2.The fluxes (F; at m −2 year −1 ) were calculated using the 36 Cl and Cl concentrations (C; at g −1 ) and the measured annual litterfall/water flux (Φ) (Supplementary data Table 3) according to Eq. ( 2): No Cl and 36 Cl stock in humus was considered as the considered forest has a mull type of humus, that is a very thin and reactive humus.Dincher 37 demonstrated for the same site that highly soluble elements were lost from the litter in less than a year.Since Pupier 36 showed that most Cl and 36 Cl in litter was in soluble form, we considered that all of it was lost to the soil within the year, the time step of our model, and hence did not consider a humus compartment.This assumption was confirmed experimentally for a mull by Svensson et al. 14 .
The uncertainties on the stocks and fluxes were calculated based on classical uncertainty propagation equations considering: (1) the analytical uncertainties for the 36 Cl and Cl concentrations; (2) 5% of uncertainty for bulk density; (3) 10% for the layer thickness, the litterfall mass and the water drainage volume; and (4) 3% uncertainty for rainfall as suggested by Météo France specifications.
Modelling approach.Description of the model.The model is a mass-balance approach that considers stocks and fluxes of Cl and 36 Cl in the different soil layers (Fig. 2).
For each box, the mass balance is evaluated at an annual time step ( t is set to 1 year) to smooth seasonal variations, and the stock of year t, for layer k, S k,t , is described by Eqs.(3) and (4): with where S k,t−1 is the standing stock of the previous year on the same layer k.The input fluxes (I k,t ) for each layer come from the vertical transfer flux (D k−1,t ) from the overlying compartment (layer k − 1), except for the upper soil layer which receives rainfall (R t ) and litterfall (L t ).X k is the fraction of the Cl and 36 Cl input that is retained in the soil at each time step.Z k is the fraction of the preexisting stock that is released.
Applied to a mature forest, Cl is considered at steady state 38 so that the total annual root absorption over the different layers is assumed to equal the annual litter fall (L t ) that corresponds to the annual litter production.
As the root distribution is classically considered as exponentially decreasing in depth 39 , we assumed an exponential decrease for the root absorption depth distribution.Other root absorption depth distributions were tested.They had no impact on the model results.
Stocks (S k,t ) in 36 Cl are expressed in at m −2 and 36 Cl fluxes (I k,t , D k−1,t , L t , R t ) in at m −2 year −1 , so that X k and Z k are in %.
In addition in a system at steady state, a residence time of Cl (T R,k ) can be calculated for each soil layer according to Eq. ( 6):

Figure 1 .Figure 2 .
Figure 1.Measured 36 Cl/Cl (unitless) (a), 36 Cl (b) and Cl (c) stocks (bars, at m −2 ), water (blue dots) and litterfall (green dots) fluxes (at m −2 year −1).These fluxes were derived from measurements made in water (rainfall, throughfall-that is the water collected below the forest canopy-and soil water drainage at 60 cm depth) in 2012-2013, and stocks from soil sampled in 2010 along a soil depth profile at Montiers, France.Colours for the different soil stocks represent the different soil layers.

Figure 3 .
Figure 3. Depth evolution of the modelled probability density of: (a) the36 Cl fraction retained in the soil (X k ); (b) the retention duration of36 Cl in the soil.The probability densities are obtained by running a few thousands of simulations, where every variable (rainfall, litterfall fluxes and the Cl and36 Cl stocks) is set randomly using a normal distribution defined by the measured variable values and their associated uncertainties.Colours represent the different soil layers.While the X k distribution is normal, the retention duration distribution is lognormal (Supplementary data Fig.1).

Figure 4 .
Figure 4. 36 Cl a tracer of the SOC.(a) The 36 Cl fraction retained in the soil layers is a function of the SOC content.(b) The yielded retention duration of 36 Cl in the soil is linearly correlated to the age of the SOC estimated by Balesdent et al. 9 for tropical forest and pasture sites.Colours represent the different soil layers as reported in Figs. 1, 2 and 3.