Carbon Dynamics in
Plaggic Anthrosols
A modelling approach
Emma Polman (10799478)Amsterdam, July 3, 2017
Supervisor Boris Jansen
Daily Supervisor Marijn Van de Broek
2
Abstract
Plaggic anthrosols are remnants of the plaggic agricultural system, which was used from about 1600 to 1900 in north-east Europe. The system aimed at making nutrient poor sandy soils fertile by adding organic matter in the form of manure enriched sods. This practice stopped 100 years ago, but high concentrations of soil organic carbon (SOC) can still be found in these soils (percentages >1%). Multiple theories on SOC decomposition exist, but quantification is still lacking. This research aimed at gaining quantitative understanding of the SOC dynamics in Plaggen Anthrosols, using a modelling approach. The model used is the DELTA-13C model, which simulates organic carbon (OC) isotope δ13C, OC content (OC%) and OC age profiles of aggrading soil profiles. The model also enables calibration of parameters using both the OC and δ13C profile as constraints, which was expected to narrowed down parameter ranges. Simulation was done for two profiles in the Netherlands of which data was already available: Nabbegat and Posteles. The DELTA-13C requires many input parameters, which were obtained via literature and available data. Four types of organic carbon were used as input: aboveground and belowground vegetation originating OC and stable and unstable sediment originating OC. The model was capable of simulating the OC% profile of Nabbegat and the δ13C profiles of Posteles and Nabbegat with the obtained parameter set, but did not simulate the OC% Posteles profile accurately. Both simulations suggest that a considerable part (>1.5%) of the OC% in the profiles consists of stable sediment originating OC. This is in line with the simulated half life of the stable sediment fraction, which is in the order of centuries or millennia. For the aboveground vegetation and unstable sediment fraction a half life of 2 years was found, for the belowground vegetation this was 2.5 years. Calibration on both the OC and δ13C profile did not result in a narrower parameter range for the calibrated parameter for the decrease of external influences with depth (rExp), because too little δ13C data points were available for the calibration. In addition, calibration of two parameters (rExp and parameter for humification h) at the same time did not result in a realistic parameter range, so it is concluded that the DELTA-13C model works best with only 1 degree of freedom in these simulations. The largest uncertainties in the simulations are caused by insufficient data on the stable/unstable in the sediment OC and the effect of manure enrichment on the OC% of the plaggen. More research is needed for better estimation of these values and more accurate simulation results and calibration.
Keywords: SOC dynamics, plaggic anthrosols, carbon modelling, model calibration, SOC
3
1. Introduction
Plaggic anthrosols are remnants of the plaggic agricultural system, which was long used by farmers to make nutrient poor sandy soils suitable for agriculture (Pape, 1970). During the winter, sods were lied as bedding in stables to get them enriched with manure. Thereafter, these sods were added to the field (figure 1). The addition of manure enriched sods lead to an input of organic matter (OM) and thus organic carbon (OC) to the soil (Blume & Leinweber, 2004; Giani et al., 2014).
The practice of this system started in the Middle Ages and lasted until the development of fertilisers in the 20th century (Van Mourik et al., 2010). One of the regions in which the plaggen agriculture was widely used is north western Europe, since these were largely covered by nutrient poor cover sands (Pape, 1970; Van Mourik et al., 2016) (figure 2a). Because of the addition of both OM and mineral material to the surface, soils have been thickened up to a metre (Pape, 1970). It was long thought that only heath sods were used to enrich the soils, but recent research pointed out that other materials, such as forest litter and straw were also used as filling material (Van Mourik et al., 2016). The input sources of the OC in plaggic soils are diverse, because of the diversity in sods, manure and vegetation already present on the deposition site. (Van Mourik et al., 2011).
Figure 1: Visualisation of the practice of plaggen agriculture (Giani et al., 2014).
Nowadays, the SOC concentrations and ages in many plaggic soils are higher than expected, since sandy soils are often pore in SOC, with concentrations below 0.75%. OC percentages above 1.5% and sometimes even 5% are measured, which is very high for sandy soils (Pape, 1970). This is striking, since the input of carbon from plaggen deposits stopped about 100 years ago. It seems that a considerable share of added SOC remains in the soil for a long time and has a residence time of centuries. However, SOC inputs, outputs and residence times, are not well quantified yet, which makes it difficult to determine the role of soils as either a carbon sink or source (Rumpel et al., 2002). In addition, little can be said about the response of soils to changes in climate or land use. This information is important for climate modelling, since globally soils contain about 2400 Pg of OC to a depth of 2 metres, which is 3.2 times as much as the atmospheric carbon pool (Sparks, 2003) and could contribute to either a decrease or an increase in atmospheric CO2 concentrations.
Quantification of soil organic carbon residence times will lead to more accurate climate models and scenarios. Plaggen soils are especially interesting for this matter, since they seem to store a large amount of slow or hardly decomposable carbon.
SOC decomposition
Multiple theories exist considering the decomposition of SOC. It was long thought that the stability of SOC was determined by the chemical composition and characteristics of the OM that was added to
4 the soil. However, many studies suggested that besides molecular structure the stability of SOC is affected by environmental aspects, such as interactions with organic and inorganic materials and low accessibility to decomposing microorganisms (Han et al., 2015). According to Schmidt et al. (2011), environmental controls, among which reactive surface minerals, water availability, acidity, redox state and climate, determine SOC persistence primarily, while molecular composition has a secondary role. Lehmann & Kleber (2015) also pose criticism on the classical humification theory, which states that some small decomposition polymers and monomers react to form humic substances, which are stable and hard to decompose. This theory relies only on molecular composition. To incorporate environmental factors a decomposition model is proposed in which carbon inputs are decomposed into smaller fractions, which can at the same time be adsorbed or desorbed by the mineral surface and form or destruct aggregates (Lehmann & Kleber, 2015). SOM decomposition is thus regarded a complicated process. It can be concluded that there is a need for better quantification of the behaviour of SOC and SOC pools. Plaggic Anthrosols are especially suitable as a testing case, since there is no theory yet explaining the preservation of large carbon stocks. In addition, these profiles have already been extensively researched, leading to knowledge on their formation history as well as a considerable data availability.
One of the methods for achieving more quantitative insight in SOC is modelling the OC inputs and outputs in developing plaggic anthrosols. The model that will be used in this research is the DELTA – 13C model, which simulates the OC concentration (OC%), OC age and stable carbon isotopes (δ13C) against soil depth (Van de Broek et al, in preparation). The use of this model is an innovative approach, since it combines the OC% and δ13C profiles. The use of more constraints in calibration is expected to increase the accuracy of the model and its input parameters. Because of the increase in constraints combined with the ability to simulate depth profiles and aggrading sediments, the DELTA-13C model is expected to give better results than the more commonly used RothC and Century carbon models which model only carbon stocks. Another advantage of the
DELTA-13
C model over other models is its ability to track the contribution of OC from different sources to the final OC stock and final OC age. The simulation of carbon profiles will be done for two soil profiles, Nabbegat and Posteles (Figure 2abc). These profiles are both located in The Netherlands and have been subject to previous research so data is already available. The model, profiles and available data will be discussed in greater detail in the methods section.
5
1.2 Research aims and hypothesis
This research aims at getting more quantitative insight in SOC residence times in plaggic anthrosols in the Netherlands while using a modelling approach. The sub-questions asked are:
1) To what extent can parameters be found that simulate the measured organic carbon and δ13C profiles accurately in plaggic anthrosols using the DELTA – 13C model?
2) What are the differences in residence time and half life between above-ground and below-ground originating organic carbon and organic carbon added by plaggen practices?
3) What are the main differences between the Posteles and Nabbegat soil profiles and what could be the cause of these differences?
In line with these research questions the following hypothesis were set up:
1) It is expected that using literature and available data, parameters can be found that simulate both profiles accurately. In addition, this accuracy is expected to increase when chosen parameters are calibrated. Calibration on both the OC% and δ13C profile is expected to give a relatively small parameter value range.
2) It is expected that OC originating from plaggen deposits will be dominant in the long residence time carbon pool and that this pool will contribute the most to the total OC%, since the soils have been subject to plaggen agriculture and have a carbon content higher than expected for sandy soils. 3) Differences between simulated carbon ages and carbon concentration between the profiles are expected, because of the different vegetation histories and deposition rates. Because of these differences, it is expected that the calibration will also yield different optimal parameter ranges. However these differences are expected to be small since the profiles are located geographically close to each other in similar environments.
To answer these research question, first the methods for simulating and assessing the OC profiles will be discussed. Thereafter, results of the simulation will be described and discussed, and concluding remarks will be given.
6
2. Methods & data
2.1 Analysed Profiles
In this research, the profiles of Posteles and Nabbegat were analysed. For both profiles, OC%, δ13C signals, OSL dating, pollen diagrams and biomarkers were available. These can be found in appendix I.
Nabbegat: The Nabbegat profile (figure 2c) consists of a Haplic Arenosol on top of a Plaggic
Anthrosol which is at its turn on top of a ploughed Umbric Podzol. Fertilisation by use of plaggen is estimated to have started around 1676 AD, based on the greater abundance of mineral grains and the OSL dating of these grains at a depth of 130 cm (Van Mourik et al., 2016). The plaggic Anthrosol was overblown by drift sand around 1803 AD. In total, it is estimated that a thickening of 60 cm occurred in 127 years. The sedimentation rates are calculated based on the OSL dating and can be found in table 1. For the 70 cm above the shallowest OSL data point, it is assumed that sedimentation took place until 2000, with a constant rate of 0.5 cm/yr. The age of the OC in the plaggen deposits was estimated at 1815 years, which is the mean age difference between the OSL and C14 ages of the humic acids fraction at the same depth points. Since the humic acid fraction is a stable and well conserved fraction, this C14 age is expected to be older than the mean age of the total OC present (Tonneijck et al., 2006). Calculations can be found in appendix IId.
Posteles: The Posteles soil profile (figure 2b) consists of a plaggic Anthrosol, which starts at a
depth of 95 cm on top of a ploughed Umbric Podzol (Van Mourik et al., 2016). The top 45 cm of the plaggic Anthrosol (Ap horizon) is actively ploughed, therefore, no OSL dating was performed on the 0-45 cm depth. The sedimentation rates of the profile are calculated using the OSL ages and can be found in table 2. It was assumed that sedimentation took place until 2000, just as the Posteles profile. The mean age of the plaggen OC was estimated at 1467 years (appendix II).
Table 1. Sedimentation rates for Nabbegat
Sedimentation rates Nabbegat
Depth interval (m) Time interval (yr) Sedimentation rate (m/yr)
0-70 1803-2000 0.0053
70-80 1770-1803 0.003
80-130 1676-1770 0.0036
Table 2. Sedimentation rates for Posteles
Sedimentation rates Posteles
Depth interval (m) Time interval (yr AD) Sedimentation rate (m/yr)
0-45 1758-2000 0.00190 45-59 1711-1758 0.00298 59-70 1651-1711 0.00183 70-82 1626-1651 0.00480 82-95 1517-1626 0.00119 95-105 2035 BC-1517 -
7
2.2 DELTA-
13C model
This research used a modelling approach to get more insight in the behaviour of organic carbon in plaggic soils over time. The model used was the DELTA – 13C model (Depth Explicit Link between Total organic carbon And δ13C) (Van de Broek et al., in preparation). This model is an adaption of the ICMB-DE (ICBM – Depth Explicit)model by Wang et al. (2015), which itself is an adaption of the ICMB model by Andrén & Kätterer (1997). The ICMB-DE model was adapted to expand the model with the capability of modelling depth profiles of stable carbon isotopes (δ13C) (Van de Broek et al., in preparation). This enabled the use of both the carbon concentration (OC%) profile as well as the isotope profile to assess the model's best fit with the measured data.
The DELTA – 13C model enables the realistic simulation of soil organic carbon which is deposited in an aggrading landscape position. Sediment and OC accumulation because of plaggen deposition are explicitly simulated by the model, so that sediments and carbon are advected from the top soil to deeper soil layers. As previously described, the generated output consists of depth profiles of OC%, stable carbon isotopes (δ13C) and C age. Moreover, the model calculates the turnover rate of SOC originating from different sources, which are above-ground (AG) litter input, below-ground (BG) litter input (roots) and external inputs (plaggen). The ICMB-DE model also simulated third recalcitrant carbon pool in addition of the original two-pool model (Andrén & Kätter, 1997), in order to model the persistent carbon found in plaggic soils. The ICMB-DE works with a pool for young and a pool for old carbon, while the DELTA – 13C consists of three carbon pools, for annual, decadal and centennial residence times respectively (figure 3). Finally, bioturbation (biodiffusion parameters) was incorporated in the DELTA – 13C model in order to give a realistic simulation of physical mixing in the soil.
Figure 3. Visualisation of the DELTA-13C model (Van de Broek et al., in preparation). A description of the parameters can be found in table 3.
8
2.3 Parameters and calibration
The DELTA – 13C model requires a considerable amount of parameters as input, of which an overview is given in table 3. This section will briefly discuss these parameters, more detailed descriptions and calculations can be found in appendix II. A complete list with all parameters used in the model code can be found in appendix III & IV.
When the model was run, first a spin-up run was performed, to create a model state on which the simulation of plaggen deposition can be started. In the simulations a spin up run of 500 years was performed, which created a soil profile with a depth of 3 metres. During the spin-up phase, forest vegetation was assumed, based on pollen diagrams of the drift sand horizons (appendix I). The spin-up was the same for both profiles and has its own parameter set. After the spin-up run the deposition phase started, in which carbon from sediment and vegetation was added to the top of the simulated soil profile. For the sake of simplicity, only one vegetation type was used for both profiles: the cultivation of Avena sativa (oat), since this species was abundant in many depths of both biomarker profiles (Appendix I). Besides parameters for the vegetation, parameters for the properties of plaggen and general model parameters were used as well as the deposition rates calculated in table 1 and 2. The deposited plaggen were estimated to have a OC content of 4.1% in total, of which 1.6% and 2.5% were estimated to be stable and unstable respectively. This was based on the OC content in a forest litter layer and an estimation of the effect of manure enrichment, a more elaborate calculation and explanation can be found in appendix II.d. The general model parameters consist of the k parameter, which gives the decay rate of carbon in a pool and the humification (h) parameter, which determines how much carbon is transferred to the next, more recalcitrant, pool. Furthermore, external influences on decomposition at the surface are captured in the r0 parameter and rExp is a parameter for the exponential decrease of r with depth. The same accounts for the D0 and Dexp, which determine biodiffusion. Values for these parameters were drawn from literature, calculated based on the measured soil profile data or on a combination of both literature and field data. When values for all parameters were found, the simulations were run with these literature parameter set, without calibration. The output of these "literature" runs was assessed to determine the values or calibration.
The aim of the calibration is creating a simulation with the smallest total relative root mean squared error (RMSE) compared to the measured data. According to Rompaey & Govers (2002), the total error of a simulation is the sum of the model error (errors caused because the model simplifies reality and does not include all processes) and input errors (errors related to the availability and quality of input data). The more variable parameters the model has, the greater the error becomes. In order to balance between these two error types and find the smallest error, only two parameters (two degrees of freedom) were used for calibration. Since literature values were most uncertain for rExp and the humification parameter (h), ranges of these two parameters were used for calibration. During calibration, all possible parameter combinations of rExp and h were simulated and the differences between measured and simulated values for both the C% and δ13C profile were used to calculate the RMSE. The combination of values and value ranges that yielded the lowest RMSE were considered most suitable. This was done for both Posteles and Nabbegat
When the most suitable parameter set was found for each profile, ranges and RMSE were assessed in order to define how well the DELTA – 13C model is able to model the studied plaggen soil profiles. Furthermore, comparisons on the different originating OC half life were made. This way, it was assessed whether external or in situ produced OC dominates in the different carbon pools. For this research, especially the persistent carbon pool is of considerable interest, since its properties are expected to explain the high carbon concentrations which are still present. Thereafter, the results of the two profiles were compared to see if any substantial differences or similarities occur.
9 Table 3: parameters in the DELTA-13C model (Van de Broek et al, in preparation)
10
3. Results
3.1 Simulation with literature values
When values for all parameters were found, the model was run without calibration, to assess how well the literature values simulate both profiles. The simulation results are displayed in figure 4ab. For the profile of Nabbegat, the simulated values and shape of the OC% approaches the measured values fairly well, with the exception of the most upper point. This point is however located in the podzol above the plaggic Anthrosol, which explains the lower OC%. Since it is not possible to vary the OC% in the deposited sediments in the DELTA-13C model, the model simulates the drift sand deposits the same as the plaggen, which is why the top 70 of the simulated Nabbegat profile should not be taken in consideration. For the δ13C profile, the values differ up to 0.6‰, but the simulation displays a slight decrease where the measured values increase with depth. The drastic change in OC% and δ13C values at the bottom of the profiles is due to the transition from the spin-up phase to the deposition phase. In the Posteles profile (Figure 5b) values deviate up to 1.5 ‰, since the simulation starts increasing 10 cm above the lowest measured point. Figure 5a displays that the simulated OC% profile deviates strongly for the measured values, except for the deepest two measured points. Especially between 0.4 and 0.8 meter depth, the measured values display a strong increase, while the simulated profile decreases slightly. Despite differences in deviation between measured and simulated values, according to both simulations, most carbon originates from the plaggen deposits, while AG and BG originating carbon only make up for about 0.25%. The largest part (about 1.5%) of the deposited carbon is stable carbon, while about 0.25% originates from deposited labile carbon.
Figure 4ab . OC% and δ13C simulation of the Nabbegat soil profile. Simulation done with literature parameters (Appendix IV).
11 Figure 5ab. OC% and δ13C simulation of the Posteles soil profile. Simulation done with literature parameters (Appendix III).
Figure 7ab. The Fraction of OC left after initial inputs for Nabbegat (a) and Posteles (b). Simulation done with literature parameters. Figure 6ab. OC age for Nabbegat (a) and Posteles (b). Simulation done with literature parameters.
12 Figure 6ab displays the OC ages of the average total OC and OC originating from AG vegetation , BG vegetation and sediment. Since the up runs were the same for both profiles, OC age from spin-up vegetation is the same for both profiles. In both profiles, Crop AG originating OC is slightly older than BG originating OC. AG and BG OC are both older for Posteles than for Nabbegat. However, the figure shows that OC in the Posteles sediment is younger than the sediment originating OC of Nabbegat . Sediment OC of Posteles is about 300 years younger at the surface and about 200 years at 1 m depth. For both profiles, the measured OC age is lower than the modeled OC age.
Figure 7ab displays the cumulative fraction of OC that is left in the soil profile in the years after the input of carbon to the soil. This concerns inputs of all previous years, so when the fraction left in year 50 is calculated, the inputs left from the years 1 to 49 are summed. The half life derived from figure 7ab is therefore higher than the half life of the individual OC inputs, which is the fraction left of the input in year 1 after x years of decomposition, displayed in table 4. For AG vegetation and labile sediment OC, the half life 2 years and is the same for Posteles and Nabbegat. The half life of BG vegetation OC is slightly higher with 2.5 years and is also the same for Posteles and Nabbegat. The fractions left after 324 years (the simulation time of Nabbegat) are all small (around 0.05), with the exception of the stable sediment OC of which about three quarters are still left after 324 years in both profiles.
Table 4. Half life and fractions left for the different OC fractions for Nabbegat and Posteles.
*higher than the simulation time, **fraction left of the OC input in year 1 after 324 years, which is the simulation time of Nabbegat. Values for Posteles were taken at t=324 to make the results comparable.
Half life (yr) Fraction left after 324 years**
AG vegetation OC Nabbegat 2 0.054
Posteles 2 0.048
BG vegetation OC Nabbegat 2.5 0.062
Posteles 2.5 0.061
Labile sediment OC Nabbegat 2 0.054
Posteles 2 0.048
Stable sediment OC Nabbegat -* 0.75
Posteles -* 0.72
Figure 7ab shows that for both profiles the belowground and aboveground vegetation associated OC decay the fastest, with a half life of about 4 years for the cumulative OC fraction in both profiles. The decay is high in the first 50 years and then stabilizes at a value of about 0.1. The sediment OC in both profiles does not reach the 0.5 fraction, so the half life cannot be read from the figure. The fractions left after 324 years are 0.83 and 0.85 for Nabbegat and Posteles respectively. The half life of the total cumulative OC fraction is 145 years for Nabbegat and 7 years for Posteles. The total OC fraction curves stabilize at about 0.42 for Nabbegat and at 0.3 for Posteles, so the total fraction of OC left after deposition is about 0.12 higher for Posteles than for Nabbegat. The increase in deposited sediment left after 100 years in the Posteles profile occurs simultaneously with a doubling in sedimentation rate.
3.2 Calibration of rExp
rExp, the parameter for the exponential decrease of r with depth was calibrated for both Nabbegat and Posteles on OC% only first, to give a better indication of its value range. This was done to limit
13 the amount of runs with carbon isotopes and kinetic fractionation, since the calculation time of these last two modules is high. Calibration of rExp was performed for values between 0.1 and 4, since values found in literature varied between 0.53 and 1.4. This yielded a value 1 with the lowest RMSE (of about 3) for Nabbegat (Figure 8a). Since values around rExp=1 do result in a relatively similar RMSE, a range of possible values for rExp is determined. This range was chosen objectively on the fit of the OC% simulations for each value of rExp with the measured values. Figure 9 displays all simulated OC% profiles for the calibration range of rExp, with the simulations of the best fit and lower and upper range highlighted. This resulted in a range for rExp between 0.9-1.7, with a RMSE range of 3-3.3 These values of rExp are considered to give an accurate simulation of the OC% profile in comparison to the measured values.
For Posteles however, unrealistically high values (20) of rExp seem to give a better fit (Figure 8b). This would mean that the influence of the r parameter would decrease to about 0 within the first 25 cm. It is assumed that since Posteles' literature run deviates strongly from the measured values, further calibration of rExp and h will not yield realistic values. We therefore chose to assume that the sedimentation OC inputs for Posteles are higher than for Nabbegat. This would give a better fit with the measured profile and enable the calibration with realistic values. Sedimentation inputs were increased by 1.6*1.6% stable and 1.3*2.5% labile OC to 2.5 % stable and 3.3 % labile OC. This was done under the assumption that the sediment input was about 1.5 larger for Posteles, but that there was relatively more stable OC present in the sediment. This resulted in the simulated profile in figure 10a, which has a smaller deviation from the measured values. Calibration of rExp resulted in figure 10bc. Despite the smaller deviation between measurements and simulation, the value of rExp tends to get a smaller RMSE at unrealistically low values.
14 Figure 10:abc Posteles profile simulation with 2.5% stable and 3.3% labile carbon (a). Calibration on rExp (b)
15
3.3 Calibration of rExp on δ13C
Calibration of rExp on the δ13C profile was only done for Nabbegat, since the calibration for Posteles did not yield a realistic parameter range. Calibration was performed on a range of 0.8-2 and is displayed in figure 11. The RMSE decreases linearly with the value of rExp and has the lowest RMSE at rExp = 0.8, at the lowest value of the calibration range and outside the 0.9-1.7 range found for the calibration on the OC% profile.
3.4 Calibration of rExp and h
rExp and h were calibrated simultaneously for Nabbegat and Posteles (Figure 12 and 13 respectively). Calibration was again first performed on the OC% profile only, to gain an indication of the optimal parameters range and save time during the calibration on the δ13C profile. Results of the calibration of Nabbegat show that the range in which rExp has a low RMSE increases from 0.9-1.7 to 1.0-4.0. In addition, the lowest RMSE is achieved for the uppermost value of the calibrated range and is thus likely to get a lower RMSE at unrealistically high values. The same accounts for the calibration of h, where unrealistically low values for h gave the lowest RMSE. A similar pattern can be derived from the calibration results of the Posteles profile. Values of rExp ranging between 0.1-4.5 all have a similar low RMSE and a lower RMSE occur at very low values for h and very high values for rExp. Since calibration of rExp and h did not result in a realistic range of parameter values for both profiles, it was chosen to not perform a calibration on the δ13C profiles.
16 Figure 12. Calibration of rExp and h for Nabbegat
Figure 13. Calibration of rExp and h for Posteles
4. Discussion
4.1. Parameters and Calibration results
As described in the results section, the calibration of rExp for Nabbegat was the only calibration which yielded a realistic parameter value with a relatively small range (0.9-1.7). Despite the relatively small deviation between the simulated and measured δ13C profile, calibration on the carbon isotopes did not narrow down this range, since the lowest RMSE value was found at the lower limit
17 of the calibration range (rExp=0.8). Extrapolating the linear decrease in the simulation, it is likely that the smallest RMSE is found at a value of rExp close to zero, which is not a realistic value. This is most likely explained by the small amount of δ13C data points. Only 5 measurements were available, of which 3 were located in the plaggic Anthrosol and 2 points were located in the buried podzol. For this research it can only be stated that the simulated isotope profiles do support the model fit qualitatively, but for quantitative support in the calibration more data points are required. A limited amount of δ13C data points will thus have a large effect on the accuracy of the calibration.
For Posteles, calibration using literature values did not result in a realistic parameter range, because of the large difference between measurements and simulation. However, even when OC inputs by sedimentation were assumed to be higher to provide a better fit, calibration did not result in an optimal parameter range. This is most likely due to the abrupt increase of OC between 0.4 and 0.6 m depth, since the model is not capable of simulating this strong increase using the given input parameters. It can thus be stated that an initial parameter set that gives roughly a good fit with the measured data is necessary for gaining realistic calibration values for parameters.
Calibration of rExp in combination with h yielded a larger range of rExp values within the same RMSE interval, and did thus not result in a smaller range of parameters. It can be stated that the performance of the DELTA-13C model is best using only one degree of freedom, when rExp and h are chosen for calibration. Additional calibration of combinations of other parameters has to be performed to further research this claim.
Parameter assumptions and uncertainties
Many assumptions were made for the input parameters used in these simulations, since literature sometimes did not give a decisive value of a parameter. The greatest uncertainty on which little literature is available is the OC% in the deposited plaggen and the contribution of manure enrichment in the stables. The carbon fraction in the plaggen originating from vegetation material, so without manure enrichment, was estimated at 3.5%. This was based on the OC content in the upper layer of a temperate forest soil (appendix II.d). However, no detailed studies on the enrichment of these sods by manure could be found. It was chosen to assume that manure added another 0.6% of OC to the sods, but this could be a too conservative assumption. In addition, assumptions have to be made for the division between labile and stabile carbon in the deposited plaggen. For the model, it was assumed that more labile than stabile carbon was present, since it consists of partly of fresh plant material and is located at the surface where it can be reached more easily by organisms and external climate factors. However, to the best of our knowledge, no studies on the composition of OC in plaggen material have been done to date. In the literature simulation 1.6% stable and 2.5 % unstable OC are used for the deposited plaggen material. To assess the sensitivity of the model to the stable/labile ratio, different ratios were simulated. The results (Figure 15) display that varying this ratio can make up for a difference of more than 1% in the OC%. These results show that additional research should be performed to determine this ratio since it is crucial in making the simulations more accurate.
Lichtaart: assess a profile with a high δ13C sampling frequency
In order to assess the influence of calibrating on a more frequently sampled δ13C dataset the soil profile of the plaggic anthrosol of Lichtaart was simulated, where δ13C was sampled with a frequency of 6 cm to a depth of 1.2 m (Figure 16ac). Since no OSL dating was available for the profile, the sedimentation was estimated to last 350 years with a rate of 0.2 cm/yr, which is the average sedimentation rate for Posteles and Nabbegat. The model simulates the OC% and δ13C profile to a good extent, with the exception of the top 40 cm of the isotope profile. This is however the ploughing layer, which explains the deviation from the simulation. In line with the Posteles en Nabbegat profiles, a considerable stable OC fraction is needed to simulate the OC% profile. Calibration of rExp on the OC and δ13C profile did however not result in a narrowed down parameter range. This could be caused by a overestimation of the sedimentation rate, since the measured OC% decreases at 1 m depth, while the simulation decreases at 1.4 m. This emphasizes that OSL dating is
18 useful and crucial in providing accurate sedimentation rate input in the simulations and gives additional evidence for the importance of accurate input data in order to gain accurate simulations.
Figure 17ab. Calibration of rExp on a) the OC profile and b) the δ13C profile for Lichtaart.
Figure 16ab. The a) OC% profile and b) the δ13C profile simulation for Lichtaart. Simulation done with literature parameters Figure 15. Different divisions in stable and labile carbon and accompanying simulation results for Posteles.
19
Soil compaction is not modeled
The DELTA-13C model does not take soil compaction into account, the BD at the surface is used throughout the whole soil profile. This could lead to an overestimation of OC% in the simulation at larger depths. BD is expected to increase with depth, due to soil compaction caused by the increased weight on top of the layer by adding new plaggen deposits. In addition, compaction occurs when organic matter is decomposed (Kirk et al., 2015). A higher bulk density with the same weight of OC will lead to a smaller OC%.
4.2 Differences between sediment and vegetation originating organic Carbon
For both profiles, sediment originating OC (carbon that is deposited with the plaggen material) had the most important share in the total quantity of SOC in the soil. For both literature runs, stable sediment associated OC accounts for about 1.5% of the simulated OC, while the labile fraction accounts for only 0.25%, which is about the same as the vegetation's contribution. The share of above ground and below ground vegetation originating OC decreases from 0.5% during the spin up run to about 0.25% during the deposition phase, due to the lower OC inputs of crops in comparison to forests. BG originating OC contributes more than AB vegetation, which can be explained by the removal of AB crop vegetation. When taking carbon age and turnover times into consideration, sediment associated OC is in general much older than vegetation associated OC, which is due to the considerable stable fraction in sediment originating OC.
In addition, stable sediment OC has a considerable higher half life than vegetation OC, but the labile sediment fraction has a half life which is the same as the vegetation OC half life of 2 years. In these simulations, 1% of labile sediment OC fraction thus has the same effect as 1% of vegetation associated OC. The stable sediment OC half life could not be quantified from the model results, but since the cumulative fraction left for both profiles was about 0.75 after 324 simulation years, it can be assumed that the half life of the stable sediment OC fraction will certainly be in the order of centuries and even of millennia. The residence times of the OC fractions could not be calculated, since the model used has a Nonautonomous, multiple pool and nonlinear structure, for which no methods are available yet to quantify residence times (Sierra et al., 2016).
It can therefore be concluded that our model results suggests that the stable sediment OC fraction forms a considerable share of preserved OC in these soil profiles. In addition, the results indicate that the simulated OC would be greatly underestimated when only locally produced OC is taken into account.
4.3 Differences between profiles
Despite the differences in OC% for the literature run and differences in calibration, a striking difference between profiles occurs in the cumulative OC fraction left after deposition: the total OC fraction left is 0.12 higher for Nabbegat than for Posteles and the half life of the cumulative OC fraction of Posteles is 7 years against 145 years for Nabbegat. This difference is striking, since the decrease in vegetation and sediment OC fractions seems almost similar for both profiles. In addition, no difference between the AG vegetation, BG vegetation and Labile sediment OC half life of both profiles was observed. A possible explanation could be the higher sedimentation rate in the last 200
20 years for Nabbegat which is 0.5 cm/yr against 0.2 cm/yr for Posteles. This could lead to the burial and better preservation of the OC in the profile. In addition, the sedimentation rate in the first 100 years of deposition is three times as high for Nabbegat, which explains the higher half life of the total OC, since more stable sediment OC with a high half life was added than was added to the Posteles profile.
The AB and BG originating OC is older for Posteles than for Nabbegat, due to the 159 year longer duration of the deposition phase for Posteles. The average OC age is however higher for Nabbegat, which is explained by the higher initial age of the OC deposited in the plaggen (1816 for Nabbegat against 1467 for Posteles). Since the stable plaggen fraction accounts for the greatest share of the total OC%, this strongly influences the age of the total OC content. However, figure 6ab displays that the modeled OC sediment age exceeds the measured C14 age. This is likely due to an overestimation of the initial sediment OC age, which was calculated as the mean age difference between the OSL and C14 ages of each profile (appendix II.d). For both profiles, two depth points with both OSL and C14 age were available, but the age difference between these two points is 1300 for Posteles and 2000 for Nabbegat. Taking the mean of these two points results in an overestimation of the initial age of the deposited plaggen carbon. More data points, frequently sampled though the soil profile are necessary for a better estimation.
Impact of current vegetation on the measured organic carbon concentration
The large deviation between the modelled and measured OC% profile for Posteles could be due to inputs by current vegetation, since the biomarker diagram (Appendix II) displays the abundance of corn (Zea mays) through the soil profile. The abundance of corn is however not displayed in the pollen diagram, which is an indication of significant corn root input. Corn was introduced in Dutch agriculture around 1950 AD, when plaggen agriculture was not practised anymore (Van Mourik et al., 2016). Since corn has a rooting depth up to 2.4 m (Canadell et al. 1996) corn roots are likely to have contributed to the total SOC content of the plaggen layer. This is however not reflected in the carbon isotope signal, which is around -12‰ for Zea Mays (Yin & Raven, 1998) and is expected to be higher than around -28‰.
OC input by corn roots does however not explain the abrupt increase in carbon (from 3 to 5%) in between 0.4 and 0.6 m depth. A possible explanation could be the that the increase starts at the bottom of the plough layer (0.45 cm depth). According to Fliessbach et al. (2007) OC% in the plough layer can decrease with 20% in 20 years for soils without fertilisers and 16% when using fertiliser. The initial OC% in the upper 45 cm could thus have been higher initially and decreased because of the ploughing. A reduction of 22% would however account for a reduction of 1% OC, which is less than the observed 2%. In addition, the research by Fliessbach et al. (2007) was done on a luvisol, which has a higher clay and loam content than a plaggic anthrosol, which could influence SOC dynamics (Schmidt et al., 2011; Lehman & Kleber, 2015). Another explanation could be that hardly decomposing black carbon from historical burning is present in the soil layer between 0.4 and 0.6 metres. However, more observations and analysis on the Posteles profile is required to provide a more decisive theory on the high OC content in the 0.4-0.6 m layer.
5. Conclusion
It can be concluded that the DELTA-13C model is able to simulate the dynamics of the carbon profile in the plaggic anthrosols of Nabbegat to a considerable extent and that calibration of rExp on the OC% profile improves the simulation. However, the parameter range of rExp is not decreased by calibration on the δ13C profile, which is likely due to the small amount of isotope data points. The use of high frequency sampled δ13C data is expected to increase the simulation and calibration accuracy. In addition, calibration on both OC and δ13C does not work for the plaggen profile at Posteles, since the model dynamics are not able to capture the shape of the OC curve using the available input data. This shows that if the simulation results do not approach the measured values,
21 calibration of rExp does not result in a realistic range and emphasizes that the performance of the model is greatly dependent on the input data. This is supported by the assessment of the profile of Lichtaart, where the absence of OSL dating creates uncertainties on the sedimentation rate input. In addition, calibration of both rExp and h did not result in a realistic parameter range too, so it can be concluded that the DELTA-13C model works best with only one degree of freedom for simulation of plaggic anthrosols. Differences between the profile occur in carbon age and the cumulative fraction left and are due to differences in the estimated initial sediment OC age and in deposition rates. Assumptions considering the impact of manure on the sods and the stabile/labile OC ratio are considered to be the largest and most important uncertainties in the model. Varying this ratio in the Posteles soil profile demonstrates that the ratio can account for a difference of more than 1% OC in the profile. Additional research to these properties of plaggen are expected to increase the quality of the input data and lead to more accurate model simulations of plaggen profiles. The simulations of Posteles, Nabbegat and Lichtaart do however indicate that the presence of a considerable stable sediment fraction is crucial for gaining the high OC concentrations measured in the soil profiles. These concentrations will be highly underestimated when only locally produced OC is taken into account. For both profiles, the half life of the OC from vegetation and labile sediment very short (labile sediment & AG vegetation 2 years, BG vegetation 2.5 years). The half life of the stable sediment fraction cannot be quantified, because the simulation times are too short, but is expected to be in the order of centuries or millennia. Taking into account this result together with the large abundance of stable sediment OC it can be concluded that this simulation suggest that the high OC content in plaggic anthrosols can be explained by a large input of a stable, sediment originating OC fraction which decomposes slowly on a time scale of at least centuries.
Acknowledgements
I would like to thank Boris Jansen and Marijn Van de Broek for their good supervision during the writing of this thesis. Their useful feedback on presentations and concept versions as well as their quick responses to questions have been valuable input while conducting the research. I would further like to thank Jan van Mourik for providing the isotope data of the soil profiles.
22
References
Andren, O. & Kätterer, T. (1997). ICBM: the introductory carbon balance model for exploration of soil carbon balances. Ecological Applications, 7(4), 1226-1236.
Blume, H. P. & Leinweber, P. (2004). Plaggen soils: landscape history, properties, and classification. Journal of Plant Nutrition and Soil Science, 167(3), 319-327.
Bolinder, M. A., Angers, D. A. & Dubuc, J. P. (1997). Estimating shoot to root ratios and annual carbon inputs in soils for cereal crops. Agriculture, ecosystems & environment, 63(1), 61-66.
Canadell, J., Jackson, R. B., Ehleringer, J. B., Mooney, H. A., Sala, O. E.,& Schulze, E. D. (1996). Maximum rooting depth of vegetation types at the global scale. Oecologia, 108(4), 583-595.
De Boer, G. (2013).The origin of soil organic carbon in the plaggic deposits of the Posteles. A reconstruction of the carbon sources based on patterns in n-alkanes and n-alcohols (biomarkers). Bsc Thesis: University of Amsterdam.
Campforts, B., Vanacker, V., Vanderborght, J., Baken, S., Smolders, E. & Govers, G. (2016). Simulating the mobility of meteoric 10 Be in the landscape through a coupled soil-hillslope model (Be2D). Earth and Planetary Science Letters, 439, 143-157.
Fließbach, A., Oberholzer, H. R., Gunst, L. & Mäder, P. (2007). Soil organic matter and biological soil quality indicators after 21 years of organic and conventional farming. Agriculture, Ecosystems & Environment, 118(1), 273-284.
Freyerová, K. & Šefrna, L. (2014). Soil organic carbon density and storage in podzols–a case study from Ralsko region (Czech Republic). AUC GEOGRAPHICA, 49(2).
Giani, L., Makowsky, L. & Mueller, K. (2014). Plaggic Anthrosol: Soil of the Year 2013 in Germany: An overview on its formation, distribution, classification, soil function and threats. Journal of Plant Nutrition and Soil Science, 177(3), 320-329.
Han, L., Sun, K., Jin, J. & Xing, B. (2015). Some concepts of soil organic carbon characteristics and mineral interaction from a review of literature. Soil Biology & Biochemistry, 94, 107-121.
Jackson, R. B., Canadell, J., Ehleringer, J. R., Mooney, H. A., Sala, O. E. & Schulze, E. D. (1996). A global analysis of root distributions for terrestrial biomes. Oecologia, 108(3), 389-411.
Jobbágy, E. G. & Jackson, R. B. (2000). The vertical distribution of soil organic carbon and its relation to climate and vegetation. Ecological applications, 10(2), 423-436.
Johnson, J. F., Allmaras, R. R. & Reicosky, D. C. (2006). Estimating source carbon from crop residues, roots and rhizodeposits using the national grain-yield database. Agronomy journal, 98(3), 622-636. Johnson, M. O., Mudd, S. M., Pillans, B., Spooner, N. A., Keith Fifield, L., Kirkby, M. J. & Gloor, M. (2014). Quantifying the rate and depth dependence of bioturbation based on optically‐stimulated luminescence (OSL) dates and meteoric 10Be. Earth Surface Processes and Landforms, 39(9), 1188-1196.
Lehmann, J. & Kleber, M. (2015). The contentious nature of soil organic matter. Nature, 528(7580), 60-68.
23 Garten, C. T. & Taylor, G. E. (1992). Foliar δ 13 C within a temperate deciduous forest: spatial, temporal, and species sources of variation. Oecologia, 90(1), 1-7.
Kirk, E. R., van Kessel, C., Horwath, W. R. & Linquist, B. A. (2015). Estimating annual soil carbon loss in agricultural peatland soils using a nitrogen budget approach. PloS one, 10(3), e0121432.
Martens, D. A. (2000). Plant residue biochemistry regulates soil carbon cycling and carbon sequestration. Soil Biology and Biochemistry, 32(3), 361-369.
Menichetti, L., Kätterer, T. & Leifeld, J. (2016). Parametrization consequences of constraining soil organic matter models by total carbon and radiocarbon using long-term field data. Biogeosciences, 13(10), 3003-3019.
O'Leary, M. H. (1988). Carbon isotopes in photosynthesis. Bioscience, 38(5), 328-336. Pape, J. C. (1970). Plaggen soils in the Netherlands. Geoderma, 4(3), 229-255.
Roberts, J., Hopkins, R. & Morecroft, M. (1999). Towards a predictive description of forest canopies from litter properties. Functional Ecology, 13(2), 265-272.
Rompaey, A. J. V. & Govers, G. (2002). Data quality and model complexity for regional scale soil erosion prediction. International Journal of Geographical Information Science, 16(7), 663-680.
Rumpel, C., Kögel-Knabner, I. & Bruhn, F. (2002). Vertical distribution, age, and chemical composition of organic carbon in two forest soils of different pedogenesis. Organic Geochemistry, 33(10), 1131-1142.
Schmidt, M. W., Torn, M. S., Abiven, S., Dittmar, T., Guggenberger, G., Janssens, I. A. Et al. (2011). Persistence of soil organic matter as an ecosystem property. Nature, 478(7367), 49-56.
Sierra, C. A., Müller, M., Metzler, H., Manzoni, S., & Trumbore, S. E. (2017). The muddle of ages, turnover, transit, and residence times in the carbon cycle. Global change biology, 23(5), 1763-1773. Sparks, D. L. (2003). Environmental soil chemistry. Academic press.
Tonneijck, F. H., van der Plicht, J., Jansen, B., Verstraten, J. M. & Hooghiemstra, H. (2006). Radiocarbon dating of soil organic matter fractions in Andosols in northern Ecuador. Radiocarbon, 48(3), 337-353.
Van de Broek, M., Wang, Z. & Gerard G. (in preparation). Assessing the potential of stable carbon isotopes to better constrain the residence time of SOC.
Van Mourik, J.M., Slotboom, R.T. & Wallinga, J. (2011). Chronology of plaggic deposits; palynology, radiocarbon and optically stimulated luminescence dating of the Posteles (NE-Netherlands). Catena, 84, 54-60.
Van Mourik, J.M, Nierop, K.G.J. & Vandenberghe, D.A.G. (2010). Radiocarbon and optically stimulated luminescence dating based chronology of a polycyclic driftsand sequence at Weerterbergen (SE Netherlands). Catena, 80, 170-181.
24 Van Mourik, J.M. & Jansen, B. (2013). The added value of biomarker analysis in palaeopedology; reconstruction of the vegetation during stable periods in a polycyclic driftsand sequence in SE-Netherlands. Quaternary International, 306, 14-23.
Van Mourik, J.M., Wagner, T.V. de Boer, G & Jansen, B. (2016). The added value of biomarker analysis to the genesis of plaggic Anthrosols; the identification of stable fillings used for the production of plaggic manure. SOIL, 2, 299–310.
Wagner, T.V. (2012). De toepasbaarheid van combinaties van lipideketens uit plaggenbodems als biomarker bij het reconstrueren van vegetatiesuccesie op de Nederlandse zandgronden. Bsc thesis: University of Amsterdam.
Wang, Z., Van Oost, K. & Govers, G. (2015). Predicting the long-term fate of buried organic carbon in colluvial soils. Global Biogeochem. Cycles, 29, doi:10.1002/2014GB004912.
Yin, Z. H., & Raven, J. A. (1998). Influences of different nitrogen sources on nitrogen-and water-use efficiency, and carbon isotope discrimination, in C3Triticum aestivum L. and C4Zea mays L. plants. Planta, 205(4), 574-580.
Yevdokimov, I. V., Ruser, R., Buegger, F., Marx, M., & Munch, J. C. (2007). Interaction between rhizosphere microorganisms and plant roots: 13 C fluxes in the rhizosphere after pulse labeling. Eurasian Soil Science, 40(7), 766-774.
25
Appendices
I. Preliminary data
Figure I.1 Pollen diagram Nabbegat (van Mourik et al., 2016)
26
Figure I.3 Reconstruction of historical vegetation inputs in the Nabbegat soil profile using Biomarkers (van Mourik et al., 2016)
Figure I.4 Reconstruction of historical vegetation inputs in the Posteles soil profile using Biomarkers (van Mourik et al., 2016)
27 Table I.1 Measured Carbon content Nabbegat (Wagner, 2012)
C% Nabbegat depth (cm) C% depth (cm) C% 55-60 0,92 100-105 1,65 60-65 1,5 105-110 1,47 65-70 1,53 110-115 1,48 70-75 1,58 115-120 1,66 75-80 1,52 120-125 1,73 80-85 1,74 125-130 1,59 85-90 1,62 130-135 1,52 90-95 1,54 135-140 1,52 95-100 1,53
Table I.2 Measured Carbon content Posteles (De Boer, 2013) C% Posteles depth (cm) C [%] 10 3,06 20 3,12 30 3,185 40 3,445 50 4,96 60 5,49 70 5,49 80 3,705 90 1,585 100 0,745
Table I.3. Radiocarbon and OSL dating and Δ13C signal of the Nabbegat profile (van Mourik et al., 2016) Nabbegat
Horizon Depth (cm) Calendric 14C ages humin Calendric 14C ages humic acicds Calendric OSL ages δ13C signal Hum δ13C signal Hac
C 70 NaN NaN AD 1803+12 NaN NaN
2An 80 AD 428+107 AD 626+45 AD 1770+11 -27,1 -27,79
2An 105 37+133 BC AD 3+101 NaN -26,99 -27,59
2An 130 1182+139 BC 811+101 BC AD 1676+14 -27,25 -27,59
3ABp 140 NaN 1299+78 BC NaN NaN -28,45
3ABp 150 NaN 1385+72 BC NaN NaN -28,18
Table I.4. Radiocarbon and OSL dating and Δ13C signal of the Posteles profile (van Mourik et al., 2016) Posteles Horizon Depth (cm) Calendric 14C ages humin Calendric 14C ages humic acicds Calendric OSL ages δ13C signal Hum δ13C signal Hac
Aan 45 NaN NaN AD 1758±14 NaN Nan
Aan 59 NaN NaN AD 1711±20 NaN Nan
Aan 70 AD 1132± 68 AD 1172±51 AD 1651±31 -28,49 -27,89
Aan 82 NaN NaN AD 1626±20 NaN Nan
Aan 95 AD 884±82 AD 861±85 AD 1571±31 -29,16 -28,24
28
II. Elaboration on methods and calculations of parameters
All preliminary data referred to in these methods can be found in appendix I.
a. Spin up run parameters
The spin up run is used to create a soil profile on which the plaggen deposits can be modelled. No sedimentation takes place during the spin up run. In the simulation, a spin up run of 500 years was performed, which created a soil profile with a depth of 3 metres.
During the spin-up phase, forest vegetation is assumed, based on pollen diagrams of the drift sand layers. The buried soils in both profiles display the abundance of Alnus (alder), Corylus (hazel) and Querques (oak) pollen. To simulate this during the spin-up run, parameters for a temperate deciduous forest are used. An overview of articles that research δ13C signals in leaf and tree rings points out that most publications consider data on oaks and pinus (Gessler et al., 2009). It is therefore chosen to use δ13C data of oak leafs and roots. The value of leafs is estimated to be -28.5 ‰ (Garten & Taylor, 1992). Roots are in general 2 ‰ more positive than leave tissue (O'Leary, 1988) and are thus estimated on -26.5 ‰.
b. Bulk density
Since no bulk densities were measured for both profiles, an estimation has to be made. This estimation is based on bulk density values of podzols and plaggic soils in literature. Pape (1970) describes two profiles of plaggic anthrosols in the Netherlands. Bulk density values in these profiles range from 1.18 to 1.48 g/cm3. Another research by Freyerová & Sefrna (2014) measured bulk densities ranging from 1.053 g/cm3 in the Ah horizon of a podzol to 1.438 g/cm3 in the Bs horizon. Based on these numbers, the bulk density for the simulations is estimated at 1.3 g/cm3. It is assumed that the deposited material has the same bulk density as the soil and that bulk density is the same for the whole profile.
c. Vegetation Stages
Biomarkers are used to establish the vegetation stages of the model, which determine the input parameters for aboveground and root vegetation. In this simulation, only one vegetation state will be used during the deposition run. This choice was made for the sake of simplicity. Biomarkers found in the plaggen layers are originating from the plaggen deposits as well as from the in situ vegetation. Since plaggen soils were used as agricultural fields, biomarkers of crops are expected to origin from in situ vegetation, while biomarkers of heather and trees are likely to origin from the plaggen.
The biomarker diagram of the Posteles profile shows a large "contamination" of Zea Mays, which was only introduced after 1950 has deep roots. It can therefore be stated that biomarkers of original vegetation and plaggen deposits are suppressed by the Zea Mays roots (Van Mourik et al., 2016). When ignoring the Zea Mays, the crop species that occurs on most depths is Avena Sativa (Oat) and its properties will therefore be used as input parameters for above and belowground vegetation.
The Nabbegat profile displays both markers of oat and rye. For the sake of simplicity it is chosen to only model oat in the simulation.
d. Properties of deposited plaggen
An estimation of multiple properties of plaggen layers has to be made for this simulation. This concerns the percentages OC in the deposits, the δ13C signal of the OC in the deposits and the age of the deposited OC. Since the biomarker diagrams of both profiles show the abundance of Quercus Robur (oak) on most researched depths, it is assumed that most deposits were forest plaggen from
29 an oak forest. The initial δ13C signal was estimated on -28.5 and -29.5 for labile (oak leafs) and recalcitrant (oak litter) OC respectively (Garten & Taylor, 1992; Roberts et al., 1999).
The Carbon content of the plaggen deposits is estimated based on the carbon content of a litter layer of a temperate deciduous forest, since most plaggen deposits are thought to be of forest origin. The OC% is calculated assuming a SOC content of 17.4 kg/m2 in 0-1 m of the soil profile, and assuming 52% of the SOC in this 0-1 m is present in the upper 20 cm. Combined with a bulk density of 1.3 g/cm3, this results in a OC% of 3.5% in the forest plaggen deposits. No information could be found considering the division between stable and unstable OC and the effect of manure enrichment on the plaggen. Therefore, it was chosen to assume that about 1/3 of the forest plaggen (1.2%) consisted of stable OC and the other 2.2% consisted of unstable OC. In addition, the enrichment by manure was estimated to be 0.6% (which is an enrichment of 1/6 compared to the 3.5% vegetation OC in the plaggen). Of this 0.6% manure enrichment, 0.3% was estimated to be stable and 0.3% was estimated to be unstable. In total, this resulted to an estimated OC content of 1.6% stable and 2.5% unstable OC in the deposited plaggen.
The initial age of the deposited plaggen was estimated on the difference between the OSL age and C14 ages of the humic acid fraction. This difference was calculated for all depth points with an OSL and C14 dating. Thereafter, the mean was calculated and used as initial age (Table II.1, II.2). Table II.1 calculation of initial plaggen OC age Nabbegat
Nabbegat
Depth (cm) OSL age C14 age OC age (yr)
80 1770 AD 626 AD 1144
130 1676 AD 811 BC 2487
Mean 1816
Table II.2 calculation of initial plaggen OC age Posteles
Posteles
Depth (cm) OSL age C14 age OC age (yr)
70 1651 AD 1172 AD 479
95 1571 AD 861 AD 2455
Mean 1467
e. Carbon input by vegetation
During the simulation, the input of OC to the soil is determined by both the above ground (AG) and below ground (BG) vegetation for which the properties of oat are used as input parameters. Since no data is was found considering the biomass production of oat on Dutch plaggic agricultural fields during the 16th-20th century, an estimation has to be made. For this estimation, AG and BG biomass data of oat in a Canadian research by Bolinder et al. (1997) are used. Since these yields were obtained using modern agricultural techniques, it is chosen to lower these by 30%, in order to simulate yields without these techniques. It is furthermore assumed that the total grain yield and 85% of the straw yield are removed from the land and to not contribute to the OC input, in line with the assumptions made in Bolinder et al. (1977). Vegetation inputs are displayed in table II.3. Table II.4 displays the carbon fractions of the AG and BG biomass.
30 Table II.3 OC input from oat vegetation during plaggen deposition phase. Data are derived from Bolinder et al. (1977). All numbers are dry biomass, in which root biomass is calculated using a Shoot:Root ratio of 2.5.
Grain yield (g m-2) Straw Yield (g m-2) Root biomass (g m-2) AG biomass (g m-2) Bolinder et al. (1997) 337 501 339 75 30% reduction 236 351 237 53
Table II.4. OC fractions of AG and BG vegetation biomass. An C content of 40% is assumed (Bolinder et al., 1997; Martens, 2000; Johnson et al., 2006).
Biomass (g m-2) C fraction C input (g m-2)
AG vegetation 53 0.4 21
BG vegetation 237 0.4 95
III. Overview of used parameter values and accompanying references for
Posteles
Spin up parameters Posteles
Parameter Matlab name Valu
e Uni t Based on Sedimentatio n rate CarbonInputData.Sed_rate_spinup 0 m yr-1 No sedimentatio n in spin up Exponential coefficient to calculate exponential decrease of r with depth
CarbonInputData.rexp_spinup 1.4 - Wang et al.
(2015) Value of r at the surface CarbonInputData.r0_spinup 0.75 yr-1 Wang et al. (2015) δ13C signal of above ground vegetation
CarbonInputData.δ13C_AGveg_spinup_input -28.5 ‰ Garten &
Taylor (1992)
δ13C signal of below ground vegetation
CarbonInputData.δ13C_BGveg_spinup_input -26.5 ‰ Garten &
Taylor (1992); O'Leary (1988)
Root biomass CarbonInputData.rootBiomass_spinup 4200 g
m2
Jackson et al. (1996)
31 rooting depth (1996) Fraction of roots above max rooting depth CarbonInputData.fractionAboveMaxDepth_sp
inup 0.95 - Jackson et al. (1996)
% of root biomass that is OC input CarbonInputData.Root_OC_perc_spinup 0.4 0-1 Bolinder et al. (1997) Turnover time of root originating OC CarbonInputData.Root_turnover_spinup 0.1 Yr-1 Gill & Jackson (2000) Above grond Vegetation Biomass CarbonInputData.Veg_biomass_spinup 1826 0 g m2 Jackson et al. (1996) % of vegetation biomass that is OC input
CarbonInputData.Veg_OC_perc_spinup 0.001 0-1 Jackson et al.
(1996) Biodiffusion coefficient CarbonInputData.Db0_spinup 0.98 cm -2 yr-1 Johnson et al. (2014) Decrease of biodifussion coefficient with depth CarbonInputData.DbExp_spinup 0.28 cm -1 Johnson et al. (2014)
General parameters Posteles
Parameter Matlab name Value Unit Based on
Amount of vegetation stages
CarbonInputData.Veg_stage_number 1 - Biomarkers and
pollen diagram % of sediment mass that is recalcitrant OC input
CarbonInputData.Sed_oc_perc 0.016 - Jobbágy &
Jackson (2000) % of sediment mass that is labile OC input
CarbonInputData.POM_oc_perc 0.025 - Jobbágy &
Jackson (2000) Initial δ13C signal of recalcitrant sediment OC
CarbonInputData.δ13C_sed_oc_init -29.5 ‰ Roberts et al.
32 Initial δ13C signal of labile sediment OC
CarbonInputData.δ13C_POM_init -28.5 ‰ Roberts et al.
(1999); Garten & Taylor (1992) Humificatio n parameter for Y to O pool CarbonInputData.h 0.125 yr-1 Andrén & Kätterer (1997) r at the surface CarbonInputData.r0 0.75 yr-1 Wang et al. (2015) Exponential decrease of r with depth CarbonInputData.rexp 1:4 - Wang et al (2015)/calibrati on K1 parameter for Y pool CarbonInputData.k1 0.8 yr-1 Andrén & Kätterer (1997) K2 parameter for O pool CarbonInputData.k2 0.006 yr-1 Andrén & Kätterer (1997) K1 for above ground vegetation CarbonInputData.k1_AGveg 0.8 yr-1 Andrén & Kätterer (1997) K1 for belowgrou nd vegetation CarbonInputData.k1_BGveg 0.8 yr-1 Andrén & Kätterer (1997) K1 for recalcitrant OC in sediment CarbonInputData.k1_sed 0.8 yr-1 Andrén & Kätterer (1997) K1 for labile OC in sediment CarbonInputData.k1_POM 0.8 yr-1 Andrén & Kätterer (1997) Humificatio n parameter for O to R pool CarbonInputData.h2 0.012 5 yr-1 Menichett et al. (2016) K parameter for R pool CarbonInputData.kR 0.002 yr-1 Menichetti et a. l (2016) Number of seperate OC inputs CarbonInputData.Number_of_pools_depos
ition 4 - Labile Sediment,
recalcitrant sediment aboveground
33 veg, belowground veg Biodiffusio n coefficient at the surface CarbonInputData.Db0_ActualRun 0.98 cm-2 yr -1 Beauchard et al. 2012 Decrease of biodiffusio n coefficient with depth CarbonInputData.DbExp_ActualRun 0.28 cm-1 Campforts et al. (2016), Johnson et al. 2014) Age of OC in sediment CarbonInputData.sedOCage 1467 yr Mean difference between OSL age and 14C age on same depth Multiple crops: yes or no CarbonInputData.multipleCrops 1 boolea n
Only ones crop
Amount of crops CarbonInputData.numberOfCrops 1 - Put priming on or off CarbonInputData.primeOldCarbon 0 - Carbon in recalcitrant pool: yes or no CarbonInputData.sedOCInRecalcPool = 1 1 boolea n Carbon from sediment goes straight in recalcitrant pool
Calibration parameters Posteles
Paramete r
Matlab name Value Unit Base
d on Range for rexp CarbonInputData.rexp_range 0.1:0.75:2.3 5 yr -1 Range for h CarbonInputData.h_range 0.1:0.25:1.1 Put calibratio n on/off CarbonInputData.CalibrateRexp 1 Boolea n Put calibratio n on/off CarbonInputData.CalibrateH 1 Boolea n
