Amazonian podzols store huge amounts of carbon and play a key role
in transferring organic matter to the Amazon River. In order to better
understand their C dynamics, we modelled the formation of representative
Amazonian podzol profiles by constraining both total carbon and radiocarbon.
We determined the relationships between total carbon and radiocarbon in
organic C pools numerically by setting constant C and
Podzols are soils characterized by the formation of a sandy, bleached horizon (E horizon) overlying a dark horizon with illuviated organic matter as well as Fe and Al compounds (spodic or Bh horizon). In wet tropical areas podzols can be very deep, with E horizons thicker than 10 m and Bh horizons thicker than 4 m (Chauvel et al., 1987; Dubroeucq and Volkoff, 1998; Montes et al., 2011). This means that they can store huge quantities of organic matter: Montes et al. (2011) estimated the C stocks in Amazonian podzols to be around 13.6 Pg C.
This C constitutes a non-negligible portion of the C stored in the Amazonian
basin. Indeed, the carbon stored in the aboveground live biomass of intact
Amazonian rainforests is estimated to be 93
Schematic of the main C fluxes in a podzol.
The main characteristics of the podzol profiles used in the study. C
stock and ages are given
Location of the studied profiles. Grey areas in the detailed map indicate hydromorphic podzol areas. Orange spots identify test areas.
In order to better understand the fluxes of C in Amazonian podzols and in particular to determine the rate of carbon accumulation in Bh horizons during podzol genesis, the size of the C fluxes to rivers via both the perched and deep water tables and the vulnerability of the podzol C stocks to potential changes in the moisture regime due to global climate change, four representative podzol profiles from the high Rio Negro basin were used to constrain a model of C fluxes. The high Rio Negro basin was chosen because it is a region that has the highest occurrence of podzol in the Amazon (Montes et al., 2011) (Fig. 2). The four representative profiles were selected from a database of 80 podzol profiles issued from 11 test areas which have been studied in detail and of which 11 have been dated; this database will be the subject of a further publication. The four profiles were used to constrain the simulations of C fluxes. We used a system dynamics modelling software package (Vensim) to simulate the formation of representative Amazonian podzol profiles by constraining both total carbon and radiocarbon with the data collected.
Four podzol profiles were selected from our database as representative both
from the point of view of the profile characteristics and the
Sketch of the studied profiles.
Soil samples were analysed for C content with a TOC-LCPN SSM-5000A, Total Organic Carbon Analyzer (Shimadzu). Radiocarbon measurements were carried out at the Poznań Radiocarbon Laboratory, Poland. We assumed that the proportion of bomb carbon in the Bh organic matter was negligible and calculated a conventional, uncalibrated age from the radiocarbon pMC (percent modern carbon) value. As the Bh organic matter is an open system mixing organic carbon of different ages, this age is an apparent age. Samples from the topsoil had a pMC higher than 100 %, which indicates that a significant part of the carbon in the topsoil is post-bomb and therefore should not be neglected. Assuming that the topsoil horizons reached a steady state before 1950, we retrocalculated the pre-1950 pMC value of these samples using a dedicated model described in Sect. 2.2.
The data given in Table 1 were calculated by linear extrapolation of values measured on samples taken at different depths: between 11 and 28 samples per profile were used for the C stocks' calculation and between 6 and 8 samples per profile were used for radiocarbon measurements.
We used an approach comparable to previous studies which dealt with carbon budgets and radiocarbon data (e.g. Baisden et al., 2002; Menichetti et al., 2016; Sierra et al., 2013, 2014; Tipping et al., 2012). The model structure, based on the schematic shown in Fig. 1, and the names of compartments and rate constants are given in Fig. 4. As the turnover time of the OM in the topsoil horizons is short relative to the average OM turnover time in the Bh, only one topsoil carbon pool was used, whereas two pools (fast and slow) were used to describe organic carbon dynamics in the Bh horizon. The C can leave the topsoil pool by mineralization, transfer to the Bh pools or via the river by the perched water table; it can leave the Bh pools by mineralization, transfer to the river by the perched water table or via the deep water table. We chose to neglect the flux of C from the fast Bh pool to the slow Bh pool in order to facilitate the numerical resolution of the system comprising equations describing both the carbon and radiocarbon contents.
The equations describing changes in the carbon content of the different
pools are presented below (see Fig. 4 to see the fluxes with which each rate
constant is associated):
The equations describing changes in the radiocarbon content of the different
pools are the following:
Model design.
Evolution of the
With regard to the apparent age of the topsoil organic matter enriched in
post-bomb carbon, we considered a single pool that reached a steady state
before 1955 (Fig. 5), which allowed the retrocalculation of the radiocarbon
fraction
An underlying assumption of this work is that soil formation processes
remained constant over time. An alternative assumption might be, for example,
that all the Bh organic matter had accumulated in a very short time, after
which the Bh was no longer subjected to external exchanges. This scenario
could also produce profile ages close to the observed
We used the Vensim® Pro (Ventana Systems inc.) dynamic modelling software to simulate the C dynamics. After setting the initial values for C pools, the model was run in the optimize mode, leaving the model to adjust the rate constants in order to minimize the difference between simulated and measured C pool values and ages. However, frequently the model did not converge when run in this way. We found that it was because of the great difference between the convergence times between the topsoil C pool and the slow Bh C pool. The long times required to model the genesis of the Bh horizons resulted in numerical errors when modelling the topsoil behaviour, because the values of exponential exponents exceeded the maximum values that the computer could handle (see for example Eq. 12 below). To circumvent this technical problem, we optimized the model separately for the topsoils and for the Bh horizons, and we found that at the timescale of the formation of Bh, the topsoil C pool and the topsoil C fluxes to river and Bh horizons could be considered constant.
Although the model structure in Fig. 4 contains two C pools in the Bh horizon, we calculated the numerical solutions of equations considering both carbon budget and radiocarbon age for a single-pool Bh in order to determine whether the model could be simplified. Furthermore, this approach allowed us to better assess the weight of the different rate constants in the long-term behaviour of a given pool. The calculation in the simplified configuration is shown in Fig. 6.
Simplified design for one pool.
In this configuration, the carbon content of the pool is given by
Considering that the C input from the topsoil to the Bh and its radiocarbon
fraction are constant with time, it comes from the two previous equations:
This section presents conceptual results on the basis of the simplified
diagram given in Fig. 6 and in which the flux leaving the Bh is described by
a single rate
Unsurprisingly, the greater the difference between input and output C fluxes,
the faster a given
When the model is constrained only by the measured values of C stocks, a
number of solutions are possible (Fig. 7). The example given in Fig. 7 is
based on data from the P7C profile (Table 1). Curves 1 and 2 describe the
evolution of
Single-pool modelling of
Single-pool modelling of both
The currently observed C stock can be reached in a shorter time, however, if
for a given input flux the value of
Results of simulation for a single-pool Bh: minimum genesis time and time to steady state.
When the model was constrained by both carbon stock and
The simulation of the minimum time required for the observed carbon stock and
The minimum time required for the C stock and
Relationship between the
Taking into account the maximum absolute error does not significantly change the simulation results: the maximum absolute error in the genesis times is lower than 1.0, 0.9, 3.5 and 2.9 % for MAR9, DPQT, UAU4 and P7C, respectively. Since such percentages do not alter the orders of magnitude and trends discussed below, the error will not be considered in the following.
Modelling the topsoil horizons.
The time taken for the Bh horizon of a given profile to form is likely
between the two values shown in Table 2 and Fig. 9. The minimum time required
for obtaining C stock and
Effect of the fast Bh pool size on the whole Bh genesis time and
the
As explained in Sect. 2.3, the topsoil horizons were modelled separately
because the time needed to reach a steady state is very much shorter for the
topsoil horizons than for the Bh horizons. The steady-state condition was
given by
Effect of constraining the output C fluxes from the Bh on the genesis time. UAU4: effect of the fast Bh output flux. MAR9 and P7C: effect of the slow Bh output flux.
The results suggest that the topsoil OM in the four profiles needed only
between 400 and 700 years to reach a steady state, if the present-day
topsoils are indeed in a steady state. The total C flux through the topsoil
(
The partitioning of the C flux leaving the topsoil between the river (rate
We therefore carried out a sensitivity analysis to determine how the main parameters (size of the fast pool of the Bh, C flux input and output C rates for the Bh pools) affected the profile genesis time and to understand the relationships between these parameters.
Modelled C fluxes,
The conclusion of this sensitivity study is that, when the size of the fast Bh pool or the C output fluxes from the Bh pools begins to grow from zero, the genesis time of the profiles increases rapidly by a factor of 5 to 20 % for the two youngest profiles and 15 to more than 60 % for the two oldest profiles.
Considering that the forest aboveground litter production is around
425 gC m
With regard to the Bh horizons, it should be noted that the total C flux
leaving these horizons can be distributed in any manner between
mineralization, transfer to depth and transfer to the river. However, at
least two pools are required for the total C flux leaving the Bh to be
sufficiently large to match the measured values. Obtaining the measured old
ages requires a long genesis time (around
Parameters used for the modelling shown in Fig. 12.
Modelling the carbon fluxes by constraining both total carbon and radiocarbon was an effective tool for determining the order of magnitude of the carbon fluxes and the time of genesis of the different carbon-containing horizons. Here modelling the upper horizons separately was necessary because of numerical constraints due to the great differences in carbon turnover time between topsoil horizons and Bh. Steady-state values obtained for the topsoil horizon could subsequently be introduced in Bh modelling. The approach we used can be applied to a wide range of situations, if necessary with simplifying assumptions to sufficiently reduce the degree of freedom of the system.
The results obtained showed that the organic matter of the podzol topsoil is
very young (
The model suggests that the Amazonian podzols are accumulating organic C in
the Bh horizons at rates ranging from 0.54 to
3.17 gC m
IGSN registration numbers of the profiles used in this paper:
The authors declare that they have no conflict of interest.
This work was funded by grants from (1) Brazilian FAPESP (São Paulo Research Foundation. Process numbers: 2011/03250-2; 2012/51469-6) and CNPq, (303478/2011-0; 306674/2014-9), (2) French ARCUS (joint programme of Région PACA and French Ministry of Foreign Affairs) and (3) French ANR (Agence Nationale de la Recherche, process number ANR-12-IS06-0002 “C-PROFOR”). Edited by: V. Brovkin Reviewed by: two anonymous referees