Introduction
Soil holds more organic carbon (OC) than there is carbon in the global
vegetation and atmosphere combined . Soil
organic matter (OM) improves plant growth and protects water quality by
retaining nutrients as well as pollutants in the soil .
Thus, understanding the soil OC dynamics is crucial for developing strategies
to mitigate the increase in atmospheric CO2 concentrations and to increase
soil fertility . The dynamic nature of the soil
carbon reservoir is the result of the dynamic equilibrium between organic and
inorganic material entering and leaving the soil .
There are contradicting views of soil carbon storage capacities. According to
, the OC stock to 1 m depth ranges from 30 Mg ha-1 in arid
climates to 800 Mg ha-1 in organic soils in cold regions; the
predominant range is 50 to 150 Mg ha-1.
consider the carbon input rate as the main
factor for influencing carbon stocks. The authors state that OC stocks
increase linearly with increasing organic input without having an upper
limit. Most current OC models using this linear relationship perform
reasonably well across a diversity of soils and land use changes
. On the contrary,
published data where soils rich in OC show little or no increase in soil OC
despite a 2- to 3-fold increase in carbon input. This motivated
and to propose that
the OC accumulation potentials of soils are limited independent of increasing
carbon input. They attributed this to the limited binding capacities of
minerals.
This concept is reflected by the model of
in which the OC input is stepwise mineralized, surpassing the form of large
biopolymeres, small biopolymers with less than 600 Da, and monomers. At
each step the possibility of interaction with mineral phases increases,
leading to different OC storage forms with differing turnover times and
degree of interaction with the mineral phase. The predominant proportion of
OM in soils is associated with the mineral phase
e.g.. Minerals have finite reactive surface
areas, and consequently finite OM binding capacities. The size of the
surface area depends on the type of mineral, and so the differences in OC
stocks among soils are due to different types and amounts of the contained
minerals. Thus, the OC input rate is only crucial as long as the mineral OC
storage capacities are not exhausted. However, the concept of limited storage
capacity has hardly been experimentally tested so far.
Allophane and imogolite-type phases are, besides Al and Fe oxides, the most
effective minerals to bind OM . They dominate the mineral assemblage of Andosols,
making them the most carbon-rich mineral soil type . Andosols are
subdivided into silandic and aluandic subgroups. Silandic Andosols have
80–120 g OC kg-1 soil, whereas aluandic Andosols can contain up to
300 g OC kg-1 soil . Differences in OC
concentrations among both subgroups are explained by differing carbon storage
mechanisms. Organic matter in silandic Andosols is mainly bound to
allophanes, imogolites, and protoimogolites (grouped as imogolite-type phases
). The OM in aluandic Andosols is mainly stored
within aluminium-organic complexes (Al–OM complexes). The Al in these
complexes can be either monomeric Al3+ ions or hydroxilated Al
species . Andosols with extremely high OC concentrations likely
present OM-saturated mineral phases, at least in the topsoil, and should
respond with no change in OC concentrations to increasing carbon input.
In order to test the concept of limited OC storage capacity in soils we took
the opportunity of a unique setting in the Ecuadorian rainforest, where a
carbon-rich Andosol (301 Mg OC ha-1 within the first 100 cm)
received an extra 1800 Mg OC ha-1 input as sawdust during a period
of 20 years. Adjacent soils without sawdust application served as controls.
We tested the following hypotheses: (i) the additional OC input did not
result in increased OC in the topsoil, but in the subsoil, because the
mineral binding capacities for OM in the topsoil are exhausted and mobile OM
is transported into the subsoil and retained there; (ii) the increase in OC
in the subsoil is due to OM binding to the mineral phase; and (iii) the total
OC stock of the soil increased significantly.
We determined total OC stocks as well as the storage forms of OM and the
mineral composition down to 100 cm depth. For determining different OM
storage forms we used the sequential density fractionation method yielding OM
fractions of different degrees of mineral interaction. We also determined
pyrogenic organic carbon (PyC) because of its significant contribution to OC
stocks in some regions of the Amazon basin
e.g.. We used ammonium-oxalate–oxalic-acid
extraction and X-ray diffraction for characterizing the prevalent mineral
species in density fractions containing organic–mineral associations.
Materials and methods
Soil sample source and handling
The study site is located in Ecuador, within the Centro de Rescarte
de la Flora Amazónica (CERFA) 3 km south of Puyo
(1∘30′50′′ S, 77∘58′50′′ E, 950 m a.s.l.). Puyo,
located in the transition zone between the Andes and the western Amazon
basin, lies in the centre of a largely homogeneous alluvial fan composed of
re-deposited Pleistocene volcanic debris of the Mera formation
. The deposited material
belongs to the andesite-plagidacite series or the andesite
andesitedacite-rhyolite series .
The climate is diurnal tropical with mean annual temperatures of
20.8 ∘C and annual precipitation of 4403 mm
. The vegetation cover is tropical rainforest and
pasture .
Before 1980 the sampling area was first used for traditional shifting
cultivation, and then pasture dairy farming. Since 1980 5 ha of pasture
were reforested by Nelson Omar Tello Benalcázar. On 3 ha, within this
area, he applied 1800 t OC ha-1 additional litter in the form of
sawdust until the year 2000 (sawdust site). About 10 m3 of sawdust
were applied approximately evenly over the site by hand every day on 5 days
a week for 20 years. The sawdust was collected on a daily basis from a
local sawmill. The 5 ha reforested area is now covered by a
37-years-old secondary rainforest .
Selected bulk properties of the studied Andosol. Soil horizon
thicknesses, pH values, bulk densities (BD), bulk organic carbon (OC)
concentrations and carbon nitrogen ratios (C / N) are given as means
with the standard error where appropriate. The Alox,
Siox, and Feox represent ammonium-oxalate–oxalic-acid-extractable aluminium, silicon, and iron and are given as the mean
concentrations and standard error. Fed is dithionite-extractable
iron analysed with samples from only one profile per site. The concentrations
of Al, Si, and Fe are normalized to the mineral part (or inorganic part) of
the dry soil assuming that the mass of OM is 2 times the mass of OC
. The row marked with n represents the number of
profiles analysed per site.
Horizon
Thickness
pHCaCl2
BD
OC
C / N
Alox
Siox
Alox/Siox
Feox
Fed
cm
–
g cm-3
g kg-1
–
g kg-1
g kg-1
molar ratio
g kg-1
g kg-1
n
5
5
2
5
5
3
3
3
3
1
control site
H1
10 (±1)
4.1
0.27 (±0.00)
212 (±16)
12
55 (±2)
12 (±2)
4.7 (±0.8)
21 (±3)
20
H2
13 (±5)
4.6
0.37 (±0.01)
128 (±9)
11
67 (±7)
23 (±8)
3.1 (±0.7)
20 (±2)
20
H3
24 (±4)
4.8
0.39 (±0.00)
75 (±0)
11
85 (±3)
35 (±4)
2.5 (±0.2)
22 (±2)
24
H4
31 (±6)
5.0
0.37 (±0.05)
76 (±1)
12
95 (±7)
42 (±4)
2.3 (±0.0)
20 (±5)
22
H5
35 (±3)
5.1
0.32 (±0.01)
66 (±1)
12
121 (±4)
53 (±1)
2.3 (±0.1)
20 (±9)
20
sawdust site
H1
12 (±1)
4.5
0.30 (±0.04)
214 (±18)
13
56 (±4)
13 (±2)
4.3 (±0.6)
25 (±1)
25
H2
16 (±6)
4.8
0.31 (±0.05)
143 (±14)
13
70 (±9)
22 (±6)
3.2 (±0.6)
25 (±1)
28
H3
20 (±3)
5.4
0.36 (±0.03)
89 (±4)
12
76 (±6)
29 (±5)
2.7 (±0.3)
28 (±2)
26
H4
29 (±3)
5.6
0.39 (±0.01)
77 (±4)
12
84 (±20)
35 (±9)
2.4 (±0.1)
28 (±7)
24
H5
28 (±9)
5.6
0.32 (±0.00)
67 (±3)
12
92 (±13)
39 (±3)
2.4 (±0.2)
31 (±8)
33
As the site was not originally designed for experimental purposes, our
research plots are not arranged in a randomized plot design. Nevertheless, we
think that it can be scientifically useful because the plot areas are fairly
large (2–3 ha at each site) and essential conditions like exposition,
inclination, climate, and geology are the same for the treated and untreated
areas. No information about changes in tree species over time and possible
differences in species due to the sawdust input are available. Therefore, no
precise information on differences in carbon input due to litterfall are
available. Using the litterfall biomass of
0.9–6.0 Mg ha-1 yr-1 carbon reported by
for tropical forests across the world, we estimated the litter input since
1980 at the study site to be 33–222 Mg ha-1. This means that the maximum
carbon input with litter represents 2–12 % of the total sawdust carbon
input and is therefore insignificant. In order to estimate the belowground
biomass as a possible soil OC source, we measured the gravimetric
root intensity. We found no significant difference between the sites (for
data see Table ).
Selected properties of bulk samples used in the sequential density
fractionation. The soil profiles were selected on the basis of having five
horizons within the upper 1 m, largest OC concentration in horizon one,
similar amount of acid–oxalate-extractable elements, and having different bulk
OC concentrations in the third horizon. Presented are bulk organic carbon
(OC) concentrations, carbon-to-nitrogen ratios (C / N), and pyrogenic
carbon (PyC) contents normalized to bulk soil OC. Alox,
Siox, and Feox are the concentrations of ammonium-oxalate-oxalic-acid-extractable aluminium, silicon, and iron. The
concentrations of Al, Si, and Fe are normalized to the mineral part (or
inorganic part) of the dry soil assuming that the mass of OM is 2 times the
mass of OC .
Horizon
Depth
pHCaCl2
OC
C / N
PyC
Alox
Siox
Alox / Siox
cm
–
g kg-1soil
–
g kg-1OC
g kg-1
g kg-1
molar ratio
control site
H1
0–8
4.0
252
12
46
57
11
5.1
H2
8–15
4.4
137
11
65
59
18
3.2
H3
15–35
5.1
75
11
67
87
36
2.4
H4
35–70
5.1
72
12
95
104
47
2.2
H5
70–100
5.1
64
12
60
119
54
2.2
sawdust site
H1
0–15
4.1
256
13
47
52
11
4.9
H2
15–28
4.4
170
12
65
60
15
3.9
H3
28–50
4.8
102
12
42
69
23
3.0
H4
50–76
5.1
63
11
147
63
26
2.4
H5
76–100
5.1
76
12
50
83
37
2.2
We classified the soil as an alusilandic Andosol, based on the
(for selected properties see
Table , for a profile example see Fig. ).
Few prominent X-ray diffraction reflections indicate simple mineral composition.
The crystalline primary minerals are amphibole, chlorite, quartz, and
plagioclase. Kaolinite and other secondary clay minerals are not present.
Contents of crystalline Fe oxides and gibbsite are little. Oxalate
extractions indicate large amounts of short-range-ordered and nano- or
micro-crystalline mineral phases.
The soil samples for this study were taken in 2014 from the upper 100 cm
at five profiles at both the secondary rainforest with sawdust application
site (sawdust site) and the adjacent forest site where no sawdust was applied (control
site). The positions of the 10 profiles were randomly selected and each
profile had a width of 1 m. We define horizons one and two as the topsoil
and horizons three to five as the subsoil. Samples were oven dried at
40 ∘C in Ecuador at the Universidad Estatal Amazónica, before
transport to the laboratory in Germany and sieving to < 2 mm. All
analyses, except for X-ray diffraction which was carried out with no
replicate, were carried out in duplicates. Results are presented as means of
replicates. Sequential density fractionation, subsequent mineralogical
analyses, and PyC analyses were carried out for one representative profile per
site (for selected soil data see Table ). The soil
profiles for these analyses were selected on the basis of having five
horizons within the upper 1 m, the largest bulk OC concentration in horizon
one, a similar amount of acid–oxalate-extractable elements, and having different
bulk OC concentrations in the third horizon. All calculations and graphs were
processed with R version 3.4.3 (The R Foundation for Statistical Computing,
2017).
Bulk organic carbon concentration and stock
OC stocks were calculated based on soil volume to the fixed soil
depth of 1 m. The equivalent soil mass approach propagated by
and was not applied,
because (i) bulk density did not vary much between sites for the same
horizons, (ii) the studied site was not cropland, and (iii) the approach
increases uncertainties in OC stocks of undisturbed soils
.
We distinguished up to five soil horizons per profile and determined bulk
density, horizon thickness, and OC concentrations. Horizon thickness and OC
concentrations were recorded at all five profiles per site. Aliquots of all
bulk samples were grounded and oven dried at 105 ∘C for 24 h
prior to OC and nitrogen (N) determination with an Elementar Vario EL III CNS
analyzer. The bulk densities were determined at two profiles per site (all
horizons). At each horizon, five replicates were sampled with 100 cm3
corers, oven-dried at 105 ∘C for 24 h, and weighed. For
calculating the OC stock of horizons in all five profiles per site
(horizoni), the mean of bulk densities (meanBD) was used
(Eq. ). The OC stock of each profile is the sum of the
respective horizons' OC stocks. The OC stocks are presented as the means with
their 95 % confidence interval. As the soils contained no material
> 2 mm in diameter, the soil particles < 2 mm represent the
total soil mass.
OCstocki[Mgha-1]=OCi[gkg-1]⋅horizonthicknessi[dm]⋅meanBD[kgdm-3]
For comparing OC stocks at different depths, we also cumulated the OC stocks
of each horizon proportionally. We choose the depths 0–25, 25–50, 50–75,
and 75–100 cm in order to represent the topsoil, horizon three
(25–50 cm), and the subsoil below horizon three. We performed two sample
t tests of mean values for comparing
bulk OC stocks and OC concentrations between
sites. The two sample t tests were performed unpaired, one sided, at the
significance level α=0.05 and the power of 1-β=0.8. The OC
concentration and OC stocks at the sawdust site were considered significantly
larger, if the t test's confidence interval did not contain zero and the
sampling number was sufficient. We calculated the minimum sample number
(nmin) and power (powerth) for the difference we wish
to detect (Δth) between the sites to evaluate the power of our
data. The Δth is either the mean difference or assumed to be
10 % of the mean at the control site.
Pyrogenic carbon analyses
Analysis of PyC was carried out in the department of soil
science at Rheinische Friedrich-Wilhelms-Universität Bonn. It followed the
revised protocol of . For quantifying the
benzene polycarboxylic acids (BPCA), 10 mg of dried and ground soil
material was treated with 10 mL 4 M CF3CO2H (99%, Sigma-Aldrich,
Taufkirchen, Germany) to remove polyvalent cations. The PyC was then oxidized
with HNO3 (8 h, 170 ∘C) and converted to BPCAs. After
removal of metal ions with a cation exchange column (Dowex 50 W X 8,
200–400 mesh, Fluka, Steinheim, Germany), the BPCAs were silylated and
determined using gas chromatography with a flame ionization detection (GC-FID;
Agilent 6890 gas chromatograph; Optima-5 column;
30 m × 0.25 mm i.d., 0.25 µm film thickness;
Supelco, Steinheim, Germany). Two internal standards citric acid and biphenyl
dicarboxylic acid were used. Carefully monitoring the pH avoided
decomposition of citric acid during sample processing as criticized by
. The recovery of internal standard 1 (citric
acid) ranged between 78 and 98 %. Carbon content of BPCA (BPCA-C) was
converted to PyC using the conversion factor of 2.3
. The analyses showed good repeatability, with
differences between two measurement parallels being
< 4.2 g PyC kg-1OC, except for the second horizon at the
control site where the parallels differed by 12.5 g PyC kg-1OC.
Sequential density fractionation scheme. SPT: sodium-polytungstate
solution, F1: fraction predominately organic matter of densities
< 1.6 g cm-3 with basically no interaction with the
mineral phase, F2: fraction predominately organic matter of densities
< 1.6 g cm-3 and weakly associated with the mineral phase,
F3: soil materials of densities between 1.6 and 2.0 g cm-3 and
holding organic matter strongly bound to mineral phases, F4: soil materials
of densities > 2.0 g cm-3 and holding organic matter
strongly bound to mineral phases.
Below the fraction labeling, subsequent analyses are listed. OC: total organic carbon concentration,
N: nitrogen concentration, XRD: X-ray diffraction, Oxalate: ammonium-oxalate–oxalic-acid extraction
Sequential density fractionation of OM
We modified the sequential density fractionation procedure
(Fig. ) described by in order to
separate four different fractions. The first light fraction (F1) contains OM
that is basically not interacting with the mineral phase, often labelled free
particulate OM. The second light fraction (F2) contains mainly particulate OM
being incorporated into aggregates, thus having little interaction with the
mineral phase. The third and forth fractions (F3, F4) are heavy fractions
mainly containing OM strongly bound to the mineral phase.
Fifteen grams of dried (40 ∘C) and sieved (< 2 mm) soil
were mixed with 75 mL of sodium polytungstate solution (SPT, TC-Tungsten
Compounds) with a density of 1.6 g cm-3 in 200 mL PE (polyethylene) bottles. To
obtain F1, the bottles were gently shaken a few times, and then the
suspensions were allowed to settle for 1 h and subsequently centrifuged at
4500 g for 30 min (Sorvall RC-5B). The supernatant was siphoned with
a water jet pump and the F1 fraction was collected on a pre-rinsed
1.2 µm cellulose-nitrate membrane filter. After rinsing with
deionized water until the conductivity of the filtrate was
< 50 µS cm-1, F1 was transferred into a 50 mL
PE bottle and subsequently freeze-dried (Martin Christ Gefriertrocknungsanlagen GmbH, models Alpha 2–4 and 1–4 LCS).
The residue was re-suspended with re-collected SPT solution and refilled with
fresh SPT solution (1.6 g cm-3) until the original sample bottle
mass was maintained. In order to release F2, the samples were dispersed by
sonication (13 mm pole head sonotrode, submersed to 15 mm depth,
oscillation frequency 20 Hz, sonication power 48.98 J s-1; Branson Sonifier 250).
The energy input was 300 J mL-1, calibrated according to
. The appropriate energy input was determined
in a preliminary experiment as that energy which released the largest amount
of largely pure OM . Temperature was kept
< 40 ∘C using an ice bath to avoid thermal sample
alteration. Thereafter, the sample was centrifuged at 4500 g for 30 min
and the floating material was separated, washed, and dried as described above for
F1.
In order to further separate the residual > 1.6 g cm-3
fraction into Al–OM complexes and imogolite-type phases, we introduced an
additional density cut off. This is sensible because the overall density of
organic–mineral associations depends on OM density, mineral density, and OM
load . The densities
of pure imogolite-type mineral phases and Al–OM complexes are similar
, but and
showed that Al–OM complexes have a higher
OM load than imogolite-type phases. The second density cut off at
2.0 g cm-3 was selected based on OC concentrations, XRD spectra, and
oxalate-extractable Al, Si, and Fe concentrations determined in a preliminary
experiment. The fraction with a density between 1.6. and 2.0 g cm-3
was found to be enriched in Al–OM complexes (F3), while the fraction
> 2.0 g cm-3 (F4) was rich in imogolite-type phases.
For obtaining F3, the residue of the previous separation step was
re-suspended in 75 mL fresh SPT solution (density of 2.0 g cm-3),
dispersed at 30 J mL-1 to ensure dispersion, centrifuged, separated,
washed, and dried as described above for F1. The final residue of
> 2.0 g cm-3 density (F4) was rinsed with deionized
water until the conductivity of the supernatant was
< 50 µS cm-1 and subsequently freeze-dried.
Aliquots of all fraction samples were oven dried at 105 ∘C for
24 h prior to OC and N determination with an Elementar Vario EL III CNS
analyzer.
To test if the separation of particles with F3 and F4 is caused by either
variations in mineral density or OM loading, we calculated the overall soil
particle densities ρsoilparticle in F3 using
Eq. () . We assume the minerals
to have densities (ρM) of about 2.7 g cm-3
and OM (xOM as 2×OC) to have
an average density (ρOM) of 1.4 g cm-3
.
ρsoilparticle=axOM+bwithb=ρOMρM-ρOManda=ρM⋅b
Acid–oxalate extraction of F3 and F4
Aluminium (Al) and silicon (Si) of short-range-ordered phases were extracted
using the ammonium-oxalate–oxalic-acid reagent proposed by
. Either 0.1 or 0.5 g of oven
dried (105 ∘C) and grounded F3 and F4 material were suspended in
0.2 M ammonium oxalate–oxalic acid at pH 3 at a soil-to-solution ratio
of 1 g : 100 mL and shaken for 2 h in the dark. The suspension was
decanted over Munktell 131 paper filters, with the first turbid effluent
being discarded. Clear solutions were stored in the dark and at room
temperature for no more than 1 day. Al and Si concentrations
were determined by ICP-OES (Thermo Scientific, iCAP 6000 series). The iron
(Fe) concentrations were measured but results are not presented here since
they were very low compared to those of Al and Si (< 30 g kg-1,
Table ). The recovery rates are calculated as the sum of
the elements' amount quantified in F3 and F4 normalized to oxalate
extractable amounts of the element in bulk samples. Since the high and
strongly varying with depth concentrations of organic matter masks the actual
contribution of oxalate extractable minerals to total mineral constituents,
we normalized the oxalate-extractable metals to the mineral soil component
(or inorganic part) instead of to the dry soil. We calculated the mineral
proportion of the samples assuming the mass of OM being 2 times the mass of
OC .
X-ray diffraction of F3 and F4
X-ray diffraction (XRD) spectra were obtained on each one F3 and F4 sample
per horizon of the samples from the sawdust site. Samples were grounded with
a ball mill and oven dried at 105 ∘C for 24 h. The random
oriented powders were analysed using a PANalytical Empyrean X-ray
diffractometer with theta/theta geometry, 1 D-PIXcell detector, Cu-Kα
radiation at 40 kV and 40 mA, at an angle range of 5–65 ∘2θ with 378 s counting time per 0.013 ∘ 2θ step
and automatically acting diaphragm. Evaluation was performed using the X'Pert
HighScore Plus V 3.0 (PANalytical) software.
The organic carbon (OC) stocks of the sawdust and the control sites.
Presented are the entire profile (0–100) and the four selected depth
increments 0–25, 25–50, 50–75, and 75–100 cm as the mean and the
95 % confidence interval. ∗ The OC stock at the sawdust site is
significantly (α=0.05, 1-β=0.8) larger compared to the control
site.
Discussion
Sequential density fractionation and oxalate extraction performance
The OC concentrations in F2 (Fig. a) are within the range of
400–500 g kg-1, which suggests pure organic matter. Thus, the
applied sonication energy of 300 J mL-1 did not cause redistribution
of mineral phases over light fractions .
However, there is also no evidence for complete dispersion of aggregates,
which state as impossible. The OC
concentrations of F2 are larger than those of F1. This may be caused by
sonication, which basically strips off all adhering mineral materials
. This “cleaning effect” leads to purer OM
fractions in F2 than in F1. As a consequence, the OC concentrations in F1 range
between 300 and 400 g kg-1.
The calculated (Eq. ) particle densities of F3 are between
1.7 and 2.0 g cm-3, which is in line with the nominal density
range of F3. This also indicates that the overall soil particle density of
the studied Andosol is largely due to OM loadings and not caused by
variations in mineral density. This contrasts the results of
.
The recoveries of acid–oxalate-extractable Al and Si were large. The lower
recovery of Si could be due to Si being present as silicic acid or Si sorbed
to ferrihydrite and other poorly crystalline phases
, and may become released during the
sequential density fractionation.
Mineral composition of F3 and F4
Peak intensities of XRD spectra in F4 are up to 2.5 times larger than in F3.
This indicates enrichment in crystalline minerals in F4 as compared to F3 and
vice versa enrichment in short-range-ordered phases in F3. This enrichment in
short-range-ordered phases is supported by the 2–7 times larger amount of
oxalate-extractable Al in F3 than in F4. The broad signals in the XRD spectra of
F3 and F4 in deeper horizons suggest the presence of imogolite-type phases.
These broad signals are less prominent in F4, because they are overlain by signals of
crystalline minerals. Despite the largely similar XRD patterns the
Al / Si molar ratios of F3 are larger than those of F4 in all horizons.
Also, the C / Al molar ratios of F3 are larger than those of F4 throughout the
profile, meaning that the organic–mineral associations in F3 have more OC. We
conclude that F3 is more enriched in Al–OM complexes than F4. Additionally,
F4 of the topsoil has Al / Si molar ratios of 4, which means Al–OM
complexes have to be present in addition to imogolite-type phases. Thus,
complete separation of Al–OM complexes and OM-loaded imogolite-type phases
was not achieved by our density fractionation method. We think that Al–OM
complexes and imogolite-type phases either form into continuous phases or
that the density ranges of the two phases may overlap. Moreover, we think
that quartz and other minerals could also be present in those fractions
because the XRD spectra of the topsoil F3 show reflections for primary minerals.
This is also supported by the extremely low Al and Si concentrations in F3 of
the topsoil samples.
Concentrations of imogolite-type phases (ITPcal), Al in
Al–OM complexes (AlAOC), and molar proportions of imogolite-type
phases on short-range-ordered phases (ITP proportion) in F3 and F4, as
calculated with Eqs. (3) to (5). Data are
presented as means, and ranges in parenthesis.
Horizon
Depth
ITPcal
AlAOC
ITP proportion
cm
g kg-1
g kg-1
mol %
F3
F4
F3
F4
F3
F4
control site
H1
4
122 (±2)
10 (±1)
55 (±0)
3 (±0)
24
32
H2
12
212 (±23)
78 (±0)
55 (±6)
15 (±1)
35
42
H3
25
452 (±11)
239 (±3)
106 (±3)
25 (±0)
37
57
H4
52
402 (±0)
332 (±4)
83 (±0)
26 (±0)
40
63
H5
85
402 (±0)
384 (±3)
60 (±0)
29 (±0)
48
65
sawdust site
H1
8
91 (±5)
26 (±4)
46 (±5)
5 (±1)
22
42
H2
22
168 (±3)
64 (±3)
64 (±1)
16 (±1)
27
36
H3
39
276 (±2)
149 (±0)
72 (±1)
27 (±1)
35
43
H4
63
396 (±5)
178 (±0)
77 (±4)
21 (±1)
42
54
H5
88
457 (±4)
254 (±2)
94 (±2)
22 (±1)
40
61
The Al / Si molar ratios decrease with depth, which indicates changes in
the assemblage of short-range-ordered phases. This allows for identifying
those phases predominating the two fractions
> 1.6 g cm-3 in the different horizons. Many authors
such as use (Al–Alpy)/Si molar ratios,
with Si and Al being oxalate-extractable Al and Si, respectively, and
Alpy being pyrophosphate-extractable Al. Pyrophosphate is
supposed to extract Al from Al–OM complexes .
We did not follow this approach, because the reliability of
pyrophosphate-extraction has been questioned . attribute this to high pH
of the extractant, which can result in dissolution of Al-containing mineral
phases.
Mass contribution of OC and short-range-ordered phases (SRO) to
fractions three (F3) and four (F4). The grouping ranges from “+++” (very
abundant, > 75 wt %), “++” (abundant, 75–30 wt %),
“+” (low, 30–3 wt %), to “tr” (traces, < 3 wt %).
The prevalent short-range-ordered species are defined according to the molar
proportion of imogolite-type phases on the short-range-ordered phases (ITP
proportion of Table ). For mean molar proportions of
< 33 wt % Al–OM complexes (AOC) prevail. For molar proportions
between 33 and 67 wt %, AOC and imogolite-type phases (ITP) are largely
balanced, with the species listed first being slightly more abundant. For
molar proportions > 67 wt % ITP prevail.
Horizon
F3
F4
OC
SRO
Prevalent
OC
SRO
Prevalent
proportion
proportion
SRO species
proportion
proportion
SRO species
H1
+++
+++
AOC
tr
tr
AOC, ITP
H2
++
++
AOC
+
+
AOC, ITP
H3
+
+
AOC, ITP
+++
+++
AOC, ITP
H4
tr
tr
AOC, ITP
+++
+++
ITP, AOC
H5
tr
tr
AOC, ITP
+++
+++
ITP, AOC
According to , oxalate-extractable Si
originates almost exclusively from imogolite-type phases. The results of the
oxalate-extraction indicate the studied Andosol to be poor in Si, and
therefore only imogolite-type phases with the minimum silicon content should
be present. Instead of relying on the pyrophosphate-extraction, we developed a
formula to determine the prevailing short-range-ordered species in the
density fractions. We estimated imogolite-type phases, Al in Al–OM
complexes, and their molar contribution to short-range-ordered phases using
Eqs. (3) to (5) (Table ).
These equations are based on the formula proposed by
, , and the assumptions
listed below. The proportion of imogolite-type phases on short-range-ordered
phases are given in mol %, because the exact composition of Al–OM
complexes are unknown.
oxalate-extractable Al (Al) is only incorporated in Al–OM complexes (AlAOC) and imogolite-type phases (AlITP)
Al–OM complexes comprise compounds which contain mainly Al–O–C bonds and scarcely any Al–O–Al bonds, because OM concentrations are high.
Therefore we assume that Al–OM complexes contain on average 1.1 mol Al per 1 mol Al–OM complexes.
oxalate-extractable Si is only incorporated into imogolite-type phases (= SiITP)
AlITP / SiITP molar ratio is 2
ITPcal calculated concentration of imogolite-type phasesITPcal=7.5⋅SiITPAlAOC=Al-2⋅SiITPITPproportion=1/2⋅AlITP1/2⋅AlITP+1/1.1⋅AlAOC⋅100withSiandAlinmol⋅kg-1
We used the imogolite-type phases (ITP) proportion data from Table to evaluate the
prevailing short-range-ordered species in each horizon. The results are
compiled in Table along with the abundance of short-range-ordered
phases with F3 and F4 on total mineral masses. The results show
that in the topsoil short-range-ordered phases are mostly present in F3 and
in the subsoils they are mostly present in F4. The prevailing short-range-ordered
species in topsoils are Al–OM complexes, since dominating F3.
In the subsoil, the presence of imogolite-type phases and Al–OM complexes is
more balanced in F4, with increasing portions of imogolite-type phases with
depth.
The characteristic broad signals indicating imogolite-type phases in the
X-ray diffractograms appear in the subsoil, but not in the topsoil
(Fig. ). Imogolite-type phases dissolve at pH values below
4.8 . The pH values in the subsoil are equal or
above 4.8, whereas pH values in the topsoil are below 4.8
(Table ). Thus, the X-ray diffractograms and pH values are
well in line with the estimated distribution of short-range-ordered species,
suggesting the presence of imogolite-type phases in the subsoil and their
absence in the topsoil. We conclude that the studied Andosol shows aluandic
properties in the topsoil and silandic properties in the subsoil.
Organic carbon stock and storage forms
The OC stocks within the first 100 cm of the studied soil are above the
mean OC stocks for Andosols (254 Mg ha-1,
) and similar to the 375 ± 83 Mg ha-1
found in an aluandic Andosol under tropical rainforest on Hawaii
. The observed low bulk densities, resulting
from large OC accumulation, are a common feature of aluandic Andosols
.
The studied Andosol has medium to high PyC concentrations compared to
Amazonian Oxisols (1–3 g PyC kg-1 soil,
140 g PyC kg-1 OC; ) and Terra Preta
soils (4–20 g PyC kg-1 soil, 350 g PyC kg-1 OC;
). The PyC, however, contributes only up to
5 wt % to the total OC concentration, thus plays only a marginal role
for the accumulation of total OM.
Only up to 20 wt % of OM in the topsoil and 2 wt % in the subsoil
is not bound to mineral phases. The low proportions of OC in the light
fractions are in line with data published by for
topsoils of Andosols under tropical rainforest (20 wt % of material with
densities < 1.6 g cm-3). Thus, the mineral phase plays the
dominant role in stabilizing OM in this Andosol, which is in agreement with
numerous previous studies e.g..
We used the data from Fig. to estimate the abundances of OC
with F3 and F4 on total OC and correlate it with the abundance and prevalent
species of the short-range-ordered phases (Table ). The
OC abundances clearly correlates with the respective abundance of mineral
phases in the two fractions. In the topsoil, OC is mainly associated with
Al–OM complexes, whereas the OC in the subsoil is mainly bonded to
imogolite-type phases. As explained in Sect. we suggest
that Al–OM complexes are in close contact to imogolite-type phases i.e. they
precipitate on their surfaces.
Organic carbon response to sawdust input
No sawdust was optically visible neither in the soil profile nor in the light
density fractions prior to grounding. Seemingly, the period since the last
sawdust application was long enough to allow for complete decay.
report that wood density and bole
diameter were significantly and inversely correlated with the decomposition
rate for dead trees in tropical forests of the central Amazon. For the
smallest bole diameter (10 cm) the authors calculated 0.26 yr-1 as
the lowest rate. A dead tree with such a diameter would then be decomposed to
99 % after 17 years. Additionally
showed that in tropical soils, the decomposition rate increases linearly with
precipitation. With sawdust being much finer in texture than a dead tree
trunk and the precipitation at the CERFA site being twice as high as at the
site of , we expect a much faster
decomposition. Moreover, the phosphorus concentrations, determined in an aqua
regia solution for samples taken from the first 20 cm (data not shown), are
significantly larger at the sawdust site than at the control site. The
additional phosphorus at the sawdust site matches the phosphorus input via
sawdust. Additionally, the C / N ratios of all fractions in the upper two
horizons are below 20, showing no difference between the two sites. From all
this, we conclude that the added sawdust, which has a C / N mass ratio of
around 110, was completely decomposed on site.
In the first horizon, the differences between both sites in bulk OC
concentrations are, with around 2 g kg-1 OC, extremely small.
Additionally, the observed variances in these forest topsoils are large. This
makes it impossible to find a significant increase in the bulk OC
concentration, even with larger sample numbers. We conclude that the OC
concentration of the first horizon did not change upon sawdust input. In the
second horizon the difference between OC concentrations are larger than in
the first horizon, but are not significant due to a large variance. For the
second horizon, a larger sample number may have revealed a significant OC
increase at the sawdust site. The results of the sequential density
fractionation show no indications for additional inclusion of OM into
macro-aggregates. This holds true despite the fact that the sample of the
second horizon used for the sequential density fractionation at the sawdust
site has a higher OC concentration than the sample at the control site. Those
results need to be interpreted with some caution in terms of site comparison,
because we conducted the fractionation only with one profile per site. The
sequential density fractionation also revealed that over 80–89 wt % of
OC is strongly associated with minerals. We therefore conclude that the
mineral phase of the topsoil, especially the first horizon, is completely
saturated with OC and so despite the massive carbon input not even the
faintest additional storage occurred.
state that in soils where percolating water
controls transport processes, a steady input of surface-reactive compounds
from overlying soil layers forces less strongly binding compounds to move
further down. With OM storage capacities in the top layers being exhausted,
increasing amounts of OM become displaced and start migrating downwards. When
reaching soil horizons with free storage capacity, these OM compounds are
retained and the respective horizon's OC concentration increases. This
process would explain the significantly larger OC concentration in the third
horizon at the sawdust site. As long as the third horizon has free storage
capacities, the OC concentration in horizon four and five will not increase,
which is in line with our results.
Over 90 wt % of the additional OM in the third horizon are recovered in
F3 and F4 together. Thus, the increase in bulk OC concentration in the third
horizon is due to OM strongly interacting with the mineral phase and likely
becoming long-term stabilized.
The increase in OC concentration in the third horizon, however, is probably
not only due to undersaturated mineral phases. The increased OC concentration
in the bulk sample of the third horizon of the sawdust site used in the
fractionation experiment correlates with a larger proportion of F3 and
slightly lower pH values than those of the third horizon at the control site.
The lower pH in the third horizon at the sawdust site was unlikely due to mineralization of the
18 Mg ha-1 nitrogen added with the sawdust, since it is not affecting the overlying horizons.
More likely, the acidification was caused by the downward movement of organic acids formed
during the decomposition of sawdust. These acids may promote the weathering
of imogolite-type phases and subsequent formation of Al–OM complexes.
The OC stock increased significantly at the sawdust site for the 25–50 cm
segment, which belongs to the subsoil and comprises mostly the third horizon.
This difference was basically due to the increase in OC concentration.
Despite this increased OC stock in parts of the subsoil, we found no evidence
for increased total soil OC stocks. We think that this is not due to the
small number of bulk density measurements because the standard error was
very low with ⩽ 0.05 g cm-3 (see
Table ). Moreover, found that
the contribution of bulk density to the OC stock variability was lower than
the contribution of OC concentration and horizon thickness and decreased with
soil depth. We rather think that the large variability in OC concentrations
in the topsoil overlay the effect of the larger OC concentration in the
subsoil at the sawdust site. Therefore, for topsoils we recommend an increased
number of samples in order to detect significant differences.
To evaluate the OC accumulation in response to the sawdust application, we
referred to the increase in OC stock within the 25–50 cm segment
(Table ). The resulting additional OC stock is
15 Mg ha-1, which represents only 0.8 wt % of the originally
added 1800 Mg OC ha-1. Thus, the OC accumulation rate is extremely
low. That is in line with who postulate in
their saturation concept that soils which are close to their maximum OC
storage capacity have low accumulation rates.
Conclusions
The massive OC input did not increase the OC concentrations in the topsoil
but in the subsoil, which also resulted in significantly larger OC stocks for
the subsoil. The OC-rich Andosol topsoils are not capable of storing
additional carbon, likely because of limited binding capacities of their
mineral phases. Seemingly, some of the additional OC migrates downwards with
the percolating water until reaching horizons where free binding sites are
available. Hence, the studied soils are saturated with OC when only
considering the topsoils but still have some capacity to host more OC in
their deeper horizons. Leaving the time and input aspect aside and imagining
that the Andosols upper horizon will one day stretch down to 100 cm, the
OC stock would have increased by 260 Mg ha-1 compared to the control
site.
The OC increase in the subsoil was exclusively due to binding to mineral
phases. Since binding to mineral phases promotes retarded mineralization,
i.e. longer turnover times, stabilization, and thus long-term storage of
the additional OC can be expected.
The additional OC was likely stored within Al–OM complexes and by binding to
imogolite-type phases. There are indications that the input of additional OC
into the subsoils induced dissolution of the imogolite-type mineral phases
and subsequent formation of Al–OM complexes. This transition from a
predominately silandic to a more aluandic mineral assemblage would increase
the subsoils storage capacity for OC. We suggest that silandic Andosols can
gradually become aluandic.
Despite the increase in subsoil OC, there was no significant change in total
OC stock in response to the massive OC inputs over a period of 20 years.
This was basically because of spatial variations that demand larger
sample numbers or larger changes than the observed ones to become
significantly different.
The results clearly show that the accumulation efficiency of the added OC was
very low. Increasing the OC stock in soils already rich in OC requires
comparably large inputs over long time periods to induce OC transport into
the deeper soil horizons. This contrasts the situation in young soils where
OC stocks build up rapidly in near-surface horizons.