Introduction
Salt-affected soils occur predominantly in arid and semiarid environments
where rainfall is insufficient to leach salts from the soil
(Mavi et al., 2012). They
form either anthropogenically as a result of agricultural mismanagement or
naturally due to the accumulation of salts from mineral weathering, dust
deposition, precipitation, or a capillary rise of shallow groundwater tables
(Essington, 2004). According to the FAO (2001),
salt-affected soils possess high salinity, high sodicity, or both features
at the same time. Salinity refers to high loads of water-soluble salts
within the soil, which is typical for Solonchaks, while sodicity is
understood to mean high levels of Na+ on the exchange sites. Sodicity
usually results in a pH > 8.5 and the dispersion of soil
particles, which in turn causes a poor soil structure with low aggregate
stability (Qadir and
Schubert, 2002; Sumner, 1993). Generally, salt-affected soils are harsh
environments for plants as high salt contents reduce the osmotic potential
and subsequently limit plant water uptake (Läuchli and
Grattan, 2007). Nutrient uptake is impeded due to ion competition and the
high pH, while the poor soil structure caused by high sodicity has adverse
effects on the soil water balance and plant development
(Qadir and Schubert, 2002). As a result, plant
residue inputs into the soil are reduced and thus lead to small soil
organic matter (OM) contents (Wong et al., 2010).
However, OM is a key component of soils as a reservoir for nutrients and
determining a soil's agricultural productivity, while at the same time it
is an important carbon (C) repository and plays a pivotal role in the course
of climate change (Lal, 2004).
According to their salinity and sodicity, respectively, salt-affected soils
can be classified with respect to their electrical conductivity (EC; in dS m-1)
and the sodium adsorption ratio (SAR) of the saturated paste extract
into saline (EC > 4 and SAR < 13), sodic (EC < 4
and SAR > 13), and saline–sodic (EC > 4 and SAR > 13; US Salinity Laboratory Staff,
1954). Both parameters exert a decisive impact on the dispersion of clay and
OM. Numerous studies showed that the desorption of OM from clay particles
increases with SAR, while a rise in EC or the proportion of divalent cations
counterbalances the dispersing effect of Na+ by inducing flocculation
(Mavi et al., 2012; Nelson and Oades, 1998; Setia et al., 2014). High soil pH is
likewise assumed to increase losses of organic C (OC) through
solubilization of OM (Pathak and Rao, 1998). Peinemann et al. (2005) concluded that
in salt-affected soils, mineral-associated OM can be rapidly lost through
dispersion and subsequent leaching as dissolved OM, while particulate OM
represents a relatively stable fraction as its decomposition is reduced due
to inhibited microbial activity. In line with this, previous work
revealed in incubation and field studies that the microbial decomposition of
soil OM is reduced at elevated salinity
(Rath and Rousk, 2015; Rietz and Haynes, 2003), while little is known about the
composition of soil microbial communities.
Baumann and Marschner (2011)
and Pankhurst et al. (2001) observed decreased fungi : bacteria ratios at
enhanced salinity, while Barin et al. (2015) found the
opposite, indicating that more research is required to reach firm
conclusions.
Though based on results from sorption–desorption experiments, previous
studies noted the sensitivity of mineral–organic associations in
salt-affected soils
(Mavi et al., 2012; Setia et al., 2013, 2014), no study to date has quantified the
amount and properties of mineral-associated and particulate OM in these
soils. This is surprising, as the occurrence of salt-affected soils is
predicted to increase as a result of climate change due to enhanced aridity
(Amini et al., 2016). Currently, these soils cover a global
area of 831 Mio. ha (Martinez-Beltran and Manzur, 2005) of
which Solonchaks constitute about 260 Mio. ha (IUSS
Working Group WRB, 2014). Thus, our objectives were to elucidate the effect
of salinity and sodicity on (i) soil OC stocks, (ii) the quantities and
properties of functionally different soil OM fractions (particulate vs.
mineral-associated OM), and (iii) the microbial community composition. We
approached this by comparing soil OC stocks, the amount and properties of
density-separated OM fractions (contents of hydrolyzable non-cellulosic
neutral sugars; δ13C and 14C activity), and the PLFA-based
microbial community composition along a transect of increasing salinity and
sodicity in the southwestern Siberian Kulunda steppe. Non-cellulosic sugars
were chosen as an OM quality parameter, as they enter the soil in large
amounts with litter, root residues, plant rhizodeposits, and as by-products of microbial and faunal metabolism;
moreover, they represent a major energy
source for heterotrophic soil microbial communities
(Cheshire, 1979; Gunina and Kuzyakov, 2015). Additionally, soil aggregate stability was
determined to assess the effect of sodicity on the structural stability of
the soils. We hypothesized that (i) soil OC stocks decrease with increasing
salinity because high salinity decreases plant growth and subsequently
lowers soil OC inputs, (ii) the proportion and stability of particulate OM
is larger in salt-affected soils compared to non-salt-affected soils
since microbial decomposition and the transformation of OM is reduced under high
salinity levels, (iii) sodicity reduces the proportion and stability of
mineral-associated OM, and (iv) the fungi : bacteria ratio is negatively
correlated with salinity.
Material and methods
Study site and soil sampling
The studied transect is located in the southwestern Siberian Kulunda steppe,
which is part of the Altay Kray (Russian Federation). Due to the
semiarid to semihumid climate in the Kulunda steppe, the proportion of the
soils subject to salinization is 19.4 % (Paramonov, 2016). The transect
belongs to the dry steppe type with a mean annual temperature of 2.6 ∘C
and a mean annual precipitation of 285 mm (climate data from
the WorldClim database;
Hijmans et al., 2005). It
ranged from a lake over a terraced hillslope to about 5 m above the lake
(52∘3′36.51′′ N, 79∘36′0.71′′ E;
Fig. 1). The groundwater table increased from ca.
140 cm next to the lake to > 300 cm at the highest point of the
transect. Soil moisture and salinity covaried with the groundwater table and
increased with decreasing distance to the lake, which is a natural
phenomenon in steppe environments. Three different soil types developed
along the transect primarily as a function of the groundwater table. At
shallow groundwater depth close to the lake, Sodic Solonchaks dominated,
while Mollic Solonchaks (non-sodic) prevailed backslope with slightly higher
groundwater at about 170–180 cm. Upslope the groundwater table reached
> 300 cm and the capillary rise did not reach the soil surface; thus,
Haplic Kastanozems and Calcic Kastanozems occurred, which were generally
grouped as Kastanozems. A detailed soil classification according to
IUSS Working Group WRB (2014) of the analyzed profiles
is given in Table S1 in the Supplement. We sampled the soils at plane areas along the terraced
slope to avoid the influence of erosion on the soil profiles. Three plots,
each with a soil profile down to the groundwater table and locations for
plant analyses, were established per soil type; only in the Kastanozems was the
groundwater too deep to be reached. Four plots were analyzed on the
footslope next to the lake where site heterogeneity was larger, but one of
the four soils was not classified as Sodic Solonchak but as Haplic
Solonchak. This soil profile was grouped together with the Mollic Solonchaks
since these soils corresponded to a lower level of sodicity and they were
referred to as non-sodic Solonchaks. Therefore, Kastanozems and Sodic
Solonchaks were represented by three soil profiles, while non-sodic
Solonchaks were characterized by four soil profiles. Composite soil samples
were taken according to generic horizons in the profiles. Plant samples
(shoots and roots) were taken within the plots at 5 m of distance from each
profile for the determination of OC, total nitrogen (TN), δ13C, and
non-cellulosic neutral sugars. The aboveground biomass was determined in
triplicate around each profile by cutting off plants in a 40 cm × 40 cm
square and subsequent drying (70 ∘C) and weighing of plant
material. The major plant species are listed in
Table 1.
Schematic representation of study sites and the experimental
design. The same colors of the soil profiles and plant samples mark the same
soils. A detailed soil type classification of the grouped soils is given in
Table S1.
Vegetation (dominant species) and aboveground biomass on each soil
type. Given are the arithmetic means and the standard error of the mean in
parentheses. Significant differences (p < 0.05) were not present and
are denoted as lowercase letters.
Soil type
Vegetation with dominant species (from most to least dominant)
Aboveground biomass
g m-2
Kastanozem
Festuca valesiaca – Thymus maschallianus – Koeleria glauca
164.8
(37.7) a
Non-sodic Solonchak
Leymus poboanus – Artemisia nitrosa – Atriplex verrucifera
133.7
(17.6) a
Sodic Solonchak
Atriplex verrucifera – Leymus poboanus – Hordeum brevisubulatum
139.5
(21.7) a
Sample preparation and basic soil analyses
Samples from generic horizons of the profiles were air-dried and sieved to
< 2 mm. Visible plant materials were removed and big clods were
gently broken to pass the sieve. An aliquot of the fine-earth fraction was
dried at 105 ∘C to determine the residual soil water content. Soil
bulk density was determined gravimetrically in triplicate for generic
horizons by using a soil sample ring. Soil pH was measured in a 1 : 2.5
(w : v) soil-to-water suspension after equilibration for 1 day. Carbonate
content was analyzed by using the Scheibler volumetric method
(Schlichting et al., 1995). The texture of the soils was
determined according to the standard sieve-pipette method (DIN
ISO 11277, 2002) and the content of oxalate- and dithionite-extractable Fe
was analyzed as described in McKeague and Day (1966). Soil
aggregate stability was measured based on a method modified from
Hartge and Horn (1989) and explained in detail in
Bischoff et al. (2016). It was calculated as the
difference between the mean weight diameter (MWD) of aggregates of a dry-
and a wet-sieving method, expressed as ΔMWD, with a high ΔMWD corresponding to low aggregate stability and a low ΔMWD
relating to high aggregate stability. The soil mineralogical composition was
analyzed to characterize the soils with respect to their composition of
water-soluble salts and the amount of expandable clay minerals. Clay
mineralogy significantly affects the physical properties of sodic soils
(Essington, 2004). The quantity of expandable clay minerals was
similar in all three soil types and cannot explain differences in the OM
dynamics between the soils. All data on soil mineralogical composition are
provided in the Supplements (S1).
Soil salinity parameters
The content and composition of water-soluble salts was determined by shaking
the soil in a 1 : 5 (w : v) soil-to-water suspension at 15 rpm for 1 h
and leaving the sample for 1 day to reach equilibrium. After measuring the
EC the extract was centrifuged at 3000 g for 15 min and filtered through
0.45 µm syringe filters (Cellulose acetate). An aliquot of the
extract was measured for Na+, K+, Ca2+, and Mg2+ with an
inductively coupled plasma optical emission spectrometer (Varian 725-ES;
Agilent Technologies, Santa Clara, USA), while another aliquot was analyzed
for Cl-, NO3-, and SO42- with an ion
chromatograph (ICS-90; Dionex Corp., Sunnyvale, USA). The concentrations of
Na+, Ca2+, and Mg2+ (mmol L-1) in the extract were used
to calculate the SAR according to Eq. (1).
SAR=Na+(Ca2++Mg2+)0.5
Determination of organic carbon, δ13C, and total
nitrogen
Ball-milled < 2 mm fractions were measured for OC, TN,
δ13C via dry combustion in an Elementar vario MICRO cube
C/N Analyzer (Elementar Analysensysteme GmbH, Hanau, Germany) coupled to an
IsoPrime IRMS (IsoPrime Ltd, Cheadle Hulme, UK) after removing inorganic C
through fumigation with HCl and subsequent neutralization over NaOH pellets
(modified from Walthert et al., 2010). The measured δ13C values were corrected by calculating response factors from
standard compounds (CaCO3, cellulose, caffeine) and expressed in the
delta notation related to the Vienna Pee Dee Belemnite Standard
(0 ‰). The complete removal of inorganic C from all
samples was confirmed by δ13C values, which are in the typical
range of soil OM (-22.5 to -28.1 ‰).
Density fractionation and 14C analysis
Density fractionation (modified after
Golchin et al., 1994)
separated the soil into a light fraction (LF), containing mostly particulate
OM, and a heavy fraction (HF), consisting of mineral-associated OM and
mineral components free of OM. As particulate OM contents are mostly very
low in the subsoil, we fractionated the soil only until the first C horizon
of each profile. In brief, 25 g of soil was weighted in duplicate into beakers
and 125 mL sodium polytungstate (ρ=1.6 g cm-3) was added and
gently stirred with a glass rod. Ultra sonification was applied with an
energy input of 60 J mL-1 for 8 min to break down aggregates. After
centrifugation at 3000 g for 20 min the LF was separated from the HF by
decanting the floating LF on polyethersulfone filters and repeating the
procedure if the separation between the two fractions was insufficient. LF
remaining on the filter was washed with deionized water to remove residual
sodium polytungstate until the washing solution had an EC < 60 µS cm-1. The HF remaining in the beaker was washed with
deionized water until the EC of the washing solution was < 100 µS cm-1 for a maximum of four times in the salt-affected soils,
as no residual sodium polytungstate was detected afterwards by ESEM–EDX
analysis, which was carried out with a Quanta 200 FEG environmental scanning
electron microscope (FEI Company, Hillsboro, USA) coupled to an XL–30 EDX
detector (Ametek Inc, Berwyn, USA). The washing solutions of both LF and HF
were collected, filtered through 0.45 µm syringe
filters (PVDF), and measured for non-purgeable OC with a LiquiTOC (Elementar
Analysensysteme GmbH, Hanau, Germany) to account for the loss of OC during
washing of the samples (mobilized OC, MobC;
Gentsch et al., 2015). The LF and HF were
freeze-dried, weighted, homogenized in a mortar, and subsequently measured
for OC and TN as well as δ13C, as described in Sect. 2.4, after the
removal of inorganic C. The mobilized OC was added to the OC content of the
LF or HF.
Three representative soil profiles were selected, one per soil type, for
analysis of the 14C activities of OM fractions at the Max Planck Institute
for Biogeochemistry Jena (Germany). As the low quantity of LF material in
the subsoil did not allow for an accurate 14C measurement at deeper
depth, we only analyzed 14C activities until the topmost C horizon of
the respective soil profile. Inorganic C was removed by using 2M HCl until pH
remained < 3.5 and samples were subsequently neutralized with 2M
NaOH to pH 6. After freeze-drying, 14C analysis was performed with a 3MV
Tandetron™ AMS 14C system (Steinhof
et al., 2011) and 14C isotope activities were converted to percent
modern carbon (pMC) according to Steinhof (2013). The
pMC was defined according to Stuiver and
Polach (1977):
pMC=ASNAabs×100%,
where ASN is the normalized sample activity and Aabs corresponds
to the activity of the absolute international standard; both activities were
background corrected and δ13C normalized. OxCal 4.2 software
(University of Oxford) was used to calculate conventional 14C ages by
selecting the IntCal13 calibration curve
(Reimer et al., 2013) if pMC was < 100 % and the calibration curve from
Hua et al. (2013) if pMC was > 100 %.
Biomarker analyses
Non-cellulosic neutral sugars
Non-cellulosic neutral sugars were analyzed in the LF and HF from the generic
horizons of each soil profile. In the LF, neutral sugars were only analyzed
in some of the topmost horizons, as its content was too low in most samples
to provide sufficient material. Additionally, neutral sugars were determined
in plant material (shoots and roots). Neutral sugars were analyzed with slight
modifications according to Rumpel and Dignac (2006), including
the EDTA purification step from Eder et al. (2010). In
brief, 600 mg of HF and 50 mg of LF or plant material was hydrolyzed in 4M
trifluoroacetic acid (TFA) at 105 ∘C for 4 h after 1.5 mL of
myo-inositol was added as an internal standard. After cooling to room
temperature, the extract was filtered through glass fiber filters
(Whatman™ GF6) and TFA was removed in a rotary evaporator. The samples
were redissolved in ultrapure water and the pH was adjusted to 4–5 by
adding NH3. Ferric Fe was complexed by adding 4 mL of EDTA and incubating
the samples in the dark for 10 min. Darkened glassware was
used thereafter to prevent photolysis of Fe(III) ligand complexes. After freeze-drying
and adding two drops of NH3, the reduction of aldoses to their
corresponding alditols (derivatization) was performed at 40 ∘C
for 1.5 h with NaBH4 dissolved in dimethyl sulfoxide. Acetylation
was carried out by adding 2 mL of acetic anhydride and 0.2 mL of glacial acetic
acid, thereby using methylimidazole as a catalyst. Ice-cold deionized water
was added after 10 min to stop the reaction. Sugar monomers were extracted
through liquid–liquid extraction with dichloromethane and subsequently measured
by gas chromatography on a 7890A GC system (Agilent Technologies, Santa
Clara, USA) equipped with an SGE forte GC capillary column (0.25 mm diameter
and 0.25 µm film thickness; SGE Analytical Science, Melbourne,
Australia) and a flame ionization detector using He as a carrier gas.
External standards were used to detect eight different sugars: arabinose,
xylose and ribose (pentoses), galactose, glucose and mannose (hexoses), and
fucose and rhamnose (desoxysugars).
Phospholipid fatty acids
Directly after sampling, sieving to < 2 mm, and removing visible
plant materials, 1.0–1.5 g of field-moist soil was weighted into cryovials and
3 mL of RNAlater® was added to prevent sample degradation
(Schnecker et al., 2012). An aliquot was
dried at 105 ∘C to determine the soil water content. The cryovials
were kept cool until they were frozen to -20 ∘C within 72 h. For
PLFA analysis we used a modified method from
Gunina et al. (2014). Briefly, samples were transferred from cryovials into test tubes and
washed with ultrapure water to remove residual RNAlater®.
Lipids were extracted twice with a chloroform–methanol–citrate buffer
(1:2:0.8 v/v/v) and separated into glycolipids, neutral lipids, and
phospholipids through solid-phase extraction with activated silica gel (Sigma
Aldrich; pore size 60Å, 70–230 mesh). Phospholipids were derivatized
into fatty acid methyl esters (FAMEs) with 0.5M NaOH dissolved in methanol
and with BF3 as a catalyst. FAMEs were analyzed with a 7890A GC system
(Agilent Technologies, Santa Clara, USA) equipped with a 60 m Zebron ZB-5MSi
capillary GC column (0.25 mm diameter and 0.25 µm film thickness;
Phenomenex, Torrance, USA) and a flame ionization detector using He as a
carrier gas. As an internal standard we used nonadecanoic acid (FA 19:0) and
17 fatty acids were used as external standards. Peak identification of the
internal standard was problematic in the salt-affected topsoils.
Therefore we could not reliably quantify individual PLFAs but only their
relative proportion in the sample. As a result the sum of all PLFAs was not
used as a proxy for the microbial biomass contents, but PLFAs were used to
characterize the composition of functional microbial groups. We applied a
principal component analysis (PCA) to the relative distribution of all 17
PLFAs to identify clusters of correlated PLFAs, which presumably derive from
identical microbial functional groups. The assignment of individual PLFAs to
certain microbial groups based on the PCA was in agreement with the
literature (Frostegård et al.,
2011; Olsson, 1999; Ruess and Chamberlain, 2010; Zelles, 1999). Thus, the
following PLFAs were used to distinguish functional microbial groups:
18 : 2ω6,9 and 18 : 1ω9c as markers for saprotrophic fungi
(SapFungi), 16 : 1ω5c to identify arbuscular mycorrhizal fungi (AMF),
i15 : 0, a15 : 0, i16 : 0, i17 : 0, and a17 : 0 were related to gram-positive bacteria,
10Me16 : 0 characterized actinomycetes (Actino), 16 : 1ω7c and
18 : 1ω7c identified gram-negative bacteria, and 14 : 0, 15 : 0, 17 : 0, and
18 : 0 related to nonspecific bacteria (NonspBact). The PLFAs Cy19 : 0 and
20 : 4ω6c were not used as markers for microbial groups as they hardly
reached the detection limit and were sometimes difficult to distinguish from
other unspecific peaks in the gas chromatogram.
Calculation of organic carbon stocks
Organic C stocks (Mg ha-1) were calculated according to
Poeplau and Don (2013) for all horizons and the entire
soil profile as well as until 1 m of depth using Eq. (3):
OCstock=∑i=1nFSMiVi×Ci×Di,
where n is the number of horizons, FSM is the fine-earth soil mass (g), V is the
volume (cm3), C is the OC content (% of soil mass), and D is the
length of the horizon (cm).
Statistical analyses
Data analysis was performed in R version 3.2.5 (R Core Team,
2016). From replicated measurements we calculated arithmetic means and
standard errors. To test for the effect of soil type on aboveground plant
biomass, a linear mixed effects model was fitted (package lme4;
Bates et al., 2012). We accounted for the nested
structure of sampling; i.e., the soil type was used as a fixed effect, while the
soil profiles (of each soil type) were included as random effects. Residuals
and random effect estimates of the fitted model were visually assessed by
Q–Q normal plots, but no deviations from normality were observed. The
difference of the response variable between the soil types was tested based
on the linear mixed effects model fit, including corrections for multiple
comparisons (analogous to the Tukey test) with Satterthwaite degrees of
freedom on the basis of the R packages lsmeans (Lenth and
Herve, 2015), lmerTest (Kuznetsova et al., 2015), and
multcomp (Hothorn et al., 2008). Soil-sample-related
parameters were analyzed descriptively, as their sample size was only 3–4
per soil type, which was insufficient for statistical hypothesis testing.
Data on PLFA and neutral sugars were analyzed by using PCA in order to consider
multiple response variables. Confidence regions (68 %) for the group
centroids of the independent factor variables were added to the biplots.
Figure 1 was drawn in Inkscape, while the other
graphs were generated using ggplot2 (Wickham, 2009).
Discussion
Soil OC stocks along the salinity gradient
Salt-affected soils, such as Solonchaks, are normally characterized by poor
plant growth resulting in small soil OC inputs and subsequently low soil OC
stocks (Wong et al., 2010).
Muñoz-Rojas et al. (2012), for example,
reported soil OC stocks in the Solonchaks of southern Spain in 0–75 cm depth
as
53.6 Mg ha-1 (coefficient of variation (CV): 60 %) under shrub and/or
herbaceous vegetation. Batjes (1996)
calculated in the framework of a global meta-analysis an average soil OC stock
in Solonchaks of 42 Mg ha-1 (CV: 67 %) in 0–100 cm of depth, while he
noted that Mollic Solonchaks had particularly larger soil OC stocks of
101 Mg ha-1 (CV: 44 %). Kastanozems, on the other hand, contained on
average 96 Mg ha-1 (CV: 50 %) in the first meter, at which Haplic
Kastanozems had soil OC stocks above that average of 138 Mg ha-1 (CV:
44 %; Batjes, 1996). Based on data
from Bischoff et al. (2016), we calculated soil OC
stocks in Kastanozems of the dry steppe type in the Kulunda steppe down to
60 cm, which amounted to 110 ± 6 Mg ha-1. All of the previously
published data confirm that salt-affected soils like Solonchaks normally have
smaller OC stocks than the non-salt-affected Kastanozems. In contrast
in our study, salt-affected soils had larger OC stocks compared to the
nearby Kastanozems. With average OC stocks of 70.9 ± 2.8 Mg ha-1
in 0–100 cm of depth for the Kastanozems, the values were clearly below those
observed by Batjes (1996) and
calculated from Bischoff et al. (2016). On the other
hand, average OC stocks of 94.2 ± 6.9 and 129.5 ± 25.6 Mg ha-1 in 0–100 cm of the non-sodic Solonchaks and Sodic
Solonchaks, respectively, were clearly above the values reported by
Batjes (1996) and
Muñoz-Rojas et al. (2012). Larger OC stocks
in salt-affected soils than in Kastanozems are also in contrast to earlier
work, which found a negative effect of salinity on soil OC stocks (reviewed
by Wong et al., 2010). Possible reasons for the
observed differences are climatic variations between the studies (strong
aridity in the Spanish Solonchaks from
Muñoz-Rojas et al., 2012) or alterations in
soil texture (finer textured Kastanozems in the study from
Bischoff et al., 2016), which may change the soil water
balance and thus plant growth and soil OC inputs. However, it appears that
the covarying moisture gradient along the salinity transect is a better
explanation for the observed differences. During sampling we observed very
dry conditions in the Kastanozems (only 4.0 ± 0.3 % soil water
related to dry soil mass), while the Solonchaks were generally wetter due to
their shallow groundwater table (15–30 % soil water;
Table 2). Overall, the water stress in the three
soil types could have been similar either as a result of osmotic or matric
stress, leading to comparable moisture conditions for plant growth.
Accordingly, plant growth (as measured by aboveground biomass) was not
reduced under high salinity along the transect
(Table 1) which is in contrast to previous work
(Läuchli and Grattan, 2007; Wong
et al., 2010). As this is expected to reduce OC stocks at elevated salinity
(Wong et al., 2010), we consider it as the most
likely reason why we did not find a negative relation between OC quantity
and salinity. Since the δ13C ratios suggested that soil OM was
mostly root derived in the studied soils (Fig. 3), one might argue that aboveground biomass is a poor proxy for soil OC
input. However, under the assumption that root residue inputs are correlated
with the aboveground biomass (evidence is given by
Titlyanova et al., 1999, who observed
significant correlations (p < 0.01, R > 0.5) between the
aboveground and belowground biomass of typical plants in Siberian
grasslands), one can conclude that both aboveground and belowground soil
OC inputs were comparable between all three soil types.
Wong et al. (2010) argued that small OC stocks in
salt-affected soils can also be the result of erosion-induced OC losses, as
sodic soils are particularly prone to erosion. Since we paid particular
attention to the fact that all soils were not affected by erosion, we can
rule out erosion as a factor that modified OC stocks in our study.
In summary, our first hypothesis has to be rejected since soil OC stocks did
not decrease with increasing salinity, which is in contrast to previous
observations from comparable soils. What is decisive for our observations is
probably the fact that the salinity gradient covaried with a moisture
gradient. This presumably led to similar water stress either due to a low
osmotic or a low matric potential along the entire transect. Hence, against
our expectation, biomass production and soil OC inputs were not reduced
under high salinity, which was initially assumed to decrease OC stocks in
salt-affected soils.
Partitioning and composition of soil OM in functionally different OM
fractions
Considering processes of soil OC stabilization, semiarid soils should have
large proportions of particulate OC, as the formation of stable
mineral–organic associations is attenuated due to low water availability and
a high soil pH (Kleber et al., 2015). However, in the
semiarid soils of the studied transect, particulate OC contributed < 10 % of bulk OC, while mineral-bound OC accounted for > 90 %
(Table 3). This contrasts with observations from the steppe
soils (mostly Chernozems) of European Russia
(Breulmann
et al., 2014; Kalinina et al., 2011), Canada
(Plante et al., 2010), and China
(Steffens et al., 2010) in which particulate OC represented
> 20 % of bulk OC. Nevertheless, our results are in line with
Bischoff et al. (2016), who reported that a maximum of
10 % OC was present as particulate OC in the Chernozems and Kastanozems of
the Kulunda steppe. Thus, we support previous observations from this region
and conclude that mineral-bound OM is the dominant OM fraction in both
salt- and non-salt-affected soils of the studied region.
In our second hypothesis we expect the proportion and stability of
particulate OM to be larger in the salt-affected than in the non-salt-affected
soils. Against this hypothesis, Sodic and non-sodic Solonchaks contained
similar proportions of particulate OC as the non-salt-affected
Kastanozems, with 4–8 % particulate OC in all three soil types
(Table 3). Comparable 14C activities in the LF
of the three soil types (small 14C activities in the non-sodic
Solonchak were probably due to contamination with HF material) indicated a
similar turnover of particulate OM, thus contradicting our hypothesis of
increased stabilization of particulate OM under high salinity levels. Based
on OC determinations in particle-size separates and analyses of lignin
components along a salinity gradient in the Argentinian Pampa,
Peinemann et al. (2005) suggested that particulate OM
is a relatively stable fraction in salt-affected soils due to a reduced
microbial transformation of the plant-derived residue inputs. This is not
corroborated by our results. The isotopic C composition (14C activity,
δ13C) and the composition of neutral sugars indicate a
comparable alteration of OM (i.e., degree of OM decomposition) between the
three soil types (Figs. 3–5). As for the first
hypothesis, a possible explanation for our observations is that soil
moisture covaried with salinity along the transect. Given that the water
stress is similar in all three soil types either due to a low osmotic or
matric potential, OM decomposition can likewise be reduced in both the
salt-affected and non-salt-affected soils. This results in a
similar proportion and stability of particulate OM and a comparable
alteration of soil OM along the transect, as indicated by the similar
composition of C isotopes and neutral sugars in the studied soils. Hence,
soil moisture can be considered a master variable in the OM dynamics of
salt-affected soils, as it controls OM input and decomposition and can thus
interfere with the effect of salinity on the quantity and quality of
soil OM.
With respect to mineral-associated OM, Peinemann et al. (2005) concluded that mineral-bound OM is relatively susceptible to losses
in salt-affected soils due to weak chemical bonding and subsequently weak OM
stabilization. Our third hypothesis was built upon this conclusion, but in
contrast the OC content of the HF of the salt-affected soils was
more than twice as large as of the non-salt-affected Kastanozems
(Table 3). Moreover, during washing of the density
separates (sodium polytungstate removal) relatively less OC was mobilized
from the HF of the salt-affected soils (3–10 % MobC) than from the HF of
the Kastanozems (16–46 % MobC; Table 3),
suggesting a lower chemical stabilization of mineral-bound OM in the
non-salt-affected soils. We explain the large contents of mineral-associated
OC under high salinity levels by considering basic chemical principles.
According to Sumner (1993), the dispersion
of clay minerals is only possible below their critical flocculation
concentration (CFC). This concept relates the dispersive effect of Na+
on the soil structure to the corresponding salt concentration of the soil
solution (Rengasamy et al., 1984; Sumner
et al., 1998). The authors classified soils into flocculated, potentially dispersive, and dispersive depending on the EC
and SAR of the soil water extract. Sumner et al. (1998)
classified soils with large proportions of non-expandable illitic
clays, while Rengasamy et al. (1984) considered soils with
expandable 2 : 1 clays similar to the smectite-rich soils of the studied
transect. According to their classification, all of the salt-affected soils
in our study fall into the category flocculated; even the A horizons of the Sodic
Solonchaks with an average SAR of 36 ± 10 remain flocculated,
presumably due to the high electrolyte concentration as indicated by a high
EC of 5350 ± 1476 µS cm-1 (Table 2). This is underpinned by the high aggregate stability of the Sodic
Solonchaks (Table 2) and the lack of clay
lessivation or OM translocation, which are processes that require the
dispersion of clay and OM. In laboratory experiments,
Setia et al. (2013, 2014) confirmed that the dispersive effect of Na+ on OM
and mineral components is only evident at low electrolyte concentrations,
particularly at low concentrations of divalent cations like Ca2+. These
studies suggest that the content of water-soluble salts in the soils of the
studied transect is large enough to provoke the flocculation of OM and mineral
components and the formation of stable mineral–organic associations.
Moreover, Nelson and Oades (1998) showed that the
solubility of Na+-coated OM is larger than that of OM coated with
Ca2+. Thus, particularly in the non-sodic Solonchaks where Ca2+ is
a dominant cation in the soil solution (Fig. S1), the solubility of OM can
be reduced. Furthermore, the Solonchaks had higher clay and silt contents
than the Kastanozems (Table 2). This may also
account for the higher HF–OC contents in the Solonchaks, as OM has an
increased affinity to sorb on minerals in the clay- and silt-sized fraction
(Kleber et al., 2015).
Interestingly, during the sodium polytungstate removal in the density
fractionation procedure, we found larger losses of HF material in the
salt-affected soils compared to the non-salt-affected Kastanozems, which
we ascribe to the leaching of water-soluble salts
(Table 3). However, the loss of MobC was much lower
in the salt-affected soils. This indicates that the water-soluble salts were
mostly not associated with OC, presumably because these salt minerals have a
fast turnover (frequent formation and dissolution as a function of the actual
soil water content) and a small number of reactive surfaces.
To sum up, in salt-affected soils particulate OM can be more labile than
previously assumed, as evidenced by its small quantity in the Sodic and
non-sodic Solonchaks together with its low 14C ages. Salinity did not
alter the proportion and stability of particulate OM, possibly due to the
covarying moisture gradient. This suggests that soil moisture is a master
variable that has to be considered when analyzing the effect of salinity on
soil OM dynamics. Mineral-bound OM, on the other hand, is stabilized in the
studied salt-affected soils as the high electrolyte concentration in the
soil solution promotes the flocculation of OM and mineral components.
Microbial community composition along the salinity gradient
Microbial communities are sensitive to environmental changes and react to
differences in the osmotic and matric potential
(Rath and Rousk,
2015; Schimel et al., 2007). Fungi and gram-positive
bacteria are thought to be particularly more resistant against drought than gram-negative
bacteria due to their ability to produce higher amounts of osmolytes
(Schimel et al., 2007). Moreover, the cell walls of
fungi and gram-positive bacteria offer better protection against water loss,
and fungal hyphae are less dependent on water-filled pore space (Lennon et
al., 2012). However, previous work on differences in the microbial community
composition along salinity gradients could not support the view that fungi
are superior to bacteria under water stress caused by high salinity levels,
as several studies even observed a negative relationship between fungal
abundance and salinity
(Baumann
and Marschner, 2011; Chowdhury et al., 2011; Pankhurst et al., 2001). This
suggests that in salt-affected soils it is not only drought that dictates the abundance
of certain microbial groups, but also that toxic effects of certain ions or
impeded nutrient uptake may exist. In our study, the fungi : bacteria ratio
was not related to the salinity gradient and was similar in the topsoils of
the three soil types (Table 4). Hence, our fourth
hypothesis has to be rejected. As with hypotheses 1 and 2, a possible
explanation is the covarying moisture gradient along the salinity transect,
which could have led to comparable water potentials (either due to low
matric or osmotic potential) along the salinity gradient.
Chowdhury et al. (2011) analyzed the effect of an alternating matric and osmotic potential on
the PLFA-based microbial community composition. They detected a decreasing
fungi : bacteria ratio with decreasing osmotic potential, while the opposite
effect was evident with declining matric potential. Thus, with respect to
our transect, both effects (decreasing matric vs. osmotic potential) could have
canceled each other out, which resulted in similar fungi : bacteria ratios
in the topsoils along the salinity gradient. Differences were only evident
in the subsoils where salt-affected soils showed higher fungi : bacteria
ratios than the non-salt-affected Kastanozems (Table 4). In the Sodic Solonchak, fungi : bacteria ratios even increased from
topsoil
to subsoil (less pronounced also in the non-sodic Solonchak), which is
contrary to what was found in previous studies of temperate soils
(Ekelund et al., 2001; Fierer et al., 2003; Taylor et al., 2002). This could indicate
larger C availability in the subsoil of the salt-affected soils
(Fierer et al., 2003), which is also suggested by the
δ13C ratios of the LF, which decrease from topsoil to subsoil in
the Solonchaks (Fig. 3).
With respect to the PLFA-based microbial community composition, PCA revealed
a higher abundance of AMF in the salt-affected soils than in the Kastanozems
(Fig. 6). Evelin et al. (2009) reviewed the role of AMF in alleviating salt stress for plants. They
concluded that AMF increased nutrient uptake, photosynthetic rate, water-use
efficiency, and improved osmoregulation in the host plant. Thus, salt stress
in plants caused by high salinity levels, such as a hampered nutrient uptake
due to ion competition or exposure to osmotic stress, can be alleviated by
symbiosis with AMF. This could explain the higher relative abundance of AMF
in the Solonchaks of the studied transect.