Soil aggregate stability to predict organic carbon outputs from soils

Geoderma 243–244 (2015) 205–213

Contents lists available at ScienceDirect

Geoderma journal homepage: www.elsevier.com/locate/geoderma

Soil aggregate stability to predict organic carbon outputs from soils V. Chaplot a,b,⁎, M. Cooper c a b c

IRD, LOCEAN, UMR 7159, Institut Pierre Simon Laplace, 4, place Jussieu, 75252 Paris Cedex 05, France Center for Water Resources Research, SAEES, Rabie Saunders Building, University of Kwazulu-Natal, Private Bag X01, Scottsville 3209, South Africa Departamento de Ciência do Solo, ESALQ/USP, Av. Pádua Dias, 11 — Cx. Postal 9, 13418-900, Piracicaba, SP, Brazil

a r t i c l e

i n f o

Article history: Received 18 August 2014 Received in revised form 15 December 2014 Accepted 17 December 2014 Available online 21 January 2015 Keywords: Water erosion Soil Multivariate analysis Soil degradation

a b s t r a c t Soil structure (e.g. aggregation) has been recognized as a key element in the stabilization of soil organic matter. While aggregate bre akdown is assumed to expose the enclosed soil organic carbon (SOC) to preferential erosion and to accelerated decomposition, the link between the stability of soil aggregates and SOC exports from soils, has either been overlooked or unaccounted for, especially when developing carbon cycle models. This study compared SOC losses in particulate (POC), dissolved (DOC) and gaseous (GOC) forms to an indicator of the soil aggregate stability, the mean weight diameter of aggregates (MWD). SOC outputs were considered at 24 locations of a typical hillslope of the South African Highveld showing clayey to sandy soils. Both POC and DOC were evaluated in-situ under natural rains using 1 × 1 m2 runoff plots while soil CO2 emissions were assessed in the laboratory from undisturbed 0–0.05 m soil samples. MWD was finally compared to selected soil and terrain attributes for predictive purpose and as a means to further the understanding of SOC outputs from soils. MWD ranged between 1.4 mm for unstable aggregates and 3.4 mm for stable aggregates. The increase in aggregate stability resulted in a significant increase in POC and DOC concentrations in the eroded sediments (r = 0.76) and in GOC losses from soils (r = 0.91 when expressed as g C-CO2 per gram of soil; r = 0.95 when as g C-CO2 per gram of soil carbon). In contrast, high aggregate stability induced low total DOC and POC losses (r = −0.81 and −0.77, respectively). The lower POC and DOC losses in the most stable soil aggregates were explained by increased soil infiltration by water and reduced transport by runoff, while the greater CO2 emissions correlated with high SOC concentration. Furthermore, there was a tendency for clayey soils which were fully covered by grass to present stable aggregates and thus to yield greater CO2 emissions but lower POC and DOC outputs than degraded sandy soils of low aggregate stability. Such a quantitative assessment of the role of soil aggregation on SOC outputs might enhance knowledge on organic matter persistence in soils, a prerequisite for developing more accurate global carbon cycle models. Finally further research is required to investigate the downslope to downstream fate of the eroded SOC and to develop land management strategies that aim at lessening carbon losses from soils while enhancing adaptation to climate change. © 2014 Published by Elsevier B.V.

1. Introduction The soil organic carbon (SOC) pool, the biggest of terrestrial ecosystems shows important exchanges with the atmosphere through photosynthesis and decomposition of organic matter. This pool has been highly reduced by past human activities (Lal, 2003) such as changes in land use and land management. It is thus thought that the current rise in greenhouse gases (GHGs) in the atmosphere can be mitigated by sequestration of organic carbon (C) in soils (Batjes, 1996; Lal, 2003). In this context, understanding the balance between SOC inputs and outputs and their mechanisms and factors of control is key to improve

⁎ Corresponding author at: IRD, LOCEAN, UMR 7159, Institut Pierre Simon Laplace, 4, place Jussieu, 75252 Paris Cedex 05, France. E-mail address: [email protected] (V. Chaplot).

http://dx.doi.org/10.1016/j.geoderma.2014.12.013 0016-7061/© 2014 Published by Elsevier B.V.

sequestration of organic C in soils while supporting important ecosystem functions such as food and biomass production and biodiversity. Soil aggregation has the potential to enhance organic matter (OM) stabilization in soils, which can be defined as the ability to increase the residence time of organic C in the soil compared with a reference situation or benchmark (e.g. Berhe and Kleber, 2013; Novara et al., 2012). While the molecular structure of OM has long been thought to determine the persistence of organic compounds in soils, Schmidt et al. (2011) recently argued that environmental and biological controls have to be considered to explain OM protection in soil against microbial decomposition for centuries to millennia of, for instance, potentially labile compounds such as sugars. Water erosion is a natural process which significantly affects the net flux of C between the soil and atmosphere. By removing OM from soils and transporting it to depositional sites and/or to the atmosphere it induces key mechanisms such as (1) the replacement of OM at eroding

206

V. Chaplot, M. Cooper / Geoderma 243–244 (2015) 205–213

sites; (2) deep burial of OM at sedimentation sites; and (3) enhanced decomposition of OM as a result of soil aggregate breakdown during either detachment or transport (Van Oost et al. 2007). In particular, the break-down of soil aggregates by raindrops and overland flow, in the process of water erosion, has been shown to induce preferential detachment and transport of soil OM due to low density of most organic compounds compared to the other soil constituents (Rumpel et al., 2009; Maïga-Yaleu et al., 2013). Moreover, aggregate breakdown enhances the liberation of protected OM and its exposure to oxidizing conditions (Schmidt et al., 2011), which in turn increases CO2 emissions. In this context, the susceptibility of soil aggregates to break-down may constitute an important soil characteristic governing OM persistence in soils. Yet, little is known on the impact of the soil aggregate stability (i.e. natural resistance to disaggregation by mechanical breakdown, slaking or dispersion; Le Bissonnais, 1996) on organic C outputs from soils. This issue is crucial not only to enhance our understanding of the controls of OM dynamics but also to improve the prediction of OM persistence in soils through the use of soil proxies, which are easy to access and relatively cheap to implement. Understanding the link between soil aggregation and C outputs from soils is also important for designing agricultural practices aiming at increasing OM stabilization. It is finally key for improving current soil C models, which are empirical in nature and mostly lacking to integrate the environmental controls of OM outputs from soils (von Lützow et al., 2008; Schmidt et al., 2011; Segoli et al., 2013). This study was conducted to test the hypothesis that aggregate stability plays an important role in controlling SOC outputs, though decomposition, which results in losses in gaseous form, and water erosion, which leads to SOC losses in dissolved and particulate forms. Aggregate stability tests following Le Bissonnais (1996) were performed from topsoil (0–0.05 m) bulk material at 24 locations of a typical hillslope of the

South African Highveld showing clayey to sandy substrate and different levels of grass basal cover. The results were compared to SOC losses in particulate (POC), dissolved (DOC) or gaseous (GOC) forms, with POC and DOC evaluated in-situ using 1 × 1 m2 runoff plots under natural rains and soil CO2 emissions assessed in the laboratory from undisturbed 0–0.05 m soil samples. Moreover, a comparison between SOC outputs and the main soil, topography and vegetation control parameters (e.g. soil texture, SOC concentration, cation exchange capacity: CEC, soil bulk density but also grass basal cover, mean slope gradient, drainage area and compound topographic index) was performed to elucidate the underlying processes much more extensively.

2. Materials and methods 2.1. Characteristics of the study area The study area is located in Potshini community of the Drakensberg region, Kwazulu-Natal province, South Africa (Fig. 1). It is marked by a temperate climate with a summer rainfall pattern with a mean annual precipitation of 684 mm per annum, a potential evaporation of 1600 mm per annum and a mean annual temperature of 13 °C (Schulze, 1997). The area is under natural rangeland with no fertilization and regular winter grass burns to prevent the rangeland from becoming woody. Features of land degradation such as a decrease of grass basal cover and the presence of patches of bare soils are common in the area. At Potshini, altitude ranges from 1381 to 1492 m.a.s.l., the relief being relatively gentle with a mean slope gradient of 16%, but with a maximum value of 70% found on the midslope, whereas bottomland and plateau are flat (slope gradient between 2 and 4%).

Fig. 1. Location of the study site in South Africa. Variations of soils and grass basal cover and position of the 1 m × 1 m microplots at the study slope positions.

V. Chaplot, M. Cooper / Geoderma 243–244 (2015) 205–213

The geology in the area exhibits a succession of fine-grained sandstone, schists, siltstone and mudstone in which dykes from the Karoo Dolerite are generally intruded (King, 2002). The soils deriving from these bedrocks are all classified as Acrisols (W.R.B., 1998). An extensive description of the soils was performed at different hillslope positions as follows: footslope (F) position shows deep (thickness N 2 m) welldrained soils. The humiferous A horizon is dark reddish brown (5 YR 3/3), blocky and friable. The Bw horizon, from 0.4 to 0.9 m, is dark reddish brown (5 YR 3/3), massive and clayey. A sandy saprolite is reached at about 1.7 m. The backslope position (B) exhibits a similar soil profile but much shorter Bw being found between 0.3 and 0.6 m and the saprolite from 0.9 m. The soils at middleslope (M) and shoulder (S) are deep. Soils at M show a dark reddish brown (5 YR 3/3) A horizon with a clear fine angular blocky structure. The Bw is red (2.5 YR 4/6) and a sandy red (10 R 4/8) saprolite is reached below. Soils at shoulder position but with a dolerite influence (SD) are marked by a 0.1–0.15 m brown (7.5YR 4/4) friable and humiferous horizon with fine angular blocky structure. The shoulder soils developed from sandstones (SS) show a yellowish-red (5 YR 5/7) and friable Bw horizon found from 0.5 to 1.4 m.

207

2, 1, 0.5, 0.2, 0.1 and 0.05 mm and were weighted to evaluate their mean weight diameter (MWD) calculated as follows: X MWD ¼

ðxi  wi Þ

ð1Þ

100

with x, the mean intersieve size and wi, the percentage of particles retained on each i sieve. The greater the MDW is, the more resistant to desegregation the soil is. 2.4. Assessment of SOC outputs from soils This study considered two forms of eroded organic C during the process of water erosion, namely particulate organic carbon (POC) and dissolved organic carbon (DOC). POC and DOC losses (POCL and DOCL, respectively) were evaluated in-situ at the 24 runoff plots and for 22 erosive events of 2012–2013, resulting in 528 measurements. After each rainfall event, the runoff stored in the totalisators placed at the bottom end of the plots had its volume evaluated. The runoff was then stirred and two aliquot samples (500 ml) were taken for the estimation of POC and DOC concentration (POCC, DOCC, respectively).

2.2. The experimental design OM outputs from soils were evaluated in conjunction with soil aggregate stability at data points from five landscape positions described above and showing different soil characteristics and grass basal cover (Table 1, Fig. 1). Three data point replicates were selected at each hillslope position, resulting in 15 data points plus nine additional data points differing by their basal cover (Cov, from 5 to 95% through 45%) at F and installed at slightly less clayey Acrisols (Table 1).

2.4.1. Determination of dissolved organic carbon (DOC) The first aliquot was filtered at 0.45 μm and analyzed for DOCC using a Shimadzu TOC-5000 analyzer with an ASI-5000 autosampler and Balston 78–30 high purity C gas generator. 2.4.2. Determination of particulate organic carbon (POC) The second aliquot was oven-dried at 50 °C and weighed and the POC concentration in the dry sediment was determined using a LECOTruMAC dry combustion furnace. The POC data were corrected by the DOC amounts because the sediment after drying also contains the DOC part. POCL from a rainfall event and plot corresponded to the product of POCC in the sediment and total mass of the sediments, while DOCL corresponded to the product of DOCC and the runoff volume. The losses from the 22 storm events were finally summed to obtain yearly erosion rates.

2.3. Estimation of the stability of soil aggregates Undisturbed 0–0.05 m bulk soil was also collected at each of the 24 data point for the estimation of aggregate stability with three replicates per point resulting in 72 measurements. In addition, forty five data points (n = 135 measurements) were randomly selected within the entire hillslope in order to investigate the controlling factors of soil aggregate stability. The stability of soil aggregates was evaluated for the resulting 207 data points following the mechanical breakdown modality of Le Bissonnais (1996) laboratory test, which reproduces destruction of soil aggregates by energy from external agents such as raindrops, sheetwash or tillage. Following Le Bissonnais (1996), aggregates (2–5 mm in size) were obtained by dry sieving of the bulk soil. The aggregates (5–10 g) were gently immersed in 50 ml distillated water for 10 min. Thereafter, they were immersed in ethanol before being rotated end-over-end 10 times. The resulting aggregates collected on sieves of

2.4.3. Gaseous SOC At the end of the rainy season, undisturbed 0–0.05 m soil samples (n = 24) were collected near the microplots using metallic cores (0.064 m in diameter and 0.05 m in height). The samples were transferred to PVC cores before incubation in 600 ml open-top jars for 20 days at 28 ± 2 °C, in the laboratory. Following Fang and Moncrieff (1998), the room was kept at a humidity of 100%. In order to avoid

Table 1 General characteristics of soils and sediments at different landscape positions (LP: footslope, backslope, midslope and shoulder): mean texture (Text); basal cover (Cov); soil suface crusting (Crust); Mean Weight Diameter (MWD), POCC, POCL, DOCC and DOCL for particulate and dissolved organic carbon concentration and loads in sediments and runoff; CO2Fgs for CO2 fluxes per gram of soil; CO2Fgc for CO2 fluxes per gram of soil carbon; ρb, soil bulk density; SOCC, top-soil (0–0.05 m) organic carbon concentration; ER, sediment enrichment in OC compared to the bulk soil; top-soil clay, silt and sand concentration. Three replicates per situation and a total of 22 rainfall events per plot were considered. LP

Texta

Cov

Crust

% Foot Foot Foot Foot Back Mid Should Should mean a

SL SL SL SCL SCL CL C SCL

5 45 95 96 52 98 94 83 71

95 55 5 4 48 2 6 17 29

S for sand, C for clay, L for loam.

MWD

POCC

DOCC

POCL

DOCL 2

mm

g/kg

mg C/l

g C/m /y

0.7 0.9 1.4 3.3 2.8 3.1 3.4 3.1 2.3

44 52 119 90 170 190 170 130 120.6

3.7 7.2 4.6 58.2 20.3 54.0 30.7 90.0 33.6

147 67 42 20 37 42 37 29 52.6

5.3 8.4 1.9 1.3 0.4 1.2 0.7 2.0 2.7

CO2Fgs

CO2Fgc

ρβ

SOCC 3

Mg C/g

Mg C/g C

g/cm

g C/kg

0.3 0.3 0.8 3.6 2.5 5.4 5.5 3.2 2.7

33.3 28.2 33.8 126.3 139.7 135.0 143.6 112.7 94.1

1.34 1.31 1.22 1.25 1.33 1.13 1.02 1.12 1.2

8.4 10.3 22.5 28.5 17.9 40.0 38.3 28.4 24.3

ER

Clay

5.2 5.0 5.3 3.2 9.5 4.8 4.4 4.6 5.2

15 15 18 28 27 40 54 31 28.5

Silt

Sand

13 13 15 22 20 29 28 19 20

72 72 67 50 53 31 18 50 52

%

208

V. Chaplot, M. Cooper / Geoderma 243–244 (2015) 205–213

significant changes in soil water concentration (due for instance to evaporation) necessary amounts of water were added to the samples to maintain constant weights during the course of the experiment (e.g. Wienhold, 2007). The jars were sealed every day with air-tight lids for 4 min to allow CO2 accumulation. Carbon dioxide emissions were assessed once every day using a PP system EGM-1 carbon CO2 analyzer and following Chaplot et al. (2012). The cumulative CO2 emissions during the incubation period of 20 days were reported both in milligram of C per gram of soil (mg C g−1) (CO2Fgs) and in milligram of C per gram of soil C (mg C g C−1) (CO2Fgc). 2.5. Selected soil and environment factors of control Several soil and environmental factors were selected to further the understanding of the processes governing organic C outputs from soils as well as to evaluate the possibility to predict and spatially map MWD and SOC outputs. The selected factors include basal grass cover, soil bulk density, soil texture, SOC concentration (SOCC), cation exchange capacity, concentration in exchangeable cations, elevation above sea level, slope gradient, specific monodirectional area and compound topographic index. Determination of grass basal cover (Cov) was done following Podwojewski et al. (2011), a method which is based on the use of a laser distance meter (Leica Disto. Pro, laser class 2–635 nm, LEICA geosystems AG, CH-9435 Heerbrugg, Switzerland) mounted 1.0 m above the plots to evaluate the occurrence of bare soil or grass at the node of a 0.05 m grid. Measurements of soil bulk density (ρb) were performed using three undisturbed soil cores (250 ml) within a radius of less than 1 m from each plot with samples dried at 110 °C for 48 h before weighting. At each data point, a composite sample was collected and sieved to obtain the clay fractions (b2 μm; AFNOR, 1992). The particle size distribution was determined by the pipette method and the SOC concentration within the soil samples was estimated using a LECO-TruMAC dry combustion furnace. Finally, the cation exchange capacity (CEC) and the exchangeable cations (Ca++, Mg++, Na+, K+, Fe++ and Al3 +) were determined by exchanging with cobalthexamine cations at natural soil pH (AFNOR, 1992). The terrain attributes of elevation above sea level (Z), slope gradient (SG), specific monodirectional area (As) and modified compound topographic index, CTI = ln (As/SG) (Chaplot et al., 2000) were extracted from an available 5 m DEM (Orchard et al., 2013) and using the Grid analyst, DEMAT and BASIN extensions of ArcView 3.2. 2.6. Statistical analysis The data were a little skewed and kurtotic, for both erosion variables and soil and environmental factors but did not significantly differ from normality (skewness and kurtosis z values neither below − 1.96 nor above +1.96) (DeCarlo, 1997). Pearson r coefficient was calculated to test the one to one relationships between variables and t-tests were applied to inform on the significance of the differences. In complement, Principal Component Analysis (PCA) was applied to the data to assess the multiple correlations between organic C outputs from soils, stability of soil aggregates and selected soil and environmental factors. PCAs convert the variables into so-called principal components (PCs) (use of Table Orthonormal option of ADE4 software, Chessel et al., 2004), which are linear combinations of the actual variables, not correlated with each other (i.e. they are orthogonal) and together explaining the total variance of the data (Jambu, 1991). Two PCAs were generated for this study. The first one evaluated the relationships between the selected soil and vegetation parameters (MWD, SOCC, clay and sand concentration, soil bulk density and Cov), which were used for PCs generation, and the organic C outputs variables. The second PCA investigated the correlations

between selected soil and terrain attributes on the one hand and MWD on the other. 3. Results 3.1. Organic C outputs from soils The average particulate organic carbon concentration (POCC) in the sediments exiting the microplots was 120.6 g C kg−1. Based on t-tests, this was significantly higher (P b 0.05) than the organic C concentration in the bulk soil of 28.4 g C kg−1; the water erosion process induced an organic C enrichment by a factor of 5.2. The C enrichment ratio (ER) was greatest at the backslope position (ER = 9.5), where the soils were sandy-clay and the basal grass cover was 50% (Table 1). In contrast, ER was smallest at the footslope (ER = 3.2), under sandy clay loams, well covered by the grass. The total POC losses (POCL) were on average, 52.6 g C m−2 y−1; with a maximum of 147 g C m−2 y−1 at the footslope under the most degraded grassland and a minimum of 20 g C m−2 y−1 at the same hillslope position but under full grass coverage. The concentration of dissolved organic carbon (DOCC) was on average of 33.6 mg C l−1 and ranged between 3.7 mg C l−1 at the degraded footslope grassland and 90 mg C l−1 at shoulder (Table 1). DOCC of over 30 mg C l−1 was found under basal covers greater than 90% at the footslope, midslope and shoulder positions. In contrast, the lowest DOCC characterized the degraded grassland at the footslope. The resulting DOC losses (DOCL) computed from the 22 erosive events of the 2012–2013 rainy season ranged between 0.4 g C m− 2 y− 1 at the backslope and 8.4 g C m− 2 y− 1 at the degraded footslope. At 2.7 g C m−2 y−1, the average DOCL was much lower than the average POCL of 52.6 g C m− 2 y−1, a 19.5 times difference (significant at P b 0.01). On average, 95.1% of the total SOC losses exported by water erosion each year from the soils were in a particulate form. The greatest POCL was 147 g C m−2 y−1 and occurred at the lowest grass basal cover, while the sandy clay loams at footslope and shoulder experienced the lowest POCL. DOCC of below 10 mg C l−1 characterized sandy loam soils and this, whatever the landscape position and grass coverage. 3.1.1. Gaseous SOC The cumulative 20-day SOC losses as CO2 emissions were on average of 2.7 mg C g−1 for CO2Fgs (i.e. emissions reported in milligram of carbon CO2 per gram of soil) and 94.1 mg C g C−1 for CO2Fgc (i.e. emissions of gram of C-CO2 per gram of soil carbon). CO2Fgs below 0.3 mg C g−1 occurred at all the sandy loam soils while rates over 3 mg C g−1 characterized the sandy-clay and clay soils with a tendency for CO2Fgs to be slightly lower at backslope position where the basal cover was 52% only. CO2Fgc ranged from between 28.2 and 143.6 mg C g C−1 with an average at 94.1 mg C g C−1. The lowest CO2Fgs corresponded to the lowest CO2Fgc and the highest CO2Fgs corresponded to the highest CO2Fgc. Interestingly, the backslope position, which showed intermediary levels of CO2Fgs (2.5 mg C g C−1), emitted one of the highest rate of CO2 per gram of SOC (139.7 mg C g C−1) (Table 1). To compare the order of magnitude of SOC losses through gaseous emissions to this of particulate and dissolved forms, the CO2Fgs data were extrapolated to an entire year. Deducing yearly emissions in the field from short duration incubation experiments in the laboratory can be highly speculative. However, assuming that the CO2 emissions during the incubation period were representative of the in-situ emissions during the rainy season period of 6 months and that little emissions occurred during the dry and cold winter season, the yearly CO2 losses from the 0–0.05 m layer would amount to 64.8 g C m−2 y−1. This would be of the same order of magnitude of the sum of particulate and dissolved losses of 55.3 g C m−2 y−1.

V. Chaplot, M. Cooper / Geoderma 243–244 (2015) 205–213

3.2. Soil aggregate stability and its correlation with OM outputs The Mean Weight Diameter (MWD) of the soils was 2.3 mm. Values ranged between 0.7 mm, which corresponded to very unstable soil aggregates (Le Bissonnais, 1996) and 3.4 mm, which corresponded to very stable aggregates. The lowest aggregate stabilities were encountered at footslope, under bare and sandy soil conditions, while the highest stabilities were found at the hillslope shoulder under clayey soils with high basal cover (Table 1). From Table 1 we also learn that the highest correlation between MWD and the C output variables occurred for CO2 emissions with a Pearson r coefficient as high as 0.91 for CO2Fgs and 0.95 for CO2Fgs. In addition, MWD significantly increased as POCC increased (r = 0.76), while POCL lessened significantly (r = −0.77). DOCC significantly increased

209

with increasing MWD (r = 0.76), while DOCL increased with decreasing MWD (r = −0.81). Fig. 2 depicts the differences derived from the mean values for each of the three MWD classes. POCC sharply increased (Fig. 2A) while POCL sharply decreased (Fig. 2B) from 108 g C m−2 y−1 at MWD b 1.5 mm to 40 g C m−2 y−1 at 1.5 b MWD b 2.5 mm, followed by a slight decrease to 35 g C m−2 y−1 at MWD N 2.5 mm. DOCC significantly increased with increasing MWD from below 20 mg C l− 1 at MWD b 2.5 mm to 55 mg C l−1 at MWD above 2.5 mm (Fig. 2C). In contrast DOCL increased with decreasing MWD from above 6 g C m2 y− 1 at MWD b 1 mm to below 1.5 g C m2 y−1 at higher aggregate stabilities (Fig. 2D). In the case of CO2Fgs, values rose from 0.02 mg C-CO2 g− 1 at MWD b 1 mm to 1.6 mg C-CO2 g−1 at 1 b MWD b 2.5 mm and to 4.5 mg C-CO2 g−1 at MWD N 2.5 mm, which corresponded to an exponential increase by

Fig. 2. Mean weight diameter (MWD) and particulate organic carbon concentration (POCC) (A), POC losses (POCL (B); dissolved organic carbon concentration (DOCC) (C), DOC losses (DOCL) (D), and CO2 losses per gram of soil (CO2Fgs, E) and per gram of soil (CO2Fgc, F). Bars correspond to standard errors. n = 528.

210

V. Chaplot, M. Cooper / Geoderma 243–244 (2015) 205–213

22,500% (Fig. 2E), while CO2Fgs increased linearly from 31 to a maximum of 133 mg C-CO2 g C−1, which corresponded to a 429% rise (Fig. 2F).

3.3. OM outputs and the selected soil and environmental factors POCC increased with rising SOCC, Clay and Cov values (r N 0.6), but decreased with increasing sand concentration (r = − 0.8) and soil bulk density (r = −0.63). In contrast, POCL lessened significantly with increasing Cov (r = −0.86), SOCC (r = −0.66) and Clay (r = −0.44). DOCC displayed a similar behavior to that of POCC, i.e. an increase with increasing SOCC (r = 0.62), Cov (r = 0.51) and Clay (r = 0.47), but a decrease with decreasing ρb (r = −0.54). Finally, DOCL increased with decreasing SOCC (r = − 0.73), Cov (r = − 0.64) and Clay (r = −0.65). CO2Fgs showed significant correlations with ρb, SOCc, clay, silt and sand concentration. The gaseous emissions increased as SOCc, clay and silt concentration increased but decreased as sand concentration and ρb increased. The multiple correlations between the OM output variables and the selected soil and environmental factors was investigated by the PCA displayed in Fig. 3A. The first two axes of the PCA explained 92% of the data variance. The first axis (Axis 1) opposed SOCC, MWD, Clay and

Cov with positive coordinates to sand and ρb with negative coordinates. Axis 1 can thus be interpreted as an axis of soil aggregate stability with a coordinate of 1 on this axis corresponding to the highest aggregate stability and a coordinate of −1 corresponding to the lowest aggregate stability. Positive coordinates on Axis 1 were found for POCC, DOCC and CO2 emissions, which was indicative of an increase in runoff content of particulate and dissolved organic C as the stability of soil aggregates increases. In contrast, negative coordinates on Axis 1 for POCL and DOCL point to a decrease in organic C losses as soil aggregate stability increases. In addition SOC losses tended to rise in conjunction with the increase in soil bulk density and sand concentration. 3.4. Relationships between soil aggregate stability and the selected soil and environmental factors Table 2 displays the r coefficients between MWD and selected soil, vegetation and terrain factors, whose basic statistics are displayed in Table 3. Soil crusting, basal cover and texture showed correlation coefficients with MWD higher than 0.55, while in the case of the terrain attribute r values were below 0.15. The multiple correlations between MWD and the factors of control are displayed in Fig. 3B. The two first axes of the PCA explained 48% of the total variability of the data. Axis 1 opposed soil crusting with positive coordinates to basal cover, organic C and clay concentration with negative coordinates. This axis was interpreted as an axis of grassland degradation. The second PCA axis, Axis 2, opposed As and CTI on the one hand to Z and SG on the second hand. It was thus interpreted as an axis of proximity to the hydrologic network with high values of AS and CTI corresponding to high proximities. The three classes of MWD, from 1 mm b MWD (class 1) to MWD N 2.5 mm (class 3) stretched along Axis 1 with the lowest aggregate stabilities (class 1) corresponding to the most degraded grassland soils, those with surface crust, low grass basal cover and soil clay concentration. In contrast, MWD classes exhibited low correlation with Axis 2 and the selected terrain attributes, which confirms the results from Table 2. 4. Discussion 4.1. Stability of soil aggregates and C outputs from soils

Fig. 3. Principal Components Analysis (PCA) showing the correlation between selected soil and environmental variables for axes generation and SOM output variables as secondary variables (A); PCA between MWD (number 3 = high values of MWD and standard deviation of MWD; number 2 = medium values of MWD and standard deviation of MWD; umber 1 = low values of MWD) and the selected environmental variables (B).

4.1.1. Soil aggregate stability and C outputs by water erosion SOC erosion by water is a natural process that has only recently been recognized to have a significant impact on the persistence of OM in soils and terrestrial ecosystems as a whole (Berhe et al., 2012). The present study showed for instance that, each year, an average of 55.3 g of organic C per square meter is eroded from soils, which corresponded to 18.9% of the 0 to 0.05 m soil stock of 0.2916 kg C m−2. When compared to the 0–0.3 m SOC stocks, using data from Dlamini et al. (2010) from the same site, the proportion of the SOC stock eroded each year would be 4.7%. This was significantly higher than the 0.01% reported in Spain and under arid conditions by Rodriguez et al. (2004), but greatly lower than the 16% reported by Roose (1977) under sandy and crusted soils in Burkina Faso. The present study pointed to high particulate and dissolved organic C losses in soils with low aggregate stability. High SOC losses can be explained by high runoff rates as pointed out at the same site by Oakes et al. (2012), because soils with unstable aggregates tend to display low soil infiltration by water (Le Bissonnais, 1996). In contrast, soils with the most stable soil aggregates produced sediments and runoff with the highest concentration in particulate and dissolved organic C, but owing to low runoff rates (Oakes et al., 2012), total organic C losses appeared to be the lowest. Soils with stable aggregates were shown to produce sediments with higher organic C concentration than this of the bulk soil. This result might be explained by the light nature of OM and the existence in soils of complex but reversible associations with soil minerals, which

−0.04 −0.08

−0.14 −0.18

0.17

−0.48 0.52 0.68 −0.97⁎ −0.83⁎ 0.49 −0.58 −0.72⁎ 0.98⁎ 0.86⁎

−0.36 0.14

0.88 −0.23

−0.71⁎ −0.79⁎ −0.14 0.29 −0.09

⁎ Significant at p b 0.05 following a t-test.

0.61 −0.59 −0.63 0.06

−0.45 0.61

0.38 −0.62 −0.81⁎ 0.79⁎ 0.80⁎

0.76⁎ 0.76⁎ −0.76⁎ −0.81⁎ 0.91⁎ 0.95⁎

−0.57 −0.45 0.63 0.62 −0.37

CTI As

0.11

SG Z

0.06 −0.84⁎ −0.80⁎

Sand

0.86⁎ 0.83⁎

Silt

0.81⁎ 0.77⁎ 0.47 −0.49 −0.65 0.94⁎ 0.79⁎

Clay Cov

0.67 0.62 0.51 −0.86⁎ −0.64 0.66 0.48 −0.37 −0.57 −0.50 −0.46 0.88⁎ 0.41 −0.57 −0.38 0.40

Crust ER

4.2. On the potential fate of the eroded POC and DOC

−0.09 0.29 −0.37 0.06 −0.14 −0.23 0.14 0.83⁎ 0.78⁎ 0.62 −0.66 −0.73⁎ 0.94⁎ 0.73⁎

SOCc

4.1.2. Soil aggregate stability and C outputs by decomposition There is consensus in the literature that aggregates protect OM from decomposers through physical disconnection (Schmidt et al., 2011). In this study the soils with the greatest aggregate stability yielded the highest gaseous emissions. A possible explanation of the greater CO2 outputs from stable aggregates could lie in the greater amount of OM and in differences in OM quality. Binding agents, generally polysaccharides from exocellular mucilages and root exudates and found in between aggregates, may constitute an active pool (von Lützow et al., 2008) potentially leading to high CO2 emissions. Fresh plant residues, biological feces, and other faunal and microbial residues, which constitute other pools of easily accessible OM to decomposers and potentially leading to higher CO2 emissions (von Lützow et al., 2008), are likely to be found in higher proportion in soils of high aggregate stability than in those of low aggregate stability.

−0.65 −0.63 −0.54 0.50 0.51 −0.79⁎ −0.50 0.44

ρb CO2FgC

0.95⁎ 0.80⁎ 0.62 −0.63 −0.79⁎ 0.88⁎ 0.91 0.79 0.63 −0.59 −0.71

CO2Fgs DOCL

−0.81⁎ −0.81⁎

−0.77⁎ −0.62 −0.57

POCL DOCc

0.76⁎ 0.38 0.76⁎

POCc MWD

211

generally result in C enrichment in the sediments compared to the bulk soil (Gregorich et al., 1998; Jacinthe and Lal, 2001; Chaplot et al., 2005). The average ER value at the study site was 5.2. This was slightly higher than the value of 3.3 observed in Kenya by Boye and Albrecht (2006) and 4.3 in a study performed in Burkina Faso (Bilgo et al., 2006). This C enrichment ratio was however much lower than this observed in a semi-arid environment with sandy soil conditions in Mali where the ER was as high as 10.0 (Drissa et al., 2004).

MWD POCc DOCc POCL DOCL CO2Fgs CO2FgC ER

Table 2 Correlation matrix between the variables of interest: Mean Weight Diameter (MWD). POCC, POCL, DOCC and DOCL for particulate and dissolved organic carbon concentration and loads in sediments and runoff; CO2Fgs for CO2 fluxes per gram of soil; CO2Fgc for CO2 fluxes per gram of soil carbon; ρb. soil bulk density; SOCC. top-soil (0–0.05 m) organic carbon concentration; ER. sediment enrichment in OC compared to the bulk soil; Crust. soil surface crusting; Cov. basal cover; top-soil clay, silt and sand concentration. Z for altitude above sea level. SG for mean slope gradient. As for multidirectional drainage area and CTI for compound topographic index. Numbers displayed within the table are coefficients of correlation.

V. Chaplot, M. Cooper / Geoderma 243–244 (2015) 205–213

Literature largely disagrees on the fate of the eroded SOC, in both particulate and dissolved forms. Smith et al. (2001) argue that only a tiny fraction of the eroded POC is decomposed and released as CO2 to the atmosphere, which contradicts the initial research findings by Schlesinger (1995) which suggested that eroded POC is oxidized to the greatest extent. Chaplot et al. (2005), in the sloping lands of Southeast Asia, pointed to large downslope redeposition of initially eroded POC with large OM accumulations at downslope position. Berhe and Kleber (2013) pointed to the fact that upon deposition, the transported light OM particles can form new physical or chemical associations with the minerals which have potential to enhance OM persistence in soils at the depositional sites. In addition to this, deep burial by successive layers of sediments is another process which renders the deposited OM less likely to become decomposed (Berhe et al., 2012). The soils with the highest aggregate stability under the current study produced the lowest runoff rates but the highest POC and DOC concentration. Because of low transport capacity, this runoff is more likely to deposit its POC in thin layers and to infiltrate DOC to nearby eroding sites and on the soil surface. The presence of deposited POC on the soil surface, where intense biological activity occurs, is likely to result in high decomposition rates. In contrast, greater overland flow kinetic energy under soils with low aggregate stability has the potential to transport the eroded OM on long distances, to reach the very bottom of hillslopes where flat topography induces the sedimentation of large amounts of eroded material. The burial of POC in depth where suboptimal conditions for decomposers exist, is likely to foster OM stabilization. We argue that the stability of soil aggregates highly impacts on the persistence of eroded OM directly through the control of OM detachment and displacement from its place of origin, and indirectly by affecting sedimentation and its environment, such as the interdependence of depth of burial, soil redox state and the presence of potential degraders in the immediate microenvironment. 4.3. Implications of the results for C output predictions and the Global C Cycle modeling The present study, which pointed to a link between OM outputs from soils and soil aggregate stability on the one hand, and several

212

V. Chaplot, M. Cooper / Geoderma 243–244 (2015) 205–213

Table 3 General statistics for Mean Weight Diameter (MWD) and the environmental factors of soil surface coverage by vegetation (Cov), soil surface crusting (Crust), soil clay concentration (Clay), above sea level (Z), slope gradient (SG), specific catchment area (As), and compound topographic index (CTI). n = 72 for Cov to Clay; n = 207 from Z to CTI. MWD

Mean SD Var. Min Q1 Median Q3 Max Skew Kurt SE CV

Cov

mm

%

2.30 0.42 0.03 0.70 0.98 2.46 3.25 3.38 −0.7 0.3 0.06 3.8

94.1 4.8 23.0 77.2 91.4 95.1 97.9 99.9 −1.2 2.2 0.61 5.1

Crust

5.8 4.8 23.4 0.1 1.6 4.9 8.7 22.9 1.2 2.1 0.61 82.7

Clay

31.5 7.4 54.8 17.8 26.8 30.6 35.6 54.3 0.4 −0.1 0.9 23.5

Z

SG

As 2

CTI

m

m/m

m /m

m/r

1449 35 1218 1383 1417 1459 1477 1492 −0.7 −0.9 4.4 2.4

12.3 8.8 77.0 0.9 5.9 10.3 14.5 43.2 1.4 2.6 1.1 71.5

178 883 7.7e + 6 0 0 1 12 6856 7 55 112 496

8.7 1.8 3.4 6.1 7.3 8.6 9.6 14.9 1.0 1.2 0.23 21.1

environmental variables on the second hand, has the potential to improve the current predictions of organic C outputs from soils. Because soil aggregate stability is relatively easy to access compared to the tedious in-situ monitoring of C losses, it could be used in routine to predict potential particulate, dissolved and gaseous C outputs from soils. Evaluating soil aggregate stability requires a certain amount of field and laboratory work, which could become tedious, especially when large areas are considered. Variables such as soil texture, SOCC or basal grass cover, which were shown to correlate to soil aggregate stability are much easier to access thanks to the use of soil maps or remote sensors. These might thus enable potential organic C outputs from soils in either particulate, dissolved or gaseous forms to be spatially mapped. Moreover, the established links between C outputs from soils, soil aggregate stability and environmental variables could be integrated in existing models of soil C for improved predictions. These models are for the most part based on the assumption that organic C losses from soils are a function of the decay rate of the different organic material pools (von Lützow et al., 2008). They would probably greatly benefit from the incorporation of a more mechanistic approach aiming not only at considering gaseous outputs but also particulate and dissolved ones through the process of water erosion. 5. Conclusions This study evaluated the link between soil aggregate stability and SOC losses due to decomposition and water erosion and revealed the following three main results: – The highest dissolved SOC concentration in runoff and particulate SOC in the sediments exported from soils during rainstorms characterized the clayey soils with the most stable aggregates. These soils exhibited, however, the lowest SOC losses by water erosion, owing to low levels of runoff rates; – The soils with the most stable soil aggregates yielded significantly higher CO2 emissions than the soils with low aggregate stability; – Soil clay and SOC concentrations and grass basal cover appeared as accurate proxies of with aggregate stability and organic C outputs from soils.

These results, which pointed to a significant impact of soil aggregate stability on organic C losses from soils, allow one to better understand the role of soil structure in the control of soil C dynamics. Such quantification of the relationships between soil C outputs, soil aggregate stability and environmental factors opens the way for land management practices aiming at mitigating against both land degradation and climate change. For instance, organic C losses from soils due to water

erosion could be significantly reduced by increasing grass basal cover, an important control of soil aggregate stability. Other soil aggregating agents such as Ca++ (Albiach et al., 2001; Spaccini et al., 2002; Garcıa-Orenes et al., 2005) could also probably lessen erosion-induced SOC outputs but the use of these would need further appraisal. More is also to be done on the use of aggregate stability as a proxy of not only SOC outputs but also on the fate of the eroded organic C from headwaters to oceans. Finally, this results on a more mechanistic approach of the dynamics of terrestrial organic C and the transfers from soils to the hydrosphere and the atmosphere could be integrated in existing models of the Global C Cycle for improved predictions. Acknowledgment The authors gratefully acknowledge the Institut de Recherche pour le Développement (IRD, former ORSTOM), the Institut National des Sciences de l'Univers (INSU, project EC2CO DESTOC) and the Water Research Commission (project K5/1904-2266) for funding this research and for the SAAES and CWRR of UKZN for providing support and assistance. We also thank the Potshini community for kindly allowing the research to be performed on their land, and for technical and logistical assistance. Final thanks for the two anonymous reviewers for their very constructive comments and for Macdex Mutema for editing the final version of the paper. References AFNOR NF X 31–108, 1992. Détermination of cations Ca, Mg, K, Na extractibles par l'acétate d’ammonium. Qualité des sols. AFNOR, Paris, pp. 59–65. Albiach, R., Canet, R., Pomares, F., Ingelmo, F., 2001. Organic matter components and aggregate stability after the application of different amendments to a horticultural soil. Bioresour. Technol. 76, 125–129. Batjes, N.H., 1996. Total carbon and nitrogen in the soils of the world. Eur. J. Soil Sci. 47, 151–163. Berhe, A.A., Harden, J.W., Torn, M.S., Kleber, M., Burton, S.D., Harte, J., 2012. Persistence of soil organic matter in eroding vs. depositional landform positions. J. Geophys. Res. Biogeosci. 117, G02019. http://dx.doi.org/10.1029/2011JG001790. Berhe, A.A., Kleber, M., 2013. Erosion, deposition, and the persistence of soil organic matter: mechanistic considerations and problems with terminology. Earth Surf. Process. Landf. 38, 908–912. Bilgo, A., Serpantie, G., Masse, D., Fournier, J., Hien, V., 2006. Carbon, nitrogen, and fine particles removed by water erosion on crops, fallows, and mixed plots in Sudanese savannas (Burkina Faso). In: Roose, E., Feller, C., Barthès, B., Stewart, B. (Eds.), Soil Erosion and Carbon Dynamics. CRC Publisher, USA, pp. 125–142. Boye, A., Albrecht, A., 2006. Soil erodibility, control and soil carbon losses under short-term tree fallows in western Kenya. In: Feller, C., Roose, E.J., Stewart, B.A., Barthes, B., Lal, R. (Eds.), Soil erosion and Carbon Dynamics. CRC Press, USA, pp. 181–196. Chaplot, V., Rumpel, C., Valentin, C., 2005. Water erosion impact on soil and carbon redistributions within the Mekong basin. Glob. Biogeochem. Cycles 19. http://dx.doi.org/ 10.1029/2005GB002493 GB4004. Chaplot, V., Mchunu, C., Manson, A., Jewitt, G., Lorentz, S., 2012. Water erosion-induced CO2 emissions from tilled and no-tilled maize plots in Africa. Agric. Ecosyst. Environ. 159, 62–69. Chaplot, V., Walter, C., Curmi, P., 2000. Improving soil hydromorphy prediction according to DEM resolution and available pedological data. Geoderma 97, 405–422. Chessel, D., Dufour, A.-B., Thioulouse, J., 2004. The ade4 package. R News 4, 5–10. DeCarlo, L.T., 1997. On the meaning and use of kurtosis. Psychol. Methods 2, 292–307. http://dx.doi.org/10.1037/1082-989X.2.3.292. Dlamini, P., Podwojewski, P., Janeau, J.L., Lorentz, S., Jewitt, G., Chaplot, V., 2010. Controlling factors of interrill erosion of sloping-land-agricultural-soils in KwaZulu Natal (South Africa). Agric. Water Manag. 98, 1711–1718. Drissa, D., Orange, D., Roose, E., 2004. Influence des pratiques culturales et du type de sols sur les stocks et pertes de carbone par érosion en zone soudanienne du Mali. In: Feller, C., Roose, E.J., Stewart, B.A., Barthes, B., Lal, R. (Eds.), Soil erosion and Carbon Dynamics. CRC Press, USA, pp. 193–207. Fang, C., Moncrieff, J.B., 1998. An open-top chamber for measuring soil respiration and the influence of pressure differences on CO2 efflux measurement. Funct. Ecol. 12, 319–325. Garcıa-Orenes, F., Guerrero, C., Mataix-Solera, J., Navarro-Pedreno, J., Gomez, J., MataixBeneyto, J., 2005. Factors controlling the aggregate stability and bulk density in two different degraded soils amended with biosolids. Soil Tillage Res. 82, 65–76. Gregorich, E.G., Geer, K.J., Anderson, D.W., Liang, B.C., 1998. Carbon distribution and losses: erosion and deposition effects. Soil Tillage Res. 47, 291–302. Jacinthe, P.A., Lal, R., 2001. A mass balance approach to assess carbon dioxide evolution during erosional events. Land Degrad. Dev. 12, 329–339. Jambu, M., 1991. Exploratory and Multivariate Data Analysis. Academic, Orlando, FL, p. 474.

V. Chaplot, M. Cooper / Geoderma 243–244 (2015) 205–213 King, G.M., 2002. An Explanation of the 1:500 000 General Hydrogeological Map. Dept. Of Water Affairs and Forestry, Pretoria, Republic of South Africa. Lal, R., 2003. Soil erosion and the global carbon budget. Environ. Int. 29, 437–450. Le Bissonnais, Y., 1996. Aggregate stability and assessment of soil crustability and erodibility: theory and methodology. Eur. J. Soil Sci. 47, 425–437. Maïga-Yaleu, S., Guiguemde, I., Yacouba, H., Karambiri, H., Ribolzi, O., Bary, A., Ouedraogo, R., Chaplot, V., 2013. Soil crusting impact on soil organic carbon losses by water erosion. Catena 107, 26–34. Novara, A., Armstrong, A., Gristina, L., Semple, K.T., Quinton, J.N., 2012. Effects of soil compaction, rain exposure and their interaction on soil carbon dioxide emission. Earth Surf. Process. Landf. 37, 994–999. Oakes, E.G.M., Hughes, J.C., Jewitt, G.P.W., Lorentz, S.A., Chaplot, V., 2012. Controls on a scale explicit analysis of sheet erosion. Earth Surf. Proc. Landf. 37, 847–854. Orchard, C., Lorentz, S., Jewitt, G., Chaplot, V., 2013. Spatial and temporal variations of overland flow at the microcatchment level. Hydrol. Process. 27, 2325–2338. http:// dx.doi.org/10.1002/hyp.9217. Podwojewski, P., Janeau, J.-L., Grellier, S., Valentin, C., Lorentz, S., Chaplot, V., 2011. Influence of grass soil cover on water runoff and soil detachment under rainfall simulation in a sub-humid South African degraded rangeland. Earth Surf. Process. Landf. 36, 911–922. Rodriguez, A., Guerra, A., Arbelo, C., Mora, J.-L., Gorrin, S.P., Armas, C., 2004. Forms of eroded soil organic carbon in andosols of the Canary Islands (Spain). Geoderma 121, 205–219. Roose, E., 1977. Erosion et ruissellement en Afrique de l’Ouest: Vingt années de mesures en petites parcelles expérimentales. Travaux et Documents ORSTOM 78. ORSTOM, Paris. Rumpel, C., Ba, A., Darboux, F., Chaplot, V., Planchon, O., 2009. Erosion budget and process selectivity of black carbon at meter scale. Geoderma 154, 131–137. Schlesinger, W.H., 1995. Soil respiration and changes in soil carbon stocks. In: Woodwell, G.M., Mackenzie, F.T. (Eds.), Biotic Feedback in the Global Climatic System: Will he Warming Feed the Warming 7. Oxford Univ. Press, New York, pp. 159–168.

213

Schmidt, M.W.I., Torn, M.S., Abiven, S., Dittmar, T., Guggenberger, G., Janssens, I., Kleberg, M., Kogel-Knabner, I., Lehmann, J., Manning, D.A.C., Nannipieri, P., Rasse, D.P., Weiner, S., Trumbore, S., 2011. Persistence of soil organic matter as an ecosystem property. Nature 478, 49–56. Schulze, R.E., 1997. South African Atlas of Agrohydrology and Climatology. TT82/96 Water Research Commission, Pretoria, RSA. Segoli, M., De Gryze, S., Doua, F., Lee, J., Post, W.M., Denef, K., Six, J., 2013. AggModel: a soil organic matter model with measurable pools for use in incubation studies. Ecol. Model. 263, 1–9. Smith, S.V., Renwick, W.H., Buddemeie, R R.W., Crossland, C.J., 2001. Budgets of soil erosion and deposition for sediments and sedimentary organic carbon across the conterminous United States. Glob. Biogeochem. Cycles 15, 697–707. Spaccini, R., Conte, P., Piccolo, A., Haberhauer, G., Gerzabek, M.H., 2002. Increased soil organic carbon sequestration through hydrophobic protection by humic substances. Soil Biol. Biochem. 34, 1839–1851. Van Oost, K., Quine, T.A., Govers, G., De Gryze, S., Six, J., Harden, J.W., Ritchie, J.C., McCarty, G.W., Heckrath, G., Kosmas, C., Giraldez, J.V., Marques da Silva, J.R., Merckx, R., 2007. The Impact of Agricultural Soil Erosion on the Global Carbon Cycle. 318, 626–629. von Lützow, M., Kögel-Knabner, I., Ludwig, B., Matzner, E., Flessa, H., Ekschmitt, K., Guggenberger, G., Marschner, B., Kalbitz, K., 2008. Stabilization mechanisms of organic matter in four temperate soils: development and application of a conceptual model. J. Plant Nutr. Soil Sci. 171, 111–124. Wienhold, B.J., 2007. Comparison of laboratory methods and an in situ method for estimating nitrogen mineralization in an irrigated silt-loam soil. Commun. Soil Sci. Plant Anal. 38, 1721–1732. W.R.B., 1998. World Reference Base for Soil Resources. International Society of Soil Science. International Soil Reference and Information Centre, Spaargaren, Rome.