Introduction

Impacts of soil erosion on the global carbon (C) cycling are profound and relatively well studied (Pimentel et al. 1995; Smith et al. 2001; Lal 2003; Berhe et al. 2007; Lal and Pimentel 2008; Kuhn et al. 2009; Quinton et al. 2010; Berhe 2012; Nadeu et al. 2012; Doetterl et al. 2012a; Berhe and Kleber 2013; Kirkels et al. 2014). The net effect of soil erosion on the global C cycle has generated significant scientific debate. For example, some studies conclude that erosion constitutes a net source of carbon dioxide (CO2) to the atmosphere (Jacinthe et al. 2001; Lal and Pimentel 2008), whereas others have estimated that erosion leads to a net terrestrial sink of CO2 (Stallard 1998; Harden et al. 1999; Berhe et al. 2007; Van Oost et al. 2007). However, numerous knowledge gaps remain, including the fate of eroded soil organic C (SOC) during and after erosional transport.

Globally, soil erosion is estimated to mobilize on the order of 75 Gt soil and 1–5 Gt C year−1 from hillslope soil profiles (Stallard 1998; Berhe et al. 2007). Subsequently, the re-distribution of the eroded SOC across the landscape leads to a lateral transfer of up to 30 % of the eroded SOC into rivers (Stallard 1998; Starr et al. 2000; Lal 2003; Mora et al. 2007; Dlugoß et al. 2012). While in transit from eroding to depositional sites (hereafter termed as “transport”), eroded SOC is selectively re-deposited downhill or downstream depending on the transport distances of the particles associated with it. Transport distance is usually not uniform, but affected by preferential deposition of differently-sized mineral particles and sediment aggregates (Basic et al. 2002; Kuhn et al. 2009; Warrington et al. 2009; Kuhn 2013). The potential transport distances of different sediment fractions are determined by their settling velocities (Beuselinck et al. 1999; Loch 2001; Tromp-van Meerveld et al. 2008), as well by sediment concentration and hydraulic conditions (Beuselinck et al. 1998, 2000). Current investigations on the fate of eroded SOC are either deduced from SOC stocks at eroding and depositional landform positions (van Hemelryck et al. 2011; Dlugoß et al. 2012; Nadeu et al. 2012), fluxes and distribution of C in different fractions at both eroding and depositional positions (Berhe et al. 2008, 2012; Berhe 2012), or calculated from mineral particle/aggregate-size distributions (Van Oost et al. 2004a; Aksoy and Kavvas 2005; Fiener et al. 2008). To our knowledge, no prior study has accounted for how differences in transport distance of C associated with different size particles influence the fate of SOC during or after erosion-induced transport.

Previous studies recognized that SOC and fine particles are eroded as individual particles or as parts of soil aggregates (Walling 1988; Beuselinck et al. 1998, 2000). Laboratory experiments also show that aggregation can potentially lead to faster settling velocities, compared to those predicted based on properties of individual mineral particles. Effects of aggregation in turn shorten the likely transport distance of eroded SOC that is associated with mineral particles (Hu et al. 2013; Hu and Kuhn 2014, 2016). After observing sediments in river systems, Starr et al. (2000) proposed that the fate of eroded SOC is a function of aggregate size, with aggregates greater than 63 µm in size being re-deposited across landscapes. Following this conceptual model, the preferential deposition of SOC-rich aggregates with fast settling velocities potentially results in SOC enrichment along hillslopes. This is in contrast against the model based on mineral particle specific SOC distribution would suggest.

In addition to skewing SOC spatial distribution, the breakdown of soil aggregates during erosion-induced transport and deposition along hillslopes exposes SOC previously protected from decomposition and thus potentially generates considerable CO2 fluxes from the sediments (Jacinthe et al. 2002, 2004; Mora et al. 2007; Hu and Kuhn 2014). The mineralization potentials of eroded SOC not only differ among soil types, but are also distinct across aggregate classes (Tisdall and Oades 1982; Six et al. 2000, 2001; Berhe 2012; Doetterl et al. 2015). For instance, physically stabilized SOC within macro- and micro-aggregates is protected from decomposition by forming physical barriers for the diffusion of microbes and enzymes and their substrates on to SOC compounds (Six et al. 2002). Breakdown of aggregates during erosion-induced transport renders previously physically protected SOM susceptible to mineralization (Berhe and Kleber 2013). Therefore, erosion-induced re-distribution of individual particles versus aggregates is likely to have different effects on the spatial pattern and stability of SOC across watersheds. Furthermore, the complex spatial distribution of SOC and its long-term fate in eroding systems is further affected by local microclimate in the geomorphic settings where the eroded SOC is deposited (Berhe 2012; Wang et al. 2014b; Doetterl et al. 2015). Recently, Fiener et al. (2015) observed that, depending on the parameterization of temporal variability and SOC enrichment processes, the vertical C flux at a depositional site could range from a C sink to a C source. They based their findings on the differences between C inputs from plant assimilates/organic fertilizer, and SOC mineralization. This highlights the need to determine how transport distance affects not just the distribution, but also the composition and stability of SOC in dynamic landscapes experiencing erosion and deposition.

Among different techniques to measure composition and stability of SOC, stable isotope compositions of organic C (δ13C) are also widely used, not only to assess the composition, but also to detect the source and stability of SOC in the soil system. By detecting the enrichment or depletion of 13C compared to 12C, the stable isotope ratio of C (δ13C) can be used to infer dynamics of the biochemical processes experienced by SOC, including changes that are associated with decomposition or mineralization (Dawson and Brooks 2001). Stable C isotope compositions have also been used to study the stability of SOC in different fractions. For example, Alewell et al. (2008, 2009) used δ13C of SOC in their field investigations on mountain grassland in the central Swiss Alps. The δ13C values were more positive on the eroding upper land than on the deposition wetland at the footslope. They partially attributed this pattern to the inherent variations of δ13C values and the potential effects from different water regimes and associated vegetation to the δ13C values. It is also accounted for the faster decomposition rates on the upper land compared to the wetland, and thus, Alewell et al. (2008, 2009) concluded that considerable degradation of eroded SOC during detachment and transport had occurred. It is still unclear, however, what role selective erosion or preferential deposition plays in the spatial pattern of δ13C signatures along hillslopes, as suggested by the studies of Hu and Kuhn (2014, 2016). A field investigation is essential, therefore, to examine these findings from the perspective of settling velocity fractionation and settling-class specific SOC distribution.

In this study, eroded soil samples from a sandy Danish soil after natural rainfall events were collected from different topographic positions along a cultivated hillslope. We hypothesized that the faster settling time of larger particles, whether individuals or aggregates, would result in erosion-induced alterations in distribution and decomposition of SOC. By analyzing settling velocity distribution, SOC amount, decomposability and δ13C composition, we aimed to determine whether the settling-velocity specific SOC distribution along hillslopes corresponds to differences in composition and/or stability of SOC and its potential transport to river systems.

Materials and methods

Soil sampling and preparation

A sandy Luvisol was sampled from a field near Foulum (56°30′N, 9°35′W), Denmark, on 4 September 2014, after a series of natural rainfall events. This field was located at the frontier of the Weichsel glaciation period (115,000–ca. 10.000 bc), which left a large amount of sand-sized materials in the soil (King 1975). The A h horizon (ca. 35 cm) of the non-eroded soil on this field was black. The SOC concentration was on average 28 mg g−1, and about 40–55 % of the soil particles on the A h horizon were of 63–250 µm (i.e., most erodible particles applied in most erosion models). The E horizon was sandier and lighter in color, indicating the loss of mineral and organic matter by leaching. The Bt horizon was accumulated with secondary clay and also slightly fingering with reddish patches, probably resulted from the effect of soil mottling. Although sandy soil is not the optimal soil type to observe the effects of aggregation in terms of skewing the transport distance of eroded SOC, it offers an opportunity to examine our hypothesis on the effects of settling-velocity specific redistribution of SOC on a field scale without a complex change in sediment aggregation. Detailed assessment on sandy soil is essential prior to conducting more extensive field investigations on other types of soils, particularly aggregated soils.

The slope of the field site ranged from 6 to 14 %. The curvature of the hillslope was convex at the top, linear in the middle, and concave at the bottom. The field site had been converted into a farmland at least 25 years ago, and is now mainly cropped with grass-clover, wheat and barley. The winter wheat was planted shortly before rainfall events, which lasted for a period of 3 days, with daily precipitation of 11, 21 and 3 mm. Such events have recurrence intervals of 0.13, 0.71 and 0.03 years, each representing a common rainfall condition in the local region (NOAA 2015) Yet, continuous rainfall over 3 days amplified the runoff and sediment connectivity (Kuhn and Yair 2004), particularly on this freshly tilled field, generating an erosion event of high magnitude. After the rainfall events, a 35-m long rill was formed along the slope, resulting in a significant deposition area (ca. 21 m long, Fig. 1). The bank on the lower end of the field, located approximately 150 m away from the lower end of the deposition fan, is slightly curved up due to tillage deposition in front of the field boundary. This formed a closed deposition fan, creating an opportunity for the collection of almost all the eroded material that would have been otherwise transported into the adjacent downstream watercourse. Eroded soils from the walls of the rill were collected every 5 m from the slope shoulder for 35 m to the start of immediate deposition at the footslope. The recently deposited sediments were collected every 1 m from the footslope to the tail of the depositional area. To form a comparison between erosion-affected soils and non-eroded soils, bulk soil (top 10 cm), where no erosion or deposition was readily observed, was also sampled alongside the eroding rill and depositional area, following the same space intervals as per the eroded sediment collection protocol. After collection, all soil samples were stored at 4 °C to preserve the soil structure and retard bioactivities. During data analysis and interpretation, consecutive sampling points along the slope were accordingly grouped into three representative topographic positions (i.e., slope shoulder, footslope and slope tail), in order to capture the topography, spatial redistribution and stability of eroded SOC.

Fig. 1
figure 1

The rill erosion (35 m long) and deposition fan (21 m long) on the farm at Foulum, Denmark. The map of Denmark (a) is from Adhikari et al. (2014), with the black dot and arrow indicating the location of the field site (b). The slope profile of the field site (c) was generated from field measurements with a leveling instrument

Full size image

Physical fractionation of soil samples

The eroded soils (30 g) at all the topographic positions were fractionated using a settling tube apparatus (Fig. 2). The settling tube apparatus consists of four components: the settling tube, through which the soil sample settles; the injection device, by which the soil sample is introduced into the tube; the turntable, within which the fractionated subsamples are collected; and the control panel, which allows an operator to control the rotational speed and resting/moving intervals of the turntable. Details of the settling tube apparatus were described in Hu et al. (2013). To enable the comparison between settling velocity distribution and conventional mineral particle size distribution, a concept of equivalent quartz size (EQS) is introduced to represent the diameter of a nominal spherical quartz particle that would fall with the same velocity as the aggregated particle for which fall velocity is measured (Loch 2001; Hu et al. 2013). Following the Stokes’ law and the concept of EQS, all the soil samples were fractionated into five EQS classes using the settling tube, which ranged from >500, 250–500, 63–250, 32–63 and <32 µm. To form a comparison with conventional mineral particle size distribution, eroded soils from selective topographic positions were also dispersed by ultrasound at energy of 60 J ml−1 (i.e., energy = output power 70 W × time 85 s/suspension volume 100 ml) (Kaiser et al. 2012). Dispersed soil fractions were wet-sieved into five size classes corresponding to the previously described EQS classes.

Fig. 2
figure 2

The settling tube apparatus (a), with soil samples falling through the water column (b), and soil fractions of different settling velocities collected at sampling trays in the water tank (c). Detailed description on the design and operation of the settling tube apparatus, please refer to Hu et al. (2013)

Full size image

Measurements of SOC and δ13C

After settling or dispersed fractionation, all the subsamples were dried at 40 °C and weight. The SOC concentrations of all the bulk soil, eroded soil and subsample fractions were determined using Leco RC 612 at 1000 °C. The SOC mass of each settling class was calculated from its SOC concentration and weighed. The δ13C of the subsamples from representative topographic positions (i.e., slope shoulder, footslope, and slope tail), after ball milling into powdery particles, were analyzed using a Costech elemental analyzer coupled to a Delta V Plus (ThermoFisher) isotope ratio mass spectrometer at the University of California, Merced. The standards were referenced to international standards Pee Dee Belemnite. The stable isotopic compositions were expressed in δ notation (‰) as: δ13C = [(Rsample − Rstandard)/Rstandard] × 1000; where Rsample is the ratio of the heavy to the light C (13C/12C) isotopes in a sample, and Rstandard is the ratio in a standard (Dawson and Brooks 2001). Prior to the measurement of δ13C, all the soil samples were pre-tested with 10 % HCl to determine whether the soils contained carbonates, but hardly any reaction with acid was observed. Hence, inorganic C content in these samples was considered negligible (<0.3 %). The total C concentration was therefore applied to represent the total organic C concentration in all the soil samples. The C:N ratios of all the soil samples were calculated from the total C concentration and total nitrogen concentration by their accurate weights after being dried at 105 °C.

CO2 emission measurements

The CO2 emission rates of soil samples were measured following the method described in Robertson et al. (1999) and Zibilske (1994). The eroded soil samples from representative topographic positions (without settling fractionation) were transferred into flasks of volume ca. 200 cm3, and incubated at 20 °C in open flasks for 33 days. Prior to CO2 emission measurements, all flasks were sealed using rubber stoppers. Gas from the headspace of each sealed flask was extracted at the beginning and end of the 1 h sampling period. The differences of CO2 concentrations between the two measurements were used to represent the soil CO2 emission rate during the period of 1 h. The CO2 emission rate measurements were repeated at day 1, 3, 5, 12, 19, 26, and 33. Concentrations of CO2 were assessed using a SRI 310C Gas Chromatograph (with Hayesep 1/8″ packed column working at 120 °C). The signals of the CO2 concentrations were detected and recorded by the PeakSimple 4.44 software. The cumulative CO2 emission rates were linearly extrapolated from hourly rates to daily rates and then further accumulated over day-intervals into 33 days CO2 emission potentials (a method adapted from Robertson et al. 1999). Prior to the incubation, all the soil samples were not dried and re-wetted to arbitrary moisture contents, but instead, were directly incubated at their respective field capacities (i.e., natural water contents after drainage 4 days after the last rainfall events). The exact moisture content of each soil sample was obtained by drying and weighing the flask after the 33 days of CO2 emission measurements. Relatively coarse-sized soil samples collected from the slope shoulder and the footslope were incubated with a gravimetric moisture content of ca. 25 %, while the finer sized soil samples collected from the slope tail were incubated at gravimetric moisture content of ca. 50 %. As water activity to permit microbial CO2 production is not only affected by soil moisture contents but also by matric effects (Rummel et al. 2014), the soil samples of distinct specific surface areas and thus different matric potentials can balance the effects of different soil moisture contents on microbial activity. While not able to separate the effects of water potential from other factors, the variations in rates of CO2 efflux were considered adequate to reflect the erosion-related differences in SOC distribution and decomposition. During the incubation period, the wet weight of each fraction was monitored every 3 days to control the soil moisture, and the variation of soil moisture was constrained within 1 %. All the data analysis and figure generation were processed by R studio software packages (R version 2.15.1).

Results

Size distributions along the slope

Size distributions of eroded soil from representative topographic positions, after settling or dispersed fractionation, are compared in Fig. 3. In general, size distribution after settling fractionation shows coarser particles than the size distribution after dispersed fractionation. Unlike the comparable size distributions of the very coarse soil fractions at the footslope (Fig. 3b), the soil samples from slope shoulder (Fig. 3a) and tail (Fig. 3c) had distinct size distributions between settling and dispersed fractionations. The settling class distributions of eroded soil collected at all topographic positions along the slope varied as a function of distance from the shoulder (Fig. 4). Nearly 90 % of the soil fractions on or near the slope shoulder were of EQS >63 µm and only 10 % were of EQS <63 µm (e.g., at 5 m represented by the bold line with open circle in Fig. 4). Gradually along the slope towards the footslope, where immediate deposition started (approximately 35 m away from the slope shoulder), the soil was markedly enriched in coarse particles of EQS >63 µm (see lines representing 35 m to 41 m in Fig. 4). The soil further downwards at the slope tail progressively showed a greater proportion of fine particles of EQS <63 µm (e.g., see lines of 46 m to 54 m in Fig. 4).

Fig. 3
figure 3

The comparison of the cumulative size distributions of eroded soil samples from the slope shoulder (a), footslope (b) and slope tail (c), after two different fractionation methods. The difference between the values of each point along a curve is the percent of material in that size class

Full size image
Fig. 4
figure 4

The cumulative equivalent quartz size (EQS) distribution of eroded soil samples from all the different topographic positions on the slope. The “start” notation in the legend denotes the beginning of obvious deposition immediately downslope of the eroding area. The difference between the values of each point along a curve represents the percentage of material in that size class

Full size image

SOC concentration and properties along the slope

The SOC concentration, δ13C, total nitrogen concentration and C:N ratio in the soil samples from representative topographic positions varied depending on their distance from the slope shoulder (Fig. 5). In general, the non-eroded bulk soil (not fractionated) had very uniform SOC, nitrogen concentrations, δ13C and C:N ratios relative to the eroded bulk sediments (not fractionated) which varied considerably. Compared to the slope shoulder, soil at the footslope where immediate deposition started was depleted in SOC and nitrogen. In contrast, SOC and nitrogen were enriched in the depositional area at the slope tail (Fig. 5a, c). Both the δ13C and the C:N ratios were greatest at the footslope, but less on the slope shoulder, and least at the slope tail (Fig. 5b, d).

Fig. 5
figure 5

The distributions of soil organic carbon (SOC) concentration (a), δ13C (b), total nitrogen concentration (c) and C:N ratio (d) of eroded soils on different topographic positions along the slope. The two dashed lines separate the topographic positions along the slope into three groups: slope shoulder, footslope and slope tail. Data points in each figure represent the consecutive sampling points along the hillslope

Full size image

Apart from the distinctive trends between different topographic positions, SOC and nitrogen distributions also differ across EQS classes (Fig. 6). The coarse EQS classes had generally lower SOC and %N concentrations than the fine EQS classes (Fig. 6a, c). The δ13C and the C:N ratios were generally greater in coarse EQS classes than those of the fine EQS classes, especially for the eroded sediment at the footslope (Fig. 6b, d). Rates of CO2 efflux from soil samples collected at representative topographic positions were lowest in the footslope and highest in the slope tail (Fig. 7). The sediment from the slope tail had about five times greater CO2 emission than that from the soil on the slope shoulder. The sediment at the footslope presented the lowest CO2 emission rates per gram of soil with values barely above detection levels.

Fig. 6
figure 6

The distributions of soil organic carbon (SOC) concentration (a), δ13C (b), total nitrogen concentration (c) and C:N ratio (d) across settling classes. Error bars indicate the standard deviation (n = 4 for four sampling points on the slope shoulder, n = 5 for five sampling sites at the footslope, and n = 3 for three sampling sites at the slope tail)

Full size image
Fig. 7
figure 7

The cumulative CO2 emission rates of eroded soil samples representing slope shoulder, footslope and slope tail over 32 days of incubation at 20 °C. Error bars indicate the minimum and maximum ranges (n = 3 for three sampling points on each topographic position)

Full size image

Discussion

Validity of settling fractionation approach

The distinct size distributions between settling and dispersed fractionation methods (Fig. 3a–c) clearly illustrate the inherent bias of data derived from dispersed mineral particle sizes taken to accurately represent the settling behavior of eroded soil along the slope. The greatest proportion (90 %) of soil deposited on the slope back had EQS >63 µm (Fig. 4), strongly suggesting that most of the eroded soil would be deposited immediately after being transported. Although repeated rainfall events may transport those temporarily deposited coarse soil fractions further downslope, the particles deposited along the extensively long slope tail were of distinctively finer sizes (Figs. 1, 4). This skewed size distribution implies two possibilities. Firstly, interrill erosion probably contributes to the downslope transportation of a greater amount of eroded sediment downslope, as interrill erosion has lower magnitude flow velocity, yet occurs at a much greater frequency than rill erosion processes investigated in this study (Kuhn et al. 2009). Second, those temporarily deposited coarse particles at the footslope would require subsequent rainfall events with much greater magnitude of overland flow velocity than in this study (35 mm over 3 days) to be transported anywhere further on the less inclined slope tail. This observation is in agreement with previous studies that suggested that 70–90 % of eroded SOC from hillslopes is likely to be deposited within the same or adjacent watersheds (Stallard 1998; Starr et al. 2000; Berhe et al. 2007; Fiener et al. 2015). Furthermore, our results are the first observation supporting the conceptual model proposed by Starr et al. (2000) in which eroded soil fractions >63 µm would very likely be deposited across landscapes, whereas eroded soil fractions <63 µm would be potentially transferred into river systems. The size distribution of eroded sediments along the slope observed in our study may also partially explain the discrepancies between the accelerated sediment flux predicted by models and the far smaller amount of sedimentation observed in global river systems (Walling 1983; Doetterl et al. 2012b). Our findings highlight the validity of the slope-scale settling fractionation approach. However, considering the variability in hillslope geomorphic variables (e.g., slope length, slope angle, surface roughness, curvature, etc.), it is hard to expect that any two natural hillslopes would serve as replicates. Therefore, we also point to the need to further examine the implications of using dispersed particles to estimate erosion-induced lateral distribution of C and its ultimate fate in aquatic systems. Future approaches could include aggregate-specific SOC fractionation, larger scale sediment tracing, modeling of sediment routing or landscape development. Moreover, the spatial distribution of eroded soil fractions observed in our study could prove to be a valuable framework for replacing mineral particle size distribution or aggregate size classes in current erosion models with actual settling velocity distributions to optimize their parameterization (e.g., Van Oost et al. 2004a; Fiener et al. 2008).

Skewing effects of settling fractionation on SOC distribution

The selective deposition along the slope not only affects the distribution of eroded soil fractions, but also determines the spatial patterns of eroded SOC associated with those fractions. As a consequence of the immediate deposition of the coarse EQS classes (Fig. 4) with low SOC concentration, SOC concentrations were depleted nearly to zero at the footslope (Fig. 5a). At the same time, the accumulation of the fine EQS classes with larger SOC concentrations led to the enrichment of SOC at the slope tail with close to four times as much SOC as in the bulk soil (top 10 cm, Fig. 5a). While regularly affected by tillage erosion (Veihe et al. 2003; Heckrath et al. 2006), the spatial distribution of SOC along the slope is consistent with the soil profiles observed during annual field investigations with thin and coarse Ap horizon at the footslope, and more than 1 m deep of dark fine Ap horizon at the slope tail (Fig. 8).

Fig. 8
figure 8

The soil profiles observed on the slope shoulder (a), footslope (b), and slope tail (c) during regular field investigation

Full size image

The settling behavior represented by the soil fractions deposited at the footslope (Fig. 4) could alternatively be valid for aggregated sediments. This means, if erosion and transport processes occur on an aggregated soil, aggregates with the same settling velocity as those coarse sandy fractions would also be immediately deposited at the footslope. Given that SOC is likely to accumulate in macro- and micro-aggregates (Tisdall and Oades 1982; Cambardella and Elliott 1994; Hu and Kuhn 2014, 2016), the immediately deposited SOC-rich aggregates would then result in SOC enrichment at the footslope. Hence, the SOC stocks at the footslope would be more than mineral particle specific SOC distribution, or SOC stock differences between the slope shoulder and slope tail, would suggest.

Relevance of settling-velocity specific SOC distribution to global carbon cycling

The δ13C composition and CO2 emission rates of soils from our study site reflected the stability of the SOC fractions on different topographic positions. The more positive δ13C values on the slope shoulder than at the slope tail observed in this study (Figs. 5b, 6b) are consistent with the studies by Alewell et al. (2008, 2009) on the mountain grassland in the central Swiss Alps. In agreement with results reported by these authors, our findings also suggest that greater decomposition rates occurred on the eroding slope shoulder position, compared to the other topographic positions. The significantly more positive δ13C of SOC at the footslope in our study (Fig. 5b) was accompanied by an accumulation of coarse EQS fractions (Fig. 4) with δ13C values 1 ‰ more positive compared to the finer fractions (Fig. 6b). Apart from the potential effects of persistent variations introduced from different water regimes and vegetation, these results also suggest that a greater degree of decomposition might have been experienced by those fractions of coarse EQS during erosion, transport and deposition. Low levels of nitrogen (Fig. 5c) may have inhibited continual mineralization of SOC, and thus resulted in relatively greater C:N ratios at the footslope (Fig. 5d). Conceivably, nitrogen addition or amendments, as often applied in agricultural land, could potentially initiate further decomposition of SOC.

The cumulative CO2 emissions of soils at the slope tail (Fig. 7) that reach almost 2500 µg CO2-C g−1 soil after 33 days of incubation can probably be attributed to the accumulation of SOC-enriched fine soil fractions and their sequent mineralization in favorable water regimes at the depositional area. Given the narrow recurrence intervals of alike rainfall events (NOAA 2015) and the commonly existing rolling topography on local farmland, similar erosion events are very likely to occur repeatedly. Consequently, this promises further mineralization after SOC fractions are rejuvenated by each rainfall event. While the CO2 emission rates from the soils at the footslope were about 3 times less than those from the slope shoulder, and even 25 times less than that at the slope tail for this Danish sandy soil (Fig. 7), their contribution to atmospheric CO2 would have been much more pronounced if they were aggregated. As discussed previously, the immediately deposited sedimentary fractions would alternatively have been SOC-rich aggregates that had equivalent settling velocities. According to our previous studies (Hu and Kuhn 2014, 2016) and other reports (Jacinthe and Lal 2001; Jacinthe et al. 2004; Polyakov and Lal 2004; Berhe 2012; Wang et al. 2014a, b; Doetterl et al. 2015), those SOC-rich aggregates at the footslope are also very likely to experience accelerated mineralization induced by exposure of previously protected SOC.

Overall, current slope-scale soil C balances are mostly calculated by coupling soil erosion rates and SOC stock archives along slopes with mineralization and dynamic replacement of eroded SOC (Lal 2003; Berhe et al. 2007; Van Oost et al. 2007; Lal and Pimentel 2008; Quinton et al. 2010; van Hemelryck et al. 2011; Berhe 2012; Nadeu et al. 2012). For instance, in Van Oost et al. (2004b), the erosion data from an almost similar study site in Denmark to ours reported a net C sequestration, which strongly contrasts against the accelerated mineralization effects observed in this study. The spatial distribution and mineralization patterns of eroded SOC identified in this and our previous studies (Hu and Kuhn 2014, 2016; Xiao et al. 2015), whilst too restricted to be extrapolated to a global context, align with some other recent reports (Billings et al. 2010; Kirkels et al. 2014; Doetterl et al. 2015; Fiener et al. 2015). They all suggest that the preferential deposition of SOC at the footslope and the potential SOC mineralization during transport, whether in the aggregated or non-aggregated form, must also be considered in order to appropriately account for the relevance of soil erosion to global C cycling.

Conclusions

Our study detected the spatial distribution of eroded soil fractions along hillslopes and further identified their mineralization susceptibility during transport. Eroded sandy soils on different topographic positions along a slope were collected after a series of natural rainfall events. The settling fractions of eroded soils clearly show a coarsening effect at the immediate deposition area at the footslope, subsequently resulting in a fining trend on the deposition area at the slope tail. The consistency between the EQS class distributions observed in this study and the conceptual model proposed by Starr et al. (2000) after sediment observations in river systems highlight the need to apply settling velocity of actual transporting fraction specific SOC distributions into soil erosion models. Furthermore, the +1 ‰ shift in δ13C values at the footslope suggests a greater degree of decomposition that has possibly been experienced during transport, while the pronounced CO2 emissions at the slope tail indicate a greater potential of CO2 emissions after deposition. Overall, this study identifies the potential for underestimating SOC loss during transport in current erosion estimates, i.e., the preferential deposition of fast settling fractions, the potential SOC enrichment and mineralization during transport or after deposition. Under estimations for deriving SOC loss from the SOC stock differences between the slope shoulder and the slope tail, or from the estimation of 20 % SOC loss to the atmosphere can lead to problems in quantifying the effects of erosion to the global C cycle (Lal and Pimentel 2008; Quinton et al. 2010; van Hemelryck et al. 2011; Nadeu et al. 2012). Our findings demonstrate the benefits of using actual settling velocity distribution of eroded fractions, aggregated or not, as a SOC erodibility parameter to better calibrate soil erosion models. As a part of the transport pathway from eroding sites to long-term depositional sites, our results observed from the hillslope also suggest that the net effect of lateral movement of SOC on the vertical fluxes requires considering the position of sampling sites in relation to transport and deposition processes as well as residence time along a hillslope.