For studying the degradation effect on the vegetation-specific tracers, five sampling sites were selected (Fig. 1) of which two are in the Southern Black Forest (Germany), two in the Lake Baldegg catchment (Switzerland), and one in the Upper Sûre Lake catchment (Luxembourg) (Table 1). The broadleaved forest site in the Southern Black Forest (site 1; BFbeech) was mainly stocked with beech (Fagus sylvatica), under which a moder-type (duff mull) humus with clearly separable Oi (undecomposed OM) and Oe (partly decomposed OM) horizons, but a diffused boundary between the O and A horizons, has developed due to the absence of a separate Oa (decomposed OM, humus layer) horizon. The second site in the Southern Black Forest is a coniferous forest where the sole tree species is spruce (Picea abies, site 2a; BFspruce) and a thick cover of moss (Sphagnum quinquefarium, site 2b; BFmoss) is growing under very acidic conditions (pH∼3) characterized by the development of raw humus with an easily separable sequence of Oi–Oe–Oa and A horizons. One of the two sites in the Lake Baldegg catchment has mull-type humus with a thin Oi layer on top of the A layer (Site 4; LBspruce). The mixed-forest site in the Upper Sûre Lake catchment has oak (Quercus robur) and spruce (Picea abies) cover with moder-type humus (site 5; LXmixed). All sampling sites have acidic soils with pH values in the range of 3–4.2. As we are aware that a range of soil pH values might have been preferable, fully developed organic horizons are not found in slightly acidic or alkaline soils.

Fresh leaves, needles, and mosses were collected in autumn. Multiple leaves and needles (15–20) from different directions at ca. 3 m above ground were collected from an individual tree and mixed to form a sample of fresh material. The O horizon was sampled and subhorizons within it were separated according to the humus form observed. The underlying A horizon was sampled to a maximum depth of 5 cm. Each sample was formed from a composite mixture of three subsamples taken at a distance of ca. 1 m from each other.

Collected samples were oven-dried at 40 ∘C for 72 h. Soils were dry sieved to a particle size <2 mm and macroscopic plant remains and stones were removed with tweezers. All the samples were roughly ground with a mortar and pestle. An aliquot of ground samples was finely powdered with a ball mill (Retsch MM 400, Retsch GmbH, Haan, Germany) for 90 s at a frequency of 24 s min-1 for analysis of bulk carbon concentration and isotopic (δ13C) composition. Powdered samples were weighed in tin capsules, and subsequently stable carbon isotope ratios (δ13Cbulk) and C content were measured with an elemental analyser coupled to an isotope ratio mass spectrometer (EA–IRMS) (Sercon Integra2, Sercon Ltd., Crewe, UK). The carbon isotope values of samples are reported in a delta notation relative to the Vienna Peedee Belemnite (V-PDB) standard. The instrumental standard deviation is lower than 0.1 ‰ for δ13C.

The particle fraction of size <63 µm was separated from four A horizon soil samples collected from various forest types (beech forest (site 1), coniferous forest (site 2a), coniferous forest with moss cover (site 3), and mixed forest (site 5)). Prior to the separation, soil aggregates were disrupted by ultrasonic dispersion using an ultrasonic probe (Branson 250 model, Branson Ultrasonics, Danbury, CT, USA). A 500 J mL-1 ultrasonic energy was applied as described in Schmidt et al. (1999). This energy is sufficient for total disruption of aggregates without redistribution and compositional changes in OM. After disruption, the fraction of size <63 µm was separated by wet sieving and water was removed by sedimentation and centrifugation.

Powdered samples were extracted using an accelerated solvent extraction system (ASE 350, Dionex Corp., Sunnyvale, CA, USA) with dichloromethane (DCM) / methanol (MeOH) (9:1, v/v) over three extraction cycles at 100 ∘C. The total lipid extract (TLE) was volume reduced by vacuum evaporation (Rocket Synergy evaporator, Genevac Ltd., Ipswich, UK). The TLE was subsequently separated into neutral, acidic, and polar fractions by solid-phase extraction using aminopropyl-bonded silica gel as the stationary phase (Jacob et al., 2005). The acidic fraction (including free fatty acids) was methylated at 60 ∘C for 1 h using 1 mL of 12 %–14 % boron trifluoride (BF3) in MeOH, and fatty acid methyl esters (FAMEs) were extracted with hexane and 0.1 M KCl. The aliphatic hydrocarbons containing long-chain n-alkanes were separated from the neutral fraction obtained earlier using a column filled with an activated silica gel by eluting with n-heptane.

Individual compounds in the FAMEs and aliphatic hydrocarbon fractions were identified using a TRACE GC Ultra gas chromatograph coupled with a DSQ II mass spectrometer (MS) (Thermo Scientific, Waltham, MA, USA). Quantification was performed using the same GC coupled with a flame ionization detector (FID) (Thermo Scientific, Waltham, MA, USA). The GC method was set to be identical during both measurements as described in Alewell et al. (2016).

The δ13C isotopic composition of individual FAMEs and n-alkanes was determined using a Trace 1310 GC with split/splitless injector and FID detector interfaced online via a GC-IsoLink II to a ConFlo IV and Delta V Plus isotope ratio mass spectrometer (Thermo Scientific, Bremen, Germany). A DB-5MS capillary column (50×0.2 mm i.d., 0.33 µm film thickness, J & W Scientific, Folsom, CA, USA) was used. The GC temperature programme for n-alkanes was 70 ∘C (held 4 min) to 320 ∘C (held 35 min) at 5 ∘C min-1. The GC temperature programme for FAMEs was 70 ∘C (held 4 min) to 150 ∘C at 20 ∘C min-1 and afterwards to 320 ∘C (held 40 min) at 5 ∘C min-1. He (purity grade 5.0) was used as a carrier gas at a constant flow of 1 mL min-1. CO2 of known δ13C isotopic composition used as reference gas was automatically introduced via ConFlo IV into the isotope ratio mass spectrometer in a series of five pulses at the beginning and four pulses at the end of each analysis during every measurement. The system was externally calibrated with standard mixtures A6 for n-alkanes (n-C16 to n-C30) and F8-3 for FAMEs (even chain fatty acid methyl and ethyl esters from n-C14:0 to n-C20:0), both obtained from Arndt Schimmelmann, Indiana University, IN, USA. Isotopically certified n-C24:0, n-C30:0 (Arndt Schimmelmann, Indiana University, IN, USA) and n-C26:0, n-C28:0 (Sigma-Aldrich, St. Louis, MO, USA) FAMEs were further added to the F8-3 mixture. The carbon isotope values of samples are reported in delta notation relative to the standard V-PDB, averaging at least three replicate measurements. The analytical uncertainty is lower than ±0.5 ‰. The reported δ13C values of FAMEs were corrected for the additional carbon atom introduced during methylation using the following formula: 1δ13CFA=δ13CFAME-1-Xδ13CMethanolX, where X is the ratio of C atoms in the FA to the C atoms in the corresponding FAME. δ13CMethanol was measured with an elemental analyser coupled to an isotope ratio mass spectrometer (EA–IRMS), using similar instrumental parameters as the bulk C isotope analysis described above.

The lipids from fresh leaves as well as O and A horizons showed even-over-odd predominance in long-chain fatty acids (>n-C22:0) and odd-over-even predominance in n-alkanes (>n-C23), a property usually ascribed to the lipids derived from higher plant biomass (Eglinton et al., 1962; Kolattukudy et al., 1976; Amblès et al., 1994a). To quantify this pattern of the chain length preference, carbon preference index (CPI) values were also calculated (Eglinton et al., 1962). Mean CPI values of fatty acids (n-C22:0–30:0) and n-alkanes (n-C23-33) were 5.7 and 5.5 respectively, indicating the input from higher plants. Similar concentrations (relative abundance with regard to total OM) of both of the compound classes were observed between fresh plant biomass and degraded OM, with a slight increase within the O horizon (Fig. 2). We also defined concentration in mineral soil with regard to total soil OM and not to soil weight for its better comparability to the fresh plant biomass and O horizon. We found a strong increase in compound concentrations in the mineral soils collected from various sites (Fig. 2), except site BFmoss. The strong increase in the concentrations from the O horizon to mineral soil (A horizon) indicates a preferential preservation of long-chain fatty acids and n-alkanes compared to soil bulk OM (Fig. 2). In general, long-chain fatty acids were relatively more abundant than n-alkanes in the plant–soil system, similar to previously reported findings (van Bergen et al., 1998; Naafs et al., 2004; Chikaraishi and Naraoka, 2006).

Interspecies variability was recorded by the molecular distribution patterns of the compounds in fresh plant biomass (Fig. S1a and b in the Supplement). The coniferous trees had relatively high concentrations of n-C29 alkane and n-C24:0 fatty acid. Fresh leaves of broadleaved trees showed relatively high concentrations of n-C27 alkane and n-C28:0 fatty acid. The sphagnum species were characterized by high n-C23 or n-C21 alkane, differing from conifers and broadleaved trees, also consistent with previously reported distribution patterns (Maffei et al., 2004; Bush and McInerney, 2013).

However, molecular distribution patterns observed in fresh plant biomass were not necessarily transferred to the O and A horizons (Fig. S1a and b). This was shown by the changes in average chain length (ACL) values of both compound classes (Fig. 3). Relatively higher ACL values of n-alkanes in the A horizon compared to overlying horizons from some of the sites might have been the result of preferential degradation of shorter-chained homologues (<n-C27) and preservation of longer-chained homologues from n-C27 and higher (Lichtfouse et al., 1998; van Bergen et al., 1998; Marseille et al., 1999). In the case of long-chain fatty acids, changes in the molecular distribution patterns (Fig. S1a) and a decrease in the ACL values (Fig. 3) from fresh matter to the A horizon were observed at most sites. This might be explained by the production of long-chain fatty acids (>n-C20:0) in the O and A horizons from the oxidation of other components of vegetation-derived lipids such as n-alkanes and n-alkanols (Amblès et al., 1994a, b; Marseille et al., 1999) and also a contribution of shorter-chained homologues from microorganisms (Lichtfouse et al., 1995; Chikaraishi and Naraoka, 2006).

The bulk stable carbon isotope values were in between -32.4 ‰ and -26.1 ‰, similar to the typical δ13C values of the C3 plant–soil system (Diefendorf et al., 2010). During degradation, the carbon isotopic composition generally showed consecutive enrichment in 13C from fresh plant biomass (leaves, needles, and mosses) via organic horizons to the mineral soil. The δ13C increment from fresh plant biomass to the A horizon varied between 3.5 ‰ (BFbeech) and 5.6 ‰ (BFmoss) (Fig. 4). Selective degradation of bulk OM with depleted carbon isotopic signature has been observed previously, but with fewer significant changes probably related to short timescales (Nguyen Tu et al., 2004). Our findings report consistent degradation (with p value = 0.72 denoting similar extent of degradation across studied forest systems) as well as consecutive increase in δ13C values through degradation stages (p value <0.001 denoting significantly different δ13C values). Hence the cautious use of bulk carbon isotope values for using it as source soil signatures is necessary, especially when the vegetation sources found are either C3 or C4 only.

Compared to bulk isotopic composition of fresh plant biomass, compound-specific isotope values of long-chain fatty acids and n-alkanes were generally lower by -3 ‰ and -3.7 ‰ respectively (Figs. 4 and 5), in agreement with previous studies (Collister et al., 1994; Huang et al., 1997; Nguyen Tu et al., 2011). The compound-specific δ13C values of long-chain fatty acids (n-C24:0 to n-C30:0) and n-alkanes (n-C25 to n-C31) showed a general increase in δ13C values from fresh plant biomass to the O horizon overlying the mineral soil (A horizon); however there were only slight or no further changes to the A horizon (Fig. 5, Table 2). Both the compound classes showed similar trends of 13C enrichment at each site. Exceptionally, some n-alkane homologues showed a strong enrichment between the O and A horizons, e.g. BFmoss (n-C25,27), LBspruce∗ (n-C29), and LXmixed (n-C27,29) sites. Fresh moss material exhibited the lowest δ13C values for both the compound classes, -37.7 ‰ (n-C26:0 FA) and -40.4 ‰ (n-C25 alkane) of Sphagnum quinquefarium (BFmoss) and -39.4 ‰ (n-C26:0 FA) and -37 ‰ (n-C27 alkane) of Thuidium tamariscinum (LBspruce∗), similar to previously reported sphagnum n-alkane isotope values (Brader et al., 2010). δ13C values of mosses are lower as they generally grow in wet and humid conditions and experience less water loss due to evapotranspiration (Sachse et al., 2006). Sites 2a (BFspruce) and 2b (BFmoss) are both located in a spruce forest with thick moss cover. Hence the strong isotopic enrichment in n-C25 and n-C27 alkanes (3.9 ‰ and 2.4 ‰ respectively) observed from the O horizon to mineral soil (A horizon) at BFmoss might have been caused by significantly higher long-term input from spruce (with a comparatively higher δ13C value) compared to the fresh and degraded moss to the mineral soil. However, supportive data of, for example, long-term biomass inputs to soil from different species present at the particular site, were not available. Also, site LXmixed has a vegetation cover of two distinct species (Table 1). The difference in isotopic composition of fresh plant biomass of these two species along with its seasonal variability might be a cause of changes in δ13C values during degradation. Also, isotopically distinct microbial input of long-chain compounds cannot be excluded (Marseille et al., 1999; Chikaraishi and Naraoka, 2006).

At sites where a single dominant vegetation is present, we observed increase up to 5 ‰ in long-chain fatty acids and up to 3.2 ‰ in long-chain n-alkanes from fresh plant biomass to the A horizon, similar to previous findings (Chikaraishi and Naraoka, 2006; Griepentrog et al., 2015, 2016). It was found that most of the degradation effect on isotopic composition took place up to the O horizon, and slight or no further changes were observed during transfer to the A horizon from the overlying O horizon. Table 2 summarizes this finding where we compared the changes in compound-specific isotope values with the level of significance. We calculated the magnitude of enrichment or depletion in 13C from the mineral soil horizon with respect to fresh biomass (Δ1) and also with respect to the overlying O horizon (Δ2) as follows: 2Δ1=δ13C(Ahorizon)-δ13C(freshplantbiomass)3Δ2=δ13C(Ahorizon)-δ13C(overlyingOhorizon). For the calculation of Δ1 (Eq. 2), we excluded the δ13C values of fresh moss to obtain the δ13C(freshplantbiomass) value, only at site 3. This was done to eliminate the overestimation of the magnitude of isotopic change, as mosses had considerably low isotope values compared to other fresh materials (Fig. 5).

The bulk δ13C values showed a continuous increase from fresh plant material through the degradation stages to the A horizon and did not exhibit the stability needed for use as an isotope tracer. Furthermore, fresh plant biomass generally exhibited a wider range of compound-specific δ13C values (7.5 ‰ for n-C25 alkane and 3.9 ‰ for n-C26:0 fatty acid) than bulk isotope values (1.6 ‰) across all the sites. Hence in terms of the stability during early degradation in soils and interspecies variability, our results confirm that the compound-specific isotope analysis is a more useful technique for sediment source attribution, and the analysis should not rely on bulk isotope values.

We compared the compound-specific δ13C values of long-chain fatty acids and n-alkanes from the bulk soil (<2 mm) to the fine silt plus clay fraction (<63 µm) (Fig. 6). None of the fatty acids showed a significant difference in their δ13C values between the two soil particle size fractions. Only for the LBspruce∗ site were δ13C values of long-chain fatty acids considerably (up to 2.1 ‰) but not significantly different among the studied fractions. Similar results were obtained for long-chain n-alkanes, where none of the n-alkanes from studied sites differed significantly between the different particle size fractions, with a difference of up to 1 ‰ (Fig. 6). Cayet and Lichtfouse (2001) reported preferential input of long-chain n-alkanes in soils via a larger (200–2000 µm) particle size fraction. However in terms of isotopic composition of soils, Quénéa et al. (2006) reported a less than 1 ‰ difference in long-chain n-alkanes δ13C among the soil size fractions. In accordance with previous findings on n-alkanes, our results simultaneously studying long-chain fatty acids and n-alkanes confirm that the compound-specific carbon isotope values of both compound classes were not significantly different between the two particle sizes. Long-chain n-alkanes were found to be even less prone to the isotopic fractionation within the soil particle size fractions. These results indicate that the analysis of bulk soil material instead of the fine soil fraction (<63 µm) might be suitable for sediment source attribution from soils to freshwater systems, which might not only increase work and cost effectiveness of the method but also reduce potential error sources due to particle size fractionation effects. In general, the results above further strengthen the necessary requirement of conservativeness of tracer signature during erosion processes like detachment, dispersion, and sediment transport, for CSIA-based sediment source apportionment.