Stream Dissolved Organic Matter Composition Reflects the Riparian Zone, Not Upslope Soils in Boreal Forest Headwaters

Abstract

Despite the strong quantitative evidence that riparian zones (RZs) are the dominant source of dissolved organic carbon (DOC) to boreal streams, there is still a debate about the potential contribution of upslope areas to fluvial carbon export. To shed new light into this debate, we investigated the molecular composition of dissolved organic matter (DOM) in four upslope-riparian-stream transects in a Northern Swedish forest catchment using absorbance (A₂₅₄/A₃₆₅ and SUVA₂₅₄) and fluorescence (fluorescence and freshness indices) metrics. Based on these metrics, our results indicate that stream water DOM molecular composition resembles that of RZs and significantly differs from that of upslope areas. The resemblance between stream and riparian DOM was most apparent for the “Dominant Source Layer” (DSL), a narrow RZ stratum that, theoretically, contributes the most to solute and water fluxes to streams. Spectroscopic characterization based on traditional interpretations of the metrics suggested that mineral upslope (podzol) DOM is less aromatic, more microbially derived, and more recently produced than organic riparian (histosol) and stream DOM. We conclude that RZs, and specifically DSLs, are the main sources of DOC to boreal headwater streams and potentially to other streams located in low-gradient, organic matter-rich catchments.

Key Points

• Stream water DOM molecular composition resembles that of riparian zones and differs from that of upslope areas in boreal hillslope transects
• The resemblance between stream and riparian DOM is most apparent for the riparian “Dominant Source Layer”
• Mineral upslope (podzol) DOM is less aromatic, more microbially derived, and more recently produced than organic riparian (histosol) DOM

Plain Language Summary

Understanding carbon cycling in natural ecosystems is critical because ongoing climate change can promote the release of previously stored carbon in forest soils to streams and rivers, with the potential to form carbon dioxide, the main greenhouse gas. This is particularly important in boreal ecosystems, which are the largest stores of terrestrial carbon in the world. In this study, we identify the near-stream area, the so-called riparian zone, as the main source of carbon from boreal forest soils to streams. We provide qualitative data to support this, which together with previous quantitative analyses, make up enough evidence to support that the riparian zone is the main source of carbon to streams. As there is still a debate about the potential contribution of other areas in the ecosystem to the fluvial carbon export, our study importantly highlights that the riparian zone should be the focus of scientific assessments and management strategies in relation to carbon exports in surface waters.

1 Introduction

Northern ecosystems, including boreal peatlands, are the largest stores of terrestrial carbon in the world (Gorham, 1991; Tarnocai et al., 2009; Yu, 2012). Large amounts of allochthonous dissolved organic carbon (DOC), the main constituent of dissolved organic matter (DOM), are transferred from those organic soils into freshwater bodies with important implications for the global carbon cycle (Aufdenkampe et al., 2011; Cole et al., 2007; Tranvik et al., 2009) and aquatic ecology (Battin et al., 2008; Tranvik et al., 2004). In boreal forest headwaters, this transfer is primarily regulated by organic matter (OM)-rich riparian zones (RZs) (Ledesma et al., 2018).

The characteristic accumulation of OM in boreal forest RZs is a consequence of the positive balance between primary production and mineralization rates (Luke et al., 2007). This is caused by the shallow groundwater tables and the low temperature for much of the year that results in hypoxia and slow decomposition. In this context, RZs have a long-term potential to transfer DOC from new and old carbon pools into stream waters (Ledesma et al., 2015). Furthermore, the characteristic exponential decrease in hydraulic conductivity with depth implies that most of the lateral DOC transfer from boreal forest RZs to streams originates from only a fraction of the total riparian volume (Schiff et al., 1998). The “Dominant Source Layer” (DSL) concept, defined as the narrow RZ depth stratum that contributes the most to solute and water fluxes to streams (Laudon & Sponseller, 2017; Ledesma et al., 2018), should be the focus of any assessment of the RZ influence in surface water chemistry and will be tested in the present study.

Despite the overwhelming quantitative evidence that identify RZs as the major source of DOC to boreal forest and other OM-rich headwaters (Boyer et al., 1997; Dick et al., 2015; Dosskey & Bertsch, 1994; Hinton et al., 1998; Knorr, 2013; Köhler et al., 2009; Strohmeier et al., 2013), there is still a debate about the potential contribution of upslope areas to fluvial carbon exports. For example, Easthouse et al. (1992), Hruška et al. (2014), and Sawicka et al. (2016) argue that there must be sources (e.g., located upslope) other than riparian sources sustaining fluvial carbon exports. Estimates of DOM quality are consequently needed to gain new insights into this debate. Only a few studies have to some extent compared DOM quality in upslope areas, RZs, and streams (Burns et al., 2016; Sanderman et al., 2009). This comparison remains lacking for boreal forest headwaters, where it is even more meaningful because DOM is a key component of surface water quality and it is present in higher concentration than in lower latitude streams (e.g., Sanderman et al., 2009). Spectroscopic ultra violet (UV) absorbance (Dahlén et al., 1996; Weishaar et al., 2003) and fluorescence (Fellman et al., 2010; McKnight et al., 2001) analyses of water samples can be used to characterize the molecular composition of DOM and to trace its origin, source, and transformation (Erlandsson et al., 2012; Knorr, 2013; Spencer et al., 2007).

Here we used a combination of absorbance and fluorescence metrics to compare the DOM character in a number of upslope-riparian-stream transects in a permafrost-free boreal catchment located in Northern Sweden. We hypothesized that the DOM character in streams resembles that of the corresponding RZs, especially at the DSL, and differs from that of the corresponding upslope soils. This hypothesis builds on our current understanding by acknowledging that (i) different soil types are present in RZs and upslope areas in boreal forest headwaters (Chesworth, 2008), (ii) there are quantitative pieces of evidence provided by previous studies suggesting that RZs can be the only sources of carbon to boreal streams (e.g., Ledesma et al., 2015), and (iii) theoretically, the DSL contributes the most to solute and water fluxes to streams (Ledesma et al., 2018). In the present study, we provide the qualitative evidence to support this paradigm.

2 Study Site Characterization and Hydrological Context

The Krycklan Catchment Study (KCS) has been long used for hydrology, biogeochemistry, and climate research in the boreal landscape (Laudon & Sponseller, 2017). The 68 km² catchment is located in Northern Sweden (outlet at 64°12′N 19°52′E), 60 km west from the Baltic Sea (Figure 1). The climate is subarctic with a mean annual air temperature of 1.9°C and mean annual precipitation of 632 mm yr⁻¹ (1981–2013), half of which falls as snow. The elevation gradient is small (114–405 m above sea level), resulting in gentle slopes. Active anthropogenic influence in the catchment is low (e.g., only very small amounts of timber are harvested each year) but most headwater streams were straightened and deepened at the beginning of the twentieth century. This was a common practice across Fennoscandia in order to increase drainage and improve forest productivity.

Figure 1

Location of the Krycklan catchment (64°12′N 19°52′E) within Sweden (a). In (b), shape of the Krycklan catchment. In (c), zoom to the highlighted square in Figure 1b showing study catchments C8, C2, and C7.

In the upper and central parts of the catchment, located above the postglacial coastline, podzols and histosols have formed over Quaternary deposits of glacial till. Three low-order (first and second) streams in this landscape were the focus of the present study, called C8, C2, and C7 (Figure 1). Land cover in the associated subcatchments is dominated by coniferous forest stands and patches of wetlands (supporting information Table S1). OM-rich (i.e., histosols) RZs of variable widths, depending on the local hydromorphology (Ledesma et al., 2015), characterize the near-stream areas, whereas well-developed mineral podzols make up the more areally extensive upslope soils. This characteristic soil transition is complemented by an increase in slope and by a vegetation change from mosses (Sphagnum spp.), Norway spruce (Picea abies), and occasional deciduous trees (Betula spp., Alnus spp.) dominating in wet, organic RZs to vaccinium shrubs (Vaccinium spp.) and Scots pine (Pinus sylvestris) dominating in drier, mineral upslope areas. Thus we conceptualize boreal forest headwater catchments as a three-compartment continuum: upslope area, RZ, and stream.

Runoff generation in the study subcatchments follows the transmissivity feedback mechanism (Bishop et al., 2004; Lundin, 1982; Rodhe, 1989). This concept explains water mobilization to streams by lateral flow exports that exponentially increase as groundwater table rises and water enters highly conductive layers in the RZ during rainfall or snowmelt events. Overland flow has rarely been observed in the area (Bishop et al., 1995) and a combination of isotopic, hydrological, and modeling approaches suggest that there is limited bypass of water flow below 1 m depth in forest RZs (Amvrosiadi et al., 2017; Peralta-Tapia et al., 2015).

3 Materials and Methods

This study focuses on four hillslope transects, including upslope, riparian, and stream locations, distributed across three subcatchments (C8, C2, and C7; Figures 1 and 2) that were studied during 2008–2014 (see Table 1 and section 3.2 for specific time periods considered for the different locations). During this period, total annual specific discharge at the subcatchments varied between 261 mm (2011) and 481 mm (2012), which are, respectively, the 28th and 88th percentiles of the long-term (1981–2014) annual values. Note that subcatchment C2 had two associated upslope-riparian locations, sampled at two different time periods and thus they are treated as two independent transects. Spectroscopic characterization of stream and soil water samples from different depths in riparian and upslope profiles was used to investigate and compare the molecular composition of DOM in the three compartments of the four transects. We compared stream and RZs both for the whole riparian profile and for only the DSL. We focused on the comparisons only on spatial variation between the three compartments (upslope area, RZ, and stream) as seasonality and hydrology have been shown to have a minor influence on stream DOM composition in the study site (Kothawala et al., 2015). As DOM is commonly quantified based on carbon content, we refer to the concentration of OM as DOC throughout the paper, whereas when referring to spectral properties we use the term DOM because absorbance and fluorescence reflect molecular structures of carbon and other elements (Kothawala et al., 2013).

Figure 2

Schematic representation of the experimental design including the four hillslope transects studied (downstream to upstream view). Note the different horizontal scale in the upper left transect, which has a wider riparian zone. Riparian zone widths were determined based on field observations using soil type changes from organic/transitional soils to podzols as the main criterion. Black squares represent ceramic cup suctions lysimeters. Lysimeter pairs are located 15, 30, 45, 60, and 75 cm below the soil surface (there is an extra pair at 90 cm in U36). Single lysimeters in the lower-right plot transect are located 10, 25, 35, 45, 55, and 65 cm below the soil surface (riparian profile S4) and 12, 20, 35, 50, 75, and 90 cm (upslope profile S22). Note the differences in vertical scaling above and below ground.

Table 1. Sampling Location Characteristics (Including Carbon Content), Sampling Periods, Number of Valid Values for Each of the Spectroscopic Metrics, and Total Number of Available Samples (Within Brackets)

Location	Depth (cm)	Type	C (%)	Sampling period	pH range	DOC range	Fe/DOC range	A254/A365 values	SUVA254 values	FI and β:α values
C8		Stream		J2013-M14	4.6–5.9 (22)	12.9–45.0 (21)		21 (21)	19 (19)	35 (35)
R2	15*	Riparian	32	J2013-M14	5.0–5.3 (6)	57.5–67.0 (6)	0.11	6 (6)	6 (6)	1 (1)
R2	30	Riparian	46	J2013-M14	5.4–5.7 (4)	34.5–40.1 (5)	0.07	5 (5)	5 (5)	1 (1)
R2	45	Riparian	50	J2013-M14	5.3–5.4 (6)	35.9–37.9 (6)	0.09	6 (6)	6 (6)	1 (1)
R2	60	Riparian	50	J2013-M14	5.4–5.5 (6)	36.0–38.2 (6)	0.10	6 (6)	6 (6)	1 (1)
R2	75	Riparian	47	J2013-M14	5.3–5.4 (6)	37.8–48.8 (7)	0.09–0.10	7 (7)	7 (7)	2 (2)
U35	15	Upslope	0.6	J2013-M14	4.8–5.0 (7)	27.8–42.7 (7)	0.03	7 (7)	7 (7)	2 (2)
U35	30	Upslope	1.4	J2013-M14	5.2–5.4 (7)	11.1–16.5 (7)	0.02–0.03	7 (7)	7 (7)	2 (2)
U35	45	Upslope	1.2	J2013-M14	5.3–5.5 (7)	8.0–13.2 (7)	0.02	7 (7)	7 (7)	2 (2)
U35	60	Upslope	1.0	J2013-M14	5.4–5.6 (7)	4.0–4.8 (7)	0.01–0.02	5 (7)	7 (7)	2 (2)
U35	75	Upslope	0.9	J2013-M14	5.4–5.6 (7)	4.1–4.4 (7)	0.01–0.02	3 (7)	7 (7)	2 (2)
C2		Stream		J2013-M14	4.6–5.4 (21)	11.4–31.8 (21)	0.02–0.06	21 (21)	19 (19)	33 (33)
R5	15*	Riparian	50	J2013-M14	4.8–5.2 (7)	19.2–48.2 (7)	0.06	7 (7)	7 (7)	1 (1)
R5	30	Riparian	44	J2013-M14	5.5–5.6 (5)	14.5–23.0 (6)	0.09	6 (6)	6 (6)	2 (2)
R5	45	Riparian	42	J2013-M14	5.5–5.8 (7)	15.0–28.9 (7)		7 (7)	7 (7)
R5	60	Riparian		J2013-M14	5.7–5.8 (7)	22.5–15.2 (7)	0.20	7 (7)	7 (7)	2 (2)
R5	75	Riparian		J2013-M14	5.8–6.0 (7)	13.9–19.2 (7)	0.24–0.26	7 (7)	7 (7)	2 (2)
U36	15	Upslope	7.8	J2013-M14	5.2–5.6 (7)	5.9–19.9 (7)	0.02–0.03	7 (7)	7 (7)	2 (2)
U36	30	Upslope	1.9	J2013-M14	5.4–5.6 (7)	1.6–3.8 (7)	0.02	4 (7)	7 (7)	1 (1)
U36	45	Upslope	0.5	J2013-M14	5.4–5.6 (7)	1.5–1.8 (7)	0.01–0.02	2 (7)	4 (7)	2 (2)
U36	60	Upslope	0.4	J2013-M14	5.5–5.7 (7)	1.0–1.9 (7)	0.02	2 (7)	4 (7)	2 (2)
U36	75	Upslope	0.3	J2013-M14	5.6–5.7 (7)	0.9–1.4 (7)	0.02	1 (7)	2 (7)	2 (2)
U36	90	Upslope		J2013-M14	5.6–5.7 (7)	1.1–1.3 (7)	0.02	2 (7)	4 (7)	2 (2)
C7	-	Stream		J2013-M14	4.4–5.2 (22)	15.9–39.6 (21)	0.02–0.09	21 (21)	19 (19)	36 (36)
R10	15*	Riparian	44	J2013-M14	5.1–5.5 (7)	12.9–20.8 (7)	0.03–0.04	7 (7)	7 (7)	2 (2)
R10	30*	Riparian	45	J2013-M14	5.4–5.7 (6)	13.4–16.3 (7)	0.07	7 (7)	7 (7)	1 (1)
R10	45	Riparian	53	J2013-M14	5.5–5.7 (5)	7.8–10.1 (5)	0.10	5 (5)	5 (5)	1 (1)
R10	60	Riparian	52	J2013-M14	5.6–5.8 (4)	7.2–9.5 (5)		5 (5)	5 (5)
R10	75	Riparian	54	J2013-M14	5.6–5.8 (6)	7.2–11.1 (7)	0.09	7 (7)	7 (7)	1 (1)
U37	15	Upslope	2.3	J2013-M14	5.6–5.8 (7)	1.1–1.3 (7)	0.02	2 (7)	5 (7)	2 (2)
U37	30	Upslope	1.2	J2013-M14	5.7–5.9 (7)	1.0–1.0 (7)	0.02–0.03	2 (7)	4 (7)	2 (2)
U37	45	Upslope	1.1	J2013-M14	5.7–6.3 (6)	0.7–2.5 (7)	0.02–0.03	3 (7)	3 (7)	2 (2)
U37	60	Upslope	0.9	J2013-M14	5.8–5.9 (7)	0.8–1.1 (7)	0.02–0.03	1 (7)	3 (7)	2 (2)
U37	75	Upslope	0.7	J2013-M14	5.8–5.9 (5)	0.7–1.1 (6)	0.02–0.03	1 (6)	3 (6)	2 (2)
C2'		Stream		J2008-D12	4.4–6.0 (133)	9.6–42.4 (131)	0.02–0.06	133 (133)	131 (131)	33 (33)
S4	10	Riparian	47	J2008-D12	3.7–4.0 (23)	36.7–91.1 (32)	0.02–0.08	23 (30)	22 (28)	1 (1)
S4	25*	Riparian	38	J2008-D12	3.8–4.1 (30)	32.9–90.7 (36)	0.01–0.08	30 (35)	29 (32)	2 (2)
S4	35*	Riparian	23	J2008-D12	4.2–4.7 (26)	19.9–61.7 (37)	0.03–0.10	36 (37)	33 (33)	1 (1)
S4	45	Riparian	7.2	J2008-D12	4.0–4.8 (29)	20.1–47.1 (38)	0.04–0.15	37 (37)	35 (35)	5 (5)
S4	55	Riparian	1.3	J2008-D12	4.6–5.2 (33)	16.1–41.7 (39)	0.06–0.19	40 (40)	36 (36)	7 (7)
S4	65	Riparian	0.6	J2008-D12	4.5–4.9 (32)	17.4–47.7 (38)	0.02–0.14	38 (38)	34 (34)	7 (7)
S22	12	Upslope	0.4	J2008-D12	5.7–6.9 (31)	1.3–3.7 (39)	0.06	2 (40)	31 (34)
S22	20	Upslope	1.3	J2008-D12	5.7–6.4 (29)	3.2–11.0 (35)		30 (37)	32 (33)
S22	35	Upslope	0.4	J2008-D12	5.7–6.4 (20)	1.3–4.7 (26)		3 (30)	22 (26)
S22	50	Upslope	0.3	J2008-D12	5.7–6.8 (32)	1.2–3.0 (37)		2 (41)	29 (36)
S22	75	Upslope	0.3	J2008-D12	5.7–6.6 (26)	1.3–5.5 (30)	0.05	4 (31)	19 (27)
S22	90	Upslope	0.2	J2008-D12	5.9–7.1 (33)	1.0–4.3 (38)		2 (41)	27 (37)

• Note. For each riparian profile, * denotes lysimeters within the “Dominant Source Layer” (DSL). Note that fluorescence index (FI) and freshness index (β:α) values originate from a single sampling campaign in October 2013 for R2, R5, R10, U35, U36, and U37 locations (lysimeter pairs were not averaged in this case); from samples collected between April 2011 and June 2012 for each of the streams; and from samples collected between March and August 2012 for S4.

3.1 Brief Logistical History of the Monitored Stream, Riparian, and Upslope Locations

A total of 18 partially nested streams have been monitored for water chemistry at the KCS during different periods, most of them from 2003 till present. For this study, three of these streams were considered: C8, C2, and C7 (note that C2 connects to two of the four upslope-riparian transects and so it will be referred to as C2 and C2′, Figure 2). Unfortunately, sampling at C8 was terminated in 2007, which is outside the period when the corresponding riparian and upslope profiles were sampled (Table 1). The nearby streams C1 and C10 (Figure 1) were used to estimate chemical, absorbance, and fluorescence data at stream C8 (see section 3.5 and supporting information).

In 2007, 13 riparian profiles were instrumented across KCS to obtain a better understanding of the influence of the RZ on aquatic chemistry (Grabs et al., 2012). Selection of the profile locations was based on an initial terrain analysis of 1 m resolution airborne light detection and ranging (LiDAR) data and subsequent field reconnaissance for identification of hotspots for water movement, runoff generation, and riparian areas that contribute the most to lateral fluxes to streams (Grabs et al., 2012). Each riparian instrumented profile (placed 2–10 m from the corresponding stream) consists of pairs of ceramic suction cup lysimeters (nominal filter pore size 1 ± 0.1 μm) at five equally distributed soil depths (15, 30, 45, 60, and 75 cm) and a perforated PVC tube equipped with an automatic recording water level logger (WT-HR 1500 Trutrack©) so that soil water chemistry and groundwater levels can be monitored (see Table S2 in the supporting information for information on groundwater tables). For the present study, three of these riparian profiles were used: R2, R5, and R10. These are located within C8, C2, and C7 stream catchment areas respectively (Figures 1 and 2).

Table 2. Tukey's Honestly Significant Difference Test of the Specified Pair Comparison for the Four Spectroscopic Metrics and the Four Transects Used in the Study

Comparison	Index	C8-R2-U35	C2-R5-U36	C7-R10-U37	C2′-S4-S22
Stream-RZ	A254/A365	N.S.	N.S.	N.S.	c
Stream-DSL	A254/A365	N.S.	N.S.	N.S.	N.S.
Stream-Upslope	A254/A365	a	N.S.	a	a
RZ-Upslope	A254/A365	a	N.S.	a	N.S.
Stream-RZ	SUVA254	b	N.S.	a	b
Stream-DSL	SUVA254	N.S.	N.S.	a	N.S.
Stream-Upslope	SUVA254	a	a	a	a
RZ-Upslope	SUVA254	a	a	a	a
Stream-RZ	FI	a	N.S.	N.S.	a
Stream-DSL	FI	a	N.S.	N.S.	N.S.
Stream-Upslope	FI	a	a	a
RZ-Upslope	FI	N.S.	a	a
Stream-RZ	β:α	N.S.	N.S.	N.S.	a
Stream-DSL	β:α	N.S.	N.S.	N.S.	c
Stream-Upslope	β:α	a	a	a
RZ-Upslope	β:α	a	a	a

• Note. RZ denotes riparian zone and DSL denotes “Dominant Source Layer.” N.S., nonsignificant.
• ^a p < 0.0001.
• ^b p < 0.001.
• ^c p < 0.01.

In 2012, three additional lysimeter profiles were established in upslope podzol soils: U35 is located 54 m upslope from R2, U36 is 17 m upslope from R5, and U37 is 16 m upslope from R10 (Figure 2). The instrumentation setup is equal to that of the riparian profiles, except that for U36, which has an extra pair of suction lysimeters placed 90 cm below the soil surface. Upslope profile locations were selected based on interpretation of the LiDAR data and local topographic slopes in order to follow the likely hydrological flowpaths to the corresponding riparian profiles.

A fourth upslope-riparian transect (S22–S4) is also located within the subcatchment C2 (hereafter C2′ for this specific transect, Figure 2). The riparian profile S4 (located 4 m from the stream) and the upslope podzol profile S22 (located 22 m from the stream) were instrumented in 1995 and have been sampled at different frequencies and for different purposes since, but were more regularly monitored from 2008 onward. Single lysimeters were placed at six different depths for S4 (10, 25, 35, 45, 55, and 65 cm) and S22 (12, 20, 35, 50, 75, and 90 cm). The profiles are also equipped with recording water level loggers (supporting information Table S2).

3.2 Stream and Soil Water Sampling for the Present Study

Stream water samples from C1 and C10 (used as proxies for stream C8), C2, C7, and C2′ were collected during the period 2008–2014 with a frequency that varied from monthly during low winter flow to 2 days-weekly during spring flood (Laudon et al., 2013). Soil water samples from the riparian profiles R2, R5, and R10 and the corresponding upslope profiles U35, U36, and U37 were manually collected from lysimeters during seven field campaigns between June 2013 and May 2014. Lysimeters from riparian profile S4 and upslope profile S22 were sampled 44 times during 2008–2012, which is approximately monthly except for some of the winter months. Note that the stream samples reported in Table 1 and used in the analyses comparing absorbance metrics were those that overlapped in time with the corresponding riparian and upslope profiles. When collecting soil water, lysimeters were first rinsed in all cases by extracting up to 50 mL of water, which was subsequently discarded before proceeding with the actual sampling. All lysimeter samples were collected with acid-washed Milli-Q rinsed Duran glass bottles using vacuum. All stream and soil water samples were kept dark and cool prior to being subsampled and chemically or spectroscopically analyzed within 1–10 days, which is within the time range in which the molecular composition of DOM is stable during storage (Peacock et al., 2015).

Only a subset of the samples was analyzed for 3-D fluorescence: (i) for each of the streams, over 30 samples collected between April 2011 and June 2012, (ii) for riparian profiles R2, R5, and R10 and upslope profiles U35, U36, and U37, samples collected in the field campaign of October 2013, and (iii) for riparian profile S4, samples collected from seven sampling campaigns carried out between March and August 2012. Data of 3-D fluorescence were not available for the upslope profile S22.

3.3 Chemical, Absorbance, and Fluorescence Analyses

All stream and soil water samples were analyzed for pH using a pH meter and for DOC by catalytic combustion using a Shimadzu TOC-5000 elemental analyzer. Both stream (Ågren et al., 2007) and soil water (Ledesma et al., 2015) in the Krycklan catchment show no statistically significant differences between DOC and total (T) OC concentrations so TOC measurements of unfiltered samples can effectively be approximated to DOC. Samples were also regularly analyzed for major cations using inductively coupled atomic emission spectroscopy (ICP-AES) and thus dissolved iron (Fe) concentrations were also available, although with less frequency (approximately for 50% of the samples analyzed for DOC).

The absorbance spectrum across the wavelength range 190–820 nm was measured using a HP-8452A diode-array spectrophotometer for all stream and soil water samples collected prior to April 2009. The detection limit for this instrument is 0.006 cm⁻¹, calculated as three times (arbitrary factor commonly used in analytical chemistry) the scattering measured in blank (Milli-Q water) samples. After April 2009, the absorbance spectrum was measured across the wavelength range 190–1,100 nm using a Varian Cary^® 50 UV-Vis spectrophotometer. The detection limit for this instrument, as indicated in its specification sheet, is 0.004 cm⁻¹. Both instruments have a photometric range of 3.3 absorbance units, which is a parameter used to set the upper reliable value of the absorbance readings.

Fluorescence scans of synchronous excitation-emission matrices (EEMs) were measured for filtered (0.45 µm) samples using a Fluoromax-2 spectrofluorometer (Horiba Jobin Yvon) and a 1 cm quartz cuvette. All EEMs were automatically corrected for blank absorption, instrument specific biases, and primary and secondary inner filter effects (Kothawala et al., 2013) using the Aqualog^® software FluorEscence™ (Kothawala et al., 2015; Lavonen et al., 2015). The fluorescence intensity was normalized to water Raman units (R.U.) following Lawaetz and Stedmon (2009). Excitation wavelengths (λ_ex) ranged from 250 to 450 nm at 2 nm increments and emission wavelengths (λ_em) ranged from 212 to 620 nm. The integration time was adjusted depending on the fluorescence intensity using a prescan run at 1 s to optimize resolution, and ranged from 1 s for the most optically dense samples, to 10 s for the most optically clear samples. After analysis, fluorescence intensity for all samples was normalized to an integration time of 1 s before further processing of the data.

3.4 Absorbance and Fluorescence Metrics

Two commonly used absorbance metrics were estimated in order to characterize the molecular composition of DOM in stream and soil water samples. First, the ratio between absorbance measured at 254 and at 365 nm (A₂₅₄/A₃₆₅) was calculated. This metric is negatively correlated to the molecular weight of DOM (Dahlén et al., 1996) and has been previously used in catchment studies to assess DOM bioavailability (Ågren et al., 2008; Berggren et al., 2009). The specific UV absorbance as DOC concentration normalized absorbance at 254 nm (SUVA₂₅₄) was also calculated. This metric is positively correlated to DOM aromaticity (Weishaar et al., 2003) and has been previously used to trace sources of DOM in catchments (Hood et al., 2006; Sanderman et al., 2009; Yang et al., 2013).

Two additional commonly used indices derived from fluorescence EEMs were also calculated for the samples that were analyzed with the spectrofluorometer. The fluorescence index (FI) is used to distinguish between terrestrial and microbial fulvic acids and it was calculated as the ratio between emission at wavelengths of 470 and 520 nm at an excitation of 370 nm (Cory & McKnight, 2005; McKnight et al., 2001). The FI typically varies between 1.2 and 1.8, with lower values (<1.4) related to higher aromatic content typical of terrestrially derived DOM and higher values (>1.6) related to lower aromaticity and greater fraction of microbially derived material (McKnight et al., 2001). The freshness index (β:α) is an indicator of the contribution of recently produced DOM and it was calculated as the ratio of emission at 380 nm (β peak) divided by the emission maxima between 420 and 435 nm (α peak), at an excitation of 310 nm (Parlanti et al., 2000). A larger β peak is related to a more recently derived DOM, whereas a larger α peak represents more decomposed material.

All sample values obtained from chemical, absorbance, and fluorescence analyses underwent a quality check in order to ensure sensible estimates. This was especially important for the absorbance values from upslope soil water samples because of their generally low DOC concentrations. Thus, many upslope absorbance values had to be discarded because they fell below the instrument detection limit. This limited the number of A₂₅₄ and A₃₆₅ values and thereby SUVA₂₅₄ and especially A₂₅₄/A₃₆₅ (as A₃₆₅ values are lower) for the upslope soils. Analogously, a few A₂₅₄ values of samples from the two upper lysimeters in riparian profile S4 had to be discarded because they were above the photometric range of the instruments (see Table 1 for details on the final number of valid samples for each sampling location and metric).

3.5 Calculations and Statistical Analyses

Multiple linear regression models were used to estimate DOC concentrations and spectroscopic absorbance at stream C8 based on data from streams C1 and C10, which are similar to C8 in terms of catchment area and land use (supporting information Tables S1 and S3, and Figure S1). Fluorescence metrics for stream C8 were estimated as mean values obtained for streams C1 and C10 because they showed relatively low variability in time and between streams (supporting information Table S4).

Values of pH, DOC, A₂₅₄, and A₃₆₅ for lysimeter pairs were averaged for each sampling campaign as to have a single value representing each depth at a specific time. Soil water samples were not available at some occasions due to, e.g., failure of some lysimeters or too little water collected from relatively dry soil layers. When only the sample from one replicate was available, the missing value from the other lysimeter was estimated by multiplying its replicate by the average of the ratios of both lysimeter measures for the other sampling campaigns, a method that followed Grabs et al. (2012) and Ledesma et al. (2013).

The DSL of riparian profiles R2, R5, and R10, defined as the depth range with the highest contribution per unit depth to 90% of the mean annual lateral DOC flux (2008–2009), were characterized by Ledesma et al. (2015) and have been considered here. The model to estimate DSLs is based on the characteristic relationship between groundwater tables at the riparian location and stream runoff (i.e., it is a consequence of the transmissivity feedback mechanism). For any given stream runoff value, lateral flows at any given riparian soil layer are proportional to the groundwater table-stream runoff curve. Analogously to the method outlined by Ledesma et al. (2015), the DSL of riparian profile S4 for the period 2008–2012 was calculated for the present study using the groundwater table-stream runoff relationship presented by Ledesma et al. (2016) and Blackburn et al. (2017) for this riparian location. DOC concentrations at S4 were linearly interpolated in space, vertically from 0 to 1 m depth, and in time, daily from 2008 to 2012. This generated a concentration profile time series that could then be multiplied by the lateral flow profile to estimate lateral DOC fluxes. This integration was made on a daily basis below the estimated groundwater table that day, as it is assumed that horizontal water flux predominantly occurs as flow through saturated layers (see Seibert et al., 2009 and related subsequent literature). From the equations describing the groundwater table-stream runoff relationship, it was possible to calculate daily groundwater tables as daily stream runoff data were available for 2008–2012. Thus, following this method, lateral fluxes of water and solutes (in this case DOC) can be directly estimated at different time and spatial (in the vertical dimension) resolutions given time series of stream runoff and solute concentration profiles (see also given references in this paragraph for more details and Table S2 in the supplement information on groundwater tables at the riparian profiles).

Pairwise comparisons based on Tukey's Honestly Significant Difference tests (Tukey HSD) were performed for each spectroscopic metric for each of the following: Stream-RZ; Stream-DSL (only values from lysimeters within the DSL depth range); Stream-Upslope; and RZ-Upslope. The Tukey HSD method is conservative and suitable for unequal sample sizes and approximately normally distributed data. The data series for these comparisons included all available values in each case (see Table 1). Some of the comparisons for the fluorescence metrics involved the use of samples from different periods. In those cases, we assumed that the time of sampling had a negligible effect on the FI and β:α variation. We justify this by the minor influence of seasonal fluctuations in hydrology on stream DOM composition studied previously (Kothawala et al., 2015). In the referred study, Kothawala et al. (2015) used the statistical approach parallel factor analysis (PARAFAC) to investigate the influence of land cover type, hydrology, and in-stream processing on the DOM molecular composition across 17 of the Krycklan streams. The PARAFAC method allows to decompose the fluorescence signal into underlying fluorescent components (Bro, 1997). Redundancy Analysis (RDA) (Legendre & Legendre, 1998) of relationships between PARAFAC components and catchment characteristics revealed that land cover type proportions, specifically wetland versus forest proportions, explained most of the variability in stream DOM composition in Krycklan. Seasonal fluctuations in hydrology and in-stream processing only had a minor influence. In accordance, the relative standard deviation (as the ratio of the standard deviation to the mean) of more than thirty FI and β:α values covering all seasons for each of the streams presented here was below 4% (samples collected between April 2011 and June 2012).

4 Results

The estimated vertical position of the DSL at the different riparian profiles covered the following depth ranges: 0–21 cm below soil surface at riparian profile R2, 0–25 cm below soil surface at riparian profile R5, 8–34 cm below soil surface at riparian profile R10, and 13–39 cm below soil surface at riparian profile S4 (the approach is exemplified by riparian profile S4 in Figure 3). This means that 90% of the total lateral DOC flux to the stream came from those depth ranges at the different riparian profiles. Thus, on average for the four profiles and following the DSL theory, a shallow layer of only about 25 cm was responsible for most of the riparian DOC flux to the streams.

Figure 3

Relative lateral dissolved organic carbon (DOC) and water flux proportions for riparian profile S4 for the 5 year period 2008–2012. The upper and lower limits of the “Dominant Source Layer” (DSL, depth range with the highest contribution per unit depth to 90% of the mean annual lateral DOC flux) and the average DOC concentrations and standard deviations at the measured depths are also displayed.

The A₂₅₄/A₃₆₅ ratios of the upslope profiles were notably more variable, both in space and time, than the stream and riparian profile ratios, which were remarkably constant (Figure 4). There was no statistical difference between the A₂₅₄/A₃₆₅ ratio in stream and RZ water samples (both varying around 4.0–4.5), except for a marginal difference between stream C2′ and corresponding riparian profile S4 (Table 2). When only the DSL was considered, no significant differences between stream and RZs were found. Upslope A₂₅₄/A₃₆₅ vertical profiles included both higher (as much as 7.0) and lower (as low as 1.0) values than those observed at the corresponding stream and riparian locations (Figure 4). Upslope A₂₅₄/A₃₆₅ ratios were statistically different from those in the corresponding stream and RZ, except for upslope location U36 and corresponding stream C2 and riparian profile R5, and upslope location S22 and corresponding riparian profile S4 (Table 2).

Figure 4

Ratio between absorbance measured at 254 and at 365 nm (A₂₅₄/A₃₆₅) for the four streams and associated riparian and upslope profiles used in the study. Black dots and error bars are, respectively, average and standard deviations of the corresponding sample values. Note that some lysimeters had only one associated value (and thus no error bars) and that when only two values were available error bars were set to the maximum and minimum of those two. Dashed lines represent the upper and lower limits of the “Dominant Source Layer” (DSL).

SUVA₂₅₄ values were consistently high in both stream and RZ samples, typically higher than 4.0 L mg⁻¹ m⁻¹ (Figure 5), indicating a percentage of aromaticity higher than 30% according to the empirical relationship presented by Weishaar et al. (2003). Upslope SUVA₂₅₄, typically varying around 1.0 L mg⁻¹ m⁻¹ (interquartile range, IQR: 0.55–1.6 L mg⁻¹ m⁻¹), was significantly lower than that of the corresponding stream and RZ in all cases. The estimated percentage of aromaticity in the upslope locations was hence typically lower than 15%. Riparian profile R5 and corresponding stream C2 were not statistically different regarding SUVA₂₅₄, but there were statistical differences between riparian profiles R2, R10, and S4 and corresponding streams C8, C7, and C2′ (Table 2). However, when considering only the DSL depths, there was no difference between R2 and C8, neither between S4 and C2′.

Figure 5

Specific ultraviolet absorbance as dissolved organic carbon normalized absorbance at 254 nm (SUVA₂₅₄) for the four streams and associated riparian and upslope profiles used in the study. Black dots and error bars are, respectively, average and standard deviations of the corresponding sample values. Dashed lines represent the upper and lower limits of the “Dominant Source Layer” (DSL).

Values of FI varied around 1.33 (ranging from 1.26 to 1.38) for all stream samples (Figure 6), which are in agreement with SUVA₂₅₄ values as they also relate to a high aromatic DOM content. The FI values were significantly higher in riparian profiles R2 and S4 than in the corresponding streams C8 and C2′ but there was no difference between R5 and R10, and corresponding streams C2 and C7 (Table 2). When focusing only on the DSL depths, there was no difference between stream C2′ and corresponding riparian profile S4. Upslope profiles U36 and U37 had FI values indicative of microbially derived DOM (i.e., higher than 1.6) and were significantly higher than the corresponding stream and RZ values (Figure 6 and Table 2). The FI values for riparian profile R2 and corresponding upslope profile U35 were similar and varied around 1.5, an intermediate value in the FI scale.

Figure 6

Fluorescence index (FI) for the four streams and associated riparian and upslope profiles used in the study. Black dots and error bars are, respectively, average and standard deviations of the corresponding sample values. Note that some lysimeters had only one associated value (and thus no error bars) and that when only two values were available error bars were set to the maximum and minimum of those two. Dashed lines represent the upper and lower limits of the “Dominant Source Layer” (DSL).

The β:α was remarkably stable in space in the riparian profiles R2, R5, and R10 (around 0.38), and not statistically different from the corresponding streams (Figure 7 and Table 2). For riparian profile S4, β:α values increased with depth and were statistically different from the corresponding stream C2′. However, when considering only the DSL depths, this difference became less significant. All upslope profiles had significantly higher β:α values than the corresponding streams and RZs, indicating a more recently derived/less decomposed DOM in the podzols than in the organic RZs and the surface waters.

Figure 7

Freshness index (β:α) for the four streams and associated riparian and upslope profiles used in the study. Black dots and error bars are, respectively, average and standard deviations of the corresponding sample values. Note that some lysimeters had only one associated value (and thus no error bars) and that when only two values were available error bars were set to the maximum and minimum of those two. Dashed lines represent the upper and lower limits of the “Dominant Source Layer” (DSL).

5 Discussion

By analyzing a set of commonly used absorbance and fluorescence metrics to characterize water samples collected from four boreal hillslopes transects, we found strong indications that the molecular composition of DOM of streams resembles that of the RZ and differs from that of the upslope soils. The resemblance between stream and riparian DOM character became visually (see Figures 4-7) and statistically (Table 2) more apparent in the DSL, the narrow layer within the riparian profile that, theoretically, contributes the most to stream solute and water fluxes.

5.1 Riparian Zones as Sources of DOC to Streams and Importance of the Dominant Source Layer

Easthouse et al. (1992) used a fractionation process which separated DOC into hydrophilic and hydrophobic fractions to study DOM composition in stream and soil water from riparian peats and upslope podzols similar to the ones presented here and suggested that the podzols were major sources of organic material to the aquatic environment. Hruška et al. (2014), Monteith et al. (2015), and Sawicka et al. (2016) used modeling and empirical approaches in UK and Czech OM-rich catchments to argue that there must be sources other than only riparian sources to sustain stream DOC exports. However, our results show that upslope DOM composition in all four transects and for all absorbance and fluorescence metrics was significantly different and more variable than that from the stream, limiting the possibility of upslope sources directly contributing to stream DOC exports. Hence, our results suggest that upslope DOC flows through the hillslope to the RZ, where there is a considerable change in DOM character before DOC enters the stream.

At the same time, we showed a remarkable similarity between stream and riparian DOM composition. There is also quantitative evidence of the importance of the RZ in regulating stream DOC dynamics given that both newly produced carbon in the RZ and storage of carbon in riparian soil pools have the potential to sustain annual riverine carbon exports in boreal forest headwaters (Ledesma et al., 2015). Altogether we argue that there is enough evidence to support the hypothesis that the RZ is the main source of DOC to these boreal streams. Because this process is ultimately the result of soil type configuration within the catchment, we believe our conclusion is characteristic of typical boreal forest headwaters with a marked soil transition from mineral podzols in upslope areas to organic soils in the RZ (Chesworth, 2008). It is an open question whether this conclusion applies to other boreal sites where there is a marked permafrost influence on catchment hydrology (e.g., Koch et al., 2013), which is commonly distinct from the characteristic hydrology presented here based on the transmissivity feedback mechanism (Bishop et al., 2004). However, our conclusion also potentially applies to forest headwaters dominated by low-gradient, OM-rich RZs in other ecoregions (e.g., Tunaley et al., 2016).

In accordance with Kothawala et al. (2015), we report a markedly low temporal variation in stream and riparian DOM composition (represented by the error bars in Figures 4-7). This is in fact sensible and consistent with the idea of a DSL within the RZ, as the vertical variation in the DOM quality metrics within the DSL was low (Figures 4-7) and consequently any seasonal change in groundwater table that would lead to changes in the contribution of flow from different sections of the DSL would in turn result in a low variation in stream DOM quality. Indeed, the estimated lateral DOC flux at riparian profile S4 was strongly related to the estimated lateral water flux (Figure 3), similar to what it was presented by Ledesma et al. (2015) for riparian profiles R2, R5, and R10. The fact that statistical (and visual) differences between stream and riparian DOM composition became less significant when only the DSL was considered further supports that the origin of stream DOC lies in the RZ and specifically within the DSL. This hydrological concept builds on the transmissivity feedback mechanism (Bishop et al., 2004), a common phenomenon beyond boreal forest soils (McDonnell et al., 1998; McGlynn & McDonnell, 2003; Seibert et al., 2003), and extends it by providing testable hypotheses to understand riparian solute transport. The DSL concept can be tested and applied in such environments, but also elsewhere where timing and position of lateral water fluxes are identified. Thus we advocate for a redefined conceptualization of RZs that includes the specific recognition and identification of the DSL.

5.2 Spatial Variability on Catchment DOM Molecular Composition

Following the results of Kothawala et al. (2015), which demonstrated little influence of seasonality and hydrology on stream DOM composition in the study site, our research question focused solely on the spatial variation of DOM character between catchment compartments. Thus, assessing temporal aspects based on different hydrological conditions was outside the scope of our approach. Previous studies have focused on the temporal variation in stream DOM and some have reported differences between base and high flow conditions caused by flowpath shifts (Broder et al., 2017; Burns et al., 2016; Neff et al., 2006). These a priori contradicting conclusions with respect to the Krycklan catchment can be a result of the different ecoregion and hydrological context of those studies. However, the observed patterns might be partially explained by a common mechanism in all cases, i.e., shifts in the relative water contributions from different land cover types within each subcatchment, which is the main factor responsible for differences in stream DOM composition in the Krycklan streams (Ågren et al., 2008; Kothawala et al., 2015). We believe this within-catchment spatial variability plays an important role in some of the observed patterns in our study and it might explain the observed temporal variation in other studies.

There were a few cases where, for a specific metric, stream values differed from those of the corresponding RZ (e.g., SUVA₂₅₄ for stream C7 and riparian profile R10; Figure 5). Consistent with what we described above, we argue that these differences can be attributed to land cover heterogeneities in the RZ within the subcatchment, as also observed by Findlay et al. (2001) in a study of DOM composition in a set New Zealand RZs that differed in land use. SUVA₂₅₄ typically increases with increasing proportions of wetlands in the catchment (Ågren et al., 2008), which are also important sources of DOC to streams (Creed et al., 2008). This can also be observed in the data presented here: streams C7 (18% wetland) and C8 (12% wetland) were characterized by higher SUVA₂₅₄ values than stream C2, which is 100% forest. Thus the higher SUVA₂₅₄ displayed by stream C7 compared to its corresponding riparian profile R10 (including the DSL) or even the apparent higher SUVA₂₅₄ in stream C8 compared to its corresponding riparian profile R2 can be a result of water inputs from RZs draining wetlands in different parts of the subcatchment. Consistently, stream C2 (and C2′), draining no wetlands, had similar SUVA₂₅₄ values than both corresponding riparian profiles R5 and S4.

Differences in solute concentrations, including DOC, can occur at different RZs within the same catchment with the same land cover (Grabs et al., 2012). Therefore, it is reasonable to find differences in spectroscopic metrics as well, and this can be exemplified here by riparian profiles R5 and S4, which are both located in the same 100% forest subcatchment and had somewhat different index values. All four metrics in S4, especially β:α (Figure 7), tend to shift from those shown by R5 in the direction that correlates with more DOM bioavailability. Differences in belowground biotic activity linked to different patterns of water table fluctuation that result in transitional oxic/hypoxic conditions (Blackburn et al., 2017) might explain small differences observed in soil water DOM composition of a priori equivalent RZ profiles. Furthermore, we believe it is unlikely that in-stream processing had a significant effect on the observed stream DOM composition because (i) of the short water residence times in these boreal headwaters (Winterdahl et al., 2016) and (ii) observations in RZs and especially DSLs were very similar. In this context, we argue that the biogeochemical processes and their variability in the terrestrial compartment, specifically in the RZ, and not the aquatic compartment, is what shapes stream chemistry in these systems, including stream DOM molecular composition.

There was only one case in which an upslope profile was similar to a riparian profile and both were different from the stream: the FI in R2 and U35 (Figure 6). There is no clear explanation for this besides the discussed potential spatial variability. Upslope profile U35 is a gley-podzol and had significantly higher DOC concentrations than the other true podzols U36, U37, and S22 (Table 1). Thus, differences in soil characteristics and associated biological activity could provide an explanation for the deviation of U35 with respect to the other upslope profiles and for its similar FI to the corresponding riparian profile R2. Moreover, there was no statistical differences between some upslope areas and corresponding RZs and streams in the case of the A₂₅₄/A₃₆₅ ratio (Table 2). This is most likely due to the uncertainty in the A₃₆₅ values of the characteristically low DOC upslope samples that results in large vertical variability in the A₂₅₄/A₃₆₅ ratio for the upslope profiles, in combination with the low statistical power. The simple statistical tests employed here were performed to formally support our hypotheses but the overall conclusions of the study are distinctly supported by simple visualization of the patterns in the metric estimates of the three compartments in the four studied transects. It is important to highlight that statistical significance does not necessarily imply ecological relevance and vice versa (i.e., ecological relevance might be important even if there is no statistical significance).

5.3 DOM Bioavailability and Implications for Spectroscopic Metrics

One unquestionable result from this study is that DOM composition is different in upslope soils as compared to RZs and streams. Except for the A₂₅₄/A₃₆₅ ratio, for which upslope locations showed a relatively wide range of high and low values (likely due to the uncertainties in A₃₆₅ values as explained above), upslope DOM appears to be less aromatic (according to SUVA₂₅₄, Figure 5), more microbially derived (according to FI, Figure 6), and more recently produced (according to β:α, Figure 7) than RZ and stream DOM. Altogether the mineral upslope areas showed a much more bioavailable DOM character than the RZs and streams. It is unlikely that differences in tree species had an influence on this as leachates from the three main species in the catchment (Scots pine, Norway spruce, and silver birch) have similar DOM chemical composition (Fröberg et al., 2011). Alternatively, mineral podzols with large organic matter adsorption capacity, similar to the upslope soils in our study, tend to reduce the aromaticity of the DOM because of a preferential adsorption of aromatic DOM fractions (Kothawala et al., 2012). This mechanism can provide an explanation for the observed upslope patterns in our study. It is worth mentioning that, in Fröberg et al. (2011) and Kothawala et al. (2012), the reported SUVA₂₅₄ and FI podzol values were, respectively, higher and lower than in our study, suggesting a further bioavailable appearance of the podzol DOM presented here. On the other hand, consistent with our results, peaty soils such as our RZs are characterized by the strong aromatic character and the low degradability rates of its DOM (Ritson et al., 2014a). This could be related to the combination of hypoxic conditions and presence of Sphagnum spp. (van Breemen, 1995). Thus we argue that soil type variability, which in its turn is a consequence of hydromorphology and climate, is what drives soil solution DOM chemical composition.

The four spectroscopic metrics employed here have been widely used in recent years for surface and drinking water DOM characterization (e.g., Lavonen et al., 2015; Matilainen et al., 2011; Ritson et al., 2014b). However, characterization of soil water DOM is far less common, especially for the wide range of DOC concentrations presented (Table 1). This is a novel aspect of the present study that, nevertheless, might involve a number of limitations. For example, high dissolved Fe concentrations can increase SUVA estimates and lead to fluorescence quenching (i.e., a decrease in DOM fluorescence intensity) (Patel-Sorrentino et al., 2002; Poulin et al., 2014; Weishaar et al., 2003). Quenching of fluorescence due to Fe interference was likely of minor importance here because the Fe/DOC ratio in our samples was low (Table 1). The potential for interferences due to pH has been reported, yet the narrow pH range in our data, and overlapping values between compartments (Table 1), implies that any low level pH interferences are normalized, rather than selectively biasing our results. Moreover, the index estimates for some of the soil water samples were outside commonly reported ranges, e.g., some upslope samples had an FI higher than 1.8. Due to their empirical nature, indices do hold some limitations, particularly as they are commonly used outside the environmental conditions they were originally developed under. As such, care needs to be taken in the traditional interpretation of results from indices. The common interpretation for surface water samples of the index values presented here relates to the DOM bioavailability, but this interpretation may actually be different for soil water samples, i.e., there might not be a direct relationship between a specific metric value and DOM bioavailability but rather a relationship with, e.g., soil properties (Shen et al., 2015), as we also discussed above. This will need to be tested in future investigations but this study provides a unique first data set on the topic that encourages further research.

6 Conclusions

We examined DOM molecular composition in three catchment compartments (upslope area, RZ, and stream) in four hillslope transects in a forest catchment in Northern Sweden. We show that the DOM molecular composition of streams resembles that of RZs and significantly differs from that of upslope areas. The resemblance between stream and riparian DOM is clearest for the DSL, a novel concept that highlights the critical importance of a narrow layer within riparian soils for solute and water flux contributions to streams. We conclude that RZs, and specifically DSLs, are the main sources of DOC in boreal headwaters and potentially in low-gradient, OM-rich catchments located elsewhere.

Acknowledgments

This study was funded by Formas (ForWater project, 2010-667-15957-55) with additional funding part of the Krycklan Catchment Study (KCS) from MISTRA Future Forest, VR (SITES), FOMA, SKB, and the Horizon 2020 Landmark program. The main author was also partially financed by NordForsk (60501) and Formas (2015-1518). All final derived data reported in the paper figures can be access through figshare at https://figshare.com/articles/Ledesma_et_al_2018_WRR_Dataset_xlsx/6106004. Part of the data (everything relative to the streams) is also available at the KCS webpage (http://www.slu.se/Krycklan). The supporting information file also contains background information relative to streams. Other parts of the data were already presented in given published references. We thank those involved in field-work and laboratory analyses as well as the Krycklan crew for their essential contribution. We specially thank Elin Lavonen for guidance in the fluorescence analyses and Kevin Bishop for helpful discussions. Finally, we thank the handling editor and three anonymous referees for their valuable reviews. Their comments and suggestions improved the quality of this paper.