Decadal carbon discharge by a mountain stream is dominated by coarse organic matter

Erosion of particulate organic carbon (POC) from the terrestrial biosphere and its transport by rivers redistributes nutrients and can contribute to atmospheric carbon dioxide drawdown (Berner, 1982; Stallard, 1998; Battin et al., 2008; Galy et al., 2015). High rates of physical erosion in mountain catchments result in elevated rates of fine POC discharge (FPOC, particles >0.2–0.7 µm and <1 mm), with biospheric FPOC (FPOC_biosphere) yields >10 t C km^–2 yr^–1 (Hilton et al., 2012; Goñi et al., 2013; Smith et al., 2013; Galy et al., 2015). As a result, mountain rivers can contribute significant amounts of FPOC_biosphere to large rivers, lakes, and the oceans (Stallard, 1998; Hilton et al., 2012; Galy et al., 2015). This carbon, recently derived from atmospheric carbon dioxide via photosynthesis, is often transported along with large volumes of clastic sediment (Hilton et al., 2012). High sediment accumulation rates in depositional settings can increase the burial efficiency of POC_biosphere and promote the drawdown of atmospheric CO₂ over geological time scales (Berner, 1982; Kao et al., 2014; Galy et al., 2015).

Despite this recognition, the organic carbon (OC) transported as coarse particulate organic matter (CPOM, particles >1 mm) remains poorly constrained, mainly because it is challenging to measure. CPOM can range in size from leaves to entire trees, and is not captured by typical river water sampling methods (e.g., Goñi et al., 2013; Smith et al., 2013; Hilton et al., 2015), while it is transported episodically during large floods when it is difficult to work in river channels (West et al., 2011; Wohl, 2013; Kramer and Wohl, 2014). CPOM also contributes to ecosystem functions because it typically contains ∼50% carbon by weight and can form the basis of the food chain in many streams (Fisher and Likens, 1973). In addition to contributing to carbon and nutrient transfers in rivers, large wood, consisting of CPOM with lengths >1 m, can affect stream morphology and hydraulics, while providing shelter for in-stream fauna and affecting breeding grounds (Wohl, 2013).

A significant challenge remains to accurately measure coarse POC (CPOC) transport in rivers across the full size range of CPOM, while linking CPOC transfer to hydrodynamic conditions in rivers; only by doing so can CPOC yields (t C km^–2 yr^–1) be accurately quantified. In addition, CPOC eroded from the biosphere is often thought to float (West et al., 2011), suggesting that it could be more susceptible to oxidation upon its delivery to floodplains (Fisher and Likens, 1973), lakes and reservoirs (Seo et al., 2008), and the oceans (West et al., 2011). However, water-logged woody debris, with a density higher than water, is a component of FPOC_biosphere in large river systems (Bianchi et al., 2007; Hilton et al., 2015). The amount of water-logged CPOC discharged by mountain rivers remains unknown. Here we use detailed measurements of CPOM transport in the Erlenbach, a 0.7 km² catchment in the Swiss Prealps. Although small, the catchment has geomorphic, climatic, and ecological characteristics that are representative of forested mountain headwater streams in a temperate climate (Schleppi et al., 1999; Smith et al., 2013).

The Erlenbach is a steep (11% slope) mountain stream with step-pool morphology and drains 0.7 km² in the Swiss Prealps (47.045707°N, 8.708844°E) (Fig. 1). The mean annual air temperature is ∼4.5 °C and the mean annual precipitation is ∼2300 mm. Approximately 40% of the total catchment area is covered by alpine forest, mainly comprising Norway Spruce (Picea abies) and European Silver Fir (Abies alba) (Schleppi et al., 1999), and a small amount of logging has been done in the upper catchment over the past 10 yr. The remaining 60% of the catchment is covered by wetland and alpine meadows. A well-developed riparian zone is generally lacking and active landslide complexes along the channel lead to strong channel-hillslope coupling typical of many steep mountain catchments. Both DOC and FPOC fluxes have been previously determined (Hagedorn et al., 2000; Smith et al., 2013). The FPOC has been partitioned into that derived from the terrestrial biosphere (FPOC_biosphere) and that from rock-derived OC using stable carbon isotopes, nitrogen to carbon ratios, and radiocarbon (Smith et al., 2013).

We use CPOM data sampled with three different methods (see the GSA Data Repository¹), each of which is suitable for a different water discharge range (Turowski et al., 2013). All sampling locations were within 30 m of a permanently installed gauge measuring water discharge (Q_w, L s^–1) at 10 min intervals (Rickenmann et al., 2012). At low Q_w (1 L s^–1 to 1000 L s^–1; most samples <250 L s^–1), Bunte traps were used (Bunte et al., 2007). These are metal frames placed on the stream bed, to which a net with 6 mm meshing is attached. At intermediate Q_w (200 L s^–1 to 1500 L s^–1; most samples >400 L s^–1) basket samplers were used (Rickenmann et al., 2012), consisting of metal cubes with 1 m edges and walls and floor made of metal mesh with 10 mm holes. The samplers automatically move into the flow when Q_w exceeds a predefined threshold value and when bedload transport is recorded. Both traps and baskets sample the entire flow depth with nearly 100% efficiency (Rickenmann et al., 2012; Turowski et al., 2013).

Woody material in the basket and trap samples was separated from clastic material in the field and weighed. Basket samples from A.D. 2011 to 2013 were separated into floating and sinking fractions in the field by dropping them into a water-filled bucket. Subsequently, the material was dried for 24 h at 80 °C, and the dry mass was obtained.

The diameter and length of large woody debris trapped in a retention basin after two extreme events (1995, 2010) complement the data at high Q_w (>5000 L s^–1). Masses were calculated assuming a cylindrical shape and a dry density of 410 kg/m³, which is typical for the Norway Spruce (Picea Abies) that is common in the catchment. The three methods were made comparable by using distributions of particle masses (Turowski et al., 2013). CPOC was calculated from CPOM using the mean OC content of 47.8% ± 3.8% (±standard deviation) measured from 37 randomly drawn subsamples.

The transport rate of CPOC (kg C s^–1) was positively correlated with Q_w and well described by a power law rating curve (r² = 0.87; Fig. 2A). CPOC transport increases much more rapidly with increasing Q_w (rating curve exponent β = 4.14 ± 0.19) than DOC (r² = 0.98, β = 1.17 ± 0.04) and FPOC_biosphere (r² = 0.88, β = 1.90 ± 0.10). The data confirm that high river power is needed to mobilize and transport CPOC (West et al., 2011; Wohl, 2013). The relationship is consistent with the difference between bedload and suspended load transport rates in the Erlenbach (cf. Turowski et al., 2009; Smith et al., 2013), suggesting that CPOC is traveling as part of the bedload. This interpretation is supported by the observation that large fractions (mean 69%, median 78%) of the CPOM were water logged and denser than water, especially at high Q_w (Fig. 2B). Water logging likely occurs during storage of CPOM in log jams in the stream, or within saturated soil and litter on the hillslopes.

To estimate the decadal rate of CPOC discharge, we fitted a linear regression in double-logarithmic space to obtain a rating curve. The data points obtained from the retention basin material were not included in the regression, but are close to the rating curve at high Q_w. We used additional data from 2013, which resulted in a different rating curve than previously published (cf. Turowski et al., 2013). The rating curve was integrated over 31 yr of Q_w measurements. During this period, 4 exceptional flood events affected the catchment (Turowski et al., 2009), with peak Q_w >9000 L s^–1 and return periods >20 yr. Not accounting for these 4 floods, the background CPOC yield was 12.3 ± 1.9 t C km^–2 yr^–1. Uncertainties were derived from analytical errors of the rating curve fits. The exceptional floods delivered between 331 and 1066 t C km^–2, with an average of 585 t C km^–2. These values are lower than the 6300–19,100 t C km^–2 of large wood carbon (LWC) delivered to the ocean during typhoon Morakot in Taiwan (West et al., 2011), but higher than the 10–24 t C km^–2 of LWC delivered from the upper Rio Chagres, Panama, in a rain storm (Wohl and Ogden, 2013). In total, the 4 floods delivered 2338 ± 1609 t C km^–2, or 75.4 ± 51.9 t C km^–2 yr^–1. When added to the background rate, the average CPOC discharge estimate is 87.7 ± 51.9 t C km^–2 yr^–1. Exceptional flood events appear to be even more important for CPOC than for FPOC_biosphere (Hilton et al., 2012), which results from the steep relationship between CPOC transport rate and Q_w (Fig. 2A; Fig. DR1 in the Data Repository).

The background CPOC yield (12.3 ± 1.9 t C km^–2 yr^–1) from the Erlenbach is a significant catchment-scale carbon transfer (Hilton et al., 2012; Galy et al., 2015) and on its own is comparable to the upper range of estimates of FPOC_biosphere yields from temperate and tropical active mountain belts (Fig. 3). Other carbon transfers from the Erlenbach, obtained using the same methods on previously collected data (Hagedorn et al., 2000; Smith et al., 2013), are lower than CPOC transfer, with a DOC yield of 11.3 ± 0.0 Mg C km^–2 yr^–1, and an FPOC_biosphere yield of 10.7 ± 0.1 Mg C km^–2 yr^–1. The background CPOC transfer thus represents ∼36% of the decadal biospheric OC discharge by this catchment. Inclusion of the exceptional events raises CPOC transfer to as much as ∼80% of the total OC (TOC) discharge (Fig. 3). We can assess the sustainability of OC export by comparing it to the net primary production (NPP) of ∼740 Mg C km^–2 yr^–1 in the Erlenbach catchment (see the Data Repository). The background rate of CPOC discharge is ∼1.7% of this NPP and is sustainable, in agreement with a global compilation of river FPOC_biosphere yields (Galy et al., 2015). However, extreme events may severely deplete the biosphere stock of carbon. The CPOC discharge during a single event appears to have the potential to exceed the catchment’s yearly production; our data suggest that on decadal time scales, exceptional events discharge ∼10% of the NPP.

The contribution of CPOC to carbon discharge by rivers is not typically quantified, and a direct comparison with data from other catchments remains challenging. Notwithstanding, it has been calculated that LWC alone contributes at least 10% and as much as 35% of the total carbon yields in mountain rivers with catchment areas as large as 2000 km² (Fig. DR2) (Seo et al., 2008). CPOM particles smaller than large wood to sizes of 1 mm were not considered in that study, but dominate CPOC in the Erlenbach (cf. Turowski et al., 2013). Based on the Erlenbach’s size, its FPOC_biophere and LWC yields are similar to those observed in other mountain regions in the world (Fig. DR2). FPOC_biosphere yields are known to be strongly linked to physical erosion rate (Fig. 3) (Galy et al., 2015), and high yields are observed in active mountain belts in temperate and tropical settings (Hilton et al., 2012). In accord with this, estimates of LWC transfer in Taiwanese catchments are larger than for the Erlenbach (West et al., 2011). Therefore, we propose that the often unmeasured CPOC fraction is a significant component of POC_biosphere export from forested mountain catchments.

To make a tentative first assessment of the global significance of CPOC transport, we assume that the Erlenbach catchment is representative for temperate mountain forests, which cover a total area of 1.2 × 10⁶ km² worldwide (Sands, 2005). While the climatic, geomorphic, and ecological characteristics of the Erlenbach support that assumption, its physical erosion rate is high (Fig. 3). Without more measurements of CPOC transport (Fig. 1) and estimation of CPOC yields (Fig. 3), a global CPOC discharge estimate remains poorly constrained. Based on the Erlenbach background CPOC yield over 31 yr (12.3 ± 1.9 t C km^–2 yr^–1), the global CPOC discharge from temperate mountain forest catchments could be ∼15 Mt C yr^–1. This is ∼10% of the recent estimate of global FPOC_biosphere discharge to the oceans by rivers of 157 +74/–50 Mt C yr^–1 (Galy et al., 2015). If extreme floods are included, CPOC discharge from temperate mountain forests could be even higher (Fig. 3). Global CPOC discharge would further increase if boreal, subtropical, and tropical mountain forests were considered. We are aware that these estimates are based on extrapolation from a very small continental area and absolute flux has large uncertainty. Nevertheless, the magnitude of the estimate demonstrates the need to better quantify CPOC transfer rates in mountain rivers and track its conveyance through large river systems.

Little is known about the onward fate and routing of CPOC through large rivers. On average ∼69% of the CPOM transported by the Erlenbach was water logged, with a density greater than water (Fig. 2B). If this observation applies to other temperate streams where channel morphology can promote transient storage of CPOM, sampling with drift nets may have missed large fractions of CPOM traveling near the stream bed. Perhaps more important, water-logged CPOM may have a different fate in fluvial networks than if it were to float. During transport in steep channels, water-logged CPOM may be ground by gravel bedload, reducing its size. The size reduction of CPOM by bedload grinding is poorly understood, but the observed magnitude of the CPOC flux means it could be an important in-stream source of FPOC_biosphere (Hilton et al., 2012).

Furthermore, a high density of CPOM may promote its burial potential in sedimentary basins. If water-logged CPOM is delivered to depositional environments as part of the bedload it is more likely to rapidly accumulate in sedimentary deposits. Observations of large terrestrial organic debris in deep-sea turbidites in Indonesia (Saller et al., 2006), woody clasts and plant debris in modern deep-sea sediments offshore Taiwan (Kao et al., 2014), and mountain rivers draining the west coast of the United States (Leithold and Hope, 1999) all suggest that CPOC can be delivered to deep-marine settings. We substantiate these arguments by estimating the contribution of CPOC to TOC in exhumed turbidite sequences in the Apennines, Italy (see the Data Repository). Despite estimated transport distances of as much as 300 km offshore, CPOC was buried and preserved for 14 m.y. and represents ∼10% of the TOC. Water-logged woody debris can be delivered by mountain rivers as CPOC (Fig. 2B), and its presence may enhance the efficiency of carbon burial and associated atmospheric CO₂ sequestration by erosion of mountain belts (Kao et al., 2014; Galy et al., 2015).

CPOC is the dominant form of OC discharge by the Erlenbach over decadal time scales, increasing the carbon loss from the biosphere by ∼250% over DOC and FPOC_biosphere. The majority of CPOC may be transported in water-logged CPOM as part of the bedload. Our observations provide new impetus to study the production, transfer, and routing of CPOC from mountain headwaters, and subsequently through large river systems to fully assess the net impact of erosion on the global carbon cycle (Battin et al., 2008; Hilton et al., 2012, 2015; Galy et al., 2015). Due to anthropogenic CO₂ emissions and global warming, extreme precipitation events may become more frequent (Rajczak et al., 2013), causing an increased number of extreme floods. CPOC transport exhibits a much stronger dependency on water discharge than FPOC and DOC transport (Fig. 1), and could therefore become more important for carbon budgets of mountain streams in the coming decades. This may have implications for forest management, food availability in stream ecosystems, and carbon mobilization by erosion of the terrestrial biosphere.

We thank the current and former members of the mountain hydrology team for help in the field and in the laboratory. M. Sieber prepared the samples for chemical analyses. F. Hagedorn shared his DOC data of the Erlenbach. A. Galy and N. Hovius assisted with fieldwork in the Apennines and subsequent chemical analyses. Discussions with A. Badoux, K. Bunte, V. Galy, N. Hovius, J. Kirchner, K. Krause, and P. Schleppi are acknowledged. We thank E. Leithold, P. Raymond, and an anonymous reviewer for their comments, which improved the work. This study was supported by Schweizerischer Nationalfonds (SNF) grant 200021_124634/1, Engineering and Physical Sciences Research Council (UK) grants EP/P502365/1 and EP/P504120/1, and the Swiss Federal Research Institute for Forest, Snow and Landscape Research.