Sources and the flux pattern of dissolved carbon in rivers of the Yenisey basin draining the Central Siberian Plateau

Frequent measurements of dissolved organic (DOC) and inorganic (DIC) carbon concentrations in rivers during snowmelt, the entire ice-free season, and winter were made in five large watersheds (15 000–174 000 km2) of the Central Siberian Plateau (Yenisey River basin). These differ in the degree of continuous permafrost coverage, mean annual air temperature, and the proportion of tundra and forest vegetation. With an annual DOC export from the catchment areas of 2.8–4.7 gC m − 2 as compared to an annual DIC export of 1.0–2.8 gC m − 2, DOC was the dominant component of terrigenous C released to rivers. There was strong temporal variation in the discharge of DOC and DIC. Like for other rivers of the pan-arctic and boreal zones, snowmelt dominated annual fluxes, being 55–71% for water runoff, 64–82% for DOC and 37–41% for DIC. Likewise, DOC and DIC exhibited also a strong spatial variation in C fluxes, with both dissolved C species decreasing from south to north. The rivers of the southern part of the plateau had the largest flow-weighted DOC concentrations among those previously reported for Siberian rivers, but the smallest flow-weighted DIC concentrations. In the study area, DOC and DIC fluxes were negatively correlated with the distribution of continuous permafrost and positively correlated with mean annual air temperature. A synthesis of literature data shows similar trends from west to east, with an eastward decrease of dissolved C concentrations and an increased proportion of DOC in the total dissolved C flux. It appears that there are two contemporary limitations for river export of terrigenous C across Siberia: (1) low productivity of ecosystems with respect to potentially mobilizable organic C, slow weathering rates with concomitant small formation of bicarbonate, and/or wildfire disturbance limit the pools of organic and inorganic C that can be mobilized for transport in rivers (source-limited), and (2) mobilization of available pools of C is constrained by low precipitation in the severe continental climate of interior Siberia (transport-limited). Climate warming may reduce the source limitation by enhancing primary production and weathering rates, while causes leading to surmounting the transport limitation remain debatable due to uncertainties in predictions of precipitation trends and other likely sources of reported increase of river discharges.


Introduction
Understanding the role of Siberian forested watersheds underlain by permafrost in global element cycling is becoming increasingly important as temperature increases and permafrost degrades at high latitudes (McGuire andAnderson 2009, Holmes et al 2011). Much attention has been focused previously on land-atmosphere exchange of C in forests of Central and Eastern Siberia (Nakai et al 2008, Dolman et al 2004, Tchebakova et al 2009 and hydrological trends linked to climate warming (Shiklomanov and Lammers 2009 and references therein), but considerably less is known about the transfer of C and other elements from terrestrial to aquatic systems in this region (Pokrovsky et al 2005, Prokushkin et al 2008. Permafrost exerts strong controls on the hydrogeochemistry of subarctic rivers by constraining water flow to surface organic-rich soil layers with low residence times and little water-rock interaction (Rember and Trefry 2004, Pokrovsky et al 2005, 2006, Holmes et al 2011, Bagard et al 2011. A deeper soil active layer (the portion of soil that thaws and refreezes annually) and shrinkage of permafrost extent with warming (Frey andMcClelland 2009, Romanovsky et al 2010) may significantly increase soil infiltration and increase retention times in the soil profile (Frey and McClelland 2009) and adsorptive capacity of soils (Kawahigashi et al 2004). These changes may also accelerate rates of biotic mineralization of DOC in soils . Longer flow path, therefore, decreases dissolved organic carbon (DOC) flux to rivers, while it increases the fluxes of inorganic carbon (DIC) and other major ions with enhancement of soil carbon mineralization (i.e., respiratory CO 2 production) and bedrock weathering at higher temperatures , Cai et al 2008. In particular, in carbonate-free soils, respiratory CO 2 will react with silicates (e.g., feldspars) to produce clay minerals and bicarbonate-ion (Amiotte-Suchet et al 1999). Thus, riverine DIC in silicate watersheds, in opposite to calcareous soils, originates mainly from soil respiration .
Other mechanisms, however, may result in an increase in fluvial export of DOC from terrestrial ecosystems under a changing climate.
These include (i) an increase in primary productivity of vegetation that increases DOC production; (ii) increased precipitation, which increases microbial mobilization of DOC from organic-rich layers, and (iii) introduction of a new source of C from degrading Crich permafrost in the western Siberia peatlands (Frey andSmith 2005, Gordeev andKravchishina 2009) and the Yedoma complexes in eastern Siberia (Neff et al 2006).
Among the apparent consequences of climate changes in high latitude basins are increased riverine discharge (Peterson et al 2002, McClelland et al 2006, northward shifts in vegetation (Sturm et al 2001, Kharuk et al 2005 and enhanced biogeochemical cycling of carbon, nutrients and major/trace elements (Frey and McClelland 2009). With approximately 50% of the world's soil organic carbon (SOC) stored in the high latitude regions (Tarnocai et al 2009) and the importance of high-carbon soils for driving riverine DOC fluxes (Aitkenhead and McDowell 2000), the arctic/subarctic river basins underlain by permafrost have an enormous potential to mobilize and transport terrestrial organic carbon (OC) to the Arctic Ocean (Guo and MacDonald 2006, Guo et al 2007, McGuire and Anderson 2009. Contemporary riverine transport of dissolved carbon to the Arctic Ocean is estimated to be 18-33 Tg C as DOC, 6 Tg C as particulate organic carbon (POC) and another 43 Tg C as DIC annually (McGuire and Anderson 2009), representing a major component of the global carbon cycle (Spitzy andLeenheer 1991, Hedges et al 1997). Recent evidence from the northern hemisphere showing increased DOC and DIC concentrations in surface waters draining upland areas and wetlands (Freeman et al 2001, Frey and Smith 2005, Cai et al 2008, highlights the importance of understanding the transfer of C between soil and freshwater systems (Hope et al 1994, Billett et al 2006. In this study, we examine the seasonal and annual variations in concentrations and fluxes of two dissolved C species (DOC and DIC) in five rivers draining the Central Siberian plateau, which drain watersheds of similar lithology that span a range of continuous permafrost distribution, air temperatures and vegetation. Our objective was to use high-resolution measurements of river water C concentrations and discharge to understand how biological, edaphic and hydrological processes interact to control DOC and DIC fluxes. Specifically, we addressed the following questions: (1) What are the differences in DOC and DIC export among rivers draining similar lithology under different climatic contexts?
(2) How do seasonal changes in flow paths (sources) control river dissolved C export? (3) How do climate parameters such as temperature and precipitation affect the export of different forms of dissolved C? and (4) How do permafrost distribution and vegetation types affect the net export of C species from river basins of the Central Siberian Plateau?

Study area
The study region is found in the central and southern part of the basaltic province located within the Central Siberian Plateau and includes a considerable part of the basins of the Nizhnyaya Tunguska and Podkamennaya Tunguska Rivers (figure 1), tributaries of the Yenisey River. Our five study rivers, Tembenchi (referred thereafter as TE), Kochechum (KO), Nidym (ND), Nizhnyaya Tunguska (NT) and Podkamennaya Tunguska (PT) drain basins that vary in watershed sizes, permafrost extent, climate and vegetation (table 1). Total watershed area of the rivers studied is about 500 000 km 2 or roughly one third of the Central Siberian Plateau. Permafrost extends throughout the studied area but ranges from continuous permafrost on the north to isolated patches in the southern part of the basins of the Nizhnyaya Tunguska and Podkamennaya Tunguska Rivers (Brown et al 1998). The mean annual air temperatures (MAAT) range among basins from −10.9 to −6.3 • C (2001-9, 0.5 • × 0.5 • grid data, CRU TS3.1 available at http: //badc.nerc.ac.uk/, Mitchell and Jones 2005). A detailed description of each river basin is given in the electronic annex (available at stacks.iop.org/ERL/6/045212/mmedia).  (Bartholomé and Belward 2005) and permafrost (Brown et al 1998).

Analyses
Repeated river water sampling was used to examine seasonal and annual dynamics of concentrations of dissolved organic and inorganic C in rivers during several hydrologic years  and to estimate annual carbon export. Sampling sites were located close to gauging stations of the Russian Federal Service of Hydrometeorology and Environment Monitoring (ROSHYDROMET) and/or river mouths if no gauging station on the river was established. Water samples were collected from just beneath the water surface at the mid-point of rivers at monthly intervals under ice conditions (October-April), at least once a week at snowmelt freshet (from mid May and beginning of June) and in summer and autumn (mid-June-September). Samples were immediately filtered (pre-rinsed 0.22 μm nitrocellulose filters, Millipore) and stored refrigerated prior to analysis. Dissolved organic carbon (DOC) concentrations were measured as non-purgeable organic carbon (NPOC) via high temperature combustion (Shimadzu TOCvcpn, Kyoto, Japan) with an uncertainty of 3% and a detection limit of 40 μM. Blanks of MilliQ water passed through the filters demonstrated negligible release of DOC from the filter material. Dissolved inorganic carbon (DIC) concentrations were obtained from alkalinity determined following a standard HCl titration procedure using an automatic Schott TitroLine alpha TA10 plus titrator with an uncertainty of ±2% and a detection limit of 5 × 10 −5 M. At the essentially neutral pH values of the river water (7.5-8.0), and 1-100 μg l −1 of dissolved boron (Bagard et  ) represents, within 10% uncertainty, the total inorganic dissolved carbon. The contribution of weak organic acids to titratable alkalinity was evaluated via (i) titration of filtered river water subjected to UV and H 2 O 2 degradation to remove (up to 80-90%) of dissolved organic matter without modifying the pH of solution and (ii) by adding organic-rich, bicarbonate-poor natural water to a standard bicarbonate solution and determining its alkalinity. In both cases, the contribution of natural DOM was found to be negligible (within ±10% of total titratable alkalinity).
Specific ultraviolet absorbance (SUVA), a measure of DOC aromatic carbon content (Weishaar et al 2003), was measured at 280 nm and normalized to DOC concentration (m l −1 mg C −1 ). The proportion of colloidal high molecular weight compounds in DOC (HMW DOC) was measured as the difference between OC concentrations in filtered water samples (0.22 μm, Millipore) and those that had passed through a dialysis membrane, representing low molecular weight substances (<1 kDa, LMW DOC). Dialysis was performed on-site using 20-50 ml precleaned dialysis bags placed in a large volume (10-20 l) of river water. The duration of this dialysis procedure was between 24 and 72 h, an exposure time based on experimental assessment of the time needed to attain equilibrium in DOC concentrations (Pokrovsky et al 2005(Pokrovsky et al , 2011. For dialysis experiments, EDTA-cleaned trace-metal pure SpectraPor 7 ® dialysis membranes made of regenerated cellulose (pore size of 1 kDa, or ∼1 nm) were thoroughly washed in 0.1 M double-distilled HNO 3 , ultrapure water, filled with ultrapure MilliQ deionized water and then placed into natural water. The efficiency of the dialysis procedure was evaluated by comparing concentration of major anion or neutral species (e.g., Cl − , H 4 SiO 4 ) not associated with colloids between the dialysis bag and the external solution. These concentrations were always identical to within ±20%, suggesting an equilibrium distribution of dissolved components. To assess the mass balance during the dialysis procedure, concentrations of OC were measured in the external solution and internal compartments and compared with <0.22 μm filtrates. In all cases, better than 95% recovery for OC was achieved indicating that the adsorption of colloids onto and inside the thin Spectra Por 7 membrane was negligible.
Daily discharges for the TE (Point ID: 6629), NT (at Tura, Point ID: 6625) and monthly discharges for the PT (at Baikit, Point ID: 6614) were obtained from the ROSHYDROMET (Krasnoyarsk) for the period 2006-10. Long-term (1939-95) monthly discharges of these rivers including the PT (Point ID: 6614) were obtained from the R-ArcticNet Database (www.r-arcticnet.sr.unh.edu/v4.0/main.html). Since there is no gauging station at Kochechum River (KO), its daily and monthly discharges were estimated using data from the TE, assumed as river-analog with a basin with similar geomorphological characteristics (Gerasimov 1964) and longterm records of individual flow measurements. In this data set, the KO to TE long-term annual discharges (29.96 and 7.97 km 3 , respectively, Gerasimov 1964), yielded the ratio 3.76 which was used to estimate KO daily discharge from TE data. The ND discharge was acquired as long-term mean (Gerasimov 1964) and then monthly separation of discharge was estimated on the basis of the NT values, demonstrating similar annual runoff separation by seasons to Taimura River (Point ID: 6630), the analog of Nidym River. However, in the latter case, daily discharges were terminated in early 1990s.
Historic data for discharges, DIC and total organic carbon (TOC, the sum of dissolved and particulate organic carbon) concentrations in NT and TE were collected by the State Service of Hydrometeorology and Environment Monitoring (USSR) for the period 1966-75 (Hydrological Yearbooks of Russian Hydrological Survey, 1954Survey, -1975Survey, 1975.

Calculation of DOC and DIC fluxes
First, for comparative analysis of all five rivers, monthly, seasonal and annual concentrations of DOC and DIC were calculated as simple averages. Additionally, monthly, seasonal and annual concentrations of DOC and DIC have been calculated for the TE, KO (estimated using Tembenchi river discharge) and NT as flow-weighted means. The calculation of mean annual concentrations for ND and PT rivers is based on the contribution of each month to the annual runoff. Then, seasonal and yearly river C loads (mass C period −1 ) were estimated using daily load (mass C day −1 ) obtained from a time series of paired river flow and constituent concentration data used to construct a calibration curve. Finally, the areanormalized DOC and DIC exports with rivers have been expressed per watershed area (gC m −2 a −1 ) and additionally, due to significant differences in annual precipitation and specific runoff, the annual DOC and DIC mobilization in the catchment have also been normalized to annual runoff (gC m −2 a −1 mm −1 ).

Statistics
Discharge and chemical data were analyzed by best fit function based on the method of least squares, Pearson correlation and one-way ANOVA with STATISTICA version 6 (StatSoft Inc., Tulsa, OK). Regressions and power functions were used to examine the relationships between discharge and river constituent concentrations or DOC parameters. Correlation coefficients were calculated to elucidate relationships between carbon concentrations in rivers and watershed attributes. ANOVA was used to test the differences in average concentrations and DOC parameters among rivers and regression slopes among watersheds.

Hydrology
River discharge was extremely seasonal in all our study sites (figure 2), demonstrating up to three orders of magnitude variation between spring freshet (peak flow in first week of June) and late winter low flows (April). Smaller rivers in the NT basin like Nidym River (ND) freeze completely in winter with zero discharge in early March. Annual hydrograph separation for the TE, NT and PT showed that 55-71% of annual discharge occurred during spring (May-June), the summer to autumn period (July-September) accounted for 18-42% of annual flow, and the winter season (October-April) contributed 3-11% (table 2). The proportion of summer flow decreased as latitude decreased, while the proportion of spring and winter flows increased. The mean seasonal distribution of discharge in our study period was similar to that from the longterm record , but with elevated winter discharges of TE (up to 5% of annual discharge) and NT (5.0-8.6%). The contribution of snowmelt during the five years of our study varied from 46 to 67% in TE (KO), and from 53 to 79% in NT. The decreased proportion of snowmelt in the annual hydrograph corresponded to an elevated annual discharge in wet years when the summer-autumn period was characterized by higher precipitation and the proportion of summer discharge rose to almost 50%. In contrast, in dry years with low annual discharge (e.g. 2006), the proportion of spring freshet reached >80%, and summer constituted only 15-18% of the annual hydrograph.
Average annual discharge at both TE and NT during 2006-10 was ca. 30% higher than the 50 year record from 1939 to 1995 (10.3 versus 7.9 km 3 and 65.2 versus 50.7 km 3 , respectively). On a seasonal basis, spring discharges increased 20% and summer discharges increased >40%. In general, there was a strong indication that annual discharge may be trending upward at these rivers at a rate of 0.333 km 3 a −1 a −1 for the NT and 0.046 km 3 a −1 a −1 for the TE (figure 3).

DOC and DIC: spatio-temporal variation
Concentrations of both dissolved C species in river waters showed high inter-seasonal and inter-annual variation depending on flow regimes of the rivers (figures 2 and 4). For each river, there was approximately an order of magnitude difference between the largest and smallest concentrations among seasons. Mean DOC concentrations were greatest during spring freshet in all rivers (table 3), although the northernmost basins (TE and KO) reached maximum DOC concentrations prior to the peak of discharges (figure 2(a)). Southernmost rivers (NT and PT) also showed an abrupt increase of DOC concentrations in the beginning of snowmelt and a gentle rise at higher discharges. As a result for all analyzed rivers, a power function (e.g., DOC = Q a ) provided the best fit to explain the relationships between discharge and river constituent concentrations or DOC parameters.
During the summer-autumn season, DOC concentrations gradually decreased, with peaks when discharges increased after rainfall events. The proportion of HMW DOC (>1 kDa) also declined after snowmelt (figure 5(a)). Aromaticity of riverine DOC (SUVA) demonstrated a similar trend which Table 2. River discharges (Q, km 3 ), C loads (GgC period −1 ) and export (gC m −2 period −1 ). followed the decrease of discharge (figure 5(b)). Interestingly, minimum DOC concentrations and SUVA index were observed in winter lowflow in April, and smallest concentrations of HMW DOC appeared at summer lowflow (July). Within the Central Siberian Plateau mean seasonal and annual DOC concentrations in rivers increased from north to south, positively correlating with MAAT (r = 0.89, p = 0.045) and forested area (r = 0.86, p = 0.07), but a negative trend found with an area of continuous permafrost distribution (r = −0.66, p = 0.22). Aromaticity of DOC (SUVA), in contrast, decreased from north to south (table 3). In contrast to DOC, dissolved inorganic C concentrations were smallest during spring and steadily increased from summer to winter (figure 2). Summer-autumn stormflows resulted in short-term drops of DIC concentration in river waters. Similar to DOC, DIC concentrations increased from the northern to the southern part of the Central Siberian Plateau and correlated positively with MAAT (r = 0.98, p = 0.004), and negatively with permafrost distribution (r = −0.93, p = 0.02).
Flow-weighted and average concentrations were calculated separately for seasons (winter, spring and summerautumn) and the entire year (table 3).
Specifically, rivers demonstrated high concordance among seasonal flowweighted and average concentrations, but annual average DIC values tended to be 1.3-2.2-fold larger than flow-weighted concentrations. DOC concentrations as seasonal averages were smaller in winter and spring (up to 10% for NT and 26% for KO) or similar in summer (1-2% difference) in comparison to flow-weighted means. Annual average concentrations were ca 20% smaller than annual flow-weighted DOC concentrations in the five rivers. Particularly, for the NT basin, average concentrations were 2.3-fold larger for DIC and 1.3-fold smaller for DOC compared to their respective flowweighted means. Similarly, average annual concentrations in KO were 1.8-fold larger for DIC and 1.2-fold smaller for DOC.

Carbon export
Averages of 111, 552, 63, 1942 and 799 Gg C (as DOC and DIC) were transported annually by TE, KO, ND, NT and PT, respectively (table 2). Riverine carbon flux normalized to watershed area demonstrated an almost two-fold variation among basins (4.3-7.3 gC m −2 a −1 ), with southern catchment tending to show larger export rates. Dissolved organic carbon constituted the major proportion of the annual total dissolved C export, but decreased in importance in the southern rivers (62-65%) compared to northern rivers (74-82%). Export of DOC followed discharge, with the peak occurring during snowmelt. Annual DOC load was dominated by the snowmelt period (64-82% of annual total), due to both the larger discharges and a 3-4-fold elevation in DOC concentrations that occurred during the snowmelt period. The DIC load generally peaked during the summer-autumn season and in spring, despite low DIC concentrations in the latter period: 28-52% and 37-41% of annual values, respectively. However, in the southernmost river studied (PT) the seasonal load of DIC varied little, ranging from 28 to 37% in the three sampling seasons. As a result, the proportion of winter load in annual fluxes increased from 7 to 36% from our northernmost to southernmost site. To characterize inter-annual variation in discharges, linear regressions of C flux versus total water yield were calculated for each flow period in 2006-10 for the KO and NT basins. Despite a general positive correlation between DOC export and water yield (figures 6(a) and (b)), a limitation of DOC export during spring was observed on the NT. DOC export did not increase as water yields exceeded 150 mm, but instead reached a maximum of 3.7 gC m −2 period −1 . Although DOC export during the summer-autumn season had a strong response to discharge, it did not reach spring values at the same discharge level. In contrast, DIC export demonstrated a weak relationship to discharge during the frost-free period with relatively stronger relationships for the NT than for the KO (figures 6(c) and (d)). Winter DIC export, however, was strongly related to discharge in both rivers.
Due to significant differences in annual precipitation and evapotranspiration among these watersheds, and corresponding differences in specific runoff (320 versus 188 mm in KO and NT), the efficiency of C mobilization from terrestrial ecosystems has also been evaluated by normalization of annual riverine DOC export to annual runoff. Figures 6(e) and (f) show that such normalized annual DOC and DIC export was 1.8 and 2.4-fold larger in NT basin relative to KO, indicating the higher efficacy of C release due to higher C stocks and/or more efficient mobilization/higher net DOC/DIC production in the southern watershed.

Seasonal patterns of riverine C flux
There are several factors that distinguish runoff hydrology and hydrogeochemistry of high latitude basins from those of temperate regions: (1) snowmelt is the major hydrological event (Finlay et al 2006), conveying to rivers about half of Table 3. Average and flow-weighted concentrations of dissolved organic carbon (DOC), dissolved inorganic carbon (DIC), and specific ultraviolet absorption (SUVA) in water of five rivers draining the Central Siberian Plateau for the period 2006-10.
DOC (mg C l −1 ) D I C ( m g C l −1 ) S U V A ( m l −1 mg C −1 ) Average ± SD Flow-weighted Average ± SD Flow-weighted Average ± SD Flow-weighted  . Such phenomena occur widely across a large geographical range in the pan-arctic and boreal zones, and they reflect a large pool of C available for leaching and transport during freshet (McGuire and Anderson 2009). The strong positive relationship between C flux and discharge also suggests that hydrological flux may cause the most important limitation of C mobilization from the highly continental climate of interior Siberia with typical summer severe droughts. However, within a watershed, timing of snowmelt and local rainstorm events in areas with different stocks of mobile C may likely cause intra-seasonal variability of riverine C sources. In particular, we documented mismatch in the timing of maximum DOC concentrations and peak of discharge during spring freshet in northernmost rivers (i.e., Tembenchi and Kochechum). These findings suggest (1) higher DOC input in the earlier phase of spring freshet (May) originates from forested C-rich southern part of basins and (2) less DOC release at the peak of discharge (June) appears as snowmelt expands to headwater areas with high coverage of tundra vegetation and limited soil C available for transport. The lower concentrations of DOC also in summer season in headwaters of the Kochechum River reported by Pokrovsky et al (2005) compared to Kochechum River downstream DOC levels corroborate this hypothesis. On the other hand, permafrost distribution on a watershed may also define basin-contributing areas, as lateral flow is confined to terrains with a shallow active layer due to their ability to restrict deep percolation. Our recent analysis of DOC chemical and isotopic tracers (Prokushkin et al 2007) revealed a negligible signal of landscape elements characterized by deep active layers (i.e. south-facing slopes) in stream water due to an insufficient amount of precipitation to move solutes from the soil to surface waters.
The increasing retention time of solutes in deeper active layers caused by seasonal permafrost thaw leads to a net decrease of DOC export to rivers in summer due to DOC sorption and/or mineralization in the mineral soil (Striegl  (2004) showed a significant decline in DOC concentrations in small tributaries to the Yenisey River along a gradient from continuous to discontinuous permafrost. This study also reported major alterations in biochemical composition of DOC in streams draining low-permafrost basins (e.g. an decrease in lignocellulose compounds and an increase in the hydrophilic DOC fraction), which confirms the significant control that is exerted by the depth of soil active layer and distribution of permafrost on the flux, composition, and biodegradability of DOC in Siberian soils during the frost-free period. In this sense, seasonal changes of DOC flux, age and biochemical composition documented for the Yukon (Striegl et al 2005, Guo and MacDonald 2006, Spencer et al 2008 and Kolyma (Finlay et al 2006, Neff et al 2006 rivers indicating a general shift from younger topsoil DOC during freshet to more degraded DOC from subsoil during summer and winter low flows reflect the likely behavior of DOC with progressive warming and degradation of permafrost.

Riverine DOC composition
The relative proportion of HMW (colloidal, 1 kDa-0.22 μm) DOC exhibited both seasonal and spatial variations. The highest proportion of HMW DOC (60-70%) is observed during the spring flood period, in accord with recent data from another subarctic, but permafrost-free river Severnaya Dvina (Pokrovsky et al 2010). During summer baseflow, the LMW (<1 kDa) fraction constitutes from 60 to 90% of DOC. These observations might be consistent with the presence of two pools of organic matter: allochthonous large size colloids formed by lixiviation from upper soil horizons and autochthonous (aquatic) small molecular-size substances, probably linked to bacterial and phytoplankton exudates. The relative proportions of both possible pools are highly variable seasonally. Allochthonous input of organic carbon strongly increases during the spring and summer-fall floods, whereas an autochthonous C typically peaks in open water season, though following changes in both light and temperature regimes that affect photosynthesis/respiration. Higher SUVA index in all rivers during spring freshet corroborates these findings, reflecting the input of fresh allochthonous (soil) organic matter from the watershed compared to autochthonous DOC with low SUVA appeared in summer and fall. Thus co-variation of colloidal material and aromaticity confirms the dominant role of soil humic substances in the riverine DOC flux in spring season and vice versa in river sources of DOC during summerfall. Further, our study of aromaticity of DOC (SUVA index) in rivers of the Central Siberian Plateau showed a decreasing trend southward, indicating release of more processed DOC even during spring (table 3, figure 5(b)).
An alternative explanation for the seasonal changes in size distribution of DOC is based on the observation that the LMW fraction in Arctic rivers is more refractory than the colloidal fraction (Guo and MacDonald 2006). As such, summertime LMW DOC in a large river may also originate from degradation of old OC buried in the permafrost, notably at the river banks. Note that such a distinction between colloidal and dissolved OC is rather conventional since it is known that fulvic acids, constituting an essential part of DOC in boreal waters, can be described as rigid oligoelectrolytes with molar mass of ∼1000 Da and a radius (∼1 nm) which is not much larger than that of hydrated metal ions (cf Buffle 1988, Buffle et al 2007.
On the other hand, increased productivity and northward shifts of vascular plants and specifically Larix spp. in headwater streams of northern rivers with warming may provide an important new source to the existing DOC pool. Indeed, the rivers we have analyzed demonstrate positive relationships between DOC concentrations and the forested area of their watersheds, which appears to serve as a useful proxy for the increased NEP and C stocks.

DIC production and transport
Weathering of basalts and associated transport of DIC is especially important in view of weathering control on CO 2 consumption from the atmosphere. Indeed, according to Dessert et al (2001), the chemical weathering rate of volcanic rocks is 5-10 times higher than the chemical weathering of granite and gneiss. In such a case, the atmospheric CO 2 consumption derived from basalts weathering represents 30% of the global silicate weathering flux and acts as an important regulator of the Earth's environment (Dessert et al 2003).
This global estimate corresponds to the CO 2 consumption associated with the weathering of volcanic rocks occurring in soils, shallow groundwaters and rivers. It thus corresponds to the flux of CO 2 that is susceptible to be involved in a climatic feedback where higher surface temperature would lead to higher CO 2 consumption by chemical weathering (Dessert et al 2003). Our study, providing the lowest values of DIC concentrations in basaltic monolithological river waters, is in full agreement with the temperature trend of atmospheric CO 2 consumption during volcanic rock weathering established in earlier studies. Indeed, there is a clear increase of DIC annual yield from 1.04 gC m −2 yr −1 for TE (MAAT = −10.9 • C) to 2.77 gC m −2 yr −1 for NT (MAAT = −7.5 • C). The other basaltic boreal rivers with similar runoff such as Kamchatka River and its tributaries (MAAT = −2.5 • C, 522±87 mm yr −1 runoff) exhibit significantly higher annual fluxes of DIC: 4.8 gC m −2 a −1 (Dessert et al 2009).

Implications for climate change
Our observation that in the Central Siberian Plateau riverine DOC concentrations increase toward the south generally agrees with data obtained for streams in a latitudinal gradient from 55 to 68 • N of Western Siberia (Frey and Smith 2005). It infers that climate warming (i.e. increase of MAAT and permafrost degradation) amplifies DOC flux. On the other hand, the negative effects of low MAAT and continuous permafrost distribution on annual and seasonal DOC and DIC concentrations in Siberian rivers oppose the patterns seen in earlier work conducted within discontinuous permafrost regions of Alaska (e.g. MacLean et al 1999, Striegl et al 2005, Petrone et al 2006. It appears that the harsh climate and continuous permafrost of the Central Siberian Plateau may constrain C release in terrestrial environments through inhibition of the biotic and abiotic processes that produce mobile organic and inorganic C. In particular, low MAAT and continuous permafrost distribution in higher latitude and higher altitude rivers of the Central Siberian Plateau show especially decreased C concentrations. To broaden our analysis of riverine dissolved C concentrations and export, we examined other high-resolution time series for watersheds within the Arctic Ocean basin. In total, this search yielded 12 estimates of annual DOC and DIC concentrations and fluxes in major Siberian and North America rivers (i.e. Ob', Yenisey, Khatanga, Anabar, Olenek, Lena, Yana, Indigirka, Kolyma, Amguema, Yukon and Mackenzie) based on data reported in a number of papers (Gordeev et al 1996, 2004, Cai et al 2008, McGuire and Anderson 2009) including those of the partners and Arctic-GRO projects (Cooper et al 2005, Holmes et al 2011. We used these data to explore spatial changes in the proportion of organic and inorganic C in the overall export of terrigenous C to the Arctic Ocean across a strong climatic gradient from west to east, and also the lithological gradient associated with different proportions of sedimentary versus  et al 1996, 2004, Cai et al 2008, McGuire and Anderson 2009, Holmes et al 2011 volcanic rocks. This compilation clearly shows a decrease of total dissolved C concentrations and exports in a northward direction for watersheds of the Central Siberian Plateau (figure 7), and a similar decline eastward for major Siberian rivers emptying into the Arctic Ocean. Despite low coverage by peatlands in the Central Siberian Plateau (0.7-2.1% of area, GLC2000, Bartholomé and Belward 2005), there are generally high DOC concentrations in all studied rivers as compared with rivers of large size watersheds of Western (Frey and Smith 2005) and Eastern Siberia (Gordeev andKravchishina 2009, Gordeev et al 2004) with greater peat soil distribution. In streams with small-size basins of Western Siberia, Frey and Smith (2005) reported elevated concentrations of DOC (up to 70 mg C l −1 ) based on measurements in late summer.
Declining DIC concentrations in both northward and eastward directions were the main cause for the observed decrease in total dissolved C with latitude. In contrast, DOC concentrations declined northward in Central Siberia (this study), but did not demonstrate an obvious trend in eastward direction for other major Siberian rivers. Nevertheless, the proportion of DOC in total C grew from 40% at Ob River to 60% at Kolyma River and reached 80% in the easternmost Amguema River. Similarly, DOC proportion was ca 80% in the northernmost Tembenchi River and dropped to 60% in the southernmost Podkamennaya Tunguska River. North American rivers (i.e., Yukon and Mackenzie) generally showed the lowest DOC concentrations and export compared to rivers of the Siberian sector of the Arctic Ocean. In contrast, they both show an extremely high proportion of DIC in total dissolved C, causing an elevated C yield. The higher concentration of DIC in North American subarctic rivers stems from the extensive coverage of easily-weathered sedimentary rocks , compared to the volcanic lithology of the Central Siberian rivers (Pokrovsky et al 2006).
Thus, there are strong indications that ongoing changes in high latitude basins may result in an increased C flux to the Arctic Ocean, as both the transport (river flow) and carbon sources (e.g. terrestrial primary production) are increasing. The recently reported projections of increases of DOC and DIC concentrations in rivers in a north-south transect within the Ob' basin (Western Siberia) due to projected near-doubling of area with −2 • C MAAT isotherm (Frey and Smith 2005) corroborate our findings. Degradation of permafrost in Eastern Siberia, characterized by large C pools conserved in Pleistocene-age loess permafrost (yedoma) has been suggested to accelerate C turnover and increase terrestrial C release to the Kolyma River (Neff et al 2006). Gordeev and Kravchishina (2009) reported recently up to 7-fold increase of C export to the Arctic ocean to 2100, as Arctic rivers demonstrate increasing runoff (Peterson et al 2002, McClelland et al 2006 and productivity of vegetation on their basins (Sturm et al 2001, Kirdyanov et al 2011. At our study sites the increase of riverine DOC concentrations and discharges for the period of 2006-10 as compared with period 1966-75 resulted in a 50% increase of DOC export in rivers of the Central Siberian Plateau. However, the source of water and consequently the longevity of increasing river discharges remain debatable (McClelland et al 2006).

Conclusions
Forested basins of the Central Siberian Plateau, underlain by continuous and discontinuous permafrost, export disproportionally to net ecosystem productivity large amounts of terrigenous C to rivers. Similar to other rivers of the Arctic Ocean basin, snowmelt loads of dissolved C dominated its annual fluxes. Dissolved organic carbon was the dominant component of terrigenous C released to riverine systems. Dissolved carbon concentrations decrease from south to north, negatively correlated with the distribution of continuous permafrost and positively correlated with mean annual air temperature. Similar trends are found for 10 major Siberian rivers from west to east, with an eastward decrease of dissolved C concentrations and an increased proportion of DOC in total dissolved C flux.
It appears that there are two contemporary limitations on terrigenous C export: (1) low productivity of ecosystems and slow weathering rates, which lessen the pools of organic and inorganic C potentially mobilized to aquatic systems (sourcelimited) and (2) low precipitation in the severe continental climate of interior Siberia limits mobilization of available pools of C (transport-limited). Thus, climate warming in high latitude watersheds may cause increased DOC and DIC production by enhanced terrestrial primary production, increased production of mobile C in soil, release of organic carbon from degrading permafrost and increased weathering rates for potential release to rivers. DIC fluxes in rivers of the Central Siberian Plateau yielded the lowest values of CO 2 consumption in basaltic terrains, in agreement with the temperature trend of atmospheric CO 2 consumption during volcanic rock weathering established in earlier studies.