Dissolved organic carbon in permafrost regions: A review

A large quantity of organic carbon (C) is stored in northern and elevational permafrost regions. A portion of this large terrestrial organic C pool will be transferred by water into soil solution (~0.4 Pg C yr−1) (1 Pg=1015 g), rivers (~0.06 Pg C yr−1), wetlands, lakes, and oceans. The lateral transport of dissolved organic carbon (DOC) is the primary pathway, impacting river biogeochemistry and ecosystems. However, climate warming will substantially alter the lateral C shifts in permafrost regions. Vegetation, permafrost, precipitation, soil humidity and temperature, and microbial activities, among many other environmental factors, will shift substantially under a warming climate. It remains uncertain as to what extent the lateral C cycle is responding, and will respond, to climate change. This paper reviews recent studies on terrestrial origins of DOC, biodegradability, transfer pathways, and modelling, and on how to forecast of DOC fluxes in permafrost regions under a warming climate, as well as the potential anthropogenic impacts on DOC in permafrost regions. It is concluded that: (1) surface organic layer, permafrost soils, and vegetation leachates are the main DOC sources, with about 4.72 Pg C DOC stored in the topsoil at depths of 0–1 m in permafrost regions; (2) in-stream DOC concentrations vary spatially and temporally to a relatively small extent (1–60 mg C L−1) and annual export varies from 0.1–10 g C m–2 yr–1; (3) biodegradability of DOC from the thawing permafrost can be as high as 71%, with a median at 52%; (4) DOC flux is controlled by multiple factors, mainly including vegetation, soil properties, permafrost occurrence, river discharge and other related environmental factors, and (5) many statistical and process-based models have been developed, but model predictions are inconsistent with observational results largely dependent on the individual watershed characteristics and future discharge trends. Thus, it is still difficult to predict how future lateral C flux will respond to climate change, but changes in the DOC regimes in individual catchments can be predicted with a reasonable reliability. It is advised that sampling protocols and preservation and analysis methods should be standardized, and analytical techniques at molecular scales and numerical modeling on thermokarsting processes should be prioritized.


Introduction
Permafrost refers to ground that remains at or below 0°C for at least two consecutive years (Permafrost Subcommittee, 1988). Due to low-temperature conditions for hundreds to millions of years, organic materials in permafrost have been well preserved and shielded from microbial decomposition. As a result, substantial amount of organic C is stored in permafrost soils. About 1832 Pg C (1 Pg=10 15 g) is stored in permafrost regions in the northern hemisphere (Mu et al., 2015), about 2.5 times as much as that in the current atmosphere . Climate warming will induce permafrost degradation and alter the C budget of permafrost regions (e.g., permafrost C will be released into the atmosphere and oceans). Dissolved organic carbon (DOC) (organic C diameter <0.45 μm), constituting the main organic C form exporting from terrestrial C to soil solution, rivers, wetlands, and lakes (lateral C flux) and oceans (longitudinal C flux) Cole et al., 2007), will substantially shift the lateral and longitudinal organic C fluxes in permafrost regions. Recent evidence suggests that climatic warming does indeed influence the DOC dynamics of northern regions. However, the lateral and longitudinal C fluxes have been less studied than atmospheric C flux, probably because it is believed the latter has a more immediate positive feedback to climate warming processes. Meanwhile, DOC is an active participant in biogeochemistry, environmental chemistry, and freshwater ecology, and as a buffer for in-stream contamination (Hope et al., 1994;Wickland et al., 2018), and a major component of organic C in northern streams/rivers (Meybeck, 1981;Molot and Dillion, 1996;Neff and Asner, 2001;Prokushkin et al., 2011). In addition, the mobility of DOC in soil solution is important for pedolization processes (Neff and Asner, 2001). Thus, DOC in permafrost regions is of importance in both boreal aquatic chemistry and pedolization processes.
In northern (arctic and subarctic/boreal) permafrost regions, C storage is far more abundant (~1300 Pg) (Hugelius et al., 2014) compared to that in elevational permafrost regions at mid-latitudes (~160±87 Pg), which therefore has received less attention (Mu et al., 2015). However, elevational permafrost regions could be more important in the short term because of their extensive areas of discontinuous permafrost sensitive to climate warming (Jin et al., 2011). Thus, more attention needs to be paid to permafrost regions at low-and mid-latitudes as well.
This paper reviews recent studies on DOC dynamics in northern and elevational permafrost regions focusing on two key processes: terrestrial production of DOC, and subsurface transfer processes of DOC. An adequate understanding of these two processes benefits the correct representation of these processes in models to simulate and predict future riverine DOC dynamics. Specific research objectives of this paper are: (1) identification of sources for aquatic DOC, especially DOC in streams and rivers; (2) understanding of DOC transfer mechanisms in subsurface and the identification of key controlling factors for aquatic DOC export, and; (3) aquatic DOC modelling and prediction under a warming climate.
These objectives are reached by a discussion of the relevant research in the following sections. The relationship between research objectives and contents in each section of this paper are illustrated in Figure 1.

Terrestrial organic C and DOC pools in permafrost regions
Terrestrial plants and soils are the main sources for DOC in aquatic systems. Knowing the size of terrestrial DOC pool in permafrost regions will help further studies on potential future DOC fluxes from aquatic ecosystems. This section reviews recent studies on terrestrial DOC pools in permafrost regions.
Precipitation directly contributes a certain amount of DOC input to the terrestrial C flux, but only marginally (~1 mg C L -1 ) (Thurman, 1985). Vegetation releases organic C actively through litterfall and root exudation, whilst throughfall and stemflow following precipitation result in the leaching of C from leaves and stems. The latter process was quantified by Likens (2013) who showed that the content of organic C in water samples collected underneath the canopy (throughfall and stemflow) (12 mg C L -1 ) was much higher than that collected above the canopy (2.4 mg C L -1 ). Prokushkin et al. (2001) reported an annual throughfall DOC export is 0.1-0.2 g C m -2 in a central Siberia larch forest. Plant residues are also a major source of soil organic matter (SOM) after a Figure 1 Relationships between research objectives and contents in each section and paper structure.
long period of decomposition. Finlay et al. (2006) observed a dominant DOC flux in a snow-melt season, and attributed it to the accumulation of litterfall over the last autumn. Some scholars even refer net primary productivity (NPP) as an indicator for fluvial DOC export (e.g., Hope et al., 1994;Tokareva et al., 2006). About 1-5% of NPP is estimated as yearly watershed DOC loss. Northern permafrost regions have a large C stock. Surface (top 0-3 m) C is estimated as 1035±150 Pg C, with 472±27 Pg stored at depths of 0-1 m, 335±81 Pg C at 1-2 m and 207 ±42 Pg C at 2-3 m (Hugelius et al., 2014). Deeper (>3 m) C in permafrost regions is mainly found in yedoma (ice complexes) regions in Siberia and Alaska (210±70-456±45 Pg C), major Arctic river deltas (91±39 Pg C) and permafrost zones with thick loose sedimentary materials (350-465 Pg C) (Schuur et al., 2013). Compared with large C pools in northern permafrost regions, elevational permafrost regions at mid-and low-latitudes have less C accumulation, but the permafrost C is more sensitive and undergoing rapid and intensive degradation (10 cm yr -1 ; 1.4×10 4 km 2 yr -1 ) (Jin et al., 2011). The Qinghai-Tibet Plateau has the largest expanse of elevational permafrost, with an estimated C storage at 160 ±87 Pg C, 80% of which is stored in deeper (3-25 m) permafrost (e.g., Mu et al., 2015). DOC in soil solutions is closely related to C storage in soil layers (Prokushkin et al., 2010). Organic layer of topsoil is a major source of terrestrial DOC, which is estimated as 24 g C m -2 (Prokushkin et al., 2010). Soil organic C in Circum-Arctic permafrost regions is 472±27 Pg C (0-1 m in depth) (Hugelius et al., 2014), with an estimate of 1% as DOC in organic soil layer on permafrost terrains (Prokushkin et al., 2009), and~4.72 Pg C of DOC in shallow soils at 0-1 m in depth. This does not include the organic C in deeper (>1 m) permafrost soils in yedoma. In addition, this soil DOC pool is continuously replenished by senescence and decomposition of terrestrial biomass. It is supposed to have large potentials for being transported to aquatic ecosystems and is labile to rapid decomposition. DOC in soil solutions is estimated at~0.3 Pg C (Prokushkin et al., 2009;Hugelius et al., 2014). Thus, there is less than 1‰ organic soil C fluxed into soil solutions. DOC concentration in soil solutions ranges from 2-30 mg C L -1 (White, 2013). DOC concentration in soils declines with depth (Prokushkin et al., 2009). Deep permafrost soils also contain a large amount of DOC. Michaelson et al. (1998) measured meltwater from permafrost soils, sampled in situ and analyzed in laboratory, with DOC concentration as high as 116 mg C L -1 . Ewing et al. (2015) also reported high DOC concentration in segregated ice thaw water (48-1548 mg C L -1 ). These results suggest that thawing permafrost may release substantial DOC. Large amount of massive ice in permafrost regions, and ice wedges in particular, also have great potential for DOC storage and release. Fritz et al. (2015) reported~45.2 Tg C (1 Tg= 10 12 g) DOC stored in ice wedges in arctic yedoma permafrost regions, with an average DOC content at 9.6 mg C L −1 (maximum at 28.6 mg C L −1 ). Tanski et al. (2016) estimated DOC concentrations at 0.3-374 mg C L −1 in arctic coastal permafrost regions. The presence of ice wedges considerably facilitates the DOC loss in permafrost regions (Vonk et al., 2013a). To conclude, in permafrost regions, main terrestrial DOC sources are vegetation, surface organic layer and deep permafrost soils.

Chemical compositions of natural and anthropogenic DOC
DOC has various sources of organic C, but mostly consists of organic acids (Aiken et al., 1992). Simple organic constituents, such as fatty and animo acids, hydrocarbons, carbohydrates, and tannins, account for 10-20% of DOC in aquatic systems. Complex organic categories include humic acids (~10%), fulvic acids (~40%), hydrophilic acids and clay-humic-mineral complex (Thurman, 1985).
In different water bodies, including soil pore water, lakes/ ponds, streams/rivers, and oceans, DOC compositions differ at the molecular scale. Lignin compounds are exclusively derived from terrestrial plant and highly chemically-stable phenolic polymers. Amon et al. (2012) studied seasonal characteristics of lignin-phenol as biochemical tracers in the six largest Arctic rivers, concluding that lignin-phenols were good indicators for a fresh plant DOC source and more than 75% lignin compounds were released in spring. Plant residues can produce large portion of labile organic matter. DOC leached from litter layer is composed of carbohydrates (~50%) and organic acids, similar to fulvic acids found in water. As the water percolates downwards, the amount and compositions of DOC would undergo significant modifications (Hope et al., 1994;O'Donnell et al., 2016;Wickland et al., 2018). Carbohydrates in soil solution are rapidly decomposed by microorganisms. Therefore, DOC in soil water, groundwater and rivers has lower contents of carbohydrates and more those of carboxylic and phenolic acids. Permafrost soils also produce a large amount of labile organic C after thawing. Work by Ewing et al. (2015) and Drake et al. (2015) show that a large portion of this DOC is comprised of lowmolecular weight organic acids (e.g., acetate) from permafrost thaw water. Selvam et al. (2017) contrasted DOC compositions of soils in both permafrost and active layer, and found that DOC from thawed permafrost soil had a higher proportion of low-molecular weight.
DOC contents in snowfall and rainfall are low (0.1-5 mg C L -1 ) (Moore, 1997), in comparison with leachates in the soil organic layer and thawed ice from organic-rich permafrost soils. DOC in precipitation mainly originates from leaching dusts and organic debris in the atmosphere, incomplete burning of C-based fuels and biomass by wildfires. During the last decades, intensifying anthropogenic activities have led to increasing usage of fossil fuels and diesel engines, and climate warming promotes increasingly more frequent wild and anthropogenic fires (Bond et al., 2013). Atmospheric particulate matter, or black/brown carbon (soot), appears as aerosols more frequently. These incomplete burning products of carbon-based materials are refractory and structurally poly-aromatic (Justi et al., 2017). It has been found that black carbon experiences chemical alterations after exposing into the atmosphere, but the extent of these alterations awaits further studies (Druffel, 2004). Aromatic and fatty acids are also found in precipitation, which largely originate from plant emissions into the atmosphere.
Because of remote locations at high-latitude and elevational permafrost regions, anthropogenic DOC of agricultural, domestic and industrial activities, which mainly include hydrocarbons and polycyclic aromatic hydrocarbons (PAHs), is to a lesser influence. Guo et al. (2014) contrasted streams derived from natural and agricultural systems in permafrost catchments and addressed alteration in DOC chemical compositions. Particulate organic carbon (POC), under certain circumstances, will mutually transform with DOC in soil and water (Hope et al., 1994). The decomposition of POC, atmospheric organic matter deposition, activation of riverine sediments and anthropogenic activities also contribute appreciably to DOC.

Biodegradability of DOC in permafrost regions
Biodegradability of DOC in permafrost regions largely depends on chemical compositions of DOC (Abbott et al., 2014). Simple organic compounds, such as amino acids, carbohydrates and fatty acids, are susceptible to rapid decomposition. Complex parts of DOC, such as humic substance, take a much longer time to degrade.
The ultimate source of soil organic C is terrestrial biomass. Michalzik et al. (2003) observed that most of leachates of plants and litterfall were composed of low-weight molecules and highly labile to microbial decomposition. Hope et al. (1994) found that DOC in groundwater (~0.7 mg C L -1 ) was old, hydrophilic and recalcitrant to decompose. Thus, DOC in groundwater may largely be derived from the stabilized organic matter in soil. Surface soil is supposed to be organic-rich and contains large amount of new C derived from leaching vegetation and decomposition of plant and animal residues and with high bio-and photo-lability (Michaelson et al., 1998;. However, an inconsistent conclusion was drawn (Prokushkin et al., 2007;Neff et al., 2006). Substantial aromatization of DOC was observed during spring freshet in a permafrost watershed, suggesting a weak microbial transformation. Lignins exclusively originated from vascular plants indicate substances from vegetation can be refractory to decompose.
Highly biodegradable DOC from permafrost soils is also reported in natural systems (e.g., Michaelson et al., 1998;Vonk et al., 2013aVonk et al., , 2013bAbbott et al., 2014;Ewing et al., 2015;Drake et al., 2015;Mann et al., 2015Selvam et al., 2017). The biolabilility of DOC in permafrost soils range from 24±1% to 71%, with a median of 52%. However, other researchers find no apparent differences in DOC lability and alterations in DOC concentration in streams, either disturbed by thermokarsting or not (e.g., Larouche et al., 2015). Stubbins et al. (2017) observed a low photo-lability of old C from permafrost, in comparison with a high photo-decomposability of new C from streams drained from thawing yedoma permafrost soils. Most laboratory experiments have confirmed the high biodegradability and large amount of DOC in permafrost soils (Michaelson et al., 1998;Vonk et al., 2013a;Ewing et al., 2015;Drake et al., 2015;Mann et al., 2015;Spencer et al., 2015;Selvam et al., 2017), or ice wedges (Vonk et al., 2013b;Fritz et al., 2015), through direct measurements of meltwater from the frozen soil core sampled in situ. Some studies report no apparent alterations of biodegradability through measuring DOC compositions in rivers and streams. This contradiction is potentially caused by the passing of labile permafrost DOC through deeper mineral soils, consuming most labile DOC, and delivery of a residual part of DOC into rivers and streams.
However, biodegradability of DOC is not only determined by autochthonous characteristics, but also it is influenced by external environmental drivers, such as sunlight (Cory et al., 2014;Ward and Cory, 2016), temperature, flow paths (Striegl et al., 2005) and soil C/N ratio (Qu et al., 2017), which affect photo-lability, bio-lability and microbial decomposition, respectively. Thus, the effective biodegradability of DOC is a combination of autochthonous properties and the external drivers.

DOC transport into streams and controlling factors for DOC fluxes
In permafrost regions, shallow soils in the active layer (overlaying the impermeable permafrost, above the permafrost table), with a potentially high hydraulic conductivity, low mineral contents, and low DOC sorption capacity, lead to rapid and lateral DOC transport to streams and rivers (Prokushkin et al., 2009). During the shifting of winter to summer seasons, the seasonal thawing of the active layer in permafrost regions and new inputs of DOC will derive from the thawed active layer and/or vegetation changes (Sturm et al., 2001;Neff et al., 2006). In discontinuous permafrost regions, further infiltration of soil solutions occurs in open taliks (an unfrozen ground body that penetrates the permafrost completely, connecting supra-and sub-permafrost water (Permafrost Subcommittee, 1988)), elongating the groundwater pathways into streams and rivers (Bense et al., 2012). Thus, it increases the frequency of DOC sorption and microbial decomposition in deeper mineral soils. This will keep more refractory DOC in deeper soils and groundwater systems (Meybeck, 1981). Climate warming deepens the active layer, results in the penetration of closed taliks and expansion of open taliks. Thus, it causes incessant permafrost C emissions from continuous permafrost regions, but enhanced infiltration and possibly reduced DOC emissions from discontinuous permafrost regions.
Storms preferably generate rapid surface runoffs in permafrost regions, often occurring through the upper soil horizons of 15-20 cm in depth (Hope et al., 1994;Bishop et al., 1993). This will increase DOC exports from top organic layer and vegetation without further infiltration. Thermokarsting can influence DOC flux by altering surface hydrological processes, especially when occurring in the riparian zone (Larouche et al., 2015).
From rainfall leaching of living plants and soil column to riverine export of DOC, many environmental factors regulate this process. These impacting factors can be grouped into three categories: controlling origins, transfer pathways and human activities (Table 1).
Organic-rich soils are more prolific in DOC production. The ratio of C/N is indicative for the abundance of SOM (Aitkenhead and McDowell, 2000). Afforestation and deforestation can change DOC flux through increasing and decreasing the leached DOC concentration, respectively. Atmospheric CO 2 concentration is also a regulator of DOC flux due to its positive correlation to vegetation growth (Freeman et al., 2004). Microbial activity can be influenced by soil temperature and humidity (Yurova et al., 2008). Deposition of atmospheric organic acids can directly increase DOC concentration in rivers, but to a lesser extent (Hope et al., 1994). DOC concentration in streams declines after wildfires (Petrone et al., 2007;Betts and Jones Jr, 2009). However, there are also studies reporting an increased DOC flux into streams after wildfires, possibly due to postfire thawing of permafrost (Schindler et al., 1997). Thawing of organic-carbon-rich permafrost will release a certain amount of DOC into aquatic systems.
Watershed geomorphology and permafrost occurrence affect DOC fluxes by modifying hydrological processes. Discharge and precipitation regulate DOC fluxes into streams through increasing the frequency and intensity of leaching in organic soil (Judd and Kling, 2002). Tate and Meyer (1983) found that prolonged spells of heavy rainfall and high runoffs would reduce DOC concentrations in streams from forested watersheds. Thus, DOC fluxes are not only determined by flushing influences, but also largely depend on former storage of terrestrial DOC.
Anthropogenic influences on surface waters have now greatly increased natural nutrient levels. Guo et al. (2014) and Sun et al. (2017) suggested agricultural activities would reduce riverine DOC export, which was potentially caused by landscape types and irrigation. Laudon et al. (2009) and Hausmann and Pienitz (2009) indicated deforestation significantly increase DOC, such as chlorophyll, concentration in streams, especially in summertime, which would result from the effect of deforestation on surface hydrology and exposed plant tissue.
In summary, riverine DOC export is a function of terrestrial DOC production and leaching patterns. Soil properties (such as C/N ratio), permafrost degradation, freeze-thaw cycles and proportion of peatlands or wetlands affect soil DOC abundance. Through controls of microbial activities, soil temperature and moisture and groundwater table affect DOC production. Vegetation, afforestation or deforestation, land-use changes, wildfires, atmospheric CO 2 level and elevation can directly and indirectly increase or decrease DOC sources from vegetation. Rainfall and discharge can mobilize DOC in soil. Hence, reservoir construction influences frequency and intensity of terrestrial DOC activation. Water-logged permafrost and geomorphology of individual catchment will produce variable flow-paths, largely affecting DOC export. Atmospheric deposition chemistry and contaminated water from industrial, domestic and agricultural activities modify riverine DOC export through aquatic chemistry.
Numerous studies in individual catchments in permafrost regions have been conducted for estimating DOC export, as those examples listed in Table 2. Individual watershed features (ecosystem types, locations, elevation, catchment size and discharge) are also listed to help understand controlling factors and their relative importance. The annual export of organic C in rivers and streams of studied catchments varies from 0.1-10 g C m −2 yr −1 and DOC concentration generally ranges from 1-60 mg C L −1 . Most annual DOC export ranges from 3-8 g C m −2 yr −1 and DOC concentration, 4-40 mg C L −1 . In Table 2, a suite of patterns can be observed. Peatlands/ wetlands and taiga landscapes are featured by organic-rich soils and abundance in plant production, which are key factors in terrestrial DOC production. Thus, they generally are flushed with higher riverine DOC concentrations (4.7-60 mg C L -1 ) than tundra and shrublands (0.8-13.38 mg C L -1 ). Therefore, landscape types can be an important regulator of DOC production (Harms and Ludwig, 2016;O'Donnell et al., 2016).
However, when specific discharge (discharge divided by catchment area) is high (710 mm; 809 mm), DOC concentration is diluted to a lower level (4.91-7.53; 2±0.2 mg C L -1 ) (Ford et al., 1990;Lyon et al., 2010;Prokushkin et al., 2011). Annual DOC export is determined by both DOC concentration and runoff. Upon large DOC concentrations (6.2-29.7 mg C L -1 ) and a specific discharge (317 mm), DOC export can be as large as 8.1 g C m −2 yr −1 (Olefeldt and Roulet, 2012). DOC concentrations demonstrate no substantial variations in rivers (Meybeck, 1981;Hope et al., 1994) and streams with a variability of 9 orders of magnitude in catchment area (0.1×10 8 km 2 to 3.4×10 8 km 2 ). This indicates that catchment size may not be a major controlling factor for DOC export. However, the areal extent of peatlands/wetlands and taiga in a catchment may greatly affect DOC concentration. Brooks et al. (1999) and Stubbins et al. (2017) estimated low riverine DOC concentrations exported from taiga and peatlands/ wetlands landscapes. This is potentially caused by smaller areal extent of these kinds of landscapes. Catchment locations and elevation can also indirectly influence riverine DOC concentration. High arctic and alpine regions are very cold and outreach the tree line. There hardly are macrophytes, such as forests. Tundra landscapes are characterized by low productivity in biomass and inclusions of lichens, bryophytes and dwarf shrubs that can survive in very cold environments. Alpine regions, featured by less vegetation and rocky terrains with thin soil layers, are unfavorable for terrestrial DOC production.

Transfer pathways
Hydrological flowpaths White et al., 2008;McClelland, 2009 Geomorphology Meybeck, 1981;Moore et al., 1998;Battin, 1999;Thompson et al., 2015 Permafrost occurrence De March, 1975;Petrone et al., 2006Petrone et al., , 2007Prokushkin et al., 2007Prokushkin et al., , 2009 Rainfall De March, 1975 a) Values in parentheses are estimated from authors' data. Arctic regions refer to areas north of the Arctic Circle (66°33′N). Subarctic regions, generally between 50°-70°N, occasionally to lower northern latitudes. Alpine regions refer to highlands above the tree line. Plateau is a flat terrain raising significantly above surrounding areas. Mountains refers to areas with elevation of hundreds of meters relatively young (δ 14 C>100‰) Raymond et al., 2007). As the active layer deepens in mid-to late spring and summer, DOC flux shows older 14 C age (δ 14 C<0‰) , and downward trends (from 33.6 to 6 mg L -1 ) of DOC fluxes are observed in late summer (Guo and MacDonald, 2006). This indicates a longer pathway and retention time in the transport process when mineral soils are exposed for seasonal thawing.

Modelling and forecasting DOC flux in areas of degrading permafrost
Many model predictions indicate that DOC fluxes are important under a warming climate and subsequent degrading permafrost. Greater climate warming is projected for high latitudes and elevations (e.g., Parker et al., 2008), where permafrost contains a large C reservoir (~1832 Pg) and is vulnerable to climate warming (Schuur et al., 2013;Mu et al., 2015). Concerns are raised as to how much permafrost C will be released as DOC, which is important for global C cycling (Meybeck, 1981;Freeman et al., 2004). Quantification of DOC dynamics can be achieved by process-based modelling or regression analysis. Regression analysis employs the correlation of DOC flux and environmental factors, such as discharge, width of water body, water pH, proportion of upland/wetland area in a watershed, and soil C/N ratio (Aitkenhead and McDowell, 2000;Jutras et al., 2011;Guo et al., 2018;Rodríguez-Jeangros, 2018). Then, fluvial DOC flux can be predicted by regression models.
Process-based models generally associate hydrological models with DOC modules. Grieve (1991) coupled the Birkenes Model (Christophersen et al., 1984), with a DOC submodel, from which DOC concentration and soil respiration rate were deduced. Physical process-based simulations are generally focused on the upper 20 cm of surface organic layer. Boyer et al. (2000) improved this DOC module and coupled with TOPMODEL and a hydrological model. Based on the CENTURY soil C model, Neff and Asner (2001) synthesized the key controllers for DOC production and consumption in terrestrial ecosystems and developed a model by involving these key controllers. Michalzik et al. (2003) developed the DyDOC model dealing with soil DOC dynamics by incorporating hydrological and sorption processes and metabolism, as well as 14 C simulation. Worrall and Burt (2005) predicted future DOC flux from a peatland catchment by inputting air temperature and precipitation data into models to deduce DOC production rate and water table.
The INCA-C, an integrated model based on the INCA-N, which can model both terrestrial and in-stream DOC dynamics, has successfully simulated nitrogen dynamics in a number of European rivers (Futter et al., 2007).
However, most above-mentioned DOC Models have not taken into account of permafrost, which is of great influences on hydrological processes and DOC flow paths in the north and on uplands. Thus, Yurova et al. (2008) simulated fluvial DOC in boreal regions underlain by discontinuous permafrost, by using convection-dispersion equations and by coupling with hydrothermal models. Kicklighter et al. (2013) used a modified terrestrial ecosystem model (TEM) by Felzer et al. (2004) to simulate terrestrial DOC loading and build association with riverine discharge and DOC export. By taking into account of the effects of water blogging of permafrost, slope aspects and seasonal frost on hydrological flowpaths, a biogeochemistry model has been proposed on the base of the ECOSSE soil C model (Smith et al., 2007). Lessels et al. (2015) coupled a modified HBV hydrology model for simulating stream discharge and DOC dynamics. Liao (2017) employed a process-based three-dimensional model (ECO3D) to investigate water and DOC dynamics in a catchment of discontinuous permafrost (39%) in interior Alaska and predict water and DOC cycles. All models mentioned above deal with the release of C from the active layer. However, the C release from permafrost is another main DOC source. There are no existing models for simulating the lateral flow of deep groundwater C release. Many recent studies for modelling and forecasting DOC fluxes demonstrate either increasing or decreasing trends upon permafrost degradation under future climate change scenarios: warm-dry and warm-wet, respectively (e.g., Moore et al., 1998;Frey and Smith, 2005;Walvoord and Striegl, 2007;Prokushkin et al., 2009;Tank et al., 2016). Elevated temperature will increase NPP, which in turn will enlarge DOC input into soil profiles (Freeman et al., 2004). A new source of DOC from older and deeper layers will be introduced by thawing permafrost (Prokushkin et al., 2009). Increased runoff will mobilize DOC export (Moore, 1997). This is much related to elevated precipitation. However, evaporation also will elevate under a warming climate. Thus, uncertainties in the future climate influence DOC flux patterns. In addition to climatic scenarios, DOC export can also be affected by catchment substrate characteristics and morphology through DOC transfer processes. Permafrost degradation and deepening active layer may increase DOC retention in deeper mineral soils.
With so many uncertainties mentioned above, some important processes for DOC dynamics on permafrost terrains are still missing from current models. Under a warming climate, the thermokarst-inducing ground ice is DOC-rich (Fritz et al., 2015) and subsequent meltwater can mobilize soil C into aquatic systems. Shirokova et al. (2013) found that with shrinking thermokarst lakes, DOC concentration and storage would increase progressively. Because DOC export is constrained by multiple controllers, and so many uncertainties still remain, making prediction of future DOC fluxes a daunting task.

Conclusions
This section summarizes recent advances in the studies on DOC in permafrost regions, points out inadequacies, and recommends on potential future work ( Figure 2).

Summary
Through comprehensive reviewing current research on aquatic DOC in elevational and northern permafrost regions, following major aspects are summarized from terrestrial DOC sources, transferring processes and impacting factors, modelling and prediction of DOC and research methods: (1) In permafrost regions, terrestrial organic C is an important source for aquatic DOC. It is mainly contributed by terrestrial vegetation and organic soil. Permafrost soil is also an important C source for aquatic ecosystems under a warming climate. Varied DOC sources have different chemical compositions. DOC from vegetation contains more carbohydrates (~50%), and leachates of permafrost soil and vegetation, with more small-molecules, are more biodegradable. However, DOC from soil and groundwater, consisting of mainly complicated compounds, such as PAHs and hydrocarbons, is more recalcitrant.
(2) Fluvial DOC export is regulated by multiple controllers of subsurface processes from DOC production to transfer processes ( Figure 3). DOC production is controlled by vegetation coverage and types, and soil properties (such as C/N ratio) and organic contents, and organic content and biodegradability of permafrost soil. Subsurface transfer processes are first influenced, with a flushing effect, by precipitation and discharge. Then, interstitial water movement in soil will undergo a series of physical and biochemical processes, such as sorption/desorption and microbial mineralization. Transfer pathways largely influence fluvial DOC export. In regions with extensive permafrost degradation, the active layer deepens, talik expands and areal extent of permafrost declines. More soil solutions reaching the permafrost table will infiltrate by the talik into the intra-and sub-permafrost waters, rather than flow as the supra-permafrost water along the permafrost table and later directly into streams. This will largely boost the chance of microbial mineralization and sorption in deeper mineral soils. Atmospheric chemical deposition will also directly affect fluvial and lacustrine DOC dynamics. Although DOC flux is controlled by multiple factors, DOC concentration and export have displayed only a small range of variations.
(3) Generally, modelling on aquatic DOC export in permafrost regions is realized by statistical or processed-based models. A statistical model is built on the basis of relationships between controlling/impacting factors and fluvial DOC fluxes. The process-based models are built on the subsurface DOC transfer mechanisms. These models can be used for predicting future DOC fluxes by using drivers, such as either climate change scenarios, or projected changes in controlling/impacting factors.
(4) Current research methods in DOC are manifolds. At micro-scales, DOC quantity, chemical characteristics and composition identification by physical (combustion oxidation, leaching experiment and non-ionic XAD) and bio- chemical methods (incubation experiment). Geochemical (δ 13 C and δ 14 C) or bio-(carbohydrates, lignins or other substances) chemical (Mass spectroscopy, gas chromatography, 13 C-NMR, specific ultraviolet absorbance and fluorescence index) tracers are powerful tools in examining DOC transferring processes. At macro-scales, GIS/RS techniques are employed to explore the relation between DOC and spatial patterns of certain controlling factors (permafrost occurrence and changes, vegetation, soil temperature and humidity and wetlands/peatlands coverage). Recently, RS techniques are introduced to monitor stream/ lakes DOC concentration by means of DOC absorbance of ultraviolet and visible light. Monitoring in situ is also necessary for aquatic DOC research.

Inadequacies
(1) There are already fruitful studies in aquatic DOC in permafrost regions. However, some conclusions are disputed in several aspects.
(i) NPP is generally regarded as an important DOC source in soil. Some researchers employed NPP as an indicator for riverine DOC export rates. Although others have found that when NPP varies in 5 orders of magnitude in different catchments, DOC export rate changes insignificantly, basically in the same order of magnitude. Some researchers noted that the soil C/N ratio was the major controlling factor for aquatic DOC flux.
(ii) Many studies have concluded that permafrost de-gradation would impact on fluvial DOC export. A particular research has measured 14 C age of DOC and POC in four rivers in permafrost regions, yielding a much younger age of DOC (390 to 1140 yr BP) than POC (4430 to 7970 yr BP) in three of the four studied basins, which suggest riverine DOC in permafrost regions are mostly new C from soil organic layer and vegetation.
(iii) Meanwhile, research conclusions on DOC flux in catchments with degrading permafrost are generally inconsistent in many regards.
The divergences among different studies are potentially resulted from that lateral DOC flux is not merely decided by single factors. Meanwhile, subsurface DOC transferring can be rather complicated. In measurements, even though there are standard methods for sample collection and preservation, most studies still prefer established methods. These may collectively lead to poor inter-comparability between research results.
(2) Development of widely adoptable larger-scale models is hampered due to inadequate understanding of the mechanistic linkage of DOC transfer processes in describing the pathway from source towards rivers/streams. Furthermore, such development is hampered by the perception of a too large complexity in controlling parameters which makes inclusion into model daunting. Most current models for simulating DOC flux focus on the C release from the active layer. This will largely underestimate the effect of permafrost soil as a main DOC source for riverine DOC export. Shrinkage of permafrost bodies will laterally release DOC  Figure 3 are collected data from this paper (Thurman, 1985;Hope et al., 1994;Moore, 1997;Mladenov et al., 2008;Prokushkin et al., 2010;Laudon et al., 2011;Likens, 2013;Zhu et al., 2014;Ewing et al., 2015;Mu et al., 2016). Red "+" refers to major DOC sources, while yellow arrow refers to DOC flow. into groundwater. Therefore, the modeling of deeper groundwater and its impact on C-release also needs to be conducted urgently.
(3) Due to the complicated procedures for isolating and identifying DOC chemical compositions, chemical compositions of DOC are not systemically studied. More recent research inclines on quantifying a group of compositions according to specific physical parameters. Although it is convenient, it is not accurate enough to trace DOC transfer mechanisms. However, some key techniques deem necessary for understanding the subsurface transfer processes.

Prospects
It is suggested that the inconsistency in sampling protocols, preservation and analysis methods should be resolved for better comparing research results from individual studies. More reliable analytical techniques should be employed to distinguish in-stream DOC sources. Geochemical and biochemical tracers are efficient tools in studying subsurface DOC transfer mechanisms. Identification of chemical compositions of DOC in molecular level is of great utility in tracing subsurface DOC transfer. There is extensive space for improving the techniques for identifying DOC at a molecular level, and molecular identification should develop towards easier and more convenient methodology.
The widespread and intensifying thermokarst processes are expected to influence more on surface hydrological processes and to release a large amount of DOC from OCrich ground ice, such as ice wedges. Further research on the mechanisms for thermokarsting deems necessary to provide reliable understanding for simulating DOC dynamics in permafrost regions. Current DOC prediction models may have underestimated DOC export into aquatic systems and thus they should formulate thermokarst processes into simulation models for a better accuracy. DOC modeling in groundwater is also necessary due to its better reflection of lateral DOC flux under a warming climate and degrading permafrost. In-stream processes of DOC are key to longitudinal DOC flux, which is less of a focus within this paper, but of importance to regional and the global C budget.