Quantitative understanding of the variation in dissolved organic carbon (DOC) is important to studying the terrestrial ecosystem carbon cycle. This study presents a process-based, dissolved organic carbon dynamics model (DOCDM 1.0) that couples the soil heat conduction, water flow, DOC production, mineralization and transport in both surface and subsurface of soil profile to quantify DOC dynamics in boreal terrestrial ecosystems. The model is first evaluated and then applied for a watershed in Alaska to investigate its DOC production and transport. We find that 42 and 27 % of precipitation infiltrates to soils in 2004, a warmer year, and in 1976, a colder year, respectively. Under warming conditions, DOC transported via overland flow does not show the expected decrease trend while the overland DOC yield shows a 4 % increase. The horizontal subsurface flow only accounts for 1–2 % of total water flux, but transports 30–50 % of DOC into rivers. Water flush due to water infiltration controls DOC transport. Snowmelt plays a noticeable role in DOC flush-out and DOC transport significantly depends on flowpaths in the study region. High soil temperature stimulates DOC production. The overland DOC export does not necessarily follow the DOC downward trend in surface water transport. Overall, this study shows that DOC export behavior is complex under changing temperature and hydrological conditions in cold-region watersheds. To adequately quantify DOC dynamics in northern high latitudes, more DOC and hydrological data are needed to better parameterize and test the developed model before extrapolating it to the region.
Dynamics of dissolved organic carbon (DOC) are a critical, yet often neglected component of the terrestrial carbon cycle. The significance of DOC has been stressed by previous studies in many aspects, including its influence on nutrient cycling (Qualls and Haines, 1991; Michalzik and Matzner, 1999), its service as a microbial energy source (Baker et al., 2000; Raymond and Bauer, 2000) and its effects on pollutant transport (Morris and Hargreaves, 1997). High latitude ecosystems, which store about one-third of the global terrestrial organic carbon (Gorham, 1991; Moore, 2002), play a vital role in determining the future terrestrial carbon cycle including DOC dynamics, under rapidly changing climatic conditions. Dramatic changes have occurred in the arctic cryosphere, biosphere, and atmosphere (Hinzman et al., 2005; Serreze et al., 2000), resulting in higher DOC concentrations in the Arctic rivers in comparison with other major river basins on the globe (Lobbes et al., 2000; Raymond et al., 2007), and the DOC flux is expected to increase in the future (Freeman et al., 2001; Tranvik and Jansson, 2002).
While the important role of DOC in regulating C transport from
terrestrial ecosystems to river systems is acknowledged, the
production, loss, stabilization and release of DOC, and the
interaction of these processes with external environmental variables,
are still not well understood. Sorption and desorption processes are
widely believed to be the dominant controlling mechanisms for DOC
transport through solid soil matrices (Quails and Haines,
1992a, b). Soluble DOC may be lost in sorption process or as a result
of microbial mineralization, and DOC sorbed to the soil column may
rejoin water during desorption process or transformed into
The DOC export from northern high latitude ecosystems has more variations due to spring snowmelt, increasing temperature, and associated permafrost thawing (Ågren et al., 2010). A large portion of the annual DOC delivery into high northern latitude rivers occurs during the spring flush (Dittmar and Kattner, 2003; Raymond et al., 2007; Holmes et al., 2008), and the main DOC source during this relatively short period is surface litter and organic matter from shallow soil layers. Meanwhile, a warmer and wetter condition is thought to stimulate microbial activity, potentially increasing DOC production (Moore et al., 1998). In contrast, Striegl et al. (2005) suggested that the thawing of frozen soils increases the thickness of the active layer and promotes more flow pathways below organic-rich soil layers, which in turn leads to less terrestrial DOC yield under warming trends.
To investigate the controlling factors for DOC export, several DOC models have been developed (e.g., Neff and Asner, 2001; Michalzik et al., 2003; Fan et al., 2010; Xu and Saiers, 2010; Mei et al., 2012). However, most of them are one-dimensional (1-D) models that are only applied to soil columns under laboratory environment, not field conditions. The 2-dimesional (2-D) overland and subsurface transport processes, which are important in DOC transport on a watershed-scale, are usually not considered. The DOC models (Boyer et al., 1996, 2000; Futter et al., 2007; Yurova et al., 2008), which are incorporated with more hydrological and biological processes, have been successfully applied at catchment scales. However none of them has fully addressed the effects of soil thermal regimes on DOC dynamics, thus may not be suitable for northern high latitudes. The aim of this study is to develop a DOC model that is applicable for arctic and subarctic watersheds. The model is then used to study the effects of climatic changes on the DOC dynamics in an Alaskan watershed.
The developed DOC modeling system fully considers the effects of frozen-soil thawing on DOC variation and explicitly treats 1-D and 2-D transport behaviors of DOC. It contains four modules (Fig. 1) including: (1) land surface processes, (2) soil heat conduction, (3) water infiltration; and (4) vegetation and DOC dynamics. The land surface module calculates the water and energy balance at the ground surface at an hourly time step. It provides ground surface temperature, throughfall and snowmelt rates, which are needed in the soil heat conduction and water infiltration modules. The soil heat and water infiltration module then estimates the soil temperature profile, soil ice content and water infiltration rate by solving soil thermal and moisture equations. Finally, the DOC production, sorption and desorption, mineralization and transport processes are modeled in the DOC dynamic module. Below we describe each module (also see the detailed information in Supplement A).
The land surface process module (Fig. 1) integrates hydrological and energy processes to simulate overland and channel flow dynamics, which drive the DOC horizontal transport. Precipitation is partitioned to rain and snow according to air temperature. Canopy layer interception is a function of leaf area index (Dickinson et al., 1986). Snow accumulation and melt are simulated when there is snow on the ground and snowfall. Snowpack is treated as a two-layer medium, and its accumulation and ablation is estimated by solving mass and energy balance equations (Andreadis et al., 2009). The energy exchange on the snowpack is modeled based on the net radiation, sensible heat carried by convection, heat advected by rainfall, evaporation, sublimation, condensation and latent heat loss or gain due to melting and refreezing. Processes such as snow accumulation and ablation, dynamics in the snow water equivalent, and meltwater yield are represented in the mass balance equation. In the snow model, snow interception (Storck et al., 2002), atmospheric stability (Anderson, 1976; Tarboton et al., 1995) and blowing snow (Bowling et al., 2004) processes are considered. Total evapotranspiration is based on the Penman–Monteith equation (Monteith, 1990) and has three components: (1) evaporation from bare soil, (2) evaporation from canopy and (3) transpiration from canopy. The vegetation module (green section of Fig. 1) provides radiation and wind speed attenuation, architectural and stomatal resistance, and roughness length to estimate snow and rain interception, evapotranspiration and root uptakes. It also provides the vegetation type information required in the DOC production estimation. The equations within the two modules are adopted from a large-scale hydrology model, the variable infiltration capacity (VIC) model (Liang et al., 1994).
Rainfall could contribute to surface or subsurface hydrological systems through infiltration, which controls the pathway of DOC transport. In arctic regions, the soil ice directly affects infiltration and heat conduction within the soil profile (Cherkauer and Lettenmaier, 1999, 2003; Cherkauer et al., 2003).
In cold regions, soil heat conduction is affected by liquid water and ice distribution; on the other hand, infiltration and water movement in soils is also influenced by changes of ice content. In this module, heat conduction and vertical water movement are resolved at a 90 s time step. When infiltration occurs, the time step is adjusted to 1 s (red section of Fig. 1). Here we made several assumptions: (1) ice is immovable and only liquid water can move in frozen soils, (2) the influences of vapor transport on water and heat conduction can be ignored, (3) liquid flows due to thermal gradients and heat conduction by convection can be neglected; and (4) unfrozen water and the subzero temperatures of frozen soils are included in the dynamic equilibrium. Under these assumptions, water movement and heat conduction only occurs in the vertical direction and can be described with one-dimensional equations.
The one-dimensional Richards equation (Richards, 1931) is used to
describe vertical water movement in both frozen and unfrozen soils:
By defining:
The presence of ice may significantly reduce water flow in the porous
medium. The scheme of Hansson et al. (2004) is used to address the ice
effect. The hydraulic conductivity for the liquid water portion of the
partially frozen soil,
The heat transfer in soils is modeled as:
A typical convection-dispersion equation is used to characterize the DOC transport (Patankar, 1980). Three processes are considered in the DOC transport: (1) one dimensional infiltration, (2) two-dimensional overland transport; and (3) two-dimensional saturated subsurface flow transport (Fig. 1). Since most of current DOC measurements were acquired for very large regions, it is difficult to get the DOC-related parameters for typical land cover types. In this study, we used the available data from previous studies to derive type-specific parameters for those types having no direct measurements (Supplement C). The DOC desorption and microbial production processes are treated as DOC sources, and sorption and mineralization processes are sink terms. Point processes including DOC production, mineralization, sorption and desorption processes as well as DOC infiltration and two-dimensional DOC transport processes are detailed below.
The equations in Yurova et al. (2008) are used to estimate the DOC
production rate (
Sorption and desorption are two key mechanisms regulating DOC
stabilization and release rates. Here soluble organic carbon includes
dissolved (DOC) and potentially soluble, but currently solid
(PDOC). PDOC can exist as suspended particulates or in soil surface
and sediment. The total concentration of DOC
(
There are three DOC or PDOC pools including DOC, PDOC1 and PDOC2. All
of them are subject to mineralization. The DOC and PDOC pools have the
following mineralization rates (
The vertical transport of DOC is modeled with the one-dimensional
convection-dispersion equation:
Overland flow can occur when the water depth on the ground surface exceeds the depression storage (Julien et al., 1995). In this study, overland flow is estimated in two dimensions using the continuity equation and the momentum equation with the diffusive wave approximation method. Channel flow is simulated in one dimension using the diffusive wave approximation method. The floodplain connection between overland cells and channels is also implemented. Water can move from the overland plane to the channel or vice versa according to the channel bank height and water surface elevation (Julien et al., 1995). Subsurface flow driven by the hydraulic gradient will enter the river when the water table is higher than water surface in channels.
The governing equations for overland flow routing are based on the
Saint–Venant equations of continuity and momentum. The full
Saint–Venant equations can be simplified by using the diffusive wave
approximation (Singh, 1996) which assumes that the net forces acting
along the given axis of interest are zero with the continuity
equation:
The channel flow routing process is modeled with one-dimensional
diffusive channel flow equation (Julien and Saghafian, 1991), which is
derived in a similar manner to its two-dimensional overland
counterpart with the exception that channel flow routing only happens
in a finite space established for a given channel section. The
one-dimensional continuity relationship can be expressed with the
following equation:
The quasi- three-dimensional saturated subsurface flow model was used in this study. Each simulation cell can exchange water with its four adjacent neighbors. Local hydraulic gradients are approximated by local surface slopes. Therefore, a given grid will receive water from its upslope neighbors and discharge to its downslope neighbors. The subsurface routing method was not used for cells that contain perched water tables.
Under non-isothermal conditions, the effect of the ice layer on the
subsurface flow should be considered. We used the soil profile of the
two neighboring cells in order to illustrate our method (Fig. 3). The
left column has one saturated zone whose thickness is H1, and its
adjacent pixels shown in the right column may have four possible zones
within the location of H1: top unsaturated layer (Z1), frozen
saturated layer (Z2), liquid saturated layer (Z3) and bottom ice layer
(Z4). One can deduce some relationships:
The rate (
In the overland transport, DOC might be transported with the flow to
a certain distance from its production sites but then re-accumulate at
this new point if overland flow is not persistent enough for the DOC
to reach a river channel. In surface overland DOC transport, we only
consider the advection process. For two-dimensional flow in the
overland plane, a continuity (conservation of mass) equation can be
written as:
The DOC transport in the subsurface system is assumed to only occur
with ground water movement, and the dispersion effect, explained
below, is not considered:
Soil parameters (Ds, Ws, b_infilt) controlling runoff, infiltration
and baseflow were adjusted to make the simulated hydrography match the
observation at the outlet of our test watershed (see the Sect. 2.6). The comparison was documented in Supplement F. Because DOC
concentrations are often measured for large river basins, it is
difficult to parameterize the model for specific ecosystem types based
on those measurements. Thus, we use the remotely-sensed DOC
concentration data to parameterize the DOCDM 1.0. First, following the
empirically-based algorithms in Griffin et al. (2011), the river DOC
concentrations are estimated by using Landsat5 Thematic Mapper data on
18 July 2003 and 8 May 2004, respectively. Second, we compare our
modeled DOC with the satellite-based estimates. Since the DOCDM 1.0
provides DOC channel concentrations on each hour, the model outputs
that are closest to the satellite passing time are used for
comparison. To spatially compare DOC concentrations from Landsat with
the model results, remotely-sensed DOC concentrations are averaged
into a 4
Although the remotely-sensed DOC concentrations are overall higher than those from the simulation, the two datasets have a good relationship (Fig. 4; Supplement F), suggesting that the DOCDM 1.0 can capture the general trend of the watershed-level DOC yield. We are aware of that many parameters in the DOCDM 1.0 may introduce uncertainties and remotely-sensed results are also suffered from cloud contamination, observing angel, and the parameters in their retrieve method. For example, the overestimates in streamflow may be one of the reasons that result in low DOC concentrations in our simulations.
To test the DOCDM 1.0, we choose a watershed in the Yukon River Basin
as a case study (Fig. 5), The watershed is 6043
The location of our study pixel is denoted as the yellow point in
Fig. 5. Its annual average air temperature for 2004 is
Soil temperature has increased (Fig. 7a) and active layer depth (ALD),
the front of no-ice zone in Fig. 7b, became deeper, from 150 to
200
The initial overland flow occurred on 8 June and 26 May in 1976 and 2004, respectively. The overland flow depths in three days after the initial overland events are shown (Fig. 8a and b). Overland flow depth was larger in 2004 than that in 1976. Also note that overland flow had a very shallow depth at nodes that contain slopes and channels. It suggests that temperature, precipitation and catchment topography are the main factors controlling temporal and spatial trend of overland flow. The warming trend makes snow melting earlier. Deeper snowpack due to more precipitation results in higher overland flow. Pixels have channels or steep slopes have higher routing capacity and thus lower overland flow depth.
The DOC concentrations in overland flow on those two days were presented in Fig. 8c and d. Although more DOC stored during the winter of 2004 (Fig. 7c), more overland flow in 2004 diluted its DOC concentrations. Analysis for DOC concentrations in overland flow, land cover, and DEM (Fig. 5) shows that the different land cover types contributed differently to DOC concentrations. Specifically, the higher concentrations in downstream forests area suggested that forestland might have released more DOC into overland flow. On the other hand, shrubs disturbed at the upstream part of the watershed yielded less DOC. Many lines of evidence show that the rate of DOC production correlates well with the organic matter content in soils (Michalzik et al., 1999; Neff and Asner, 2001; Yurova et al., 2008). In the DOCDM 1.0, rooting depth is used to determine the amount of soil organic matter. Therefore, woody vegetation with deeper root has more DOC production than those with shallow root, such as shrubs and grass. The spatial pattern of overland DOC also indicated that topography is another important factor to consider. Pixels with larger compound topographic index (not shown), which is used to quantify the water convergence, tend to have higher DOC concentrations in the overland flow. The steep slopes make short ponding time, implying short contact time with organic-rich surface soil horizon, thus less soil DOC is transferred to overland flow. Meanwhile, flat areas will result in water-logged, thus anaerobic conditions that slow mineralization process, resulting in more DOC to be extracted from soils. No flush events before snow melt make a slow DOC accumulation in soils during winter. The maximum DOC concentration usually occures at the very beginning of the spring flood (See Fig. 8e and f) when long-term stored DOC is initially flushed. Since DOC is mainly distributed in surface soils and snow melting dominates overland water flow, more than half of annual DOC export is delievered during a brief snowmelt period (around 4–7 weeks). After snow melting, rainfall barely resulted in overland events and its DOC concentration is low for the study region.
In the DOCDM 1.0, DOC can leave the watershed from three sources: overland (O), subsurface (S) and soil bottom (B). Water can be further removed by evapotranspiration (ET) in addition to the above three pathways. Water and DOC in overland and subsurface flow first route to channel and then to the river outlet eventually. Water and solute leak from soil bottom will enter the deeper ground layer and may reroute into channel system later. The groundwater in this study refers to both subsurface flow and bottom flow. Here we analyze how water and DOC distribute via those three pathways under changing environment conditions.
We define the annual infiltration ratio as the infiltration to precipitation in a given year. The higher infiltration ratio suggests the more dominant role of the groundwater system. Our simulations showed that the ratio increased from 0.27 to 0.42 during the study period (Fig. 9a). In other words, rainfall and snowmelt water have a larger tendency to join the subsurface, mainly due to increasing air temperature and deepening active layer (Fig. 7). Degrading permafrost increases hydrologic conductivity in soils, facilitating water infiltration.
Despite the increased groundwater, the ratio of overland DOC yield,
defined as
The percentages of water flow via overland flow, subsurface flow,
bottom flow and ET are shown in Fig. 9a. Decreasing ice due to warming
led to increasing soil hydraulic conductivity and increased water leak
from soil bottom. The large leak subsequently limited water supply to
overland and subsurface flows. There was 43 % of water left the
study region via belowground pathways in 2004, while the counterpart
in 1976 was only 30 %. Thawing fosters soil moisture, but deeper
active layer depth and larger infiltration make more water storage in
the depth below the rooting system. Thus, ET is limited by moisture
supply in shallow layers and low overland flow. As a result of these
water flows, the DOC delivered through subsurface and bottom flow
decreased from 68 to 61 %. The subsurface flow only took
1–2 % of the total water flux, but carried 30–50 % of DOC
away. These dynamics are determined with two processes: (1) our
horizontal subsurface flow takes place in active layer in which roots
are distributed and have high DOC production. Therefore, subsurface
flow usually has high DOC concentrations; and (2) the mixing zone in
our model is set as 2
We also listed the ratios of DOC sorption, mineralization and yield to DOC production for 1976 and 2004 in Table 2. The DOC yield here refers to the amount of DOC leaving the study watershed via both channel flow and bottom flow. Note that the DOC mineralization here includes the decomposition of both DOC and absorbed DOC. The fact that the absorbed DOC pool has ten times storage of the annual DOC production, suggesting it is necessary to take the sorption and desorption into account. The increasing soil moisture (liquid water) in the thawing months (May–September, not shown here) from 1976 to 2004 reduced DOC production, while the ratio of DOC mineralization to its production changed from 2.32 to 1.64, suggesting that the warming stimulated DOC production in soils. Meanwhile, the ratio of DOC yield to its production also showed a decreasing trend. The thawing soil and the associated hydrological alteration actually slowed down the DOC transport, resulting in more DOC accumulated in soils.
A catchment-scale DOC model is developed and applied for a watershed in Alaska. DOC flux and concentrations are found to be closely coupled with hydrological dynamics. We find that: (1) the snowmelt period (May–June) dominates the catchment's DOC export, (2) the permafrost thawing and enhancement in groundwater circulation contribute to river discharge in arctic and subarctic basins, (3) permafrost-driven changes in subsurface flow paths and water fluxes influence the flux of DOC from terrestrial to aquatic ecosystems; and (4) Increasing DOC production and more percolation water are two counterbalanced mechanisms in affecting DOC fluxes and concentrations under warming conditions. These dynamics might vary in different watersheds.
There are several limitations in this study. First, the initial
condition of the model can induce uncertainty, especially when the
simulation period is long. There is no available data as model initial
conditions for large temporal scale simulations. Instead, in this
study, we first run the VIC model for 10
DOCDM 1.0 is written in C
The code is available under request for academic and non-commercial
use. The code is also archived in our lab website
(
This research is supported with projects funded to Q. Zhuang by NSF (DEB-0919331), the NSF Carbon and Water in the Earth Program (NSF-0630319), the NASA Land Use and Land Cover Change program (NASA-NNX09AI26G), Department of Energy (DE-FG02-08ER64599), the NSF Division of Information & Intelligent Systems (NSF-1028291), and DOE/Lawrence Berkeley National Laboratory IMPACTS Program. This research was also in part supported by the Director, Office of Science, Office of Biological and Environmental Research of the US Department of Energy under contract no. DE-AC02-05CH11231 as part of their Earth System Modeling Program.
The parameters and constants used in models of DOC production, minearlization and sortpion and desortion.
The ratios of DOC prodction to its sorption, mineralization and transport in 1976 and 2004.
DOCDM 1.0 framework of hydrology, vegetation and DOC dynamics. See Supplement A for more details of description. The processes of DOC dynamics (yellow part) are further illustrated in Fig. 2.
Schematic diagram of the pools of soluble organic matter (DOC and PDOC) and production, mineralization and adsorption and desorption processes (the yellow part in Fig. 1).
The two scenarios for the saturated subsurface flow between adjacent neighbors. Grey, blue and white represent ice, liquid saturated and unsaturated regions in soil column, respectively. H1: saturated layer in soil column. Z1: possible top unsaturated layer. Z2: possible ice layer. Z3: possible liquid saturated layer. Z4: possible bottom ice layer.
The DOC validation results for days of 18 July 2003
The land cover map for the test watershed (HUC 19040405). The red star stands for the river outlet and the red soild cicrle is the demonstration point. The resolution for this map is 1
The model outputs on the demonstraion point in 1976 and 2004:
The model outputs in 1976 and 2004:
The overland flow depth on 10 June 1976
The change of water