Environ. Sci. Technol. 2005, 39, 8057-8063
Model Assessment of Biogeochemical Controls on Dissolved Organic Carbon Partitioning in an Acid Organic Soil DAVID G. LUMSDON,* MARC I. STUTTER, RICHARD J. COOPER, AND JOHN R. MANSON Biogeochemistry of Catchments, The Macaulay Institute, Craigiebuckler, Aberdeen, AB15 8QH, U.K.
A chemical model (constructed in the ORCHESTRA modeling framework) of an organic soil horizon was used to describe soil solution data (10 cm depth) and assess if seasonal variations in soil solution dissolved organic carbon (DOC) could be explained by purely abiotic (geochemical controls) mechanisms or whether factors related to biological activity are needed. The NICA-Donnan equation is used to describe the competitive binding of protons and cations and the charge on soil organic matter. Controls on organic matter solubility are surface charge and a parameter, γ, that accounts for the distribution of humic molecules between hydrophobic and hydrophilic fractions. Calculations show that the variations in solute chemistry alone are not sufficient to account for the observed variations of DOC, but factors that alter γ, such as biological activity, are. Assuming that DOC in organic soils is derived from soluble humic material and that γ is modified seasonally due to biological activity (with monthly soil temperature used as a surrogate for biological activity) we are able to model the observed seasonality of soil solution DOC over a 10-year period. Furthermore, with modeled DOC coupled to other geochemical processes we also model soil solution pH and Al concentrations.
periods when the DOC was more hydrophilic in nature, with greater fluxes attributed to post-production autumn leaching (10). At the same site, Tipping et al. (11) showed by using organic-rich soil microcosms that artificial warming and drying accelerated the production of DOC, but concluded that leaching was a complex interaction of factors related to flushing events and adsorption in mineral horizons. For forested ecosystems in Germany (12, 13) and the Northeast United States (14, 15) studies have similarly related seasonality in DOC concentrations to temperature. Both Kaiser et al. (12) and Scott et al. (16) provide evidence for the compositional seasonality of DOC as seen in the distribution of macromolecular and low-molecular-weight components. It is clear that in natural systems climatic influences affect DOC by impacting on both biological and hydrogeochemical controls (10, 11, 16). Organic matter partitioning between the soil solid and solution phases is a fundamental process needed for modeling water quality and contaminant transport. Recent progress in modeling the chemical behavior of soil organic matter, e.g., WHAM (2) and the NICA-Donnan model (3), has allowed ion binding by soil organic matter to be predicted for a wide range of chemical factors (pH, competition, ionic strength). The WHAM model also includes a “fulvic acid” partitioning sub-model where fulvate solubility depends on the charge and hydrophobicity of the fulvic acid. The approach used in WHAM to account for fulvate partioning has recently been used with the NICA-Donnan model to allow prediction of DOC in soil solution (17). Both models provide a means to interpret how geochemical factors could control DOC concentrations in soils. Even though biological processes can exert some control on DOC, physicochemical partitioning mechanisms will also influence DOC solubility. In this paper we apply a process-based model to an independently obtained data set to assess the relative importance of geochemical control mechanisms as opposed to processes influenced by biotic factors. The model is used to explain the variations in soil solution pH, DOC, and dissolved aluminum over a 10-year period in a peat soil at the Environmental Change Network (18) monitoring site in NE Scotland.
Methods and Data Collection Introduction Mechanistic soil chemical models describing a wide range of processes including solution speciation, mineral solubility, and adsorption reactions, particularly those for soil organic matter (1-3), have made significant advances in recent years. With growing interest in the global carbon cycle and release of dissolved organic carbon (DOC) and associated pollutants (metals, acidity, and nutrients) to surface waters (4) there is a need to develop and test models further. Particularly important is the fact that the observed seasonal variation and long-term rising trends of DOC have been linked to climate and climate change (5-7). In the Northern hemisphere, tundra, moorland, and forested ecosystems have accumulated vast stores of carbon, estimated at 182 × 1015 g C by Jenkinson et al. (8). Measured fluxes of CO2 (9) and DOC (10) from soil suggest seasonal trends in soil carbon decomposition, related to temperature and wetness. For upland organic soils (Northern Pennines, U.K.) DOC concentrations were greatest in warmer summer * Corresponding author telephone: +44-1224-498200; fax: +441224-311556; e-mail:
[email protected]. 10.1021/es050266b CCC: $30.25 Published on Web 09/21/2005
2005 American Chemical Society
Sampling Site and Soil Data. Soil solution and climate data from upland Calluna moorland at Glensaugh, NE Scotland (57° N, 3° W) were recorded at the Environmental Change Network site (18). The site, with mean annual rainfall of 1040 mm and temperature of 7.6 °C, comprises humus iron podzols derived from an acid schist drift and is typical of many organic soils. Six soil tension lysimeters (Prenart Super Quartz), all at a depth of 10 cm, were evacuated (50 kPa) to collect soil solutions from the Oa horizon, across a 6 × 6 m plot on a 20° slope (300 m altitude). Filtered samples (1. In this case solubility is not only driven by geochemical controls but also biotic processes that cause γ to vary. In previous studies (17, 26) γ has been fitted to batch equilibration data consisting of DOC measurements as a function of pH. Values for γ ranging from 0.8 to 3 have been reported for organic soils. However, it was impossible to match the seasonal DOC trends with model 1, using a unique value of γ. Model 1 failed with a fixed value of γ because there was insufficient variation in other DOC control variables, e.g., pH, between the seasons to cause the magnitude of DOC change observed. Therefore the following approach was used to parametrize eq 6 to allow calculation of the temperature-dependent value of γ (γt). First model 1 was fitted with R equal to 1 (i.e., no temperature dependency for γ), tref was fixed at 273 K, and γref was fitted to the DOC concentration observed for the winter months. Values of γref of about 2.2 gave calculated DOC in the region of 5-10 mg/ L, close to those observed for the winter months. Next for the model 2 scenario γref was fixed at the value derived for model 1 and R was adjusted to allow a temperaturedependent value of γt to be derived. A value of R of about 1.4 resulted in change of γt from 1.8 at 2 °C to 1.3 at 13 °C, sufficient to bring about a change of predicted DOC from about 7 to 35 mg/L. Accordingly, R was fixed with a value of 1.4 for the model 2 scenario. The results of the model 1 and 2 calculations compared to the observations are shown in Figure 3. For pH there is no difference between model 1 and model 2, because solid phase and solution phase organic matter contribute equally to the solution acidity. The acidic episode seen in January 1998 where soil solution pH decreased to 3.8 was described by the model. Examination of the rainfall chemistry revealed that this acidic episode was caused by a pulse of NaCl. The general trend of increasing pH over the time period 1993 to 2003 has been attributed to a decline in sulfate acidity at the site (30, 31) and will not be considered any further. Using model 1, in which geochemical controls determine solubility, only a weak seasonal pattern of DOC was predicted. However, the amplitude of the oscillation was small relative to the observed data, at the most pronounced being about 5 mg/L (Figure 3). The introduction of a temperaturedependent value of γ (model 2) produced a marked improvement in the description of the seasonal pattern of DOC (Figure 3). The improved model simulation can be explained in relation to the sensitivity of calculated DOC in relation to γ as a function of pH (Figure 2). The figure shows that variations in γ result in greater changes of calculated DOC than can be obtained by the range of pH (ca. 3.8-4.4) observed in the data. Another factor for consideration is the assumed source of the DOC. In our simple model DOC results from the chemical equilibrium between solid-phase humic substances and the soil solution, i.e., the origin of DOC is from humic substances. Hagedorn et al. (28) provides evidence that DOC in forest soils mainly originates from recalcitrant organic matter and that much of the fresh seasonally produced organic matter (e.g., root exudates) is lost from the soil as CO2 gas. Our modeling approach can be contrasted to soil C dynamic models (e.g., DyDOC) which simulate metabolic transformations of carbon between differing carbon pools
FIGURE 3. Model-predicted and observed (a) pH, (b) DOC, and (c) total dissolved aluminum in soil solution in an organic (Oa horizon) soil at 10 cm depth. The dashed line shows the prediction in which geochemical factors control DOC (Model 1). The continuous line (Model 2) shows the effect using a temperature-dependent value of γ, calculated according to eq 6. using first order kinetic processes (32). In DyDOC and our model, DOC originates from the humified organic matter pool. In DyDOC, the humic pool size varies according to first order kinetic processes, whereas in our model we assume γ varies, thereby changing solubility of the humic pool. Our approach recognizes that biotic processes which are seasonally (stimulated by increasing temperature) driven will metabolize organic matter, leading to changes in the hydrophilic character of the organic matter. A significant difference between our model and the current DyDOC approach concerns the coupling of humic solubility to changing geochemical conditions. At present the metabolic model in DyDOC is not linked to geochemistry, in our model it is. Therefore we can assess how varying solute chemistry changes organic matter solubility. In relation to general model applicability, DyDOC has been applied to two sites, both forested with Norway Spruce (Picea abies), and requiring similar model input parameters (32). Our model, though applied to only one soil, modeled 10 years of data for a typical organic soil. An important feature of DyDOC which we do not address concerns the tracking of 14C, from its input to its loss as DO14C in drainage water, enabling information about C dynamics to be obtained. The geochemical com-
ponent of our model is mechanistically based, but the empirical linkage between γ and temperature, while being simple to implement, requires further testing. A significant observation in relation to the application of models such as WHAM (2, 26) is the use of γ as an intrinsic soil property. Our data (Figure S1) would suggest that the value of γ required to model batch equilibration data depends on the time of year the soil is sampled. The behavior of aluminum in the Strichen O horizon can be analyzed using the model. The observed total concentrations of aluminum in soil solution and other chemical components allow calculation of Al3+ activity (aAl3+) and the distribution of aluminum between soluble inorganic and organic complexes. The ion activity product (IAP) defined by log {aAl3+}/{aH+}3 ranged from ca. 6.4 at pH 3.8 to
