Reactivity continuum of dissolved organic carbon decomposition in lake water

Authors


Abstract

[1] We determined microbial decomposition of dissolved organic carbon (DOC) over 3.7 year long dark bioassays of six Swedish lake waters. The overall lost DOC fraction was similar in clearwater lakes (34.8 ± 2.4%) and in brownwater lakes (37.8 ± 1.9%). Reactivity continuum modeling revealed that the most labile DOC fraction, degrading at rates >0.01 d−1, was larger in the clearwater lakes (11.1 ± 1.2%) than in the brownwater lakes (0.8 ± 0.1%). The initial apparent first-order decay coefficientk was fivefold larger in the clearwater lakes (0.0043 ± 0.0012 d−1) than in the brownwater lakes (0.0009 ± 0.0003 d−1). Over time, k decreased more steeply in the clearwater lakes than in the brownwater lakes, reaching the k of the brownwater lakes within 5 months. Finally, k averaged 0.0001 d−1 in both lake categories. In the brownwater lakes, colored dissolved organic matter (CDOM) absorption decayed with an initial k twice as large (0.0018 ± 0.0008 d−1) as that of DOC. The initial kwas inversely correlated with initial specific UV absorption and CDOM absorption and positively correlated with initial tryptophan-like fluorescence as proxy for autochthonous DOC. Exposure to simulated sunlight at the end of the incubations caused loss of color in the clearwater lakes and loss of DOC in the brownwater lakes, where subsequent mineralization was also stimulated. The DOC lost in the absence of photochemical processes fell below previously reported watershed-scale losses in Sweden by 25% at most. This suggests that a major part of the in situ DOC loss could potentially be attributed to dark reactions alone.

1. Introduction

[2] Inland waters receive large quantities of terrestrial dissolved organic carbon (DOC) from their catchments (“allochthonous DOC”), a substantial fraction of which is mineralized during passage toward the sea. On a global scale, inland waters are estimated to emit 1.2–1.4 Pg CO2-C yr−1 to the atmosphere [Tranvik et al., 2009; Aufdenkampe et al., 2011]. Primarily owing to stimulated terrestrial vegetation cover in response to climate change, this flux is predicted to further increase in boreal regions [Larsen et al., 2011].

[3] DOC is a complex mixture of compounds that are poorly characterized [Thurman, 1985; Dittmar and Paeng, 2009]. Its turnover is largely determined by the bioavailability of the different constituents, and spans timescales of minutes to millennia [Del Giorgio and Davis, 2003]. In a closed system, DOC bioavailability decreases with time because microbial communities selectively consume the more labile substances first [Middelburg, 1989]. DOC decomposition may be influenced by extrinsic factors (e.g., temperature, microbial composition, nutrient and oxygen availability), intrinsic molecular properties (e.g., size, complexity, aromaticity, aliphaticity) [Del Giorgio and Davis, 2003; Bastviken et al., 2004a], photochemical reactions [Wetzel et al., 1995] and DOC concentrations [Søndergaard and Middelboe, 1995].

[4] On short timescales in laboratory experiments, allochthonous DOC is usually less bioavailable than internally produced “autochthonous DOC.” For example labile DOC, which was defined as DOC microbially mineralized in dark bioassays within 1 to 2 weeks, averaged 14% of the total DOC in 26 clearwater lakes but only ∼1–2% in a brownwater lake and river [Søndergaard and Middelboe, 1995]. However, similar DOC fractions were mineralized during a weeklong incubation of oligotrophic lake water with different influence of allochthonous DOC [Tranvik, 1988]. Also, allochthonous DOC mineralization can be extensive on longer timescales. For example, 30–50% of brownwater DOC from a mountain bog pool decomposed during 91 days [Satoh and Abe, 1987], and half of the DOC from vascular plant leachates decomposed during 2.5 year long dark incubations [Vähätalo and Wetzel, 2008]. During inland water passage, on average 46% of the total organic carbon exported into lakes and rivers of major Swedish catchments were mineralized [Algesten et al., 2003]. Hence, allochthonous compounds are an important part of bioavailable DOC.

[5] Photochemical processes also contribute to DOC loss since solar irradiation readily degrades chromophoric allochthonous DOC into more microbially labile compounds [Lindell et al., 1995; Wetzel et al., 1995] or inorganic carbon [Granéli et al., 1996]. It has been suggested that the main factor for autochthonous DOC mineralization is its biological reactivity whereas mineralization of intrinsically more recalcitrant allochthonous DOC is to an important degree mediated via its photochemical reactivity [Cole, 1999; Moran et al., 2000; Farjalla et al., 2009]. However, the actual depth-integrated contribution of photochemistry to overall inland water CO2emissions remains largely speculative. Moreover, it has been hypothesized that DOC mineralization could increase with increasing concentrations, which could possibly be related to differences in half-saturation constants [Søndergaard and Middelboe, 1995]. Contradicting this hypothesis, however, bacterial growth rates and efficiencies in lake batch cultures did not depend upon DOC concentrations at levels relevant for most natural aquatic systems [Eiler et al., 2003].

[6] Short-term bioassays are a useful operational tool to assess DOC bioavailability. However, incubations which last just a few weeks may poorly reflect the timescale of DOC processing at the landscape scale. In this study, we determined DOC decomposition in samples from boreal clearwater and brownwater lakes during 3.7 year lasting dark bioassays, and subsequently assessed the photochemical reactivity of the remaining DOC. Microbial DOC mineralization was described using reactivity continuum modeling, and DOC quality was assessed using UV-visible absorbance and fluorescence spectrometry. Our study shows that, while the DOC from the brownwater lakes decomposed initially slower, differences between mineralization rates diminished within just 5 months and overall bioavailability was similar in both clearwater and brownwater lakes. Hence, substantial DOC mineralization in boreal lake systems dominated by allochthonous organic matter does not necessarily require light-induced photochemical decomposition.

2. Materials and Methods

2.1. Sampling Area and Lakes

[7] Six boreal lakes were sampled in south-central Sweden (Table 1). In this region, annual mean temperatures average 2–6°C with a mean annual precipitation of 600–900 mm (1961–1990, Swedish Meteorological and Hydrological Institute; see http://www.smhi.se/klimatdata). The catchments are dominated by coniferous forest except for lake Valloxen with a high agricultural share. In addition, we used an aged reverse osmosis concentrate from a brownwater lake (Lilla Björntjärn) to study DOC concentration effects on decomposition.

Table 1. Location and Characteristics of the Seven Boreal Clearwater and Brownwater Lakesa
LakeLilla BjörntjärnLilla SångarenLjustjärnLumpenSiggeforasjönStensjönValloxen
CategoryBWCWCWBWBWBWCW
Position64°07′N, 18°47′E59°54′N, 15°23′E59°55′N, 15°26′E59°58′N, 17°17′E59°58′N, 17°09′E60°03′N, 17°49′E59°43′N, 17°50′E
Surface area (km2)0.010.240.120.250.760.082.90
Maximum depth (m); mean depth (m)8.3; 4.617; 611; 41.9; 1.311; 4.22.2; 1.39; 3.8
Theoretical water residence time (years)NA1.184.270.450.490.061.70
DOC (mg C L−1)b14.63 ± 1.326.65 ± 0.224.60 ± 0.6924.82 ± 1.2515.09 ± 1.2520.00 ± 2.9715.84 ± 1.98
pH5.30 ± NA6.80 ± NA6.73 ± 0.167.00 ± 0.176.74 ± 0.147.00 ± 0.178.02 ± 0.05
Total phosphorus (μg P L−1)28.00 ± 3.4611.5 ± 1.248.84 ± 1.1820.34 ± 2.379.71 ± 0.7520.14 ± 2.9347.10 ± 2.86
Total nitrogen (mg N L−1)0.46 ± 0.030.28 ± NA0.22 ± 0.021.15 ± 0.070.63 ± 0.071.20 ± 0.250.93 ± 0.10

2.2. Experimental Design

[8] We conducted a factorial lake sampling to investigate the relationship between DOC quality and microbial mineralization, with three randomly chosen lakes each at the levels “clearwater” and “brownwater.” Lakes were assigned to the factor levels on the basis of specific UV absorption at 254 nm and colored dissolved organic matter absorption (250 to 500 nm) which averaged threefold and sixfold larger in the brownwater compared to the clearwater lakes (both P < 0.004; see Table 2). Furthermore, manipulative split-plot experiments were conducted within each factor level to investigate responses of mineralization rates to addition of inorganic nutrients (nitrogen and phosphorus combined) and labile organic C (glucose). Each treatment within the nested design was carried out in three subsample series, which were averaged before statistical analyses. To test for relationships between DOC quantity and mineralization we conducted a separate split-plot experiment in which preconcentrated water from one lake was incubated at six different initial concentrations (referred to as “DOC concentration experiment”; carried out in two subsample series and including N + P and glucose addition treatments). All incubations lasted 3.7 years, time series of DOC concentrations and UV-visible absorbance were measured, and initial fluorescence scans were taken. Additional split-plot experiments were conducted at the end of the incubations to assess direct and indirect effects of irradiation with simulated sunlight, and effects of microbial reinoculation on residual color and DOC concentrations.

Table 2. Mean Initial DOC Concentrations, CDOM Absorption, SUVA254, and Intensities of the Five Identified Fluorescence Componentsa
 Clearwater LakesBrownwater Lakes
  • a

    Means are ±SE, n = 3. Ex/Em, maximum excitation/emission wavelength. If applicable, common peak names of the fluorescence components are added [Stedmon and Markager, 2005b].

DOC (mg C L−1)8.62 ± 3.6225.01 ± 3.74
Colored dissolved organic matter absorption (cm−1)23.02 ± 10.85156.67 ± 25.81
Specific UV absorption at 254 nm (L mg C−1 m−1)4.16 ± 1.088.17 ± 0.15
Fluorescence component 1 (unitless; Ex/Em 332/448, peak C)0.81 ± 0.020.62 ± 0.04
Fluorescence component 2 (unitless; Ex/Em 306/404, peak M)0.57 ± 0.040.32 ± 0.02
Fluorescence component 3 (unitless; Ex/Em 245/438, peak A)0.24 ± 0.060.41 ± 0.01
Fluorescence component 4 (unitless; Ex/Em 240, 413/482)0.20 ± 0.010.20 ± 0.01
Fluorescence component 5 (unitless; Ex/Em 281/362, peak T)0.25 ± 0.100.03 ± 0.01

2.3. Water Sample Preparation and Bioassays

[9] Lake water samples were filtered through a 45 μm mesh and bubbled with synthetic air (20.5 ± 0.5% O2 in N2, Alphagaz Auto IV, Air Liquide, Sweden), reaching initial oxygen concentrations of 9.8 ± 0.2 mg L−1 across treatments. One batch of this water received 480 μg N L−1 of potassium nitrate and 100 μg P L−1 of disodium hydrogen phosphate, and one remained untreated (control). Subsequently, the water was distributed into precombusted 40 mL glass vials with Teflon coated septa and sealed headspace free. To minimize residual gas exchange across the septa, which were not completely gas tight, the vials were submersed in pure water in containers with closed lids. These were kept at 20°C in the dark. By incubating a blank series with only synthetic lake water (see below) we verified that no C contamination from the environment occurred during the course of the experiment. In this blank series, mean DOC concentrations averaged 0.55 ± 0.05 mg C L−1on seven sampling dates spread across the experimental duration. To measure the decomposition of DOC over time, vials were sacrificed weekly during the first month, and afterward in months 2, 6, 10, 21, 29 and 44 for the brownwater lakes, and in months 3, 8, 13, 27 and 43 for the clearwater lakes. To subsample series of both treatments, fresh DOC was added after a period of incubation (0.5 mg glucose-C L−1after 3.5 months to the clearwater lake samples, and 2 mg glucose-C L−1 after 2 months to the brownwater lake samples), and the series was measured on the same days as the control up to month 43 for the clearwater lakes, and up to month 21 for the brownwater lakes. Flocculation was observed at later times during the experiment but seemed negligible (i.e., too little to be analytically quantified). Therefore, we call the overall DOC loss “microbial decomposition” or “mineralization” throughout the text.

[10] For the DOC concentration experiment we used DOC of lake Björntjärn which had been concentrated by reverse osmosis to 230 mg C L−1 and washed extensively with pure water by tangential flow ultrafiltration (1 kD cutoff) to remove inorganic nutrients and small organic molecules [Kragh et al., 2008]. At the start of the experiment, the preparation of the concentrate was 10 years ago after which it had been stored at 4°C. Prior incubation, we filtered the concentrate through a A/E Gelman filter (142 mm) and a membrane filter (0.2 μm, Supor filter series, Pall Corporation, Port Washington, New York, USA), adjusted the pH from 3.2 to 7 by addition of NaOH, and diluted the concentrate with organic-free synthetic lake water [Lehman, 1980] to six different initial concentrations (∼2, 4, 8, 12, 20 and 30 mg C L−1). These were inoculated with unfiltered lake water (2% of the sample volume) and oxygen saturated as described above. One batch of each sample series received 1 mg glucose-C L−1, and all samples received N and P addition at the same concentrations as above. Incubations in dark water baths at 20°C proceeded as explained above. Measurements were taken on days 0, 4, 67, 552, 785, and in months 43 and 45. Flocculation was somewhat more pronounced than in the lake water incubations but remained too little to be quantitatively determined.

2.4. Total Organic Carbon and Oxygen Concentrations

[11] Total organic carbon (TOC) concentrations were measured using a Sievers 900 TOC Analyzer (General Electric Analytical Instruments, Manchester, UK) which determines TOC in a range from 0.03 ppb to 50 ppm with a precision of <1% relative standard deviation and an accuracy of ±2% or ±0.5 ppb (whichever is greater) [General Electric Analytical Instruments, 2005]. Since the bulk of TOC is dissolved we use the term DOC throughout the text. Oxygen concentrations were measured at the start and end of the incubations using a PSt1 micro-optode (Microx TX3, Precision Sensing GmbH, Regensburg, Germany).

2.5. UV-Visible Spectroscopy

[12] Absorbance spectra (200–600 nm, 1 nm resolution) were recorded using a 1 cm quartz cuvette on a UV-VIS spectrophotometer (PerkinElmer Lambda 40, Waltham, Massachusetts, USA). All spectra were blank subtracted with pure water (Milli-Q Water Purification System, Millipore, Billerica, Massachusetts, USA). On the basis of the Beer-Lambert law, Napierian absorption coefficientsa (m−1) were calculated as

equation image

where A is absorbance (dimensionless) and L is optical path length (m) [Kirk, 1994]. Total colored dissolved organic matter (CDOM) absorption was calculated by applying the trapezoid rule to a in the range 250 to 500 nm [Moran et al., 2000]. Specific UV absorption (SUVA254, L mg C−1 m−1) was calculated by dividing a at 254 nm by the DOC concentration. Please note that (even in the N + P addition samples) total N was less than 0.5 mg N L−1, which is well below the concentration in which nitrate causes interference with absorbance measurements at wavelengths greater than ∼230 nm [Weishaar et al., 2003]. Exponential spectral slopes were not calculated since model assumptions (i.e., normality of residuals and homogeneity of variance) were severely violated with S-shaped residuals, as has been found and discussed also by others [Twardowski et al., 2004].

2.6. Fluorescence Spectroscopy

[13] On the first day of the incubations, fluorescence excitation-emission matrices (EEM) were measured in a 1 cm quartz cuvette on a fluorescence spectrophotometer (SPEX FluoroMax-2, Horiba Jobin Yvon, Unterhaching, Germany; excitation (Ex): 240–450 nm with 5.1 nm increments, emission (Em): 300–600 nm with 2 nm increments). Samples were run in signal mode and corrected for lamp excitation reference intensity. Excitation and emission slit widths were set to 5 nm and integrated over 0.1 s. All EEMs were blank subtracted using pure water, and corrected for instrument-specific lamp biases using the manufacturer supplied factors, and for the inner filter effects [McKnight et al., 2001]. Fluorescence intensities were converted to Raman units by division with the Raman area of pure water (integrated at Ex = 350 nm, from Em 380–420 nm). EEM intensities were further normalized to the Ex:Em point 240:438, which exhibited the maximum peak height in most spectra. Individual fluorescing EEM components were generated applying parallel factor analysis (PARAFAC) [Andersson and Bro, 2000], using the DOM-fluor toolbox (MATLAB® 7.7.0, The MathWorks, Natick, Massachusetts, USA, 2008) and associated tutorial [Stedmon and Bro, 2008]. Before modeling, we removed first-order scatter (starting at 298 nm), second-order scatter (starting at 500 nm), and cut the nonfluorescing region of EEMs with an emission above 560 nm. Model validation was conducted using ten runs of split-half random initializations.

2.7. Reactivity Continuum Modeling of DOC Mineralization

[14] Time series of DOC mineralization were described using a reactivity continuum model which is an integration of first-order exponential decay functions over the reactivity distribution (also termed “continuum quality model”). The initial distributions of reactivity were described with gamma distributions, whose exponential parts account for the more reactive fractions while the power parts describe the degradation of the more recalcitrant compounds. Accordingly, relative DOC concentrationsDOCt/DOC0 (unitless) at time t (day) are defined as

equation image

where α is a rate parameter (i.e., average lifetime of the more reactive DOC compounds; days) and υ is related to the shape of the distribution near a decay coefficient of zero (unitless). Hence, υ depicts the relative preponderance of the more recalcitrant compounds. For υ = 1 all DOC compounds would have the same reactivity whereas if 0 < υ< 1 the mixture is dominated by refractory compounds. The initial apparent first-order decay coefficientsk are the expectation values of the gamma distribution defined as υ/α (d−1) [Aris, 1965; Boudreau and Ruddick, 1991; Burnham and Braun, 1999; Manzoni and Porporato, 2009; Dolgonosov and Gubernatorova, 2010]. The decrease of k over time was calculated as υ/(a + t) [Boudreau et al., 2008]. Apparent initial DOC residence times were calculated as the inverse of the initial k.

[15] For comparison with other studies, we also fitted an exponential decay model with residual pool to our data [Westrich and Berner, 1984] as

equation image

where DOCL and DOCR are the initial concentrations of the “labile” and “residual” DOC, respectively (mg C L−1), and k is the decay constant of DOCL (d−1). We did not expand the model to a three-pool version (i.e., five fit parameters) because it yielded convergence problems and for reasons discussed insection 4.1.

[16] Parameter estimation and statistical testing were conducted using nonlinear mixed effects (NLME) models [R Development Core Team, 2010]. This approach was chosen since the use of least square estimation for data with grouped structure, such as longitudinal data, results in inefficient parameter estimation and biased standard errors [Moulton, 1986]. In the NLME models, factor level (i.e., clearwater versus brownwater lakes, N + P addition, glucose addition) was defined as fixed effect, and lake or DOC concentration level as random effect. Significance of the fixed effect was evaluated using analysis of variance [Crawley, 2009]. In the models for the DOC concentration experiment, the factor level (i.e., initial DOC concentration) was included both as fixed and random effect. Differences between DOC concentration levels were assessed with all-pairwise multiple comparisons of the fit parameters in whichP values were multiplicity adjusted [Hothorn et al., 2008].

2.8. Bacterial Abundance, Irradiation, and Microbial Reinoculation

[17] To quantify final bacterial abundances, subsamples were preserved with sterile-filtered borax-buffered formaldehyde (4% final concentration) and dyed with the nucleic acid stain SYTO 13 before using flow cytometry (CyFlow space, Partec GmbH, Münster, Germany) [Del Giorgio et al., 1996]. To gauge the presence of photoreactive DOC after long-term decomposition in the dark, vials of the DOC concentration experiment and of the clearwater lakes were exposed to 14 h of irradiation with simulated sunlight (Q-Sun 1000 Xenon test chamber, Q-panel Lab Products Europe, Bolton, UK). The solar simulator was set to an intensity of 0.65 W m−2at 340 nm. The vials absorbed most of the UV-B but were transparent at wavelengths >320 nm (i.e., <1% of UV-A and visible light absorbed). Inside the vials, the photon density of photosynthetic active radiation was 0.18μE cm−2 s−1 (measured using a spherical light meter, Biospherical Instruments Inc., San Diego, California, USA), which corresponds to about fourfold the photon density received in southern Sweden during July and August (Swedish Meteorological and Hydrological Institute; see http://www.smhi.se/klimatdata). One sample series was immediately cooled to 7°C after irradiation and analyzed for DOC concentrations and UV-vis absorbance. A second sample series was further dark incubated at 20°C for 6 weeks before analysis. Finally, to test the possibility that mineralization was hampered owing to deterioration of the bacterial community confined in the vials, another subsample set was reinoculated with bacteria by adding freshly sampled lake water (100μL) with a Hamilton precision syringe (Hamilton Bonaduz AG, Bonaduz, Switzerland). These samples were analyzed after 6 weeks of further incubation.

2.9. Statistical Analyses

[18] Independent t tests were used to assess differences in initial parameter values (DOC concentrations, absorbance and fluorescence properties). Linear least square regressions were used to describe relationships between k and explanatory variables, and model significance was assessed by regression analysis of variance. For simultaneous inference testing about the influence of optical properties on k, a linear mixed effects (LME) model was set up which included SUVA254, CDOM absorption and the fluorescence components as fixed effects, and the factor level (N + P addition) as random effect. To avoid variance inflation due to multicollinearity of the fixed effects explanatory variables x these were mean centered and standardized prior to modeling [x* = (x − mean(x))/standard error(x)]. Pvalues were multiplicity adjusted using a single-step method [Hothorn et al., 2008]. LME models, set up as described in section 2.7, were used to test effects of irradiation with simulated sunlight and microbial reinoculation on DOC concentrations and absorbance properties. If residual plots revealed nonnormality in the residuals or nonhomogeneity of variance, right-skewed data were logarithmic transformed and analysis was repeated [Crawley, 2009]. In all analyses, differences were considered significant if P value < 0.05. Mean values in the text are given with ±1 standard error, and with ±95% confidence intervals in regression equations. Analyses were conducted using R 2.11.1 [R Development Core Team, 2010].

3. Results

3.1. DOC Concentration Experiment

[19] We compared dark microbial mineralization at six initial DOC concentrations from 2 to 30 mg C L−1, prepared by dilution of a reverse osmosis DOC concentrate with organic-free synthetic lake water. Across this gradient, initial CDOM absorption increased eightfold with increasing DOC concentration (P < 0.001), and initial SUVA254 decreased by 35% (P = 0.0264). DOC mineralization rates were similar among concentration levels (Figure 1) and reactivity continuum model parameters (equation (2)) did not differ. Initial apparent first-order decay coefficientsk averaged 0.0010 ± 0.0001 d−1. On the basis of the probability distributions of reactivity, 1.2 ± 0.6% of the initial DOC would decay at rates >0.01 d−1, 24.1 ± 0.7% at rates between 0.001 and 0.01 d−1 and the remaining 74.7 ± 1.1% at rates <0.001 d−1. Glucose addition decreased the average lifetimes of the more labile compounds (parameter a) in the lowest concentration level (P = 0.008; see Figure 1a). However, it did not significantly affect a in the other concentration levels, and also not the relative preponderance of the more recalcitrant compounds (parameter υ; see Figure 1). DOC concentration time series did not differ among DOC levels in the glucose addition treatment either (Figure 1). Microbial reinoculation at the end of the experiment and 6 additional weeks of dark incubation did not affect DOC concentrations, SUVA254 or CDOM absorption.

Figure 1.

Mean (±, n = 2) time series of DOC (subscripts denote day of the incubation) during dark incubation of dilutions of preconcentrated DOC from the brownwater lake Björntjärn at six different initial concentrations: (a) 2.12 ± 0.02 mg C L−1, (b) 3.97 ± 0.01 mg C L−1, (c) 7.84 ± 0.17 mg C L−1, (d) 11.85 ± 0.07 mg C L−1, (e) 19.98 ± 0.01 mg C L−1, and (f) 31.47 ± 0.02 mg C L−1, without (open circles and solid lines) and with initial glucose addition (solid triangles and dashed lines). Lines are concentrations predicted by the reactivity continuum model. Gray vertical lines mark endings of years after the start of the experiment.

3.2. Clearwater and Brownwater Lakes

[20] Initial DOC concentrations were threefold smaller in the clearwater than the brownwater lakes (P = 0.037; see Table 2). The DOC proportion mineralized during the first incubation month was four times larger in the clearwater than the brownwater lakes (P = 0.013), but the overall DOC fraction mineralized over 3.7 years did not differ between lake categories (see Figure 2a and Table 3). Hence, although the most labile fraction was smaller in the brownwater lakes, long-term recalcitrance over 3.7 years was similar. According to the reactivity continuum model, the parametersa and υ were smaller in the clearwater than the brownwater lakes (both P < 0.0057). Combined, this resulted in larger initial apparent k, and in smaller initial apparent DOC residence times in the clearwater compared to the brownwater lakes. Ratios between theoretical water residence and apparent initial DOC residence times were larger but not significantly so in the clearwater than the brownwater lakes (Table 4).

Figure 2.

Mean (±, n = 3) time series of (a) relative decrease of DOC (subscripts denote day of the incubation), (b) CDOM absorption (250 to 500 nm), and (c) specific UV absorption at 254 nm (SUVA254) during dark incubation of samples from clearwater (open circles, solid lines) and brownwater (solid triangles, dashed lines) lakes. In Figure 2a the N + P addition series of the brownwater lakes is also plotted (solid inverse triangles, dot-dashed lines). For clarity the N + P addition series of the clearwater lakes, in which DOC loss was unaffected, is not plotted. Gray vertical lines mark endings of years after the start of the experiment. Gray squares show landscape-scale aquatic DOC losses to mineralization as reported from a mass balance study for 21 major Swedish catchments covering 79,536 lakes and rivers [Algesten et al., 2003]. In Figure 2a and for the brownwater lakes in Figure 2b, curves show values predicted by the reactivity continuum model. For the clearwater lakes in Figures 2b and 2c, curves show values fitted using local polynomial regressions.

Table 3. Mean Proportions of Initial DOC Concentrations Which Decomposed During the First Month and During 3.7 Years and Proportions of the Initial DOC Pool Decayinga
 Clearwater Lakes, DOCBrownwater Lakes
DOCCDOM Absorption
ControlN + P AdditionControlN + P AdditionControlN + P Addition
  • a

    Means are ±SE, n = 3. At k > 0.01 d−1 (reactivity class 1), 0.001 < k < 0.01 d−1 (reactivity class 2), and k < 0.001 d−1 (reactivity class 3) based on the probability distributions of reactivity for the control and N + P addition samples of the clearwater and brownwater lakes. For the brownwater lake samples the respective values for loss of CDOM absorption are also given.

Proportion lost during the first month (%)8.0 ± 0.99.8 ± 2.22.2 ± 1.02.4 ± 0.36.3 ± 1.76.8 ± 2.2
Proportion lost during 3.7 years (%)34.8 ± 2.436.5 ± 2.639.0 ± 2.541.2 ± 3.035.7 ± 5.038.4 ± 2.6
Reactivity class 1 (%)11.1 ± 1.211.8 ± 0.90.8 ± 0.12.2 ± 0.35.1 ± 0.85.6 ± 0.3
Reactivity class 2 (%)17.9 ± 1.417.6 ± 1.022.3 ± 1.925.6 ± 2.117.8 ± 2.220.6 ± 0.8
Reactivity class 3 (%)71.9 ± 2.670.6 ± 1.976.9 ± 2.072.2 ± 2.477.1 ± 3.073.8 ± 1.1
Table 4. Mean Parameters Derived From the Reactivity Continuum Model Which Was Fitted to the Decrease in DOC Over Time for the Clearwater and Brownwater Lakesa
 Clearwater Lakes, DOCBrownwater Lakes
DOCCDOM Absorption
Average lifetime of the more reactive DOC compoundsb (days)25.2 ± 5.8251.3 ± 68.163.4 ± 23.6
Relative preponderance of the more recalcitrant compoundsc (unitless)0.11 ± 0.020.23 ± 0.040.11 ± 0.02
Apparent initial first-order decay coefficient (d−1)0.0043 ± 0.00120.0009 ± 0.00030.0018 ± 0.0008
Apparent initial Residence time (years)0.7 ± 0.23.1 ± 1.01.6 ± 0.7
Theoretical water/apparent initial residence time3.69 ± 1.020.11 ± 0.040.22 ± 0.09

[21] The probability distributions of initial reactivity show that, in both lake categories, roughly three quarters of the initial DOC pool had a reactivity below 0.001 d−1 (see Figure 3a and Table 3). In the clearwater lakes, more than 10% of the initial DOC was likely to decay at rates of 0.01 d−1 or faster (see Figure 3a and Table 3), and k decreased sharply during the first incubation months (Figure 3b). The probability distribution of the brownwater lakes showed a steeper increase toward larger reactivity (that is, reached a probability of 1 quicker), and hence only a minor proportion of the initial DOC was likely to decay at rates >0.01 d−1. Moreover, k declined less pronouncedly over time. Within 5 months, the k of both lake types converged, and finally averaged 10−4 d−1 at the end of the incubations (Figure 3b). The probability distribution of reactivity and the respective decrease in k over time were markedly different from what was previously reported from sediment [Boudreau and Ruddick, 1991], visualizing the more labile nature of water column DOC compared to sediment OC (Figure 3).

Figure 3.

(a) Probability distributions of initial reactivity for DOC from the clearwater (solid blue curve) and brownwater (solid red curve) lake samples, as well as for CDOM absorption from the brownwater lake samples (solid orange curve). For the brownwater lakes the probability distribution is also shown for the N + P addition samples in which DOC mineralization was stimulated (dashed red curve). For comparison, the probability distribution of initial reactivity of marine sediment OC from Long Island Sound, USA, is added (solid green curve) [Boudreau and Ruddick, 1991]. Vertical dashed lines mark k = 0.001 and k = 0.01 d−1. Please note that the y axis commences at p= 0.4. (b) Apparent first-order decay coefficient over incubation time for DOC from the clearwater (solid blue curve) and brownwater (solid red curve, control; dashed red curve, N + P addition) lake samples for CDOM absorption from the brownwater lake samples (solid orange curve) and for the marine sediment OC (solid green curve). Gray vertical lines mark endings of years after the start of the experiment.

[22] N + P addition did not affect DOC mineralization in the clearwater lakes, but decreased the average lifetimes of the more labile DOC compounds by 32% in the brownwater lakes (a = 169.8 ± 35.3 days, P = 0.0223; see Figure 3a). This resulted in larger mean initial k (0.0013 ± 0.0003 d−1) which remained, however, threefold smaller than in the clearwater lakes (Figure 3b). Accordingly, mineralization in the brownwater lakes depended to some extent on inorganic nutrient concentrations (Table 3). In the clearwater lakes, glucose-DOC added after 3.5 months was rapidly mineralized but final DOC concentrations remained 18 ± 4% higher in the glucose addition compared to the untreated samples (P < 0.0001; see Figures A1a–A1c), suggesting that either glucose hampered decomposition or glucose-derived OC remained throughout the incubation. In the brownwater lakes, DOC concentrations did not differ between the control and glucose addition series (Figures A1d–A1f). At the end of the incubations, oxygen concentrations averaged 8.7 ± 0.1 mg L−1 (n = 20, randomly selected across treatments), suggesting that the observed declines in DOC mineralization rates were not due to oxygen depletion. Microbial abundances after 3.7 years were 0.38 ± 0.14 × 106 cells mL−1 in the clearwater lakes, and 0.74 ± 0.16 × 106 cells mL−1 in the brownwater lakes. Microbial reinoculation and 6 additional weeks of dark incubation did not affect DOC concentrations, SUVA254 or CDOM absorption in the clearwater lakes (not available for the brownwater lakes apart from the concentration experiment of lake Björntjärn; see section 3.1).

[23] Describing the loss of DOC over time with an exponential model (equation (3)) gave larger decay constants in the clearwater (0.0042 ± 0.0012 d−1) than the brownwater lakes (0.0013 ± 0.0002 d−1, P = 0.014). The most labile pool size was smaller in the clearwater (2.58 ± 1.19 mg C L−1) than the brownwater lakes (10.47 ± 1.50 mg C L−1, P < 0.0001), corresponding to 31% and 39% of initial DOC, respectively. Also the residual pool sizes (that is, the remaining DOC fractions) differed (P = 0.0096). The Akaike information criterion (AIC), which takes the goodness of fit and the number of estimated parameters into account, was smaller for the reactivity continuum model (−267.3) than for the exponential model (164.7). This shows that the reactivity continuum model was statistically superior.

3.3. Fluorescence and Absorption

[24] Five fluorescence components (C) were identified from PARAFAC analysis. Apart from C4, all these correspond to common fluorescence peaks described in earlier studies [Coble, 1996; Stedmon and Markager, 2005a, 2005b]. Initial fluorescence intensities of C1 (peak C) and C2 (peak M) were larger in the clearwater than the brownwater lakes (both P = 0.0185), and initial C3 (peak A) showed a tendency to be larger in the brownwater than the clearwater lakes (P = 0.0880; see Table 2).

[25] In the clearwater lakes, CDOM absorption increased by on average 37% during the first 3 incubation weeks, and then stabilized (Figure 2b). SUVA254 increased by on average 28% during the first month and remained similar thereafter (Figure 2c). In the brownwater lakes, CDOM absorption decreased steadily and most pronounced by 20.5 ± 2.4% during the first incubation year (until day 290), and decreased more gradually by a further 15.3 ± 1.4% in the subsequent incubation time (Figure 2b). According to the reactivity continuum model both v (P = 0.0370) and a (P = 0.0002) were smaller than for DOC loss. Consequently, the initial k was larger for CDOM absorption loss than for DOC loss, while apparent initial residence times were shorter for CDOM absorption than for DOC loss (i.e., water color turned over faster than bulk DOC; see Table 4). SUVA254 decreased slightly during the first 1–2 incubation months to an approximately constant level and increased somewhat again during the third and fourth incubation year (Figure 2c).

3.4. Explanatory Variables for Initial DOC Bioavailability

[26] First, we explored univariate relationships between the initial k and optical water properties. In the six lakes, mean initial k decreased linearly with increasing mean initial SUVA254 (Figure 4a) and exponentially with mean initial CDOM absorption (i.e., linearly on a logarithmic y scale; see Figure 4b). Mean initial k was positively related to C2 (peak M) fluorescence (Figure 5a), inversely to C3 (peak A; see Figure 5b), and positively to C5 (peak T; see Figure 5c). When including all optical parameters simultaneously in a mixed effects model, only SUVA254 and C5 were significant explanatory variables for k (Table 5). In the DOC concentration experiment, mean initial k was not related to mean initial SUVA254 or CDOM absorption (Figures 4d and 4e).

Figure 4.

Linear regressions (±95% confidence intervals are shown as dashed lines, and prediction intervals are shown as dotted lines) between mean initial (a) apparent first-order decay coefficientsk and specific UV absorption at 254 nm (SUVA254; y = 0.0070 (±0.0013) − 0.0007 (±0.0002) x, R2 = 0.97, P = 0.0005) and (b) the logarithm of k and CDOM absorption (250 to 500 nm; log(y) = −2.294 (±0.266) − 0.0048 (±0.0023) x, R2 = 0.89, P = 0.0045) for the samples from the clearwater (open circles) and brownwater (solid triangles) lakes. Here k was not related to DOC concentrations. (c) The analyses are based on the control samples; however, the relationships were qualitatively the same for the N + P addition samples as well. (d–f) No relationships between the respective variables were found in the DOC concentration experiment. Uncertainty bars are standard errors, which are not available for k in Figures 4a–4c since these were calculated from the random effects estimates of mixed effects models (see section 2.7).

Figure 5.

Linear regressions (±95% confidence intervals are shown as dashed lines, and prediction intervals are shown as dotted lines) between mean initial apparent first-order decay coefficientsk and fluorescence intensities with (a) component 2 (peak M; y = −0.0038 (±0.0027) + 0.0146 (±0.0058) x, R2 = 0.92, P = 0.002), (b) component 3 (peak A; y = 0.0088 (±0.0035) − 0.0190 (±0.0105) x, R2 = 0.86, P = 0.007), and (c) component 5 (peak T; y = 0.0009 (±0.0013) + 0.0126 (±0.0065) x, R2 = 0.88, P = 0.006) for the samples from the clearwater (open circles) and brownwater (solid triangles) lakes. The analyses are based on the control samples. For the N + P addition samples the relationships in Figures 5b and 5c were qualitatively the same, but the relationship in Figure 5a was not significant (P = 0.055). Uncertainty bars are standard errors, which are not available for the initial apparent k since these were calculated from the random effects estimates of mixed effects models (see section 2.7).

Table 5. Parameter Estimates and Test Statistics of the Linear Mixed Effects Model Used to Assess the Influence of Optical Parameters on the Initial Apparent Decay Coefficient ka
ParameterEstimateStandard Errort ValueUnivariate P Valuez ValueMultiplicity-AdjustedP Valueb
  • a

    C1–C5 are intensities of the fluorescence components 1–5; explanatory variables were mean centered and standardized prior to modeling.

  • b

    Multiplicity adjustment was conducted using the single-step method provided in the R package multcomp [Hothorn et al., 2008].

Intercept0.00300.000310.790.00218.68<0.01
CDOM absorption0.00020.00030.510.6450.880.898
SUVA254−0.00330.0012−2.820.067−4.88<0.01
C10.00020.00050.320.7690.560.986
C2−0.00080.0007−1.150.333−1.990.220
C3−0.00030.0007−0.480.664−0.830.919
C40.00010.00020.390.7230.680.966
C5−0.00210.0008−2.690.074−4.66<0.01

[27] Initial optical water properties, in turn, were related to theoretical water residence times (WRT). Specifically, mean initial SUVA254 (L mg C−1 m−1) decreased linearly with WRT (years) (SUVA254 = 8.64 (±1.00) − 1.95 (±0.51) WRT, R2 = 0.97, P < 0.001), and mean initial CDOM absorption (cm−1) decreased exponentially with WRT (log(CDOM absorption) = 4.21 (±0.41) − 0.38 (±0.21) WRT, R2 = 0.87, P = 0.007). Accordingly, mean initial k (d−1) was inversely related to WRT as well (k = 0.0017 (±0.0005) WRT, R2 = 0.93, P = 0.0005).

[28] Finally, we tested for relationships between initial DOC quantity, quality and k. In the six lakes, mean initial k was not related to initial DOC concentrations (Figure 4c). The mean fraction of DOC decaying at rates >0.01 d−1 (DOCL, %) decreased with increasing mean initial DOC concentrations (mg C L−1) (DOCL = 13.81 (±7.34) − 0.47 (±0.38) DOC, R2 = 0.75, P = 0.026), as did the fraction of DOC mineralized during the first incubation month (DOC1, %) (DOC1 = 9.85 (±4.63) − 0.28 (±0.24) DOC, R2 = 0.73, P = 0.030). The overall proportion of lost DOC was not related to DOC concentrations (not shown). In the DOC concentration experiment, mean initial k was not related to DOC concentrations (Figure 4f), and neither the fraction of the most reactive DOC (k > 0.01 d−1) nor the fraction of DOC mineralized in the first two incubation months correlated with DOC concentrations.

3.5. Effects of Light Exposure

[29] In the clearwater lakes, final DOC concentrations did not show an immediate response to irradiation (i.e., after 14 h of exposure to simulated sunlight) but CDOM absorption and SUVA254 declined by 32.4 ± 3.4% (P = 0.0325) and 26.5 ± 6.6% (P < 0.0001) compared to the pretreatment samples, respectively; that is, the DOC was bleached but not mineralized (Table 6). After 6 weeks of subsequent incubation, no further change in DOC concentrations, CDOM absorption or SUVA254 occurred. In the brown water of lake Björntjärn (samples from the DOC concentration experiment), final DOC concentrations decreased immediately by 9.7 ± 1.5% through irradiation (P < 0.0001), but CDOM absorption and SUVA254 did not change (Table 6). After 6 weeks of dark incubation following exposure to simulated sunlight, DOC concentrations had further declined by 19.3 ± 1.6% (P < 0.0001), together with a 26.5 ± 3.3% decrease of CDOM absorption (P < 0.0001) and a 7.4 ± 1.4% decrease in SUVA254 (P < 0.0001).

Table 6. Mean DOC Concentrations, CDOM Absorption, and SUVA254 at the End of the 3.7 Year Long Dark Incubations and After Exposure to 14 h Irradiation As Well As Irradiation With Subsequent 6 Week Long Incubationa
 Clearwater LakesBrownwater Lake (Björntjärn)
Pretreatment14 h Irradiation14 h Irradiation + 6 Weeks IncubationPretreatment14 h Irradiation14 h Irradiation + 6 Weeks Incubation
  • a

    The mean is ±SE, n = 3 for the clearwater lakes, and n = 6 for the brownwater lake; that is, samples of the DOC concentration experiment.

DOC (mg C L−1)6.23 ± 1.836.10 ± 1.836.05 ± 1.837.58 ± 2.056.83 ± 2.055.35 ± 2.05
CDOM absorption (cm−1)24.11 ± 6.0516.07 ± 6.0416.67 ± 6.0371.31 ± 22.5766.43 ± 22.5746.38 ± 22.57
Specific UV absorption at 254 nm (L mg C−1 m−1)5.41 ± 0.664.09 ± 0.664.17 ± 0.6711.95 ± 0.4312.23 ± 0.4311.30 ± 0.44

4. Discussion

4.1. Reactivity Continuum Modeling

[30] Owing to difficulties to chemically disentangle the complex array of molecules in natural OC [Dittmar and Paeng, 2009] current mathematical models generally describe the behavior of bulk OC instead of compound-specific mineralization. We chose to analyze our DOC mineralization data using a probability approach called reactivity continuum model, which accounts for the chemical diversity of DOC by assuming a continuous distribution of compounds [Boudreau and Ruddick, 1991]. To our knowledge, we compare here for the first time the reactivity continuum of DOC in different lake categories, and also first apply it to color loss of DOC. DOC loss proceeded initially faster in the clearwater compared to the brownwater lakes, brownwater CDOM absorption was lost faster than their bulk DOC and the DOC from both lake categories was similarly bioavailable within ∼3.7 years. These patterns are reflected by the probability distributions of reactivity which depict (1) a larger proportion of fast decaying DOC compounds in the clearwater compared to the brownwater lakes, (2) a shift toward higher lability of brownwater color (CDOM absorption) compared to DOC, and (3) an intersection point of both DOC reactivity distributions at a probability of ∼68%, which is similar to the DOC fraction which in both lake categories still remained after the 3.7 year incubations (see Figure 3a and Table 3). The reactivity continuum model captured differences in DOC decomposition dynamics between clearwater and brownwater lakes as well as between the bulk and colored fraction of brownwater DOC, and was consistent with our experimental data.

[31] Multiexponential models, in which DOC is expressed as the sum of discrete pools of different decomposability [Westrich and Berner, 1984], have often been used to describe mineralization in other studies. For our data an exponential model, including a labile pool and a residual pool that is not degraded, gave decay constants comparable to the initial k of the reactivity continuum model. However, the exponential model was statistically vastly inferior to the reactivity continuum model (i.e., yielded a much larger AIC). In a recent DOC decomposition study, the exponential k did not show a systematic pattern across lakes, rivers and marshes, while linear slope estimates (i.e., linear degradation rates) during the initial and a later stage in the decomposition curves did [Guillemette and Del Giorgio, 2011]. The exponential k was not systematically related to the linear slopes, therefore the authors suggested to use all parameters estimates jointly as a set of indicators for DOC bioavailability. However, this approach would not cure the general disadvantages of multiexponential models. Specifically, while dividing OC into pools may be conceptually useful [von Lützow and Kögel-Knabner, 2010], they are unrelated to measurable entities. This makes them largely theoretical constructs introduced to approximate OC heterogeneity [Bruun et al., 2010], but unlikely reflecting its natural heterogeneity [Boudreau and Ruddick, 1991], and difficult to independently validate [Benbi and Richter, 2002]. Consequently, the derived reactivity and quantity of certain pools in multiexponential models are just fit parameters rather than representing true or apparent decay constants and fractions [Middelburg, 1989].

[32] Advantageous compared to multiexponential models, the reactivity continuum model (1) does not require a priori assumptions about the number of pools, but rather assumes a continuous distribution of OC reactive types (Figure 3a); (2) captures the fact that reactivity is not constant but decreases with time (see Figures 2a and 3b); (3) allows a more parameter-parsimonious description of the data; (4) exhibits greater functional flexibility; and (5) shows greater parameter robustness as well as validity (extrapolative power) beyond an observed data range, and even on geological timescales [Boudreau and Ruddick, 1991; Boudreau et al., 2008; Bruun et al., 2010; Dolgonosov and Gubernatorova, 2010; Vähätalo et al., 2010]. In addition, the reactivity continuum model is computationally as simple as multiexponential models (i.e., a nonlinear fitting procedure) but gives considerably more information. The size of conceptual reactivity classes may easily be estimated by evaluating the probability distribution of reactivity within threshold values (Table 3). Even though, as discussed above, this is operational and somewhat arbitrary, it may provide a defined, reproducible and easily visualized summary of initial bioavailability patterns. Currently, C turnover models usually conceptualize DOC as one or a few functionally homogeneous compartments which decompose following first-order kinetics. We suggest that, similarly as recently discussed for soil OC modeling [Manzoni and Porporato, 2009], DOC reactivity transport modeling would become more flexible and realistic by describing k as a random variable which decreases over time instead of basing it on deterministic and constant kinetic compartment structures.

4.2. Factors Regulating DOC Bioavailability

[33] We aimed to control for the extrinsic factors that could limit DOC mineralization. Oxygen availability was ample, and inorganic nutrients (N + P) had a minor stimulating effect on DOC mineralization only in the brownwater lakes (Figure 2a). We found no evidence for deteriorating decomposition capacity of microbes throughout the incubations, since (1) reinoculation with a fresh microbial community after 3.7 years did not stimulate decomposition and (2) a spike of labile OC (glucose) added after 2 to 4 months was rapidly mineralized (Figure A1). Since the added glucose did not stimulate degradation of aged DOC compared to the control there was no evidence for “priming,” a process which is commonly described from soil systems and has been observed in a few aquatic experiments [Guenet et al., 2010]. Taken together, the selective removal of intrinsically more labile DOC was likely the most prominent cause of mineralization to slow down in the laboratory incubations.

[34] Owing to the larger proportion of more labile compounds in the clear water, mineralization proceeded initially faster than in the brown water. The initial increase in SUVA254 (Figure 2c) indicated that these most labile fractions in the clearwater lakes were mostly noncolored as typical for autochthonous DOC [Steinberg, 2003]. The importance of autochthonous DOC was corroborated by the positive correlation of the initial kwith peak T fluorescence as proxy for algal-derived, proteinaceous OM [Stedmon and Markager, 2005b], which has similarly been observed in earlier studies [Fellman et al., 2008; Guillemette and Del Giorgio, 2011]. The larger proportion of peak M fluorescence and its stimulating effect on initial k gives further support, since this component is related to plankton productivity [Coble et al., 1998] and accumulates during microbial processing of algal-derived DOC [Stedmon and Markager, 2005b]. Peak A, which retarded initial k in our study similar as in earlier ones [Fellman et al., 2008; Guillemette and Del Giorgio, 2011], was the dominant fluorescence component in soil solutions from temperate forest and wetlands [Fellman et al., 2008], suggesting that terrestrial compounds were more important in the brownwater than the clearwater lakes.

[35] We add to the evidence that DOC quality is a better predictor of initial bioavailability than DOC quantity. Positive correlations between DOC concentrations and labile DOC proportions have been reported [Søndergaard and Middelboe, 1995], but no relationship was found in another meta-analysis [Del Giorgio and Davis, 2003] or between DOC quality parameters and concentrations [Jaffé et al., 2008]. Accordingly, DOC concentrations constrained bacterial growth only at levels lower than typical for most natural lakes [Eiler et al., 2003]. DOC concentrations could also not predict the initial k, neither in the lake samples in which concentrations covaried with quality differences (Figure 4c) nor in the DOC concentration experiment in which quality was kept constant (Figure 4f). DOC quality, as reflected by spectral properties, successfully predicted DOC bioavailability, corroborating findings from other studies. For instance, SUVA254 correlated positively with percent DOC aromaticity [Kalbitz et al., 2003; Weishaar et al., 2003] and negatively with (1) portion of oxygen-containing functional groups [Kalbitz et al., 2003], (2) bacterial production and growth efficiencies [Berggren et al., 2009], (3) proportion mineralized DOC in 3 months [Kalbitz et al., 2003], and (4) bioavailable DOC determined by several methods [McDowell et al., 2006]. Since UV-vis absorbance and fluorescence spectral properties integrate intrinsic DOC molecular properties they are good proxies for initial DOC mineralization rates.

[36] DOC that remained throughout 3.7 years of dark incubation, and degraded very slowly toward the end of that period, was rapidly lost from the brownwater samples upon exposure to simulated sunlight while this was not the case for the clearwater lake DOC. As indicated by the fluorescence properties, the clearwater lakes were more strongly influenced by autochthonous DOC, and they probably had also been subject to more extensive photochemical decomposition of allochthonous DOC prior to incubation [Molot and Dillon, 1997]. Our study therefore supports that DOC source is an important predictor for photochemical reactions, with allochthonous DOC being particularly sensitive to photodegradation [Moran and Covert, 2003; Obernosterer and Benner, 2004]. Moreover, the responses to light detected at the end of the incubation suggest that extensive microbial processing of brownwater DOC (with a high initial proportion of allochthonous DOC) does not neutralize the capacity for photochemical transformation.

4.3. Landscape-Scale Comparison and Integration

[37] In Swedish catchments, DOC loss during inland water passage increases with water residence times up to ∼2–5 years, and takes an asymptotic course toward longer times [Algesten et al., 2003]. The importance of time in influencing DOC mineralization was also evident in our study, specifically (1) even though the modeled apparent first-orderk in the brownwater lakes fell below that in the clearwater lakes during the initial 5 months (Figure 3b), the proportional DOC mineralization had caught up in the brownwater lakes within 2 more years (Figure 2a); (2) water residence times correlated with initial optical water properties which in turn were related to the initial k (see section 3.5); and (3) ratios of water residence to initial apparent DOC residence times were very small in the brownwater lakes (Table 4), meaning that the DOC degradation potential cannot be realized during its residence in the lake. The DOC fraction microbially mineralized in our laboratory bioassays was comparable to that mineralized in about 40% of the major Swedish catchments, and fell below the in situ estimates by maximally ∼25% in the remaining studied watersheds (Figure 2a) [Algesten et al., 2003]. The extensive long-term dark decomposition, particularly of colored DOC, suggests that a substantial fraction of landscape-scale mineralization could theoretically be explained by microbial activity independent of DOC photodecomposition, in contrast to suggestions that photochemistry place a crucial role for the loss of DOC in lakes [Molot and Dillon, 1997; Cole, 1999; Farjalla et al., 2009]. We are still lacking understanding of the contribution of light to inland water DOC cycling.

5. Conclusions

[38] Allochthonous DOC is generally considered more recalcitrant than autochthonous DOC, as observed in experiments at short timescales of days to weeks [Søndergaard and Middelboe, 1995]. However, we reveal by reactivity continuum modeling that the apparent first-order decay coefficientsk of both lake categories converged within just a few months of exposure to microbial decomposition. Hence, initial differences in bioavailability may not exert a strong control on the mineralization of DOC at the landscape scale, considering that water residence times are typically in the order of years. Consequently, brownwater DOC with a higher proportion of fresh allochthonous components may, on timescales exceeding a couple of months, be effectively even more labile than clearwater DOC with a strong influence of autochthonous components, considering that it was similarly bioavailable at these timescales but in addition also more sensitive to photomineralization. This is in contrast to the current paradigm that substantial mineralization of allochthonous DOC is tightly constrained by solar irradiation.

Appendix A

[39] We conducted 3.7 year long laboratory incubations using samples from three clearwater and three brownwater lakes in south-central Sweden. The incubations were conducted in the dark at 20°C with ample oxygen, and the loss of dissolved organic carbon (DOC) concentrations over time was assessed. After a period of incubation, we added fresh DOC to subsample series (0.5 mg glucose-C L−1after 3.5 months to the clearwater lake samples, and 2 mg glucose-C L−1after 2 months to the brownwater lake samples), and measured DOC concentrations on the same days as the control up to month 43 for the clearwater lakes, and up to month 21 for the brownwater lakes. In the clearwater lakes, the added glucose-DOC was rapidly mineralized but final DOC concentrations remained higher in the glucose addition compared to the untreated samples (Figures A1a–A1c), suggesting that glucose hampered decomposition or glucose-derived OC remained throughout the incubations. In the brownwater lakes, DOC concentrations did not differ between the control and glucose-addition series (Figures A1d–A1f). Since the added glucose did not stimulate mineralization of aged DOC compared to the control there was no indication of “priming,” a process which is commonly described for soils and has been observed in a few aquatic experiments [Guenet et al., 2010].

Figure A1.

Mean (±, n = 3) time series of DOC (subscripts denote day of the incubation) during dark incubation of samples from clearwater lakes (a) Lilla Sångaren, (b) Ljustjärn, and (c) Valloxen, as well as from brownwater lakes (d) Lumpen, (e) Siggeforasjön, and (f) Stensjön, with control (open circles) and glucose addition samples (solid triangles). Gray vertical lines mark endings of years after the start of the experiment.

Acknowledgments

[40] We thank Jan Johansson for his skillful advice and help during field sampling and laboratory analyses; Hannes Peter, Cristian Gudasz, and Sebastian Sobek for their advice and assistance during measurements; and the editors and two anonymous referees for their thorough and constructive suggestions. The study was funded by the “Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning” (FORMAS), and was part of the research environment “The Color of Water.”