Corresponding author: F. Guillemette, Groupe de Recherche Interuniversitaire en Limnologie, Département des Sciences Biologiques, Université du Québec à Montréal, CP 8888, Succ. Centre-Ville, Montréal, QC H3C 3P8, Canada. (firstname.lastname@example.org)
 It has often been hypothesized that the dissolved organic carbon (DOC) pool of algal origin in lakes is more bioavailable than its terrestrial counterpart, but this hypothesis has seldom been directly tested. Here we test this hypothesis by tracking the production and isotopic signature of bacterial respiratory CO2 in 2 week lake water incubations and use the resulting data to reconstruct and model the bacterial consumption dynamics of algal and terrestrial DOC. The proportion of algal DOC respired decreased systematically over time in all experiments, suggesting a rapid consumption and depletion of this substrate. Our results further show that the algal DOC pool was used in proportions and at rates twice and 10 times as high as the terrestrial DOC pool, respectively. On the other hand, the absolute amount of labile terrestrial DOC was on average four times higher than labile algal DOC, accounting for almost the entire long-term residual C metabolism, but also contributing to short-term bacterial C consumption. The absolute amount of labile algal DOC increased with chlorophyll a concentrations, whereas total phosphorus appeared to enhance the amount of terrestrial DOC that bacteria could consume, suggesting that the degradation of these pools is not solely governed by their respective chemical properties, but also by interactions with nutrients. Our study shows that there is a highly reactive pool of terrestrial DOC that is processed in parallel to algal DOC, and because of interactions with nutrients, terrestrial DOC likely supports high levels of bacterial metabolism and CO2 production even in more productive lakes.
 The organic C present in the bulk dissolved organic carbon (DOC) pool of lakes exceeds the amount present in other detrital pools and in the biomass of aquatic organisms by over an order of magnitude [Prairie, 2008]. Consequently, any process that may induce even small variations in this pool of carbon and energy has the potential to greatly influence the functioning and the role of lakes both at the local and landscape levels. The consumption and degradation of DOC by heterotrophic bacteria is one of the main processes influencing the DOC pool of lakes, with profound ecological and biogeochemical consequences. For example, bacterial DOC consumption contributes to the sustained widespread net heterotrophy observed in many freshwater ecosystems [Duarte and Prairie, 2005; Karlsson et al., 2007, McCallister and del Giorgio, 2008] and serves to transfer a portion of this organic pool to higher trophic levels of aquatic food webs [Azam et al., 1983; Berggren et al., 2010b].
 The bulk DOC pool of lakes is composed of a highly heterogeneous and complex mixture of organic compounds with different chemical attributes and availability to bacterial metabolism [Benner, 2003]. Because bacteria preferentially utilize the more labile components first while leaving the more recalcitrant (or unreactive) molecules behind [Mateles and Chian, 1969; Middelburg, 1989], the overall bioavailability of bulk DOC typically declines over time when isolated from fresh inputs, and this pattern can be modeled as a reactivity continuum (RC) [Boudreau and Ruddick, 1991; Vähätalo et al., 2010; Koehler et al., 2012]. The degradation dynamics of the bulk DOC is thus the direct reflection of the contribution of various pools that coexist within this bulk DOC and which differ in terms of bioavailability and degradation dynamics [Westrich and Berner, 1984]. Identifying the intrinsic properties and dynamics of these major pools will undoubtedly improve our understanding of the controls on overall DOC decomposition in natural aquatic ecosystems.
 There is both theoretical and empirical evidence that the consumption dynamics of specific DOC pools should differ based on their molecular size [Chrost and Faust, 1983; Amon and Benner, 1996], chemical composition [Weiss and Simon, 1999; Amon et al., 2001], and elemental stoichiometry [Sun et al., 1997; Hunt et al., 2000]. In turn, these properties are presumably dependent on the respective origin of these specific DOC pools [Benner, 2003]. In lakes, DOC originates from local primary production [Bertilsson and Jones, 2003] and is also imported from the terrestrial environment [Aitkenhead-Peterson et al., 2003]. In this regard, it has traditionally been assumed that algal DOC should be more readily consumed by bacteria than terrestrial DOC, owing to the presence of simple, low molecular weight carbon compounds that turn over very rapidly in the former [Sundh, 1992; Chen and Wangersky, 1996], and the presence of more complex and aromatic compounds [McKnight and Aiken, 1998] that have undergone previous aging in soils in the latter [Berggren et al., 2009]. These assumptions lead to a scenario wherein algal DOC should be preferentially degraded on short time scales, followed by a slower utilization of terrestrial DOC pool on longer time scales. Recent reports challenge this scenario, however, by showing a fast bacterial utilization of a highly labile pool of low molecular weight compounds of terrestrial origin in boreal streams [Ågren et al., 2008; Berggren et al., 2010a], leading to a second scenario wherein terrestrial DOC could fuel significant bacterial C consumption on both short and longer time scales [Holmes et al. 2008; Ward et al., 2013].
 In practice, however, tracking the bacterial consumption dynamics of specific pools within complex mixtures has been a major challenge, and the above scenarios are based on either a priori judgment of the chemical attributes of each sources, or on experiments that have assessed C consumption of these pools in isolation. There have been a handful of studies that have attempted to trace the bacterial utilization of different DOC sources in lake water [Kritzberg et al., 2004; Karlsson et al., 2007; McCallister and del Giorgio, 2008], based on the isotopic signature of metabolic products (bacterial respiratory CO2 or biomass), which have confirmed that bacteria consume both terrestrial and algal DOC when grown on natural lake DOC. These studies have provided important insight into the contribution of each source to bacterial metabolism but do not allow one to reconstruct the actual degradation dynamics of terrestrial and algal DOC since the differential utilization of these sources within bulk DOC has never been followed over time. As a consequence, the consumption dynamics of these major C categories in natural complex mixtures of lake DOC remains to be described.
 The aim of our study was to describe the degradation dynamics of algal and terrestrial DOC within natural bulk DOC pools across a diversity of northern lake types, and to link the resulting patterns of degradation to environmental factors. The main challenge associated with this question is the ability to differentiate, at relevant time scales, the consumption of algal and terrestrial C that occurs simultaneously within a complex natural DOC mixture. In this study, we have used the approach developed by McCallister et al. [2006a], based on tracking the production and isotopic signature of bacterially produced respiratory CO2, and have used this as a proxy to reconstruct the degradation of algal and terrestrial DOC over a period of 6, 13, and 20 days. Based on the respiratory CO2 isotopic signature and on a coupled two-source (algal and terrestrial) mixing model, we then apportioned the amount of CO2 originating from the degradation of algal and terrestrial DOC at each time interval. Finally, we reconstructed and modeled the complete degradation dynamics of these two DOC pools. Our study shows that algal and terrestrial DOC follow very different degradation dynamics in lake water: algal DOC was essentially processed over short time scale whereas terrigenous material was degraded over both short and longer time scales, but demonstrated slower overall degradation kinetics. In addition, phosphorus played a key role in modulating the amount of bioavailable terrestrial DOC, and more importantly, the degradation kinetics of terrestrial DOC, suggesting that the biological degradation of these sources in lakes is not simply a function of their respective concentrations and intrinsic properties, but is also strongly influenced by nutrient dynamics.
2 Materials and Methods
2.1 Study Lakes and Sampling Scheme
 We sampled five northern temperate lakes located in the Eastern townships region of south-eastern Québec (45.24°N, 72.12°W), Canada, between 2004 and 2007, and sampled an additional lake (Lac à la Truite) in 2006 located in the Laurentian region north of Montréal (46.01°N, 74.15°W). The watersheds of these lakes are characterized by temperate mixed-wood forest and low density of human population, and are underlain by the sedimentary St. Lawrence Lowlands (Eastern townships) or by the Canadian Shield bedrock (Laurentian). The sampled water bodies represent a moderate gradient in lake productivity, both in terms of chlorophyll a (Chl a) and nutrient levels, and also of DOC content and water color (Table 1). As a result, the influence of autochthonous and allochthonous C to bacterial metabolism in these lakes differed [Guillemette and del Giorgio, 2012].
Table 1. Limnological Characteristics of the Six Northern Temperate Lakes Sampleda
aAbbreviations: Chl a, chlorophyll a; TP, total phosphorus; TN, total nitrogen; A440, colored dissolved organic matter absorption at 440 nm; DOC, dissolved organic carbon; LA, lake surface area; WRT, theoretical water retention time; MD, mean depth; nd, not determined. Mean biological and chemical data reported with ± SD, N = 2.
À la Truite 2006
0.9 ± 0.1
6.3 ± 0.4
0.13 ± 0.02
3.20 ± 0.01
0.17 ± 0.08
5.7 ± 1.1
15.0 ± 1.0
0.32 ± 0.04
5.54 ± 0.10
1.70 ± 0.15
1.5 ± nd
11.2 ± 0.8
0.27 ± 0.02
6.90 ± nd
2.23 ± 0.13
3.0 ± 0.3
7.5 ± 1.0
0.28 ± 0.02
5.06 ± 0.01
2.01 ± 0.15
2.9 ± 0.3
2.7 ± 0.1
0.24 ± 0.02
4.37 ± 0.06
1.31 ± 0.20
2.5 ± 1.7
6.5 ± 0.4
0.27 ± 0.05
4.78 ± 0.01
1.07 ± 0.18
 Integrated (<3 m) water samples (60 L) from the epilimnion of these lakes were collected using a plastic hose mounted to a diaphragm pump, stored in acid-washed (10% HCl) polycarbonate bottles, and kept cool in the dark prior to processing (<3 h). In the laboratory, 8 L of lake water was set aside for the determination of ambient nutrient and chlorophyll a concentrations. The remaining water was filtered through precombusted Pall AD glass fiber filters (3.0 µm; Port Washington, NY, USA) to isolate the bacterial communities from other planktonic components, and to serve as an inoculum for the different bacterial respiratory CO2 recovery incubations described below. Pretests showed that 92.6 ± 4% of the initial bacterial abundance remained in the filtrate while keeping bacterial grazers (i.e., flagellates) to a minimum level (<10% initial [Guillemette and del Giorgio, 2011]). We further filtered 25 L of this 3.0 µm filtered water through a Gelman filter capsule (Pall; 0.2 µm) to prepare the bacteria-free (<1% initial) lake water media used in the incubation experiments. Samples for DOC analysis were taken from this filtered water and poisoned with 5 N sulfuric acid.
 In the field, zooplankton samples, which were used to approximate the isotopic signature of phytoplankton (see below), were collected in parallel to lake water samples, by pumping a large amount of lake water (200 L) through a 50 µm mesh size net. Back in the laboratory, organisms were stored overnight at 4°C to void their gut contents. The following day, we handpicked over 100 individual Cladocerans, represented by the genus Daphnia (Daphnia mendotae and Daphnia catawba) and Copepods, dominated by Diacyclops bicuspidatus, Mesocyclops edax, and Leptodiaptomus minutus, and the organisms were placed in smooth-walled tin capsules, fumed with 10% HCl, and dried overnight at 45°C pending stable isotopic analysis.
2.2 Bacterial Respiratory CO2 Recovery Experiments
 We used a modified version of the procedure described in McCallister et al. [2006a] to quantitatively recover the respiratory CO2 produced by bacteria grown in culture medium consisting of 0.2 µm filtered lake water. The procedure involves an initial acidification and helium bubbling of the culture medium to remove background dissolved inorganic carbon (DIC) concentration (<1% of initial), the neutralization and oxygenation of the acidified lake water to initial conditions, the inoculation with the natural lake bacterial assemblage and incubation of the regrowth medium in an airtight, 20 L glass system at room temperature (20°C ±1), and finally, the collection of the bacterially produced CO2 with dedicated cryogenic traps after several days of incubation by subsequent acidification and helium bubbling.
 Our project involved recovering the CO2 produced by bacteria at several points during a time course, so as to be able to reconstruct the DOC consumption dynamics. Rather than setting up several parallel systems with the same water and harvesting them for CO2 at different times, we chose to reset the same system after the initial CO2 harvest, in order to recover the signature of the gas actually produced within that period rather than the cumulative signature, which would be more difficult to interpret. We repeated this reset procedure twice, such that we had three time points for CO2 production and its isotopic signature for each sample, in addition to the initial time point. While the initial incubation was performed over 6 days, the subsequent incubations (experiments 2 and 3) were extended for 7 days before performing a second respiratory CO2 harvest in order to account for declining rates of bacterial metabolism in the samples along the time course. The entire time course thus covered a period of 20 days.
 At each subsequent incubation, we collected 1.4 L of the incubation medium containing the bacteria grown in the previous incubation before harvesting the respiratory CO2 to serve as an inoculum (1 L) for the following incubation, and to perform various measurements (e.g., bacterial abundance, production, DIC and DOC concentrations, and nutrients; 0.4 L). Upon the complete collection of the respiratory CO2, the incubation medium was reneutralized and oxygenated before the injection of the inoculum. The inoculum was prepared according to McCallister et al. [2006a] with slight modifications. We first used a Pellicon Mini-ultrafiltration unit (Millipore, Billerica, MA, USA) with a 1000 kDa cartridge to reduce 1 L of the collected incubation medium down to 100 mL final volume. At this stage, a 1.5 mL sample of the filtrate was taken for bacterial abundance to ensure the effectiveness of bacterial concentration, which was >70%. Then, the bacterial inoculum was reduced to 30 mL on an acid-cleaned (10%) 0.2 µm Anodisc 47 mm filter (Whatman, Springfield Mill, UK). The concentrate volume was subsequently exchanged with a 0.1 L sample of DIC-free lake water also collected from the incubation vessel in order to avoid introducing atmospheric CO2 during the inoculation and to remove any DOC that could have been concentrated during the ultrafiltration process. Finally, a last 30 mL sample was collected from the incubation vessel, exchanged with the inoculum, and reintroduced in the incubation vessel to start the experiment. Thus, exactly 1.5 L was removed from the incubation vessel at the end of each experiment (i.e., after the first and second experiments), so that at the beginning of the first, second, and third experiments, there was 20 L, 18.5 L, and 17 L of lake water present in the incubation vessel, respectively. Note that the initial inoculum used to start the first experiment was prepared according to this procedure, but using the 3.0 µm filtered lake water instead.
 The traps containing bacterial respiratory CO2 collected on days 6, 13, and 20 were mounted to a vacuum extraction line to isolate the gas from residual moisture, and the purified CO2 was quantified manometrically (Baratron, MKS Instruments, Andover, MA, USA; 0.5 µmol sensitivity) and transferred to breakseals (i.e., torch-sealed Pyrex tubes; Corning, Corning, NY, USA) pending isotopic analysis. In addition, we followed the activity of the incubated bacterial communities over the course of the experiments by measuring bacterial production (BP) every 1–2 days using the 3H-leucine incorporation technique as detailed in del Giorgio et al. .
2.3 Isotopic and Chemical Analyses
 The δ13C isotopic signature of zooplankton was determined using a Finnigan MAT (Bremen, Germany) Deltaplus dual-inlet continuous flow isotope ratio mass spectrometer (IRMS) with online sample combustion at the G.G. Hatch lab, Ottawa University (Ottawa, Canada). A subset of zooplankton samples were assessed for analytical precision and run in duplicate; the relative standard deviation was <0.3%. Respiratory CO2 breakseals were transferred into Exetainers and analyzed for δ13C with a continuous flow GasBench peripheral (Thermo Finnigan) interfaced to an Isotope Ratio Mass Spectrometer Delta XP (Thermo Finnigan), which has an analytical precision of 0.10‰ (G.G. Hatch Lab). DOC concentration and δ13C determination were performed on a modified 1010 TIC TOC analyzer (O.I. Analytical, College Station, TX, USA) coupled to a Finnigan MAT DeltaPlus IRMS with a Conflo III continuous flow interface (Thermo Finnigan) as described in St-Jean . Stable isotope values are reported hereafter in standard δ notation as:
where R is 13C:12C.
 Total phosphorus concentrations were determined in unfiltered lake water using the standard blue molybdenum colorimetric method. Total nitrogen was measured as nitrate after alkaline persulfate digestion using the cadmium method [Cattaneo and Prairie, 1995]. Ambient chlorophyll a concentrations were measured spectrophotometrically from ethanol extracts. Colored dissolved organic matter absorption coefficient (A440) was calculated by dividing the optical absorbance at 440 nm by the path length in meters and multiplying by 2.303 [Cuthbert and del Giorgio, 1992].
2.4 Algal and Terrestrial Contribution to Bacterial Respiration
 The relative contribution of algal and terrestrial C to bacterial respiratory CO2 was estimated by a two-source mixing model:
where δ13CCO2 corresponds to the isotopic signature of bacterial respiratory CO2, and f1 and f2 are the relative contributions of terrestrial (CTerrestrial) and algal (CAlgal) sources to respiratory CO2, respectively. Uncertainties in the algal and terrestrial C contributions to respiratory CO2 (expressed as ± SE) were constrained using IsoError [Phillips and Gregg, 2001], which accounts for variations in the isotopic signature of the algal and terrestrial end-member as well as in the mixture (i.e., respiratory CO2). The terrestrial end-member was set to the commonly accepted value of −27.0‰ typically found for terrestrial C3 plants [Lajtha and Marshall, 1994; Boschker and Middelburg, 2002], but we further allowed for a small deviation (SD = 0.2‰) from this assumption, in line with our own observations in a forested stream from our study site (−27.2 ± 0.1‰, N = 2). The algal δ13C end point was constrained using the isotopic signature of zooplankton according to recent studies [Karlsson et al., 2007; Marty and Planas, 2008; McCallister and del Giorgio, 2008]. We further assumed a 16% terrestrial C content in zooplankton biomass, which corresponds to the mean zooplankton allochthony values reported for other Canadian temperate lakes (N = 25) [Mohamed and Taylor, 2009] and used the range (9–23%) reported by these authors as a measure of uncertainty. Finally, since we could only perform one measurement of the respiratory CO2 isotopic signature, we used the standard deviation (0.2‰) of a past duplicate measurement performed in the same lake [McCallister et al., 2006a] as a measure of variation for the mixture in the calculation.
2.5 Modeling Algal and Terrestrial C Consumption
 To assess the individual degradation dynamics of algal and terrestrial DOC, we first reconstructed the consumption curve of these different DOC pools as follows. We converted the amount of CO2 recovered at the end of each time interval into DOC concentration units (mg C L−1) by dividing the mass of CO2 (i.e., mass carbon) collected by the volume of water incubated (i.e., 20 L, 18.5 L, and 17 L, for experiments 1, 2, and 3, respectively). We then subtracted the quantity of DOC consumed previously (in terms of concentration) from the starting DOC concentration of each time step. We finally apportioned the initial DOC concentrations and the concentrations of DOC consumed at each time step into algal and terrestrial C using the δ13C isotopic signature of DOC and CO2 and their coupled mixing models (the mixing model described above for respiratory CO2 was also used for DOC), respectively. The amount of DOC consumed within each pool was calculated as the initial concentration minus final, and the fraction of DOC consumed, as the difference between the initial and final divided by the initial.
 Following the above reconstruction procedure, we then applied a reactivity continuum (RC) model, based on the gamma distribution, to the algal and terrestrial DOC consumption curves in order to derive first-order decay coefficients, k, for these two pools. The theoretical basis of the model is that the overall decay of DOC corresponds to a continuous distribution of an infinite number of reactive types, which can be described by a variable function of the decay coefficient, k [Boudreau and Ruddick, 1991]. Here we used a similar approach as in Koehler et al.  and modeled the relative decrease in algal or terrestrial DOC concentrations over time (DOCt/DOC0) as:
where a is the average lifetime of the more reactive components (days) and v is the shape of the distribution near k = 0 (dimensionless), which corresponds to the relative predominance of the most refractory compounds. At v = 1, every compound has a similar reactivity, whereas a v approaching zero indicates the predominance of refractory compounds. In the gamma distribution, the initial apparent first-order decay coefficients are described by the two model parameters (a and v) and calculated as v/a (day−1). The decrease in decay coefficients over time was calculated as v/a + t [Boudreau et al., 2008]. Model parameters were estimated for the different algal and terrestrial DOC time courses using the nonlinear regression tool in JMP 10 statistical software (SAS Institute, Cary, NC, USA), which uses a Gauss-Newton algorithm to perform model fit.
 Mean differences between the amount (and proportion) of algal and terrestrial DOC consumed, and between the initial apparent first-order degradation coefficients calculated for algal and terrestrial DOC pools were assessed using the Wilcoxon rank-sum test as the data violated the normality distribution assumption. We explored potential relationships between model output parameters and environmental variables using simple linear regression models. All statistical analyses were considered significant if P < 0.05 and were performed using the JMP 10 statistical software (SAS Institute).
3.1 Patterns in Algal and Terrestrial DOC Degradation Dynamics
 Bacterial DOC degradation resulted in an average production of 0.56 ± 0.14 mg of CO2 (C mass) during the first experiment, declining to an average of 0.24 ± 0.18 mg by the final time point (Table 2). BP measurements carried out during the incubations confirmed that there was a roughly 2 day lag phase at the beginning of each new experiment (hereby defined as <10% of maximum BP, Figure S1). This pattern has been observed in the past and attributed to the lack of dissolved inorganic C in the incubation system, which may inhibit some metabolic pathways involving the anaplerotic B-carboxylation reaction [McCallister et al., 2006a; Overbeck, 1979]. Thus, we removed 2 days from each time interval upon which respiratory CO2 was allowed to accumulate in the incubations to more accurately reflect the actual period of time that DOC was consumed in our incubations. In all subsequent sections, including the modeling of DOC consumption, we use these corrected time intervals, i.e., 4, 9, and 14 days.
Table 2. Bacterial Respiratory CO2 δ13C Isotopic Signature Measured in the Different Lake Water Incubations Over Short and Long Terma
bMean ± SD δ13C of the algal end-member derived from zooplankton isotopic signature assuming a 16% ± 10% terrestrial C content in biomass according to the range reported for other Canadian temperate lakes [Mohamed and Taylor, 2009] (see section 2.4 for details).
cThe terrestrial end-member was set to the commonly accepted value of −27.0‰ [Boschker and Middelburg, 2002; Lajtha and Marshall, 1994], and we further assume a 0.2‰ deviation based on the isotopic signature of a forested stream entering lake Fraser (δ13C of −27.18‰ ± 0.01, N = 2).
À la Truite 2006
−31.8 ± 0.7
−27.0 ± 0.2
−32.0 ± 0.6
−27.0 ± 0.2
−33.1 ± 0.7
−27.0 ± 0.2
−33.1 ± 0.7
−27.0 ± 0.2
−32.0 ± 0.6
−27.0 ± 0.2
−31.4 ± 0.5
−27.0 ± 0.2
 The isotopic analysis of the recovered gas indicated a marked change in the δ13C isotopic signature of the respiratory CO2 collected over time. For the initial incubation, the δ13C of respiratory CO2 ranged from −32.1‰ to −27.9‰ across experiments (Table 2). The isotopic signature of respiratory CO2 was neither as depleted as the algal δ13C nor as enriched as the terrestrial isotopic signature (Table 2), suggesting that at the beginning of the incubations, bacteria were consuming a mix of algal and terrestrial C. Results from the mixing model indicated that between 18% and 47% of bacterial respiratory CO2 was derived from algal C at the initial stage (Figure 1). The proportion of algal C consumed did not remain constant over time, however, as the isotopic signature of respiratory CO2 systematically shifted toward more enriched δ13C values as the incubations progressed (Table 2). There was in fact a rapid and consistent decline in the % algal C respired over time, and by the final time point, the DOC respired appeared to be almost entirely terrestrial, regardless of the initial proportion of algal DOC consumed (Figure 1). The only exception to this pattern was observed in lake Simoneau, where a significant proportion of algal C was still respired by the end of the incubation. This general pattern we observed appeared to be robust despite the uncertainties in the assumed δ13C values of the algal and terrestrial end-member (Table 2), as these uncertainties resulted only in small variations in the estimated contributions of algal C to bacterial respiratory CO2 across experiments (5%–7%, average = 6%; Figure 1).
 Following the reconstruction of the algal and terrestrial DOC degradation time courses (see section 2.5 for details), we observed systematic differences in both the amount and the proportion of algal and terrestrial DOC consumed during the incubations, as well as in their respective decay coefficients, k. The labile algal DOC pool (0.012 ± 0.008 mg L−1) was on average significantly smaller than that of the terrestrial DOC pool (P = 0.005), which averaged 0.044 ± 0.007 mg L−1 (Figure 2a). However, the proportion of algal C that was bioavailable within the time frame of our incubations was on average twofold greater (2.1 ± 0.8%; P = 0.045) than that of terrestrial origin (1.0 ± 0.3%; Figure 2b). The shifts in the amount of algal and terrestrial DOC consumed, estimated from the reconstructed time courses, were in all cases well described (R2 > 0.99; see Figure S2) by the RC model. According to the model, the average lifetime of the more reactive constituents (parameter a) of the algal DOC pool was 1 order of magnitude shorter on average than that of terrestrial DOC (P < 0.001), while the v parameter (relative preponderance of the more refractory compounds) did not show any systematic variation between algal and terrestrial DOC (P = 0.69; see Table 3 for model parameters values). The initial apparent first-order decay coefficients were on average 1 order of magnitude higher for the algal substrate (0.024 ± 0.012) compared to its terrestrial counterpart (0.0023 ± 0.0001; P = 0.005; Figure 2c). However, the initial decay coefficients of the algal DOC pool rapidly decreased overtime (within a few days) and converged toward the terrestrial degradation coefficients in all incubations (Figure 2d), suggesting a rapid depletion of the more labile algal compounds when new inputs are cut off.
Table 3. Estimated Parameters of the Reactivity Continuum Models Fitted to the Algal and Terrestrial DOC Degradation Time Coursesa
aAbbreviations: nd, not determined. All fitted models had coefficients of determination (R2) > 0.99 (see Figure S2). Model parameters reported with ± SE.
bAverage lifetime of the more reactive compounds [Boudreau and Ruddick, 1991].
cRelative preponderance of the more refractory compounds [Boudreau and Ruddick, 1991].
dMean initial apparent first-order decay coefficient calculated as v/a (see section 2.5 for details).
À la Truite 2006
0.46 ± 0.30
0.0095 ± 0.0019
0.021 ± 0.014
1.76 ± 0.13
0.0057 ± 0.0002
0.0033 ± 0.0003
0.35 ± 0.25
0.0079 ± 0.0017
0.023 ± 0.001
14.11 ± 0.30
0.0170 ± 0.0003
0.0012 ± 0.0001
0.07 ± nd
0.0017 ± nd
0.023 ± nd
5.27 ± nd
0.0081 ± nd
0.0015 ± nd
0.87 ± 0.37
0.0204 ± 0.0033
0.023 ± 0.001
1.86 ± 0.04
0.0035 ± 0.0001
0.0019 ± 0.0001
0.74 ± 0.10
0.0061 ± 0.0003
0.008 ± 0.001
0.45 ± 0.05
0.0014 ± 0.0001
0.0033 ± 0.0004
0.08 ± 0.09
0.0036 ± 0.0009
0.045 ± 0.051
1.47 ± 0.064
0.0042 ± 0.0001
0.0029 ± 0.0001
3.2 Links Between Degradation Dynamics and Environmental Gradients
 Our results show not only systematic differences in the average degradation dynamics between algal and terrestrial DOC pools, but also variations in the size and degradation dynamics within each of these pools along environmental gradients. Not surprisingly, we found that the absolute amount of labile algal DOC increased as a function of Chl a concentrations (R2 = 0.97, N = 6, P < 0.001; Figure 3a), although the rate at which this DOC was degraded (k) did not vary systematically along the same gradient (Figure 3b), nor with any of the other environmental variables tested (see Table 1 for the list of variables). Similarly, there was a positive relationship between the absolute amount of labile terrestrial DOC and TP (R2 = 0.79, N = 6, P < 0.05; Figure 4a), but in this case, there was also a negative relationship between its associated initial apparent first-order decay coefficients and TP (R2 = 0.81, N = 6, P < 0.05; Figure 4b). In fact, TP appeared to concomitantly increase the average lifetime of the more labile compounds of terrestrial DOC (R2 = 0.99; N = 6, P < 0.001; Figure S3a) and decrease the predominance of the more refractory compounds (R2 = 0.88, N = 6, P < 0.01; Figure S3b) across lakes.
 The biological recalcitrance of terrestrial C, as compared to its algal counterpart, is a longstanding assumption based on the notion that this substrate is intrinsically more structurally complex, having undergone significant decomposition in soils before reaching downstream water bodies [Breger, 1970; Hobbie, 1988]. Surprisingly, this simple and widely assumed hypothesis has seldom been empirically tested, likely because of the challenges associated with disentangling the C sources that are consumed within complex natural DOC mixtures. Here we used the isotopic signature of bacterial respiratory CO2 to trace the sources of DOC being degraded and were able to recreate and describe the degradation dynamics of the algal and terrestrial DOC pools of lakes, and thus to directly test this fundamental hypothesis.
4.1 Degradation Dynamics of Algal DOC
 Our results show that algal and terrestrial DOC followed very different degradation dynamics, with the algal pool being more rapidly degraded compared to its terrestrial counterpart (Figure 2c). These results confirm previous work suggesting that algal DOC is a preferred substrate for bacteria [Kritzberg et al., 2004; Kritzberg et al., 2005; McCallister et al., 2006b] and suggest that algal DOC is a major component of the highly reactive pool within bulk DOC. The amount of C that can be extracted from bulk DOC by bacteria has been shown to also increase with lake trophy [Ostapenia et al., 2009], suggesting that as the relative proportion of algal versus terrestrial DOC increases, the amount of labile C within bulk DOC also increases. Our results provide evidence for the mechanistic underpinning to this pattern, by showing that the amount of algal DOC that can be utilized by bacteria actually increases as a function of lake productivity (Figure 3a), thus potentially driving the overall DOC lability.
 The amount of labile DOC of algal origin varied several-fold across lakes and increased with lake productivity, yet the mineralization rates of this pool (as expressed by k) did not vary systematically with lake productivity (Figure 3b). This suggests that the intrinsic quality of algal DOC rather than its quantity (or other external factors) influence its degradation dynamics in lake water. Culture work has shown that algal exudates derived from different phytoplankton species may be degraded at very different rates (0.19 to 0.49 day−1) for a similar proportion of labile algal DOC (~10%) [Chrost and Faust, 1983], suggesting that in lakes, the quality of algal DOC and its associated bioavailability may vary as a function of the dominant assemblages of phytoplankton. Because we sampled our lakes at different periods of the year, and over several years, we cannot discard the possibility of a temporal change in phytoplankton community structure as a factor to explain the lack of a systematic pattern of variation of the algal k constant along environmental gradients.
 Interestingly, while algal DOC was degraded more rapidly (higher average initial k values; Figure 2c) and appeared to be proportionally more labile than terrestrial DOC in our incubations (Figure 2b), our results suggest that this algal DOC pool has nevertheless a significant refractory component. In fact, the proportion of refractory C in the algal pool was as high as of that of the terrestrial DOC pool (Table 3). The exudation of refractory, high-molecular weight algal DOC [Chrost and Faust, 1983; Sundh, 1992], and the biological [Romera-Castillo et al., 2011; Guillemette and del Giorgio, 2012] and photochemical [Tranvik and Bertilsson, 2001] conversion of labile algal DOC into biologically inert compounds are probable mechanisms leading to the recalcitrance of algal C in freshwaters. Refractory algal DOC can then be transported downstream, flocculate and settle within the system, or be utilized on longer time scales than those addressed here.
4.2 Degradation Dynamics of Terrestrial DOC
 It has been often assumed that bacteria exhaust the algal DOC pool first before utilizing terrestrially derived material in lakes [Cole et al., 2002; Jansson et al., 2003]. Recent modeling based on diverse scenarios of selective consumption of algal and terrestrial DOC suggest that this assumption is not realistic, as the amount of algal C present in lake water is generally insufficient to sustain the observed levels of bacterial C consumption [Kritzberg et al., 2005; Kritzberg et al., 2006]. Our study provides experimental evidence to that latter contention: We observed that even at the initial stages of decomposition, the isotopic signature of respiratory CO2 was dominated by terrestrial DOC (~70%; Figure 1), suggesting that at least a fraction of terrestrial DOC is highly reactive and also processed on short time scales even though terrestrial DOC was associated with low decay coefficients (Table 3 and Figures 2c and 2d). A rapid utilization of low molecular weight compounds of terrestrial origin in aquatic ecosystems has been recently reported [Covert and Moran, 2001; Berggren et al., 2010a], and because bulk DOC is dominated by terrestrial inputs in our lakes (80%–94% based on the δ13C of lake DOC), it is likely that similar terrestrial compounds may have readily fueled bacterial C consumption concomitantly with algal DOC in our incubations.
 The shifts in isotopic signature of the resulting CO2 toward the terrestrial end-member suggest, however, that the labile algal pool is generally exhausted well before the labile terrestrial C pool is depleted (Figure 1). In this regard, there was a remarkable long-term convergence toward terrestrially dominated substrates in all incubations (except for Simoneau where a significant fraction the algal was still process), thus pointing to a widespread terrestrial support of a residual or baseline metabolism in these lakes, independent of the composition of the initial labile pool. The concept of a baseline metabolism that is weakly connected to contemporary primary production and fueled by residual terrestrial DOC has been postulated before [del Giorgio and Williams, 2005; McCallister and del Giorgio, 2008], but to date poorly documented. Our results further suggest that this baseline metabolism may be more rapidly reached, on the order of days (Figure 1), whenever local primary production is interrupted, for example, during the ice-covered period of lakes at wintertime. Hence, the slow degradation of the major portion of the terrestrial DOC may potentially be a very important component of C cycling in lakes on an annual basis, sustaining processes such as lake net heterotrophy and buffering the overall ecosystem metabolism to environmental fluctuations.
4.3 Phosphorus Modulation of Algal and Terrestrial Degradation Dynamics
 The degradation dynamics of terrestrial and algal DOC appeared to be independent from each other, as there was no relationship between algal and terrestrial k values (P > 0.05). Nevertheless, our results suggest that they both may be influenced by phosphorus concentration in these lakes (Figure 5). On the one hand, phosphorus may indirectly increase the amount of labile algal DOC in lake water by stimulating primary production (Figure 3a). On the other hand, our results show that the actual amount of labile terrestrial C increases with TP (Figure 4a), which is reflected in the lower predominance of refractory C in the terrestrial pool at higher TP concentrations (increase in the v parameter with TP; Figure S3b). These patterns suggest that phosphorus may directly influence the terrestrial DOC degradation dynamics by enhancing the amount of C that can be extracted from this pool on the short term, likely through an increase in bacterial growth or enzymatic activity [Zweifel et al., 1993; Wikner et al., 1999]. The observations that the initial decay coefficients tend to decline, and the average lifetime of the more reactive terrestrial pools tends to increase with TP, would suggest that phosphorous is enhancing the use of a more refractory DOC pool, but is not necessarily influencing the consumption of the highly labile terrestrial DOC pool. This in turn would imply that phosphorous may play a role in determining baseline metabolism in these ecosystems by modulating the size of a terrestrial DOC pool that can be accessed by bacteria. Thus, lakes that are rich in TP relative to DOC may have, for example, relatively higher winter and hypolimnetic metabolism (V. Ducharme-Riel et al., The contribution of winter under-ice and summer hypolimnetic CO2 accumulation to the annual CO2 budget of temperate boreal lakes in Québec, submitted to Ecosystems, 2012), not only because of the increase in the local production of algal DOC but also because of a P-mediated increase in the bioavailability of terrestrial DOC. Significant increases in nutrient delivery to lakes have been recently reported [Elser et al., 2009; Hessen et al., 2009; Jeppesen et al., 2009], and our results suggest a scenario where proportionally more terrestrial DOC could be processed in freshwaters as a whole, with a resulting increase in overall aquatic CO2 emissions to the atmosphere and a decrease in terrestrial C delivery to coastal waters and the ocean.
4.4 Assessment of Potential Errors
 The approach we have used here, while providing new insights into the degradation dynamics of these major DOC pools, has limitations that need consideration. In particular, while there was enough CO2 recovered at each time step to be measured as C mass by mass spectrometry, both the total amount (expressed as DOC concentration; (0.0554 ± 0.015 mg L−1) and the proportion (1.1 ± 0.3%) of total labile DOC (algal + terrestrial) that we derived from these measurements are lower than reported values for similar incubation time frames in freshwaters [del Giorgio and Davis, 2003], including for this same set of lakes [Guillemette and del Giorgio, 2011]. We believe that these discrepancies may be mainly attributed to the difficulty in associating a specific time frame to the final amount of CO2 (and thus to DOC consumption) harvested in the successive regrowth incubations, particularly in terms of incorporating the initial lag phase. We used bacterial production measurements to better constrain this initial lag phase (approximately 2 days; Figure S1), but there is clearly uncertainty in the assignment of the effective time associated to CO2 production. Had this lag phase been assumed to be somewhat longer, it would have resulted in higher apparent rates of DOC consumed, much closer to literature values. However, bacterial production measurements show that after 2 days, there is a steep increase in bacterial activity (Figure S1), and thus, we decided to standardize all experiment spans according to this “apparent” time frame, in order to reconstruct and compare the degradation dynamics of algal and terrestrial DOC on a similar basis. Regardless, the objective of this study was not to quantitatively describe the degradation kinetics of the ambient bulk DOC, as this has been the focus of a previous study targeting the same lakes [Guillemette and del Giorgio, 2011]. Rather, here we have focused on the potential differences in the degradation patterns of the algal and terrestrial DOC pools within this bulk DOC, and we surmise that while the absolute values may be somewhat biased by the approach, the actual patterns of degradation of algal and terrestrial DOC within a complex mixture that we describe are likely to reflect DOC consumption processes occurring in natural lake water.
 Our study presents one of the first empirical demonstrations of systematic differences in the degradation dynamics of algal and terrestrial DOC in natural lake waters. Algal DOC tends to be proportionately more labile, and consumed more rapidly, than its terrestrial counterpart. In contrast to current assumptions that terrestrial C should be consumed once the more labile algal C is depleted, we show that the consumption of these pools is not sequential, but rather proceeds in parallel. We show that overall, there is more labile terrestrial C across all lakes and that this terrestrial C pool contributes significantly to the highly labile pool that fuels the short-term bacterial C consumption and is also responsible for essentially the entire long-term residual (baseline) C metabolism. The degradation dynamics of terrestrial and algal DOC across lakes appeared to be independent from each other, and there was no relationship between algal and terrestrial degradation coefficients or lability. As a consequence, the resulting degradation dynamics of the bulk DOC is not simply a reflection of the relative proportions of each of these pools in the ambient waters but also of their respective patterns of degradation. In addition, our results reveal that far from being uniform, the respective degradation dynamics of the algal and terrestrial DOC pools vary significantly across lakes, and that at least a portion of this variability may be related to interactions with nutrient and lake trophy. Consequently, our results show that the degradation of the algal and terrestrial DOC pools does not appear to be solely governed by their respective chemical properties [Benner, 2003] but is also driven by interactions with nutrients.
 Marianne Lefebvre and Anny Gagné provided precious help in the lab and in the field, and C. Beauchemin provided support for water chemistry analyses. We extend our thanks to P. Middlestead of the G. G. Hatch Lab of the University of Ottawa for isotopic analysis, and to the editors and the two reviewers for their valuable and constructive inputs. This work is a contribution of the Carbon Biogeochemistry in Boreal Aquatic Systems (CarBBAS) program co-funded by the National Science and Engineering Research Council of Canada (NSERC) and Hydro-Québec, and NSF-DEB 0820725 to S.L.M.