SEARCH

SEARCH BY CITATION

Keywords:

  • lakes;
  • boreal;
  • carbon dioxide;
  • dissolved organic carbon;
  • lake-watershed interactions;
  • regional models

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and Methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgments
  9. References

[1] In this study, we examine the magnitude and temporal variability of surface water pCO2 in a set of lakes in boreal Québec, and explore the links between lake and catchment properties. The study lakes were consistently supersaturated in CO2, with the mean lake pCO2 ranging from 400 to over 1800 μatm. There was significant interannual variability in pCO2, apparently driven by regional patterns in precipitation. The best multivariate model of average pCO2 included dissolved organic carbon (DOC), lake area and chlorophyll as independent variables, suggesting that external carbon (C) loading to lakes plays a central role in lake CO2 dynamics and that lake trophic status may modulate the influence of external C loading. We show that even if the key drivers of lake pCO2 are similar, they interact differently among regions and the resulting models may be dramatically different. In particular, we show that although pCO2 is invariably correlated to DOC, the shape of this relationship varies greatly among regions, suggesting large-scale regional differences in C delivery, quality, and in-lake processing. As a consequence, current models cannot be extrapolated across regions unless we apply region-specific variables.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and Methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgments
  9. References

[2] Inland waters play a regional role that is disproportional to their surface coverage, acting as both funnels and reactors of materials derived from the surrounding terrestrial systems; lakes in particular both store and process large amounts of carbon they receive from land [Cole et al., 2007]. It has now been established that the vast majority of lakes worldwide are supersaturated in CO2 and hence, sources of carbon dioxide and methane, [see Bastviken et al., 2004] to the atmosphere [Kling et al., 1991; Cole et al., 1994; del Giorgio et al., 1999; Prairie et al., 2002; Jonsson et al., 2003; Hanson et al., 2004; Sobek et al., 2005; Kortelainen et al., 2006]. Several lines of evidence suggest that CO2 supersaturation in lakes is at least in part the result of biological and abiotic in-lake processing of allochthonous organic matter [Dillon and Molot, 1997; del Giorgio et al., 1999; Jonsson et al., 2003; Duarte and Prairie, 2005]. Indeed, factors linked to external carbon loading, such as lake and drainage basin area and particularly the dissolved organic carbon (DOC) content of the lake water, have been identified as the main driving factors of lake CO2 concentration [Hope et al., 1996; Kelly et al., 2001; Prairie et al., 2002; Hanson et al., 2004; Sobek et al., 2003]. It is also clear that the links between CO2 supersaturation and DOC, lake size and drainage basin parameters are complex and characterized by a large degree of variability. For example, for any DOC concentration or lake surface area, lake pCO2 can vary by nearly 2 orders of magnitudes, both within a given region, and even more so between regions [Sobek et al., 2003]. This variability suggests that the regulation of lake pCO2 is complex and influenced by multiple factors, other than simply DOC amount and lake morphometry.

[3] Regardless of the underlying processes driving this gas exchange, the CO2 and CH4 flux from lakes and rivers represents a net loss of carbon from the landscape, and there is increasing evidence that this loss may be a significant fraction of terrestrial primary production and carbon storage [Algesten et al., 2003; Hanson et al., 2004]. This carbon (C) loss through the aquatic system is seldom included in regional carbon models, in part because most regions lack the data and empirical tools needed to actually incorporate lakes and other aquatic systems into whole landscape carbon budgets.

[4] The circumboreal regions are a good example of this, because the extraordinary abundance of lakes and rivers that characterizes this biome [Downing et al., 2006; Teodoru et al., 2009], contrasts with the very limited number of studies that have explored the functioning and regional role of these systems, particularly in terms of carbon dynamics. For example, the boreal landscape of Québec is characterized by an extremely high freshwater density, with an estimated 1.2 million lakes over a total surface area exceeding 1.2 × 106 km2. These lakes span a wide range in morphology, from less than a few hectares to over 800 km2, and cover an estimated 9% of the total surface area [Teodoru et al., 2009], rendering them a major component of the landscape of the region. There are compelling reasons to understand the functioning of these aquatic systems, particularly in terms of carbon dynamics: The boreal biome as a whole is now thought to play a critical role in the global carbon budget [Chapin et al., 2000], and there are large efforts worldwide to determine the net role of these ecosystems as carbon sinks or sources, and to predict how this role might shift under the rapidly changing climatic conditions [Houghton, 2003].

[5] The main objectives of this paper are (1) to examine the magnitude and variability of pCO2 in a wide range of lakes of the boreal zone of Québec; (2) to isolate the main controlling factors and to develop simple empirical models of boreal pCO2 based on DOC, nutrients and lake morphometry; and (3) to compare the empirical patterns obtained in this boreal region to relationships established in other regions of the world, to explore the importance of regional factors in explaining the large range and variability of pCO2 encountered worldwide.

2. Materials and Methods

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and Methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgments
  9. References

2.1. Study Region and Lake Sampling Scheme

[6] The study area is located in the Eastmain River region (51–52°N, 75–76°W), of northern Québec, Canada. This boreal part of the Canadian shield is relatively homogeneous in geology, topography and climate. The vegetation is dominated by black spruce (Picea mariana), and to a lesser extent, white spruce (Picea glauca), American larch (Larix laricina), Jack pine (Pinus banksiana) and the balsam fir (Abies balsamea). Deciduous trees, which dominate early forest successional stages include the quaking aspen (Populus tremuloides), the black cottonwood (Populus trichocarpa) and the paper birch (Betula papyrifera). The region has an average altitude of 250 m and is characterized by an average temperature varying between 0 and −2.5°C, with 600 to 1000 mm of annual precipitations. Freshwaters cover over 15% of the territory, with extensive bogs and peatlands in addition to complex networks of rivers and lakes.

[7] A total of 78 lakes were sampled over the course of this 3-year study. The lakes were sampled during the ice-free period during three sampling series, in June, July–August, and September. We sampled a total of 34 lakes in 2005, 53 lakes in 2006, and 20 lakes in 2007. Forty-two of these lakes were sampled only once in the course of the entire study, 26 lakes were sampled in each of the three sampling campaigns in a given year, and 10 of the lakes were sampled three times per year in each of the three sampling years; the latter were chosen as reference lakes and used to assess both seasonal and interannual variations in the variables measured. The study area is remote and completely undeveloped, and none of the lakes have road access, so sampling was carried out using helicopters fitted with floats or hydroplanes. Since the morphometry of these lakes has not been studied, the team took several depth readings around the central area and water samples and in situ measurements were taken in the deepest point recorded in each lake. All subsequent samplings (if any) were carried out in this same spot. A complete depth profile of temperature, pCO2, and O2 was carried out at this central point, and water samples taken at various depths for chemical analyses. Here we report only the measurements made at 0.5 m depth.

2.2. In Situ Measurements

[8] In situ pCO2 was measured using a PP-Systems EGM-4 infrared gas analyzer (IRGA) receiving air equilibrated with the water by pumping lake water through a Liqui-Cel® MiniModule contactor membrane. The water is circulated through the exchanger at a constant flow rate of 100 L min−1 using a peristaltic pump, and the gases are continuously recirculated into the EGM-4 for instantaneous pCO2 measurements. Measurement precision of the EGM-4 is estimated to be within 1%, and the half-equilibration time of gases from the MiniModule at a flow rate of 100 L min−1 is of a few seconds only. Ambient dissolved O2, temperature, conductivity, and pH were measured with a YSI sonde (Model 600XLM), equipped with a rapid pulse DO probe. The instrument was calibrated in water-saturated air each day prior to sampling.

2.3. Laboratory Analyses

[9] Total phosphorus samples were analyzed spectrophotometrically following potassium persulfate digestion. Total nitrogen samples were analyzed as nitrates following alkaline persulfate digestion and measured on an Alpkem Flow Solution IV autoanalyzer. Chlorophyll samples were analyzed spectrophotometrically following filtration on Whatman (GF/F) filters and hot ethanol (90%) extraction. Dissolved organic carbon (DOC) concentration was measured in 0.2 μm filtered water samples in an OI-1010 Total Carbon Analyzer using wet persulafate oxidation.

2.4. Lake and Watershed Morphology and Data Analysis

[10] The lake morphological variables (perimeter, area) and watershed area were determined using the ArcView 3.2 and ArcGIS v.9 software applied on the DEM derived from 1:50000 maps.

[11] Statistical analyses were carried out using JMP© 7 (SAS Institute). Data were log10 transformed when necessary, to satisfy assumptions of homogeneity and/or normality of residual variance.

3. Results

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and Methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgments
  9. References

3.1. General Lake Characteristics

[12] The sampled lakes in this boreal region of Québec spanned a wide range in morphometry, from 0.007 to 56 km2 (mean 2.1 km2), and from 1 to 26 m (mean 7.3 m) in maximum depth. There was also a wide range in lake chemistry and trophic status: TP varied from 4 to 32 μg L−1 (mean 9.4 μg L−1), DOC from 4 to 24 mg L−1 (mean 7.9 mg L−1), and pH from 4.8 to 8.5 (mean 6.5). Chlorophyll a concentrations ranged from 0.4 to 5.6 μg L−1, yet most lakes were clustered around an overall average of 2.7 μg L−1. The sampled lakes cover much of the range in lake morphometry and chemistry that can be found in this boreal region. There are missing data owing to analytical or sampling problems, so the total numbers of observations may be lower than 78, depending of the analysis.

[13] Mean air temperature was significantly different (p < 0.0001) between campaigns and decreased from 24.9 ± 0.8°C in July to 11.10 ± 0.7°C in September. Mean water temperature was also significantly different between campaigns and decreased from 21.9 ± 0.29°C in July to 11.37 ± 0.30°C in September. Average wind speeds varied between 0.3 and 7.5 m s−1 over the whole summer. Maximum wind speeds varied between 0.4 and 9.4 m s−1 over the summer.

[14] The overall mean pCO2 for all the lakes for the 3 years was 631 ± 222 μatm, ranging from slight undersaturation (340 μatm) to over sixfold supersaturation (2400 μatm), with over 60% of all lakes within 500 and 800 μatm. We could not detect any significant seasonal trend in surface water pCO2 in the 10 reference lakes that were sampled three times over the ice-free period for each of the study years (data not shown). In contrast, there were significant differences in mean annual pCO2 between years for these same reference lakes (Figure 1a): the average pCO2 was significantly higher in 2007 (616 μatm) relative to both 2006 (515 μatm) and 2005 (544 μatm). There was a similar pattern in mean annual DOC concentration in the reference lakes (Figure 1b), with higher average DOC in 2007 (7 mg L−1) and lowest in 2006 (6.6 mg L−1), but these interannual differences were not statistically significant. Not all of the reference lakes showed the same level of interannual variability: the small lakes tended to have much greater year-to-year variability in average pCO2 than the larger lakes (Figure 2), suggesting that interannual variations in hydrological regime and material loading are more buffered in large lakes with longer water retention times.

image

Figure 1. (a) The interannual variation of average surface water pCO2 in the 10 reference lakes. Box plot shows the 75% quartiles and median for all lakes. Dashed line shows the average total wet precipitation (mm) for each of the sampling years in the region. (b) The interannual variation of average dissolved organic carbon (DOC) concentration in the 10 reference lakes. Box plot shows the 75% quartiles and median for all lakes.

Download figure to PowerPoint

image

Figure 2. The difference in average pCO2 between consecutive years (2005–2006 and 2006–2007) as a function of lake size for the 10 reference lakes.

Download figure to PowerPoint

3.2. Patterns in pCO2 Across Lakes

[15] Mean lake pCO2 was significantly negatively correlated to lake area (Figure 3a) and positively related to DOC (Figure 3b); numbers in brackets are the standard error of the estimates:

  • equation image
  • equation image
image

Figure 3. (a) Surface water pCO2 as a function of lake area. Each point corresponds to a lake, and multiple samples taken over the season or in different years for any given lake were averaged. Data are log transformed, and parameters of the regression model are presented in text (equation (1)). (b) Surface water pCO2 as a function of DOC concentration. Each point corresponds to a lake, and multiple samples taken over the season or in different years for any given lake were averaged. Data are log transformed, and parameters of the regression model are presented in text (equation (2)).

Download figure to PowerPoint

[16] These patterns have been showed before for other regions, and suggest that lake pCO2 tends to be on average higher in small compared to large lakes, and also higher in high DOC compared to low DOC lakes. The positive relationship between pCO2 and DOC did not apply only on a cross-lake basis over large DOC gradients, but was also evident within a given lake over different years. The interannual change in average pCO2 was significantly positively related to the interannual change in average DOC (Figure 4) in the 10 reference lakes, calculated for two sampling intervals, 2005–2006, and 2006–2007. Lake area and DOC were weakly but significantly negatively correlated to each other (r2 = 0.08, p < 0.02), but nevertheless both are still both highly significant in a multivariate regression model of average lake pCO2 (numbers in brackets are the standard error of the estimates):

  • equation image

This model confirms that smaller lakes tend to be more supersaturated in CO2, but that for any given lake size, lakes with higher DOC concentrations tend to have higher pCO2. The multivariate model that explains the most variability in average lake pCO2 contains lake area, DOC and chlorophyll concentration (numbers in brackets are the standard error of the estimate):

  • equation image

This model implies that lake trophic status, here expressed in terms of chlorophyll a, and the subsequent differences in primary productivity, modulate the influence of external DOC and DIC loading on lake CO2. For any given lake size and DOC concentration, more productive lakes tend to have lower surface water pCO2, presumably due to its assimilation by the phytoplankton. While the ratio of lake to drainage basin area was also significant together with DOC and Chla in multivariate models of pCO2, the model containing lake area explained more of the overall variability in pCO2.

image

Figure 4. The change in average surface water pCO2 per lake between consecutive years as a function of the average change in DOC over the same periods for the 10 reference lakes.

Download figure to PowerPoint

3.3. Relationship Between O2 and CO2 Dynamics

[17] The in situ mean seasonal concentration of dissolved oxygen was 9.42 ± 0.82 mg L−1. There was a significant difference in concentration between campaigns with an observed increase from 8.26 ± 0.25 mg L−1 in July to 10.23 ± 0.13 mg L−1 in September. Surface waters tended to be undersaturated in O2, with an overall average of 94 ± 5% saturation, (range from 85 to 115%). There was overall coherence between surface water O2 and CO2 dynamics in terms of departures from saturation: O2 undersaturation generally coincided with CO2 oversaturation (Figure 5), which would suggest that a least a portion of the excess CO2 in these lakes may originate from biological mineralization of organic matter in the water column. While the sign and overall magnitude of CO2 and O2 saturation are consistent, there is a wide scatter in the individual values shown in Figure 5, suggesting there may be processes other than metabolic CO2 production, including methodological or sampling variability.

image

Figure 5. O2 and CO2 departure from saturation in lake surface waters for all individual measurements made between 2005 and 2007.

Download figure to PowerPoint

4. Discussion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and Methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgments
  9. References

[18] The boreal region of northern Québec is vast and has among the highest densities of lakes in the world [Downing et al., 2006; Teodoru et al., 2009]; only a handful of the over 1.2 million lakes in this region have been studied, and this mostly been due to the remoteness of the area, lack of accessibility and the resulting logistical challenges. Our study shows that almost without exception, these boreal lakes are consistently supersaturated in CO2, and thus act as sources to the atmosphere, as has been shown for other boreal regions [Jonsson et al., 2003; Sobek et al., 2003; Kortelainen et al., 2006].

[19] Over the 3-year follow-up of the 10 reference lakes we recorded an average interannual difference of 10% in mean seasonal pCO2 between consecutive years. This interannual variability was substantially greater for smaller lakes (up to 30%) than for the larger lakes, but all lakes followed the basic pattern, suggesting regional-scale influences, possibly induced by climatic differences. In this regard, measurements from 3 weather stations within 100 km of our study area, collectively show some major differences climatic between years: 2005 was an exceptionally hot year, whereas 2006 and 2007 were more average. There were also major differences in total wet precipitation (Figure 1a), 2006 having been a very dry year, 2006 had roughly average precipitations for the region, whereas 2007 was well above average in terms of precipitation. The interannual pattern in average pCO2 tends to follow this precipitation pattern: Lakes collectively had the highest average pCO2 during the wet year (2007), lowest in the dry year, and intermediate during the average precipitation year. Although we would require a longer time series to quantitatively establish the connection between precipitation patterns and lake CO2 dynamics, these preliminary data would suggest a direct link between water load and CO2 dynamics. Kelly et al. [2001] also noted a positive relationship between total precipitation and average lake pCO2 in northern Ontario lakes, and Rantakari and Kortelainen [2005] found a similar pattern in large Finnish boreal lakes. Although there was a similar overall pattern with DOC, this was much weaker than that of pCO2, suggesting that hydrology may influence pCO2 not only through the delivery of DOC (which may be then converted to CO2 within the lakes), but perhaps through increased delivery of DIC directly from the drainage basin. In this regard, the loading of inorganic and organic C to boreal lakes has been unequivocally linked to water inputs [Dillon and Molot, 1997; Schiff et al., 1997], but how these interact to determine net lake CO2 emissions is less clear.

[20] Current models predict a general warming of the boreal and taiga regions of Northern Québec, as well as increased overall precipitation [Lemmen et al., 2008], although the shift in precipitation is expected to vary greatly across the northern landscape, with localized declines [Plummer et al., 2006]. This climate change scenario should lead, on the one hand, to increased terrestrial primary production, soil respiration and DOC mobilization, and on the other, increased discharge, which combined should result in increased loading of both DOC and DIC to lakes [Eimers et al., 2008] and to increased CO2 fluxes. There will be other, more indirect effects of climate change of lake CO2 dynamics. For example, in a companion paper [Marchand et al., 2009], we have shown that in this boreal landscape there is a negative relationship between the extent and age of fire in the basin and lake respiration and CO2 flux; since fire frequency is expected to increase overall in these Northern landscapes, this indirect effect will further enhance lake C emissions beyond the positive effect of climate-induced hydrological changes.

[21] The best individual predictors of pCO2 in the boreal lakes of Québec were found to be DOC and lake area. Like in other regions of the world where lake CO2 dynamics have been studied, we found a negative relationship between lake size and average surface water pCO2 [Kelly et al., 2001; Sand-Jensen and Staehr, 2007]. Previous studies, for example that of Kelly et al. [2001], had covered an extremely wide range of lake sizes, from less than 0.1 to over 1000 km2. In contrast, the boreal landscape in the Eastmain region is dominated by small lakes, and in our data the largest lake is 40 km2, and all but 2 were smaller than 10 km2, which suggests that the scaling of surface pCO2 to lake size is not a pattern that emerges only when extreme ranges are considered. This relationship between average lake pCO2 and lake area is important because it implies that regions that have similar total lake area might have very different total lake CO2 emissions depending on the size distribution of lakes.

[22] In this boreal region, freshwaters cover from 6 to 20% of the total area; lake size ranges from a few hectares to several hundred km2, but the size frequency distribution has a distinct peak in the very small lake size classes (median is 2.5 ha). Although these small lakes, which have the highest average pCO2 and consequently the highest CO2 emissions, often account for less than 10% of the total lake area, they account for a much larger fraction of the total lake emissions in the area. It is thus important to consider lake size when assessing the regional role of these systems at the whole landscape level. Lake area is important not only in determining the average levels of pCO2, but also the seasonal and interannual variability of these average values. Our results show that while all lakes followed the same interannual trends, the larger lakes had much smaller ranges of variation, and appear buffered from shifts in hydrology and C loading, most likely through longer water turnover times. This suggests that landscapes that are dominated by small lakes will not only have higher total CO2 emissions, but also much higher temporal variability in these emissions.

[23] In a companion paper we have described the patterns in CO2 dynamics in streams and rivers in this same region [Teodoru et al., 2009], and have further estimated the relative contribution of lakes and running waters to both the total aquatic area and C fluxes in a large block (950 km2) of boreal territory. In the case of lakes, we combined the lake area-pCO2 relationship presented here with GIS-based estimates of lake number and size, and conservative assumptions concerning gas exchange, to derive average lake C fluxes in the order of 77 mg m−2 d−1, which translate to annual C emissions of 1–2 g C per m2 of watershed. These preliminary estimates of lake C emission are in the same order of magnitude as other key components of the regional C budget, including soil C accumulation [Banville et al., 2009], and C emission due to fire [Bachelet et al., 2005].

[24] Given the significance of these lake C emissions, the density of lakes in the boreal landscape, and the extent of the circumboreal biome, it is thus imperative to more effectively scale up these lake fluxes at the regional scale. In the calculations discussed above we have applied the simplest empirical model, that containing lake area, and the latter can be obtained with relative ease from digital maps. Better predictions could be obtained by applying the multivariate model that includes DOC and Chl in addition to lake area, but this approach would have to involve a combination of GIS analysis of lake distribution and area, and remote sensing to estimate both DOC and chlorophyll at the regional scale. There have been some attempts to determine boreal lake DOC concentrations from satellite images [i.e., Kutser et al., 2005], but clearly more work in this are area is needed before remote sensing of boreal lake properties can be applied on a routine basis.

[25] DOC concentration was the best single predictor of average surface water pCO2 in these boreal lakes, as has been previously noted in the literature for other regions [Hope et al., 1996; Riera et al., 1999; Hanson et al., 2003; Sobek et al., 2003; Jonsson et al., 2003; Prairie et al., 2002]. The influence of DOC on lake pCO2 is evident not only over a large range of DOC concentrations across lakes, but also within a given lake in time. We have shown that the interannual variability in our set of reference lakes was linked to the interannual variability in DOC (Figure 4). The strong relationship between DOC and pCO2 may reflect a direct link, through the decomposition of terrestrial DOC and subsequent production of CO2, or an indirect link, where DOC acts a proxy for C loading to lakes. The overall coherence observed between surface water O2 and CO2 dynamics (Figure 5) would point to the former, and suggest that the respiration of terrestrially derived DOC may be a significant source of CO2 supersaturation in these lakes. Isotopic studies in other boreal and temperate lakes have shown that a significant fraction of total water column respiration is based on terrestrial organic matter [Karlsson et al., 2007; McCallister and del Giorgio, 2008], in agreement with previous studies [del Giorgio et al., 1999; Jonsson et al., 2003; Rantakari and Kortelainen, 2005]. Other studies, however, have reported a much stronger relationship between O2 and CO2 [e.g., Hanson et al., 2003; Kortelainen et al., 2006] than the one we found in our study lake, which, while coherent, was weak, suggesting either methodological problems in the measurements themselves, time lags between the respective processes, or a relatively large degree of uncoupling. The latter may result from nonmetabolic processes such as injection of CO2 from the surrounding watershed, as has been observed by Riera et al. [1999], or from anaerobic processes [Torgersen and Branco, 2008]. This is further supported by the departures from saturation of O2 and pCO2, where above approximately 20 μM, CO2 and O2 tend to decouple with much higher CO2 supersaturation relative to O2 undersaturation. The lakes that showed the largest departures from a 1:1 relationship were those that were smallest in size, perhaps indicating a strong input of inorganic carbon, through either groundwater [Prairie et al., 2002], or riverine inputs [Riera et al., 1999]. Overall, these results highlight the dual role of the lakes in these boreal regions, as “reactors” mediating the biological and abiotic decomposition of terrestrial DOC, and as “chimneys,” venting CO2 that originates in terrestrial processes.

[26] The multivariate model that best predicts average lake pCO2 in this region contains lake area, DOC and chlorophyll as independent variables, and explained approximately 60% of the total variation in average surface pCO2. The fact that lake area and DOC are both highly significant suggests an effect that is beyond the correlation that they may have between each other. Small lakes do tend to have higher DOC concentrations, but the relationship between lake area and DOC is only marginally significant in our study region. Sobek et al. [2005] also found no overall relationship between lake area and DOC, although locally lake area explained some of the variance in DOC. Rather, these two variables reflect two distinct components of the regulation of lake pCO2: Lake area is a proxy for both water retention time and for contact surface with the drainage basin, with small lakes having the highest water turnover and lowest volume to perimeter ratio, both of which amplify the influence of the surrounding terrestrial system. DOC in turn, is directly related to the source of C that generates the CO2 flux, as discussed above.

[27] The multivariate model further suggests that lake trophic status, reflected in either the chlorophyll (or TP concentration, model not shown), and the associated primary production, modulates the influence of the catchment on lake pCO2. Both DOC and TP interact to determine the net metabolic balance of lakes [del Giorgio et al., 1999; Hanson et al., 2003; Prairie et al., 2002; Prairie, 2008]. As reported by Prairie et al. [2002], we find a significant negative relationship between O2 departure and TP or chlorophyll concentrations (r2 = 0.25; p < 0.01) indicating that the less productive lakes tend to be more heterotrophic, and also more undersaturated.

[28] DOC concentrations have recurrently emerged as the primary predictor for lake CO2 concentrations in a variety of regions [Hope et al., 1996; Riera et al., 1999; del Giorgio et al., 1999; Kelly et al., 2001; Prairie et al., 2002; Sobek et al., 2003]. Sobek et al. [2003] further reported that this pattern extended across a latitudinal gradient and was valid for a variety of climatic conditions within the Scandinavian boreal zone. The question remains, however, as to whether the patterns in pCO2 relative to DOC are similar across landscapes and geographic regions (e.g., boreal versus temperate). In Figure 6 we have plotted several published DOC versus pCO2 models for different northern regions, together with our own for boreal Québec. While the relationship is positive and significant in all studies, there are large differences between their respective parameters. An ANCOVA showed that both slopes and intercepts were not homogeneous among regions (p < 0.01), thereby precluding the use of a single relationship for all regions. The intercepts vary threefold, are lowest in northern Finnish lakes (for example 112 μatm) [Rantakari and Kortelainin, 2005] as well as in boreal Québec lakes (150 μatm, this study), intermediate in boreal southern Québec and Sweden (200 to 270 μatm), [Sobek et al., 2003; Jonsson et al., 2003], and highest in southern Québec lakes (around 330 to 350 μatm) [del Giorgio et al., 1999; Prairie et al., 2002]. These intercepts probably reflect a baseline level of lake CO2 that is related to the regional water chemistry, hydrology and material loading, and Figure 6 would suggest that this baseline may in fact be a regional feature. On the other hand, the log-slopes also varied greatly between studies and regions, from very low slopes in temperate Québec lakes, to very high slopes in northern Scandinavian lakes. This pattern would suggest that changes in DOC loading or input to lakes have very different consequences in terms of surface water pCO2 in lakes of different regions, and this in turn could be related to fundamental differences in the nature of the DOC among regions.

image

Figure 6. Published empirical models of lake surface water pCO2 as a function of DOC concentration for different regions of the world. The variables were log transformed and the data fitted to a power model (pCO2 = a × DOCb), and each individual study is coded with a different color.

Download figure to PowerPoint

[29] The consequence of these differences in both intercepts and slopes is an overall threefold variation in the predicted pCO2 for any given DOC concentration. For example, at a DOC concentration of 6–7 mg L−1, typical of many temperate and boreal region lakes, the pCO2 predicted by the different empirical models varies from 400 to 1150 μatm. The models presented in Figure 6 comprise the open water season, mostly late spring and summer, so temperature alone cannot be invoked to explain these differences. What are then the factors that determine these regional differences in the pCO2 – DOC relationship? For any given level of DOC, nutrients may influence how a given level of DOC will translate into dissolved CO2 concentration. On one hand, increasing productivity should decrease pCO2 via photosynthetic inorganic carbon assimilation [del Giorgio et al., 1999; Prairie, 2008]. On the other hand, there is evidence that nutrients can stimulate the degradation of the organic matter present [Smith and Prairie, 2004]. The net balance between these processes has been difficult to estimate, such that the net effect of nutrients is still unknown over the large scale. Our own results support the former scenario, and suggest that nutrients, particularly P, may modulate the influence of the drainage basin on lake CO2 dynamics through their influence on authochthonous production (via chlorophyll concentration, equation (4)). However, the influence of chlorophyll (or TP) is modest (based on the proportion of variance explained): For any given combination of lake area and DOC concentration, a doubling in chlorophyll would yield less than a 15% decline in pCO2 in the observed surface water pCO2. Interestingly, not all studies have reported a negative relationship between pCO2 and nutrients. For example, Huttunen et al. [2003], Kortelainen et al. [2006], and Rantakari and Kortenainen [2005] reported a positive relationship between lake pCO2 and TP, and explained this as resulting from nutrient enhancement of DOC mineralization. These contrasting reports suggest a fundamentally different regulation of lake pCO2 in different regions, although clearly the regional differences in pCO2 to DOC cannot be due to differences in nutrients (or chlorophyll) alone.

[30] Lake DOC concentration is often viewed as a proxy for allochthonous C loading, but the ambient DOC is the net result of DOC loading, in situ production and removal, which are in turn dependent on water residence time, lake trophy, the intensity of photochemical processes, the chemical nature of the external DOC itself [Curtis and Schindler, 1997], all of which can vary with relative independence of one another. It is possible that there are systematic regional differences in one or several of these features, such that any given amount of DOC may result in very different amounts of CO2 produced. One factor that may profoundly alter the relationship between DOC and lake pCO2 are the pathways of delivery of this organic carbon to lakes [Mattsson et al., 2005], which influence not only the amount but also the nature of the DOC inputs [Schiff et al., 1997]. Hydrology mediates the delivery of DOC to lakes, and the interactions between lakes and their watersheds are complex owing to the extremely heterogeneous nature of hydrologic networks [Jenerette and Lal, 2005]. The cross-system differences that exist in the pCO2/DOC relationship may be primary linked to regional differences in hydrologic pathways and the resulting differences in the amount and nature of the C that is delivered to lakes.

5. Conclusions

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and Methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgments
  9. References

[31] The lakes in this boreal region of Québec are all supersaturated in CO2, and there is a clear lake size scaling of pCO2 and of the resulting CO2 fluxes. The cross-system patterns that we have described here are superimposed to region-scale interannual variation, that seems to be directly linked to climate and more specifically to precipitation. The best model for surface water pCO2 in these boreal region was shown to incorporate DOC concentrations, lake area and chlorophyll, the latter having only a marginal effect. The weak influence of trophic variables is probably the result of a relatively narrow range in nutrient and chlorophyll concentrations in these boreal lakes, and the virtual absence of eutrophic lakes where local primary production might exert a larger role. Nutrients appear to be more important variables in temperate regions where there is a greater differentiation between watersheds and land use patterns. Our results further highlight the importance of the regional landscape conformation in terms not only of total lake surface but the contribution of different lake categories in determining the overall lake fluxes. Comparison with previously published models shows that while we can predict relatively well the average surface water pCO2 using a few basic parameters in any one region, the models cannot be extrapolated across regions unless we apply region-specific variables. Even if the key drivers of lake pCO2 appear to be similar, they are expressed differently between regions and the reasons underlying these regional differences remain to be explored.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and Methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgments
  9. References

[32] The authors would like to thank Hydro-Québec, and especially Alain Tremblay, for logistical and financial support through the Hydro-Québec/UQÀM Eastmain-1 Research project. We would also like to thank Annick St. Pierre, Martine Camire, Alexandre Blain, Delphine Marchand, Martin Genest, and Simon Barette for field assistance; Catherine Beauchemin for laboratory analysis; and two anonymous reviewers for helpful comments.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Materials and Methods
  5. 3. Results
  6. 4. Discussion
  7. 5. Conclusions
  8. Acknowledgments
  9. References