Water Research 38 (2004) 53 n hy ) i , G Jan mainly due to mineralization of allochthonous organic flooded vegetation produces large quantities of CO2 [9]. ARTICLE IN PRESS Factors that affect the amount of carbon emitted from reservoirs include: input of allochthonous organic carbon, the amount and type of organic carbon deposits in the flooded land, age of the reservoir, and water temperature [11,12]. *Corresponding author. Tel.: +46-90-786-95-44; fax: +46- 90-786-67-05. E-mail address:
[email protected] (A.-K. Bergstr .om). 0043-1354/$ - see front matter r 2003 Elsevier Ltd. All rights reserved. doi:10.1016/j.w Keywords: Emission; Carbon dioxide; Mineralization; Sediments; Morphometry 1. Introduction Lakes where organic carbon mineralization by hetero- trophic organisms exceeds CO2 fixation by phototrophic organisms are net heterotrophic. The water columns of net heterotrophic lakes are supersaturated with respect to CO2, which results in a net flux of CO2 from the lake surface to the atmosphere [1]. Most of the lakes worldwide can be expected to be net sources of CO2, carbon [2–4], and the degree of net heterotrophy in unproductive lakes is proportional to their input of allochthonous organic carbon [5–7]. Hydroelectric reservoirs have special prerequisites for net heterotrophy and should in many cases be regarded as anthropogenic sources of atmospheric CO2 [8–12]. The flooding of land areas means that the photosynth- esis of the terrestrial vegetation stops and the break- down of large organic carbon sources in peat and Received 15 May 2003; received in revised form 16 October 2003; accepted 20 October 2003 Abstract Carbon balances were calculated for the summer stratification period of 2001 for the hydroelectric reservoir L. Skinnmuddselet (created in 1989) and the natural L. .Ortr.asket, and estimated on annual basis for both lakes. The reservoir and the lake have similar chemical characteristics and are located in adjacent catchments in the northern part of Sweden. Our main hypothesis was that the CO2 production and emissions from the reservoir, L. Skinnmuddselet, would be greater than in the natural L. .Ortr.asket, due to the decomposition of flooded vegetation and peat. The carbon balances showed that the total production of CO2 per unit lake surface area during the summer was very similar in the natural lake and the reservoir (31.3 gCm�2 in L. .Ortr.asket and 25.3 gCm�2 in L. Skinnmuddselet). The sediments were the major CO2 source in the reservoir, while most of the mineralization in the natural lake occurred in the water column. On annual basis the natural L. .Ortr.asket produced and emitted more CO2 per unit of lake surface area than the reservoir L. Skinnmuddselet since mineralization proceeded during winter when L. Skinnmuddselet was emptied for electricity production. Therefore, the potential for CO2 emission was not greater in the reservoir than in the natural lake. r 2003 Elsevier Ltd. All rights reserved. Department of Ecology and Environmental Science, Umea University, Umea SE-901 87, Sweden A comparison of the carbo (L. .Ortr.asket) and a (L. Skinnmuddselet Jan (Aberg, Ann-Kristin Bergstr .om* Mats atres.2003.10.035 1–538 balances of a natural lake droelectric reservoir n northern Sweden rete Algesten, Kenneth S .oderback, sson The climate within both catchments is characterized CO2 concentration was measured as follows: 50mL of headspace gas (outdoor air taken about 2m above ground) was equilibrated with the water, in 1.2 L glass bottles, by vigorous shaking for 1min. The headspace gas was transferred to a 50mL plastic syringe, after which two measurements of CO2 were made using an infrared gas analyzer (PP-systems EGM-3). DIC was analyzed by adding 5mL 25% HCl to the glass bottles and then measuring CO2, as described above. pH was measured with an Orion 230A+ meter (electrode: Orion 9107BN). Water for absorbance (ABS) and DOC analysis was filtered through glass fiber filters (Whatman GF/F) and 0.2 mm membrane filters (PALL—Gelman Sciences Supors-200), respec- tively. The filtered water was stored in a refrigerator until analysis in September 2001, when absorbance (wavelengths 250 and 430 nm) and DOC were measured ARTICLE IN PRESS J. A˚berg et al. / Water Research 38 (2004) 531–538532 by a mean annual temperature of 0–2�C and an annual precipitation of 600–700mm per year [14]. The bedrocks of both catchments are dominated by granite and gneiss, mainly covered by till (75%), peat (15%), and coarse- grained glaciofluvial deposits (5%) [15]. Coniferous forest and mires account for most of the vegetation (70% and 15%, respectively). Less than 1% of the catchments are used for agriculture. 2.2. Sampling and analyses Sampling in the lakes, the inlets and the outlets was carried out every third week during the study period of 28 May—19 August, for analysis of carbon dioxide (CO2), dissolved inorganic carbon (DIC), dissolved In general, hydroelectric reservoirs tend to emit larger quantities of CO2 per unit area than natural lakes [11]. However, few studies have included detailed compar- isons between natural lakes and hydroelectric reservoirs. We therefore conducted a study in which we compared the carbon balances of a natural lake (L. .Ortr.asket) and a hydroelectric reservoir (L. Skinnmuddselet). The lakes have equal volumes, similar chemistry and are located close to each other in the same geological and climatological region in northern Sweden. Our hypothesis was that the CO2 emissions from the artificial L. Skinnmuddselet should be greater than those from the natural L. .Ortr.asket, due to the decomposition of flooded vegetation and peat. 2. Material and methods 2.1. Study area L. Skinnmuddselet (63�590N, 18�260E) and L. .Ortr.asket (64�100N, 18�550E) are two large humic lakes, located in northern Sweden along the main courses of River Gide.alven and River .Ore.alven, respectively. L. Skinnmuddselet is an artificial hydroelectric reservoir created in 1989. Before flooding, the land area where L. Skinnmuddselet is located was a low relief moraine landscape covered by forest areas (55%), mires (30%) and small lakes (15%). The tree vegetation was removed before damming (1987–1988). Branches thicker than 1 cm were gathered and burnt on site. Mires and lakes were left untouched (Sundberg, pers. comm.). L. .Ortr.asket is a glacial trough, which has been water filled since the deglaciation about 9300 years ago. There is practically no macrobiological activity in the littoral zones in the reservoir due to the continuous fluctuations of the water levels and ice erosion during winter, and in the natural lake, the littoral zones only constitutes a few percent of the total lake area [13]. Selected physical characteristics of both lakes are presented in Table 1. organic carbon (DOC), pH and absorbance (ABS). All samples were collected with a Ruttner sampler (2 L). In each lake, water was collected at five stations, which were evenly distributed along the length axis of the lakes and located over the deepest parts possible. Wind speed, air temperature, and absolute air pressure were mea- sured with a portable meteorological tool (Silva Alba Windwatch) on all sampling occasions at all stations. Water temperatures in the inlets and the outlets were measured with a Ruttner sampler thermometer. A temperature probe (WTW LF196) was used to define the depths and temperatures of the epilimnion, meta- limnion and hypolimnion. Lake water was collected from the upper epilimnion, lower epilimnion, upper hypolimnion and lower hypolimnion. Water from the greatest depth at each station was collected 1–3m above the sediment. In the major inlets (R. .Or(an, R. Varg(an and R. Gig(an) and the outlets (R. .Ore.alven and R. Gide.alven) surface water (0.5m depth) was collected. During transport to the laboratory all bottles were stored in the dark in thermally insulated bags. CO2 and DIC were analyzed 1–5 h after sampling. Table 1 Physical and hydrological characteristics of Lake Skinnmudd- selet and Lake .Ortr.asket in the summer of 2001. Lake Skinnmuddselet Lake .Ortr.asket Catchment area (km2) 1400 2210 Mean lake area (km2) 27 8 Mean lake volume (Mm3) 170 170 Maximum depth (m) 20 64 Mean depth (m) 6.5 23 Mean water inflow (m3 s�1) 20 33 Whole lake water exchange time (days) 110 61 Epilimnion water exchange time (days) 94 22 Assumption 1: The internal net supply of DIC (INS) ARTICLE IN PRESS J. A˚berg et al. / Water Research 38 (2004) 531–538 533 using a Hitachi U-1100 spectrophotometer and a Shimadzu TOC-5000 analyzer, respectively. 2.3. Carbon flux calculations Carbon fluxes (F ) were calculated using the formula given by Cole and Caraco [16]: F ¼ kððpCO2w � gassatÞKHÞ; ð1Þ where pCO2w is the partial pressure of the gas in the surface water calculated from ppm-values using the ideal gas law and Henrys law, KH is Henry’s constant calculated according to Weiss [17], and gassat is the concentration of the gas the water would have at equilibrium with the overlying atmosphere (365 matm). The piston velocity, k (cmh�1), was calculated accord- ing to Cole and Caraco [16] as: k ¼ 2:07þ ð0:215 U1:7Þ; ð2Þ where U is the wind speed. Since both lakes are elongated in a southeast–northwest direction and quite narrow the differences in actual wind speed induced by morphology were very low. We therefore estimated a mean daily wind speed (2.5m s�1) for the observation period (cf. [18]). Wind speed data from weather stations located nearby the study areas were obtained from the Swedish National Road Administration. 2.4. Lake carbon balances Lake carbon balances for DOC, DIC and CO2 for the whole study period of 28 May–19 August were calculated for both lakes. The calculations were based on carbon concentrations in the inlets, the outlets and lakes (measured on four occasions during the study period), and daily inflow and outflow data for L. Skinnmuddselet (courtesy of Graninge AB Energy Company) and L. .Ortr.asket (Bergstr .om, unpubl. data). The study period was divided into four 3-week periods (P1; P2; P3 and P4), with sampling dates in the middle of each period. The sampling occasion in every period yielded the area weighted emissions, EP1; EP2; EP3; EP4 (mgCd�1) and the carbon component concentrations, CP1; CP2; CP3; CP4 (for inlets, outlets and water columns), which were assumed to be representative averages for P1; P2; P3 and P4; respectively. Diffusive inflows were assumed to correspond to the carbon concentrations of R. Gig(an (L. Skinnmuddselet) and R. .Or(an (L. .Ortr.asket). The amounts of carbon that were transported to or from the lake (Componentin, Componentout and Emis- sion) were calculated as Componentin ¼ ðCP1inVP1inÞ þ ðCP2inVP2inÞ þ ðCP3inVP3inÞ þ ðCP4inVP4inÞ; ð3Þ is a valid estimate of the internal mineralization of organic carbon. Considering the carbonate-poor soil and bedrock of the studied catchments, inorganic inputs of DIC are unlikely to be significant. Assumption 2: The mineralization of the DOC-pool in the water column of L. .Ortr.asket (mw .o) and L. Skinnmuddselet (mwS) during our study period was assumed to be the same as in Lake .Ortr.asket during the summer of 1994 (equivalent to ca. 10% of the lake carbon pool) [13]. This was calculated as: mw .O ¼ 0:1DOCpool .O: ð7Þ Assumption 3: Since the littoral zones have a minor importance for the whole lake carbon mobilization in the lakes (cf. Study area), the difference between the internal net supply of DIC (INS) and the amount of carbon mineralized in water columns was assumed to represent the amount of carbon mineralized in the sediments in each lake (ms .O; msS): ms .O ¼ INS .O � mw .O: ð8Þ Componentout ¼ ðCP1outVP1outÞ þ ðCP2outVP2outÞ þ ðCP3outVP3outÞ þ ðCP4outVP4outÞ; ð4Þ Emission ¼ nðEP1 þ EP2 þ EP3 þ EP4Þ; ð5Þ where VPxin is the total inflowing water volume of the period, VPxout is the total outflowing water volume of the period, and n is the length of each period (i.e. 21 days). The retention of DOC, DIC and CO2 was calculated, in each case, as the difference between input and output. Lake Carbon pools were calculated from the volume- weighted concentrations in the different strata and from the volumes of each stratum. The internal net supply (INS) of the carbon components was calculated as: INS ¼ DPool�Retention; ð6Þ where DPool represents the mass difference between 28 May and 19 August in the lake’s pool of the different carbon components. DPool was calculated using vo- lume-weighted CP1 and CP4 concentrations in the different strata and volume data from each lake. 2.5. Estimations of mineralization in the water columns and sediments The calculated internal net supply of DIC, combined with data from a study conducted in L. .Ortr.asket in 1994 [13], allowed the carbon mineralization in both the water columns and the sediments of L. Skinnmudd- selet and L. .Ortr.asket to be calculated. The estimations were based on the following assumptions: 2.6. Surface pCO2 and sediment area in contact with the epilimnion-L. Skinnmuddselet The lake was divided into five sub-areas and the five sampling locations were located approximately in the middle of each sub-area. The proportion of the sediment area that was in contact with the epilimnion in each sub- area was estimated from depth level-sediment area relationships. For each sub-area, the relative amount of the total sediment area in contact with the epilimnion was then plotted against the surface pCO2. The calculation was performed for the period late June to late July, when the lake was stratified. 3. Results 3.1. Hydrology Both lakes had similar hydrological regimes, char- acterized by a peak inflow during the spring flood in April/May and occasional high flow events during the summer (Fig. 1a). The outlet discharge (Fig. 1b) differed due to the fact that the spring-flood water to a large ARTICLE IN PRESS Mar Apr Jun Jul Aug 0(a) J. A˚berg et al. / Water Research 38 (2004) 531–538534 0 50 100 150 200 D isc ha rg e (m 3 s- 1 ) Outlet-L. Skinnmuddselet Outlet -L. Örträsket Mar Apr Jun Jul Aug(b) Fig. 1. Discharge from (a) the inlets and (b) the outlets of Lake Skinnmuddselet and Lake .Ortr.asket. extent was retained in L. Skinnmuddselet to replenish the large volume of water (approximately 90% of the lake’s volume) that had been released for electricity production during the preceding winter. Inlet - L. Skinnmuddselet Inlet River Örån - L. Örträsket 50 100 150 200 D isc ha rg e (m 3 s- 1 ) 3.2. Thermal stratification A weak and unstable metalimnion at 12m depth was observed in L. .Ortr.asket in early June. At the same time, the shallow water column of L. Skinnmuddselet was almost completely mixed. During the second sampling occasion in late June, both lakes had a distinct, 5m deep epilimnion. Thereafter, the depth of the epilimnion increased in both lakes. In Lake .Ortr.asket, the depth of the epilimnion reached >13m in mid-August. At this point most of the water column in L. Skinnmuddselet was mixed. The proportion of the total sediment surface in contact with epilimnion water was always o45% in L. .Ortr.asket. In L. Skinnmuddselet the corresponding proportion was approximately 60% during late June, and >80% during early June, July and August. 3.3. Water chemistry and gas fluxes The summer means and ranges of DIC, CO2, pH, ABS, DOC and nutrient concentrations in the two lakes are presented in Table 2. The absorbance and DOC concentrations were somewhat higher in Lake .Ortr.asket and there were clear differences in concentrations of CO2 and DIC between the epilimnion and hypolimnion in both lakes. The hypolimnetic concentrations of DIC and CO2 in L. .Ortr.asket increased >50mM during the study period, and a similar tendency was observed in the sporadic samples from the hypolimnion of L. Skinn- muddselet (Table 2). Both lakes had similar pH-values and nutrient concentrations. A strong positive correlation between surface pCO2 and DOC concentration in the epilimnion was found in L. .Ortr.asket (Fig. 2a) (r2 ¼ 0:80; po0:0005), but not in L. Skinnmuddselet (Fig. 2b). During the period of stable stratification, a strong positive relationship between surface pCO2 and the proportion of the total sediment area in contact with epilimnion water was found for the reservoir, L. Skinnmuddselet (Fig. 3) (r2 ¼ 0:92; po0:0005). The ABS250nm: DOC-ratios of the lakes were not significantly different (two-sample t-test; n ¼ 15þ 15; p ¼ 0:05), indicating that differences in pCO2 between the lakes were not caused by differences in DOC quality or bacterial availability of DOC. The chl-a concentra- tion was very similar in the lakes (Table 2), and the diel variations of pCO2 in the surface waters were less than 60 matm, with pCO2 peaking during daytime. This indicates that photosynthesis had a minor impact on the pCO2 in the lakes. The average area-weighted emission of CO2 during the study period was 1095mg CO2m �2 d�1 (25mmolm�2 d�1) in L. Skinnmuddselet and 900mg CO2m �2 d�1 (21mmolm�2 d�1) in L. .Ortr.asket. The CO2 emission did not differ statistically between lakes during summer (two-sample t-test, n ¼ 15þ 15; p ¼ 0:4). ARTICLE IN PRESS sket nion 9–253) 115 (110–121) 167 (137–197) J. A˚berg et al. / Water Research 38 (2004) 531–538 535 Table 2 Chemical characteristics of Lake Skinnmuddselet and Lake .Ortr.a are given with ranges in parentheses) Lake Skinnmuddselet Epilimnion (n ¼ 39) Hypolim DIC (mM) 120 (114–128) 223 (19 3.4. Lake carbon balances and mineralization Whole lake carbon balances are presented in Table 3. The carbon balance for L. .Ortr.asket was checked against the carbon balance calculated by Jonsson et al. [13] for the summer of 1994, where several independent methods were used to estimate the carbon mineraliza- CO2 (mM) 56 (50–62) 141 (128–157 DOC (mgL�1) 8.5 (7.9–10.8) 8.2 (7.6–10.7 ABS (430 nm. 5 cm cyv) 0.15 (0.13–0.21) 0.16 (0.14–0. pH 6.3 (6.2–6.5) 6.1 (5.9–6.3) Total-N (mgL�1)a 460 Total-P (mgL�1)a 15 Chl-a (mgL�1)a 4 aValues originate from one sampling occasion in the summer of 20 500 7 9 11 1000 1500 p C O 2 (µa tm ) y = 71x + 187 R2 = 0.80 500 1000 1500 7 9 11 13 15 DOC (mg L-1) DOC (mg L-1) p C O 2 (µa tm ) (a) (b) Fig. 2. Relationship between surface pCO2 and DOC in (a) Lake .Ortr.asket and (b) Lake Skinnmuddselet. ) 50 (34–65) 118 (88–153) ) 10.5 (8.4–13.6) 10 (9.2–10.5) 21) 0.19 (0.15–0.31) 0.19 (0.18–0.23) 6.4 (5.2–6.7) 5.9 (4.4–6.3) 532 17 2.2 01 (Algesten unpublished data). 1500 2000 (µa tm ) during the study period of 28 May–19 August 2001 (mean values Lake .Ortr.asket (n ¼ 6) Epilimnion (n ¼ 38) Hypolimnion (n ¼ 38) tion in L. .Ortr.asket. Jonsson et al. [13] estimated the whole lake carbon mineralization to be 27 gCm�2 for the summer period (range 22–33 gCm�2). In our budget the whole lake mineralization is represented by INS (Table 3) and was 31.3 gCm�2. In 1994, the amount of carbon mineralized in the sediments during summer (84 days) was estimated to 80 t [13]. This amount was similar in our budget: 86 t (Table 3). Based on the close agreement between our balance and the previously calculated balance for L. .Ortr.asket, we consider that the carbon balances calculated in this study accurately describe the carbon fluxes in both L. .Ortr.asket and L. Skinnmuddselet. 4. Discussion The carbon balances demonstrate the strong influence of the internal net supply of CO2, i.e. mineralization of organic carbon in water and sediment on carbon cycling y = 11.0x + 309 R2 = 0.92 0 500 1000 30 40 50 60 70 80 90 100 Relative proportion of the total sediment area in contact with the epilimnion (%) p C O 2 Fig. 3. Relationship between surface pCO2 and relative proportion of the total sediment area in contact with the epilimnion in Lake Skinnmuddselet during summer stratifica- tion. Points represent five sub-areas of L. Skinnmuddselet. Each sub-area was sampled on June 29 and on July 17. ARTICLE IN PRESS nd La issio .9 (64 .9 (64 .3 (17 .3 (17 of th Inte J. A˚berg et al. / Water Research 38 (2004) 531–538536 in both lakes (Table 3). The largest uncertainties within our study, which could have affected the final results in the carbon balances, probably lies within the infrequent sampling (every third week). But since our carbon Table 3 Lake carbon balances (g Cm�2) for L. Skinnmuddselet (L. S) a (values in parentheses are given in tonnes C for the whole lake) Poola DPoolb Inputc Output Outletd Em L. S. DOC 53 (1430) 12.2 (330) 46.7 (1260) 36.7 (990) L. S. DIC 10 (270) 2.2 (60) 8.5 (230) 7.8 (210) 23 L. S. CO2 5.2 (140) 1.1 (30) 3.3 (90) 3.7 (100) 23 L. .O. DOC 205 (1640) �10 (�80) 350 (2800) 350 (2800) L. .O. DIC 35 (280) 7.5 (60) 47.5 (380) 50 (400) 21 L. .O. CO2 22.5 (180)) 6.3 (50) 15 (120) 18.7 (150) 21 aLake water pool of carbon. bChange in lake water pool from days 0 to 84. c Input of carbon via the inlets. dOutput of carbon via the outlet. eOutput (emission) of carbon over the water surface. fTotal output of carbon (outlet+emission). gRetention (input-total output). hTotal Internal Net Supply (DPool-Retention). i Internal Net Supply from water column mineralization (10% et al. [13]). j Internal Net Supply from sediment mineralization (Total mineralization). balance for L. .Ortr.asket was very well matched with the earlier more detailed study (where sampling was conducted every second week) (cf. [13]) our carbon balance calculations seem reasonably accurate. Another possible source of error is the assumption that 10% of the DOC-pool in the water column of the lakes was mineralized during summer. However, this share was carefully chosen, based on the earlier study in L. .Ortr.asket [13], and on the observed range reported for humic lakes (6–14%; average 10%, [19]). Moreover the water chemistry of both lakes was very similar (Table 2) and DOC availability should not differ much between the lakes (cf. [19]). Therefore the potential uncertainties should not have caused large errors in the conclusions of this study. The summer means of pCO2 in the surface water of the reservoir (1200matm) and the natural lake (1050matm) were 2.5–3 times greater than the atmo- spheric average (365matm), and similar to the estimated global mean for natural lakes (1000matm) [2]. Conse- quently, net fluxes of CO2 to the atmosphere occurred from both lakes and the mean emission rate was similar in the reservoir and in the natural lake (1095 and 900mg CO2m �2 d�1, respectively). These rates are higher than the average for natural lakes (700mg CO2m �2 d�1) [11], but lower than reported summer means for hydroelectric reservoirs in temperate regions (1300mg CO2m �2 d�1) [12]. The total production of CO2 per unit lake surface area during the summer was similar in both lakes (31.3 gm�2 ke .Ortr.asket (L. .O) during the period 28 May–19 August 2001 Retentiong Internal net supply ne Totalf Totalh Wateri Sedimentj 36.7 (990) 10 (270) 2.2 (60) 5) 31.7 (855) �23.1 (�625) 25.3 (685) 5) 27.6 (745) �24.3 (�655) 25.3 (685) 5.3 (143) 20 (542) 350 (2800) 0 �10 (�80) 0) 71.3 (570) �23.8 (�190) 31.3 (250) 0) 40 (320) �25 (�200) 31.3 (250) 20.5 (164) 10.8 (86) e DOC-pool (Pool); based on results from the study by Jonsson rnal Net Supply—Internal Net Supply from water column in L. .Ortr.asket and 25.3 gm�2 in L. Skinnmuddselet; Table 3). Thus, both lakes had a similar capacity for mineralizing organic carbon and producing CO2. How- ever, the CO2 produced in the natural lake and the reservoir originated from different pools. The surface pCO2 in L. .Ortr.asket was positively related to the DOC concentration in the epilimnion (Fig. 2a), indicating that epilimnetic mineralization of DOC in the lake water was the major source of the emitted CO2 during the summer, in agreement with previously reported results from L. .Ortr.asket [13,20]. In L. Skinnmuddselet there was no relationship between DOC concentration in the epilim- nion and surface pCO2 (Fig. 2b), but a significant relationship between sediment area in contact with epilimnion and pCO2 in surface water (Fig. 3). This result, together with the indications from the mass balance calculations that most of the CO2 production occurred in the sediments in L. Skinnmuddselet (Table 3), demonstrate that the sediment rather than the lake water was the major CO2 source in L. Skinnmuddselet. One reason for the importance of sediments in L. Skinnmuddselet is the high sediment surface area to lake volume ratio (approximately 160) compared to the corresponding ratio in L. .Ortr.asket (approximately 50). However, these ratios do not explain why the CO2 production rate in the sediment was twice as high in L. 2 the reservoir on an annual basis, because mineralization previous net accumulation of carbon in the terrestrial CO2 production was related mainly to differences in ARTICLE IN PRESS J. A˚berg et al. / Water Research 38 (2004) 531–538 537 Skinnmuddselet (20 gCm�2) as in L. .Ortr.asket (10.8 gCm�2) (Table 3). This difference could have been an effect of substrate availability [20] and/or temperature [21,22]. In L. Skinnmuddselet, a majority of the sediments (>80%) were in contact with epilimnion water with temperatures around 15�C during the summer. In contrast, in Lake .Ortr.asket >55% of the sediments were in contact with hypolimnion water with temperatures around 5–7�C. Thus, most of the difference in CO2 production rate in the sediments between the lake and the reservoir could be explained by assuming a Q10 value of 2–3 for sediment microbial processes [22]. Therefore, our results do not provide clear evidence that the flooded bottom contained more available organic substrates for heterotrophic bacteria than natural lake sediments. We also compared the two lakes on an annual basis. During the period from January to May, 90% of the water volume of L. Skinnmuddselet is emptied for hydroelectric power generation. If we assume that the accumulated CO2 in the reservoir represented by the supersaturation levels during late autumn is emitted to the atmosphere when the reservoir is emptied and the water is transported 100 km downstream to the sea, another 3.8 g Cm�2 must be added to the amount of CO2 emitted during the summer (23.9 g Cm �2, Table 3). In L. .Ortr.asket, CO2 is produced throughout the winter, resulting in an accumulation of CO2 in the unstratified water column (from 70mM in autumn to 115mM in late winter, Algesten unpubl. data). During spring circulation the accumulated CO2 (45mM) will probably, to a large extent, be emitted to the atmosphere [23], which then corresponds to an emission of 11.5 g Cm�2. During autumn circulation L. .Ortr.asket will probably have a larger emission than L. Skinnmuddse- let, due to the small volume of hypolimnetic water in L. Skinnmuddselet. The difference in supersaturation between late summer and autumn circulation corre- sponds to an emission of 11.4 gCm�2. Consequently, on an annual basis the total amount of CO2 emitted from L. .Ortr.asket can be estimated to amount to 44.2 g Cm�2, which is more than in L. Skinnmuddselet (27.8 g Cm�2). Thus, the difference in emission between the reservoir and the natural lake on an annual basis was mainly due to the differences during autumn and spring turnover, when approximately 50% of the amount of carbon was emitted from the natural lake. Our results do not indicate any dramatic differences in CO2 production and emission between the reservoir and the natural lake during the summer, which can be explained as an effect of river regulation. However, the CO2 produced in the natural lake and the reservoir originated from different pools. This variation in CO2 production was related mainly to differences in mor- phometry (i.e. the surface to volume ratio), which caused CO2 production to be higher in the lake water in L. morphometry (i.e. the surface to volume ratio) and not to regulation. On an annual basis, the natural lake produced and emitted more CO2 per unit of lake surface area than the reservoir, since mineralization proceeded during the winter when the reservoir was emptied for electricity production. Acknowledgements We thank Bjarne Sundberg and Lars-Erik Dahlen for their assistance, at the courtesy of Graninge AB Energy Company, during the fieldwork and for providing data. Financial support was provided by the Swedish Energy Agency. References [1] del Giorgio PA, Cole JJ, Caraco NF, Peters RH. Linking planktonic biomass and metabolism to net gas fluxes in northern temperate lakes. Ecology 1999;80:1422–31. system now covered by water. This total effect is difficult to quantify since we do not know the net carbon balance of the terrestrial system, but it is possible that the most pronounced effect of the reservoir is a decrease in CO2 fixation in an area of the catchment corresponding to the surface of the reservoir rather than an increase in respiration within that area. 5. Conclusions The total production of CO2 per unit of lake surface area during the summer was similar in the natural lake and the reservoir. The sediments were the major CO2 source in the reservoir, while most of the mineralization in the natural lake occurred in the water column. This variation in processes also occurred in the winter in the natural lake. However, an important aspect when discussing the consequences of building reservoirs, such as L. Skinn- muddselet, is that, before flooding, the land area now covered by water was probably a net sink for CO2 [9]. The total effect of the reservoir on the catchment carbon balance is therefore equivalent to the present emission from the reservoir (750 t of CO2 per year) plus the .Ortr.asket and higher in the sediment of L. Skinnmudd- selet. Therefore, our results do not support the hypoth- esis that emission from the reservoir should be higher than from the natural lake. On the contrary, it appears that the natural lake was a greater source of CO than [2] Cole JJ, Caraco NF, Kling GW, Kratz TK. Carbon dioxide supersaturation in the surface waters of lakes. Science 1994;265:1568–70. [3] del Giorgio PA, Cole JJ, Cimbleris A. Respiration rates in bacteria exceed phytoplankton production in unproductive aquatic systems. Nature 1997;385:148–51. [4] Prairie YT, Bird DF, Cole JJ. The summer metabolic balance in the epilimnion of southeastern Quebec lakes. Limnol Oceanogr 2002;47:316–21. [5] Sobek S, Algesten G, Bergstr .om A-K, Jansson M, Tranvik LJ. The catchment and climate regulation of pCO2 in boreal lakes. Global Change Biol 2003;9:630–41. [6] Hope D, Kratz TK, Riera JL. Relationships between pCO2 and dissolved organic carbon in northern Wisconsin lakes. J Environ Qual 1996;25:1442–5. [7] Jansson M, Bergstr .om A-K, Blomqvist P, Drakare S. Allochtonous organic carbon and phytoplankton/bacter- ioplankton production relationships in lakes. Ecology 2000;81:3250–5. [8] Rudd JMV, Harris R, Kelly CA, Heckey RE. Are hydroelectric reservoirs significant sources of greenhouse gases? Ambio 1993;22:246–8. [9] Kelly CA, Rudd JWM, Bodaly RA, Roulet NP, St. Louis VL, Heyes A, Moore TR, Schiff S, Aravena R, Scott KJ, [12] Duchemin E, Lucotte M, St Louis V, Canuel R. Hydro- electric reservoirs as an anthropogenic source of green- house gases. World Resource Review, 2003;14:334–53. [13] Jonsson A, Meili M, Bergstr .om A-K, Jansson M. Whole- lake mineralization of allochthonous and autochthonous organic carbon in a large humic lake ( .Ortr.asket, N. Sweden). Limnol Oceanogr 2001;46:1691–700. [14] Raab B, Vedin H. Climate, lakes and rivers—national atlas of Sweden. Stockholm, Sweden: Almqvist and Wiksell International; 1995. [15] Fred!en C. Geology—national atlas of Sweden. Stockholm, Sweden: Almqvist and Wiksell International; 1994. [16] Cole JJ, Caraco NF. Atmospheric exchange of carbon dioxide in a low-wind oligotrophic lake measured by the addition of SF6. Limnol Oceanogr 1998;43:647–56. [17] Weiss RF. Carbon dioxide in water and seawater: the solubility of a non-ideal gas. Mar Chem 1974;2:203–15. [18] Kling GW, Kipphut GW, Miller MC. The flux of CO2 and CH4 from lakes and rivers in arctic Alaska. Hydrobiologia 1992;240:23–36. [19] Tranvik LJ. Availability of dissolved organic carbon for planktonic bacteria in oligotrophic lakes of differing humic content. Microb Ecol 1998;16:311–22. [20] Jansson M, Bergstr .om A-K, Blomqvist P, Isaksson A, ARTICLE IN PRESS J. A˚berg et al. / Water Research 38 (2004) 531–538538 fluxes of greenhouse gases and methyl mercury following flooding of an experimental reservoir. Environ Sci Technol 1997;31:1334–44. [10] Duchemin E, Lucotte M. Production of the greenhouse gases CH4 and CO2 by hydroelectric reservoirs of the boreal region. Global Biogeochem Cycles 1995;9: 529–40. [11] St Louis VL, Kelly CA, Duchemin E, Rudd JMV, Rosenberg DM. Reservoir surfaces as sources of green- house gases to the atmosphere: a global estimate. BioScience 2000;50:766–75. microbial food carbon dynamics and structure in Lake .Ortr.asket. Arch Hydrobiol 1999;144:409–28. [21] den Heyer C, Kalff J. Organic matter mineralization rates in sediments: a within- and among lake study. Limnol Oceanogr 1998;43:695–705. [22] Wetzel RG. Limnology: lake and river ecosystems. San Diego, USA: Academic Press (NJ); 2001. [23] Striegl RG, Kortelainen P, Chanton JP, Wichlna KP, Bugna GK, Rantakai M. Carbon dioxide partial pressure and 13C content of north temperate and boreal lakes at spring ice melt. Limnology Oceanogr 2001;46:941–5. Dyck B, Harris R, Warner B, Edwards G. Increases in Jonsson A. Impact of allochthonous organic carbon on A comparison of the carbon balances of a natural lake (L. ørtrðsket) and a hydroelectric reservoir (L. Skinnmuddselet) in north Introduction Material and methods Study area Sampling and analyses Carbon flux calculations Lake carbon balances Estimations of mineralization in the water columns and sediments Surface pCO2 and sediment area in contact with the epilimnion-L. Skinnmuddselet Results Hydrology Thermal stratification Water chemistry and gas fluxes Lake carbon balances and mineralization Discussion Conclusions Acknowledgements References