Global Biogeochemical Cycles
  • Open Access

Carbon evasion/accumulation ratio in boreal lakes is linked to nitrogen



[1] The role of lakes in landscape carbon (C) cycling and primary drivers behind freshwater C balance have remained poorly known, although lakes are an important landscape component and cover 10% of Finland's total area. We studied CO2 evasion and average Holocene C accumulation in 82 boreal lakes (0.04–1540 km2; max depth 1–93 m) located between the latitudes 60°N and 69°N. Both CO2 evasion and C accumulation correlated with numerous drivers and were closely linked to lake area and maximum depth. The total aquatic C retention (C evasion + C accumulation) was largely determined by lake area (r2 = 0.96, p < 0.0001). Carbon in boreal lakes is mostly terrestrially fixed and our data demonstrate that boreal lakes are important conduits for transferring terrestrially fixed C to the atmosphere. C evasion/accumulation ratio (Cev/Cacc) ranged from 4 to 86 (on average 30) and correlated positively only with lake water NO3-N concentration and maximum depth, and negatively with sediment N pool. Numerous possible important drivers like land use and climate as well as lake physical and chemical characteristics were compared to C balances. Our study might thus be valuable in trying to understand links between aquatic and terrestrial C cycling over boreal landscapes. Although climatic drivers generally play a key role in short-term C balance fluctuations, our data indicate nitrogen/fertility to be a key contributing factor to long term C balance—increasing Cev/Cacc and thus decreasing the role of lakes in landscape C sequestration in boreal zone.

1 Introduction

[2] Regional studies [e.g., Richey et al. 2002; Algesten et al., 2003; Hanson et al., 2004; Kortelainen et al., 2004; Roehm et al., 2009] and global estimates [Cole et al., 2007; Williamson et al., 2008; Tranvik et al., 2009] have indicated that freshwater ecosystems are metabolically active sites contributing significantly to landscape carbon (C) balances. Lakes can exert a disproportionately large impact on C mass balances and cycling rates [Downing, 2010], which is indicated in the increasing body of evidence that although dissolved organic matter exported from terrestrial to aquatic ecosystems is partly refractory and buried in sediments, a considerable part of it can be mineralized to CO2 [e.g., Striegl et al., 2001; Algesten et al., 2003; Hanson et al., 2004; Rantakari and Kortelainen, 2005; Kortelainen et al., 2006a; Roehm et al., 2009].

[3] Carbon input to lakes from surrounding catchments and C outflow from lakes play an important role in lake C budgets both in boreal [e.g., Jonsson et al. 2001; Sobek et al., 2006; Einola et al., 2011] and hard water lakes [Finlay et al., 2010], although variability in the contribution of organic and inorganic C fluxes can be significant. In the boreal zone, the majority of organic carbon export is in a dissolved form; POC (particulate organic carbon) proportion is < 6% both in small headwater streams [Kortelainen et al., 2006b] and in large rivers flowing to the Baltic Sea [Mattsson et al., 2005]. Furthermore, in predominantly acid boreal lakes, surrounded by forests and peatlands, TIC (total inorganic carbon) concentrations are generally minor [Rantakari and Kortelainen, 2008]. Large latitudinal differences in the shares of particulate vs. dissolved C and organic vs. inorganic C [e.g., Meybeck, 1993] can significantly contribute to C budgets. For example, average Holocene C burial rates are minor in boreal Finnish lake sediments (on average 2 g C m–2 [Pajunen, 2004; Kortelainen et al., 2006a]), in Greenland lake sediments (6 g C m–2 [Anderson et al., 2009]), in European lake sediments (< 6.5 g C m–2 in 75 % the lakes) [Kastowski et al., 2011], and in boreal northern Québec sediments (3.8 g C m–2 [Ferland et al. 2012]) compared to heavily human impacted landscapes. In heavily loaded agricultural impoundments, water POC and TIC concentrations are very high reflecting both rapid weathering from surrounding catchments and a high primary production. In such lakes carbon accumulation has been shown to be very rapid [Downing et al., 2008], whereas pH is generally high resulting in relatively small CO2 evasion [Balmer and Downing, 2011].

[4] Although carbon sequestration in terrestrial ecosystems has been intensively studied, the role of upland forests, wetlands, and agricultural land in C sequestration/C release in changing climate is still highly uncertain. Boreal lakes have traditionally been totally neglected in landscape energy/element cycles, although areal C pools for example in Finnish lakes are significantly larger than in upland forest soils, 19 kg C m–2 vs. 7.2 kg C m–2, respectively [Kauppi et al., 1997; Kortelainen et al., 2004]. The traditional focus of limnology on individual lakes has significantly delayed the synthesising of the fragmented, and often unrepresentative, freshwater data to provide regional or global estimates. Recent global syntheses [Cole et al., 2007; Williamson et al., 2008; Tranvik et al., 2009] have significantly improved the knowledge of the role of freshwater ecosystems in global perspective.

[5] This paper focuses on: (1) The role of climate vs. land use in contributing to lake chemistry in randomly selected boreal lakes south of the Arctic Circle vs. north of the Arctic Circle and (2) identifying key drivers (climate, land use, topography, lake chemical, physical and morphological characteristics) primarily controlling the role of boreal lakes in landscape C balances. Our main hypotheses were: (1) land use is a more important driver for lake chemistry than climate, because land use patterns were important drivers for boreal river water chemistry [Mattsson et al., 2005] and (2) Cev/Cacc is significantly linked to numerous drivers, because both CO2 evasion in small boreal lakes [Kortelainen et al., 2006a] and Holocene C pools in lake sediments [Kortelainen et al., 2004] significantly correlated with lake morphology, lake water chemistry, catchment land use, and lake temperature.

[6] Finland is the only country where both Holocene C pools in lake sediments [Kortelainen et al., 2004] and average annual CO2 and CH4 evasion from lakes to the atmosphere [Kortelainen et al., 2006a; Juutinen et al., 2009] have been studied based on randomly selected databases. The 82 lakes that were included both in C gas and C stock studies cover a wide climatic gradient between the latitudes 60°N and 69°N and provide thus a unique opportunity to compare the impacts of climate/deposition and land use on lake water chemistry and lake C balances. Our data demonstrate that although both lake chemistry and carbon pools and fluxes reflect topography, climate, and land use, Cev/Cacc was primarily linked to nitrogen (N) and maximum depth of the lake reflecting catchment topography. These spatially representative databases thus indicate N to be a key driver for the role of lakes in long-term C balances in boreal N limited ecosystems.

2 Material and Methods

[7] Subpopulations of 177 and 122 lakes were randomly selected from the Nordic Lake Survey database for CO2 evasion [Kortelainen et al., 2006a] and Holocene C pool [Kortelainen et al., 2004] studies, respectively. The catchment areas of the Nordic Lake Survey lakes were determined from the topographic maps, and the catchment boundaries were digitalized and combined with land use data based on satellite images using the Arc View georeferencing software. Lake area, catchment area, catchment to lake area ratio, latitude, and the proportions of peatland, forest on mineral soil, agricultural land, water (consisting of upstream water bodies and the lake itself), and built-up area in the catchments were determined for each lake. Altitude, slope and elevation were measured for most of the lakes, however, excluding the largest lakes/catchments. Thus, these variables could not be included in stepwise multiple regression analyses.

[8] The 177 lakes (Figure 1a) randomly selected for C gas studies were sampled four times per year: before and after ice melt, at the end of summer stratification, and during a fall overturn. Half of the lakes were sampled in 1998 and the rest in 1999. A depth profile (1 m from the surface of the lake, middle of the water column, 1 m from the sediment surface, and 20 cm from the sediment surface) at the deepest point of the lake was sampled for physical and chemical characteristics. However, samples from 1 m from the bottom were excluded from tables presenting water chemistry, because in the shallowest lakes they represent both surface and bottom water.

Figure 1.

(a) Location of the 177 randomly selected C evasion lakes and (b) location of the 82 lakes included in C balance calculations (Cev/Cacc).

[9] Altogether 22 water quality parameters were measured; however, in this paper we concentrate on those that are most closely linked to C balances, i.e., TIC, TOC, dissolved oxygen, alkalinity, conductivity, pH, color, total nitrogen (TN), nitrate nitrogen (NO3-N), ammonium nitrogen (NH4-N), total phosphorus (TP), phosphate phosphorus (PO4-P), total iron (Fe), and manganese (Mn). Water chemistry [National Board of Waters, 1981] was analyzed from unfiltered samples in the accredited laboratories of the Regional Environment Centre. The acidified TOC samples were bubbled with nitrogen to remove inorganic carbon (CO2), and TOC was determined by oxidizing the sample by combustion and measuring inorganic C by IR-spectrophotometry. TIC samples were acidified in the field and gas concentrations were analyzed with a headspace equilibrium technique [McAuliffe, 1971]. Color (mg L–1 Pt) was determined by comparison with standard platinum cobalt chloride disks. The analytical methods are thoroughly described in Kortelainen et al. [2006a].

[10] The potential C evasion from the 177 randomly selected lakes, smaller than 100 km2 (0.04–63 km2) [Kortelainen et al., 2006a] and the 37 large lakes (101–1540 km2) [Rantakari and Kortelainen, 2005] was estimated using the boundary layer technique according to Cole and Caraco [2001]. The CO2 saturation in the surface water was determined and compared with the equilibrium value defined by the CO2 pressure in the air above the surface (cf. equations ((1))–((4)); Kortelainen et al. [2006a]). Rantakari and Kortelainen [2005] demonstrated that the annual CO2 evasion from largest Finnish lakes closely followed the annual precipitation pattern being largest during rainy years. The mean annual precipitation in Finland during our study years, 1998 and 1999, was 574 and 686 mm a–1, respectively. The average value, 630 mm a–1, is close to the long-term mean annual precipitation in Finland, 660 mm a–1 [Climatological Statistics in Finland 1961–1990, 1991] indicating that our CO2 evasion estimates during the study years can be considered to represent average long-term precipitation patterns in Finland.

[11] The Holocene sediment C pool and average Holocene C accumulation was estimated for 122 lakes including all large lakes (> 100 km2 ) in Finland (range 101–1540 km2) and a subset of randomly selected small lakes (0.04–88 km2) [Pajunen, 2004; Kortelainen et al., 2004]. Holocene sediment volumes were estimated from information collected with an acoustic sounder in lakes larger than 2 km2 and by coring in small lakes. Acoustic sounding profiles were used to estimate the area covered by different lake bottom types and the amount of gyttja (freshwater mud containing an abundance of organic matter) systematically at intervals of 150 m in large lakes, while the thicknesses of the gyttja layers was determined by coring and field observations in the case of the small lakes. Sediment samples were taken from the area of maximum sedimentation and C content was analyzed from cores representing the entire period since the last glaciation. At least one long core extending from the sediment surface to the glacial clay or till was taken from each lake, and two or three such cores were taken from the largest lakes if they consisted of several basins. The chemical analyses were carried out on cores divided according to their stratigraphy, mostly into pieces of length 10–30 cm, and the C and N in the cores was analyzed using a LECO CHN-600 elemental analyzer (LECO Corporation, St. Joseph MI, USA). The age was determined by radiocarbon dating, paleomagnetic dating or deglaciation/isolation of the basin. Dating, dry density, sediment quantity, accumulation rate, and C stock calculations are thoroughly described in Pajunen et al. [2000].

[12] Lake specific Cev/Cacc was calculated based on average annual Holocene C burial (g C m–2 LA a–1) (LA, i.e., lake area) [Kortelainen et al., 2004] and average annual C evasion (g C m–2 LA a–1) based on the measurements carried out during the years 1998–1999 [Kortelainen et al., 2006a]. The Cev/Cacc in those 82 lakes, which were included both in sediment and C gas evasion studies (Figure 1b), was compared to lake chemistry, morphology, latitude, sediment properties, and catchment characteristics (peatland %, field %, forest %, water %) using Pearson's correlation coefficients and stepwise multiple linear regression analyses using SAS 8.2 for Windows software (PROC REG, SAS Institute Inc., Cary, NC, USA). The variables were ln or square root transformed to improve the normality of their distribution. In figures, however, unmodified variables were presented to show the actual values. In the stepwise linear multiple regression models, all cases with an absolute value of the Studentized residual exceeding 3 were excluded and only the independent variables with P-values less than 0.05 were included.

3 Results

3.1 Lake Water Chemistry

[13] South-north differences (reflecting predominantly climate and atmospheric deposition) in the average lake water TOC, TP, Fe, and CO2 concentrations were slightly larger than land use driven differences (reflecting land use, vegetation and soil chemistry) among the southern lakes (Table 1) indicating that climatic drivers and/or atmospheric deposition contribute significantly to lake water chemistry. Winter TOC concentrations in the north were less than summer TOC concentrations. The variation in the average TOC concentrations from the lake surface to the bottom was, however, small both during winter and summer, under different land use patterns and under a wide climatic gradient (Table 2). Highest TOC concentrations were found in the southern lakes surrounded either by a high proportion of peatland or agricultural land. Catchment land use in our lakes represents typical patterns in the boreal zone; lakes are predominantly surrounded by forests and peatlands with a lower proportion of agricultural land (Table 3).

Table 1. Mean Water Chemistry (4 Seasons, 3 Depths) of the 177 Randomly Selected Finnish Lakes
 Northa n = 23South 1b n = 73South 2c n = 81South 1/ NorthSouth 2/ South1
  1. a

    Lakes north from the Arctic Circle.

  2. b

    Lakes south from the Arctic Circle with land use similar to northern lakes (agricultural land % < 3 and peatland % < 40).

  3. c

    The rest of the southern lakes (agricultural land % > 3 or peatland % > 40).

  4. South 1/North and South 2/South 1 ratio reflect predominantly differences in water quality due to climate and land use, respectively.

TOC (mg L–1)6.510111.51.2
TIC (mg L–1)
CO2 (μM)1101701801.51.1
O2 %7766640.91.0
TN (µg L–1)3405006701.41.4
NH4-N (µg L–1)4252631.21.2
NO3-N (µg L–1)19491102.52.2
TP (µg L–1)1221281.81.4
PO4-P (µg L–1)
Fe (µg L–1)680110010001.61.0
Mn (µg L–1)721062021.51.9
Table 2. Mean Water Chemistry (177 Randomly Selected Lakes) for Selected Variables in Depth Profile Samples
 WinterSouth 1b n = 70South 2c n = 83SummerSouth 1b n = 70South 2c n = 83
Northa n = 24Northa n = 24
  1. X = max depth of the lake (see Table 1 for other abbreviations).

TOC (mg L–1)      
1 m4.810128.09.312
X-20 cm5.512147.89.411
TIC (mg L–1)      
1 m4.
X-20 cm8.
CO2 (µM L–1)      
1 m230250290365755
X-20 cm44046046084300250
O2 (%)      
1 m626158898686
X-20 cm171820754044
1 m6.
X-20 cm6.
TN (µg L–1)      
1 m290550750410460630
X-20 cm510730920450610760
NH4-N (µg L–1)      
1 m404975207.08.1
X-20 cm25024024033120100
NO3-N (µg L–1)      
1 m621402405.94.513
X-20 cm5478180112256
TP (µg L–1)      
1 m8.51522141726
X-20 cm154847183947
PO4-P (µg L–1)      
1 m3.
X-20 cm5.417134.31414
Fe (µg L–1)      
1 m330710810530530670
X-20 cm35003700250075020502200
Mn (µg L–1)      
1 m314790243563
X-20 cm56039475848237471
1 m0.81.11.312.816.116.0
X-20 cm3.53.83.411.611.011.7
Table 3. Average, Minimum, and Maximum Values for Catchment and Lake Area, Maximum Depth, Catchment Land Use, Molar C/N Ratio in Water (1 m Samples) and in Sediment, and C Evasion/Accumulation Ratio for the Lakes (n = 82) Included in C Balance Studies
Catchment area (km2)43000.3353,000
Lake area (km2)1500.041540
Max depth26193
Agricultural (%)6.90.139
Peatland (%)160.152
Water area (%)130.8943
Urban (%)0.670.0213
Forest (%)633891
Water C/N233.356
Sediment C/N126.522

[14] The variability of TIC concentrations from the lake surface to the bottom was larger than the variability of TOC (Tables 2 and 4). Highest TIC concentrations at all depths were recorded during winter ice cover period, whereas lowest TIC values in bottom water samples were recorded during the spring and autumn overturn periods, reflecting the seasonal distribution of CO2 concentrations in Finnish lakes [Kortelainen et al., 2006a]. In winter, TIC depth profile patterns in the south and north were almost identical (Table 2).There was a wide range in the average TOC/TIC ratio calculated over different seasons and depths, reflecting primarily the variability in TIC accumulation in the water column. The ratio was highest (> 6) in surface water samples during the open water period and lowest (1.6) in winter close to the sediment resulting from the accumulation of CO2 during the ice cover period (Table 4).

Table 4. Mean TOC and TIC Concentrations and TOC/TIC Ratios (Calculated Separately for Each Lake, the Average Value Presented in the Table) in the Depth Profile Samples Before (Winter) and After (Spring) Ice Melt, at the End of Summer Stratification and During Autumn Overturn in 177 Randomly Selected Lakes
  1. X = max depth of the lake.

TOC (mg L–1)    
X-20 cm129.69.89.9
TIC (mg L–1)
X-20 cm8.
X-20 cm1.

[15] Sediment processes have been shown to significantly contribute to lake water chemistry in Finnish lakes, especially in agricultural landscape [e.g., Ekholm and Mitikka, 2006]. Also, in our data O2 percentage in winter close to the bottom was low (< 20%) both in northern and southern lakes, reflected in elevated Fe, Mn, NH4-N, TIC, CO2, TN, TP, and PO4-P concentrations. During summer, O2 percentage close to the bottom was higher, over 70% in northern lakes and over 40% in southern lakes. Correspondingly, the summertime concentrations of Fe, Mn, NH4-N, and TIC close to the bottom were higher in the southern lakes than in the northern lakes (Table 2). Also, TOC, TN, TP, PO4-P, and NO3-N were somewhat higher in bottom water in summer, whereas pH was lower reflecting accumulated CO2.

[16] In the south, the longer open water period and stronger temperature depth gradient enable more favorable conditions for the development of stronger summer stratification. In the north, summer temperature gradient from the lake surface to the bottom was small. Consequently, lake chemistry in the northern lakes was more evenly distributed from the surface to the bottom.

3.2 Landscape C Balance

[17] Both average annual C evasion and average Holocene C stocks in sediments were largest in small, shallow lakes, which have large catchment areas and low percentage of water in the catchment. C evasion and C accumulation correlated also with each other (Figure 2), although the relationship was not very strong. The variability in lake area (from 0.04 to 1 540 km2) was larger than in previous studies. Consequently, lake area was an important driver both for CO2 evasion and C accumulation and explained as much as 96% of the total C retention (C evasion + C accumulation) in our lakes (Figure 3).

Figure 2.

The relationship between average annual CO2 evasion and average annual Holocene C accumulation.

Figure 3.

The relationship between lake area and (a) average annual CO2 evasion, (b) average annual Holocene C accumulation, and (c) total C retention (C evasion + C accumulation) in the lakes.

[18] Water chemistry was also linked both to C evasion and C accumulation. Color, turbidity, Fe, Mn, and TOC had strong positive correlation both with C evasion and C accumulation, whereas low pH and O2 resulted in increasing C evasion and C accumulation (Table 5). The primary driver behind these patterns is, however, difficult to determine, because lake area is an important driver for water chemistry; small lakes often have higher color, TOC, Fe, and Mn concentrations but lower pH and O2. Furthermore, upstream lakes modify downstream lake chemistry, and lake percentage was strongly linked both to C evasion and C accumulation.

Table 5. Correlation Coefficient Values Between Cev/Cacc, C Evasion, C Accumulation, N Accumulation and Sediment C/N Ratio With Lake and Catchment Area, Maximum Depth, Catchment Land Use, Water Chemistry, C Evasion, C Accumulation, N Accumulation, and Sediment C/N Ratio
 ln Cev/Caccln CO2 Evasionln Sediment C Accumulationln Sediment N Accumulationln Sediment C/N Ratio
  1. Significant coefficients *p < 0.05, **p < 0.01 and ***p < 0.001 are shown, ns = not significant.

Catchment and lake
ln Lake areans–0.711 ***–0.656 ***−0.631 ***–0.670 ***
ln Max depth0.224 *−0.638 ***–0.695 ***−0.600 ***–0.531 ***
ln Catchment areans−0.662***−0.691***−0.604***−0.617***
√ Field%nsnsns0.348 **ns
√ Peat%nsnsnsns0.245 *
√ Water%ns– 0.579 ***−0.489 ***−0.445 ***−0.440 ***
Water chemistry     
ln TOCns0.452 ***0.410 ***0.335 **0.345 **
ln NH4nsns0.250*ns0.266 *
ln NO30.290 **−0.234 *−0.328 **−0.242 *−0.279 **
ln TPns0.279 *0.249 *0.325 **ns
ln Fens0.471 ***0.490 ***0.432 ***0.321 **
ln Mnns0.406 ***0.513***0.543 ***0.286 *
ln Colorns0.608 ***0.580 ***0.462 ***0.510 ***
ln Conductivitynsnsnsns−0.271 *
pHns−0.587 ***−0.406 **−0.216 *−0.609 ***
ln Alkalinitynsnsnsns−0.350 **
O2ns−0.553***−0.396***−0.326 **−0.505 ***
ln Turbidityns0.488***0.516***0.615 ***ns
ln CO2 evasion0.270 *1.00   
ln C accumulation−0.551***−0.655***1.00  
ln N accumulation−0.538***0.571 ***0.913***1.00 
ln Sediment C/N ratio−0.251*0.568 ***0.642 ***0.361 ***1.00

[19] In contrast, when lake specific Cev/Cacc, which in our data ranged from 4 to 86, was compared with numerous drivers like lake chemistry, lake morphology, catchment land use, latitude, and sediment properties, the only significant relationships (if C-related variables were not used as predictors) were found with sediment N pool (Figure 4), lake water NO3-N concentration (Figure 5), and maximum depth of the lake (Table 5). If C accumulation/evasion ratio (instead of Cev/Cacc) was compared with all possible background drivers excluding C-related variables, the only significant correlations were found with the same predictors and correlation coefficient values were very similar as with Cev/Cacc, but had the opposite direction; sediment N pool was positively linked, lake water nitrate and maximum depth negatively linked.

Figure 4.

The relationship between sediment nitrogen pool and C evasion/accumulation ratio. Four outliers were excluded.

Figure 5.

(a) The relationship between lake water NO3-N concentration (autumn overturn samples from 1 m depth) and C evasion/accumulation ratio. (b) Mean Cev/Cacc in lakes with NO3-N concentration ≤ 100 µg L–1, 100 < NO3-N ≤ 200 µg L–1 and NO3-N > 200 µg L–1. Two outliers were excluded.

[20] When stepwise linear multiple regression models were used to explain Cev/Cacc using water chemistry, lake morphometry, latitude, catchment, and sediment data as predictors, the highest explanation power (r2 = 0.58, n = 68) was achieved with a model that selected nitrogen (N) accumulation, percentage of the catchment covered by water and urban area, lake water pH, and alkalinity as predictors

display math(1)

[21] Excluding catchment characteristics resulted in a comparable explanation power of the model (r2 = 0.60, n = 70) with N accumulation, lake water conductivity, pH, and alkalinity as predictors

display math(2)

[22] Excluding all N variables but including water chemistry, lake morphometry, catchment characteristics, and latitude as predictors resulted in a significantly lower explanatory power of the model (r2 = 0.21, n = 76)

display math(3)

[23] These correlation and regression analyses indicate N to be a key factor contributing to the role of lakes in long term C balances in boreal catchments. Also, land use, maximum depth (reflecting catchment topography), lake acidity and ionic strength significantly improved (up to 33%) the explanation power of the models. pH and alkalinity contribute to the inorganic carbon speciation and are thus closely linked to lake CO2 concentrations/evasion, whereas conductivity is a similar kind of bulk parameter for inorganic substances as DOC for organic substances.

3.3 N Accumulation

[24] Similar to C accumulation, also N accumulation was largest in small, shallow lakes, which have large catchment areas and low percentage of water in the catchment. Furthermore, N accumulation increased with increasing proportion of agricultural land in the catchment. Among water chemistry the same variables, i.e., turbidity, Mn, color, and Fe concentrations, had the highest correlation coefficient values both with C and N accumulation (Table 5).

[25] When stepwise linear multiple regression models were used to explain average annual N accumulation the highest explanatory power (r2 = 0.62, n = 72) was achieved with a model that selected lake area, the percentages of field and urban area in the catchment, lake water Fe content, and catchment to lake area ratio

display math(4)

[26] Despite the large variability in lake area (from 0.04 to 1 540 km2), maximum depth (from 1 m to 93 m), latitude (from 60°C to 69°C), land use, and topography (lakes from headwater catchments downstream to large drainage basins), C accumulation and N accumulation were very closely linked to each other (Figure 6). The stepwise linear multiple regression models both for C accumulation [Kortelainen et al., 2004; equation (1)] and N accumulation (this study; equation (4)) selected the following common predictors: lake area, field percentage, and lake water Fe content. Small lakes that are surrounded by a high proportion of agricultural land, reflected in high lake water Fe content, have thus highest C and N stocks in boreal Finnish lake sediments.

Figure 6.

(a) The relationship between sediment carbon and nitrogen pool and (b) the relationship between average annual carbon and nitrogen accumulation. One outlier was excluded.

[27] Sediment C/N ratio significantly correlated with numerous drivers with the highest negative correlation coefficient values with lake area, catchment area, lake water pH, and maximum depth. Highest C/N ratios in lake sediment were thus recorded in small, shallow, low pH lakes (Table 5).

4 Discussion

4.1 Lake Water Chemistry and C Pools and Fluxes Reflect Climate, Topography, and Land Use

[28] The study lakes are located over a wide climatic gradient between the latitudes 60°N and 69°N. Besides climatic drivers, the spatial, seasonal, and interannual variation in lake chemistry reflects catchment land use, atmospheric deposition, and sediment water interaction, the latter being closely linked to catchment topography and lake morphology. In the boreal zone, climate is closely linked to atmospheric deposition. For example in Finland, N, P, and TOC deposition generally decreases from south to north [e.g., Kortelainen et al., 1997]. Impacts of atmospheric deposition and climate are thus difficult to separate; year-to-year variations in atmospheric deposition are closely linked to the ones in climate [Weyhenmeyer, 2008]. Despite these patterns only a few water quality variables indicated a strong south-north gradient throughout the year (Table 1). Average pH was higher in the north reflecting both lower acid deposition, lower organic acid export from northern catchments and more fertile soils in the south [Kortelainen et al., 1989]. Also, NO3-N concentrations were higher in the south reflecting higher N deposition and more intensive land use in the south, i.e., higher field percentage.

[29] The average TOC, TP, and Fe concentrations had slightly stronger links to climate and/or deposition compared to land use (Table 1), presumably reflecting significantly longer snow cover period in the northern catchments and longer ice cover period in the northern lakes resulting in decreased terrestrial export of TOC, TN, TP, and Fe [Kortelainen et al., 1997]. Also, CO2, NO3-N, and PO4-P had somewhat stronger links to climate and/or deposition than to land use (Table 1). The average length of the ice cover period ranges from about 5 months in southern lakes to over 7 months in northern lakes [Laasanen, 1982] although during many recent mild winters both the snow and ice cover periods have been shorter. Autochthonous production in predominantly oligotrophic boreal lakes is minor and majority of organic matter is transported from surrounding catchments [Jonsson et al., 2001; Larmola et al., 2004; Einola et al., 2011]. Majority of the annual organic matter load from boreal catchments is transported during spring and autumn high flow periods and many element fluxes, including TOC, generally follow river runoff patterns [e.g., Kortelainen et al., 1997; Laudon et al., 2004; Lepistö et al., 2008; Räike et al., 2012]. Although changing climate and hydrological conditions can alter the lateral transport of carbon through numerous mechanisms, recent studies have indicated the important role of precipitation as a key driver for C fluxes [e.g., Rantakari and Kortelainen, 2005; Raymond and Oh, 2007; Roehm et al., 2009; Ojala et al., 2011; Butman and Raymond, 2011; Sadro and Melack, 2012], whereas regional pCO2 patterns were shown to be closely linked to altitude [Lapierre and del Giorgio, 2012].

[30] Striegl et al. [2001] demonstrated that Finnish lakes had generally greater pCO2 and lighter d13CDIC compared to lakes in Minnesota and Wisconsin, which indicates respiration to be the primary CO2 source in Finnish lakes. Nevertheless, the relatively stable TOC patterns over different seasons and depths compared to more variable TIC patterns (Table 4) do not support the degradation of TOC in the water column to play a major role to TIC/CO2 concentrations in Finnish lakes. The variability in the average TOC concentrations from the lake surface to the bottom was minor over different seasons, land use patterns, and a wide climatic gradient. In contrast, there was a large range in the average TOC/TIC ratio over different seasons and depth profiles, reflecting primarily the variability in TIC accumulation in the water column. In boreal dimictic lakes sediment interaction can be expected to be pronounced during winter and summer stratification when lakes are more isolated from the surrounding catchment than during spring and autumn high flow periods. This was presumably also reflected in the seasonal variation of TIC in Finnish lakes: (1) TIC and CO2 concentrations were highest in winter and linked to low O2 and high Fe, Mn, and NH4-N concentrations—presumably reflecting metabolism of organic matter in the sediment-water interface—whereas (2) during spring and autumn high flow periods TIC and CO2 concentrations were lower and more evenly distributed in the water column. Furthermore, summer stratification is stronger in the south resulting in lower O2 content and higher Fe, Mn, NH4-N, and TIC concentrations close to the sediment (Table 2). Elevated concentrations of Fe, Mn, NH4–N, TIC, and CO2 might also be due to groundwater input. However, Finnish Lake Survey in 1987, based on randomly selected lakes, demonstrated drainage lakes to be the major lake type (70 %) followed by headwater lakes (17%). Minority of the lakes were classified as seepage (10%) or closed (3%) [Kortelainen, 1993], indicating groundwater contribution in Finnish lakes to be generally small compared with many other regions in the world.

4.2 Nitrogen Is a Key Driver for Cev/Cacc

[31] In single lakes located in variable terrains, Cev/Cacc can show a large variability [e.g., Sobek et al., 2006; Finlay et al., 2010; Tranvik et al., 2009; Einola et al., 2011; Ferland et al. 2012], reflecting variable contributing drivers. Also, in our data the Cev/Cacc ratio varied from 4 to 86. Mean Cev/Cacc C was 30, slightly higher than the average estimate for Finnish lakes weighted by lake area (21) [Kortelainen et al., 2006a]. However, instead of focusing on most important drivers for Cev/Cacc in single lakes we use our data to study the role of lakes in landscape C balances.

[32] Terrestrial C balances have been intensively studied for numerous years, whereas freshwater C balance studies have traditionally been based on a few lakes neither with a focus on landscape patterns nor based on representative lake populations. Nevertheless, lakes contribute significantly to landscape C balances, evading C gases to the atmosphere and accumulating C in sediments. Finnish lakes are supersaturated both with CO2 and CH4 throughout the year, thus releasing C gases continuously to the atmosphere during the ice-free period and accumulating high concentrations of CO2 and CH4 in the water column during the winter ice cover period [Kortelainen et al., 2006a; Juutinen et al., 2009]. Although average annual C sequestration in Finnish lake sediments during the Holocene is small compared to average annual C evasion to the atmosphere, areal C stocks in sediments are larger than in upland forest soils [Kauppi et al., 1997; Kortelainen et al., 2004]. Our data demonstrate that boreal lakes play an important role in landscape C balances. The average annual CO2 emission from Finnish lakes to the atmosphere was estimated as 1.4 Tg C, approximately 20% of the average annual C accumulation in Finnish upland forest soils and tree biomass [Kortelainen et al., 2006a; Liski et al., 2006].

[33] In 177 randomly selected boreal lakes smaller than 100 km2, O2, NH4-N, and Mn were important drivers for CO2 evasion [Kortelainen et al., 2006a], whereas C pool in sediments was linked to lake water Fe and NO3-N concentrations [Kortelainen et al., 2004, equation (1)]. Terminal electron acceptors (i.e., O2, NO3-N, Mn, Fe oxides) thus played a key role in small boreal lakes regulating both C gas evasion to the atmosphere and average long-term C accumulation in sediments. As much as 79% (n = 2 740, p < 0.0001) of the variation in CO2 departure from the saturation in these lakes could be explained by O2 departure from saturation [Kortelainen, et al., 2006a]. Recently, Weyhenmeyer et al. [2012] demonstrated that O2 and hydrological patterns were key drivers for CO2 concentrations both in boreal lakes and streams, whereas mean elevation was shown to play a key role in regional CO2 patterns in boreal lakes [Lapierre and del Giorgio, 2012]. Altitude and slope have been shown to be important predictors both for spatial NO3-N and DOC distributions [D'Arcy and Carignan, 1997; Kortelainen et al., 2006b; Helliwell et al., 2007; Sobek et al., 2007], although nitrate and DOC distributions often show opposite topographical patterns.

[34] Both lake area and maximum depth showed larger variability (0.04–1540 km2 and 1–93 m, respectively) in our data than in previous studies. Consequently, lake area and depth were important predictors both to C evasion and C accumulation. Both CO2 evasion to the atmosphere and areal C stock in the sediment were highest in small, shallow lakes, which can be considered as biogeochemical “hot spots” within the terrestrial landscape. Furthermore, the total C retention (C evasion + C accumulation) was largely explained by lake area (Figure 3). Both C evasion and C accumulation correlated also with water chemistry: positively with color, turbidity, Fe, and Mn concentrations and negatively with O2 and pH. The primary driver behind these relationships is difficult to predict, since small Finnish lakes are predominantly colored and have high TOC, Fe, and Mn concentrations, but low pH.

[35] When lake specific Cev/Cac was compared with lake morphometry, lake chemistry, sediment characteristics, catchment land use, and climatic drivers, only very few statistically significant correlations were found: negative correlation with sediment N pool (Figure 4) and positive with lake water NO3-N concentration (Figure 5). Furthermore, maximum depth was one of the few variables significantly correlated to Cev/Cacc. Depth of the lake is linked to topography, which was shown to play a key role in regional CO2 patterns [Lapierre and del Giorgio, 2012]. Furthermore, sediment C/N ratio was negatively linked to Cev/Cacc.

[36] Although N deposition and climate are closely linked in the boreal zone, none of the other drivers (including latitude and lake water temperature) showed significant correlations with Cev/Cacc. Although lake specific precipitation/runoff/deposition data was not available, atmospheric deposition in Finland decreases from south to north, precipitation is somewhat larger in the south, whereas runoff increases to the north due to higher evapotranspiration in the south. Precipitation, runoff, and deposition are thus linked to latitude.

[37] Molot and Dillon [1996] studied lateral C mass balances of 20 small catchments and seven lakes in central Ontario and suggested that the ratio of evaded/accumulated C should increase with decreasing alkalinity. In our data including lakes from headwater catchments downstream to large drainage basins reflecting thus variable land use patterns and topography, Cev/Cacc correlated significantly neither with alkalinity nor with pH (Table 5). However, stepwise multiple regression models selected alkalinity and/or pH as predictors, but only after maximum depth (equation (3)) or N and catchment land use or conductivity (equations ((1))–((2))). Excluding N variables as predictors resulted in significantly lower explanation power of the model: maximum depth, pH, and alkalinity explained only 21% of the variation in Cev/Cacc (equation (3)). Our boreal data thus indicate N to be a key contributing factor contributing to the role of lakes in long term C balance in boreal zone.

[38] C and N accumulation rates and sediment C and N stocks were closely linked to each other (Figure 6). Furthermore, Cev/Cacc was positively linked to lake water NO3-N concentrations) indicating more effective C sequestration (i.e., lower respiration) in lakes with low NO3-N concentrations. This is in agreement with the study by Wickland et al. [2012] showing the dominating role of inorganic nitrogen in controlling the biodegradability of DOC in the Yukon River. Furthermore, Booth et al. [2005], conducting a synthetic analysis of 15N pool studies reported in terrestrial literature, found that soil N content exerted the strongest impact on soil N mineralization. A meta-analysis of 15N pool dilution studies of gross ammonification revealed that also the C/N ratio exerts a significant negative influence on gross ammonification indicating the higher N yield per unit of degraded soil organic matter at low C/N ratios.

[39] Most of the organic C in Finnish lakes originates from surrounding forests and peatlands resulting in low NO3-N concentrations and high C/N and C/P ratios in headwater streams and downstream lakes. Terrestrially produced organic matter has typically higher C/N ratio compared to organic matter produced in lakes. In our data average C/N ratio was lower in the sediment compared with lake water (12 vs. 23) (Table 3), presumably reflecting differences in the source, mineralization, and age of organic matter. Furthermore, C/N ratio reflects upstream-downstream position in the landscape; in boreal headwater streams significantly higher C/N ratio values have been recorded (average 48 [Kortelainen et al., 2006a]). Both DOC and CO2 in these headwater streams have been shown to be young [Billett et al., 2012]. Meyers and Takemura [1997] studied Lake Biwa in Japan and concluded that during organic matter degradation organic carbon is converted to CO2 or CH4. These two gases diffuse out of the sediment, but organic N converts to NH4, which binds to clay minerals in the sediment. These contrasting fates of C and N lead to gradually smaller C/N ratios with greater time of burial.

[40] Phosphorus limited primary production in freshwater ecosystems has been shown in many regions and also CO2 flux from Quebec lakes was shown to be associated with total P concentrations [del Giorgio and Peters, 1994]. Recently, Lapierre and del Giorgio [2012] demonstrated that TP:DOC ratio was important predictor for regional pCO2-DOC patterns. Nevertheless, recent studies focusing on boreal lakes have given evidence both to N limited production [Bergström and Jansson, 2006] and degradation [Kortelainen et al., 2006a] contrary to generally accepted view on P limitation. The synthesis by Dodds and Cole [2007] showed that N regularly stimulates widely heterotrophic and autotrophic activities in freshwater and coastal ecosystems. Consistent with these findings NO3-N and NH4-N concentrations showed the highest seasonal variability among all water quality parameters: during summer inorganic N concentrations were significantly lower than during winter, whereas PO4-P concentrations showed minor seasonal variability (Table 2). Similarly, in Swedish lakes and streams NO3-N concentrations showed the highest growing season variability among all water quality variables [Khalil and Weyhenmeyer, 2009], whereas in the Yukon River and its tributaries inorganic nitrogen was shown to regulate the biodegradability of DOC [Wickland et al., 2012].

[41] Majority of organic matter in boreal freshwater systems is terrestrially fixed and accumulation of C and N in forests occurs through the same mechanisms, production of dead organic matter and microbial turnover, i.e., the net accumulation is regulated by the balance between production and decomposition. Globally, factors influencing decomposition rates may actually play a more important role in sequestration of organic carbon in soils than productivity [Cebrian and Duarte, 1995]. Nitrogen has been shown to be a key element regulating the production, structure, and function of terrestrial ecosystems. Meyer et al. [2006] demonstrated strong N limitation in the northern taiga and southern tundra, and Wallenstein et al. [2009] suggested that enzyme activity was linked to N availability in arctic soils, being low in summer presumably due to N limitation. Furthermore, Magnani et al. [2007] showed net C sequestration to be overwhelmingly driven by N deposition in temperate and boreal forests; Gross primary production, ecosystem respiration, and Net ecosystem production only weakly correlated with temperature, and had no correlation with annual precipitation or latitude.

[42] Although N deposition may have increased C storage in northern forests to some extent [Pregitzer et al., 2008], ecosystem carbon storage in arctic tundra was reduced by long-term nutrient fertilization [Mack et al., 2004]. Root biomass was lower when using fertilizers, fertilization thus seemed to decrease below ground C sequestration. Minkkinen et al. [2007] demonstrated that the annual CO2 fluxes from peatlands drained to forestry significantly increased from nutrient-poor to nutrient-rich sites. Khan et al. [2007] analyzed results of long-term experiments in agricultural soils in the USA, and concluded that 40 to 50 year of inorganic fertilization caused a net decline in soil C content, despite increasingly massive residue C incorporation. Fertilization was of little, if any, benefit for soil C sequestration; addition of N or P was more effective for stimulating mineralization of soil organic C.

[43] We have shown that the role of lakes in long-term C balance in N-limited boreal landscape is linked to N. Nevertheless, changing climate and extreme weather conditions significantly contribute both to seasonal and annual C budgets thus modifying the role of lakes in landscape C balances. Einola et al. [2011] demonstrated that C budgets in a chain of boreal lakes showed highly variable patterns during a dry and a wet year. The variability in annual precipitation regulated TOC and TIC input and output from lakes and significantly contributed also to CO2 fluxes from lakes to the atmosphere. Extreme weather conditions have been shown to modify terrestrial C budgets rapidly. Ciais et al. [2005] demonstrated a European-wide reduction in primary productivity caused by the heat and drought in 2003; a strong anomalous net source of CO2 to the atmosphere reversed the effect of four years net ecosystem carbon sequestration. Piao et al. [2008] demonstrated net CO2 losses from northern ecosystems due to increasing respiration during autumn warming. Also, peatland C sequestration rates have been shown to be highly sensitive even to minor climatic fluctuations, with wet periods correlating rapidly with increasing peat accumulation [Yu et al., 2003]. Climate change scenarios predict increasing temperature and precipitation for the northern latitudes [Denman et al., 2007]. The increasing frequency of heavy precipitation events and wet/dry periods might further increase the variability in the timing and magnitude of the loads. Shorter ice cover and soil frost period with earlier melting periods in winter and delayed soil frost in late autumn might result in significant feedbacks in C and N cycling, which are difficult to separate from the patterns of atmospheric deposition.

5 Conclusions

[44] Although an increasing number of studies have recently focused either on CO2 evasion from lakes to the atmosphere or C burial in sediment, the role of lakes in regional C cycling and major drivers behind the balance between CO2 evasion and C burial, i.e., the role of lakes in catchment C balance have remained largely unknown. Identification of primary drivers of landscape C cycling among numerous contributing factors is challenging and the response of freshwater ecosystems to increasing temperature and/or changing precipitation patterns is still highly uncertain. In single lakes Cev/Cacc can show significant variability (also in our data from 4 to 86). Thus, results from small-scale and/or short term experiments can even be misleading when trying to identify major drivers behind long term patterns over topographically variable regions, thus delaying overall understanding on the most important drivers at landscape scale.

[45] Carbon in boreal lakes is predominantly terrestrially fixed, and freshwater ecosystems thus reflect and integrate terrestrial C cycling. Our randomly selected lakes include major lake types and land use patterns in the boreal zone located between the latitudes 60°N and 69°N. TOC concentrations were higher than TIC concentrations in all seasons and depths. The average TOC/TIC ratio was highest (> 6) in summer surface water and lowest (1.6) in winter close to the sediment resulting from the accumulation of CO2 during winter ice cover period.

[46] Although landscape C cycling integrates physical, chemical, and biological processes and linkages between atmospheric, terrestrial, and aquatic C cycling, our data indicate that the role of lakes in long-term C balance in boreal landscape is primarily linked to N. Both lake chemistry and carbon pools and fluxes were linked to several drivers reflecting topography, climate, land use, and deposition, whereas Cev/Cacc was significantly correlated only with N and maximum depth of the lake, which reflects catchment topography. In northern boreal zone atmospheric N deposition is generally low and biological processes are N limited owing to low primary production and slow mineralization of organic matter in cold, nutrient-poor, acidic conditions. Lake water NO3-N and NH4-N concentrations showed highest seasonal variability among all water quality parameters further indicating inorganic N to be a key limiting nutrient. Our results are in agreement with recent signals from boreal forests, peatlands, and agricultural land indicating N limited respiration to be a key driver for long term C balance. Climate change scenarios predict increasing precipitation and temperature for Northern Europe, which together with patterns of N deposition might significantly contribute to freshwater biogeochemical cycles and landscape C balances.