Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
Corresponding author: A. Abnizova, Alfred Wegener Institute for Polar and Marine Research, Telegrafenberg A6, D-14473 Potsdam, Germany. (email@example.com)
 Although ponds make up roughly half of the total area of surface water in permafrost landscapes, their relevance to carbon dioxide emissions on a landscape scale has, to date, remained largely unknown. We have therefore investigated the inflows and outflows of dissolved organic and inorganic carbon from lakes, ponds, and outlets on Samoylov Island, in the Lena Delta of northeastern Siberia in September 2008, together with their carbon dioxide emissions. Outgassing of carbon dioxide (CO2) from these ponds and lakes, which cover 25% of Samoylov Island, was found to account for between 74 and 81% of the calculated net landscape-scale CO2 emissions of 0.2–1.1 g C m−2 d−1 during September 2008, of which 28–43% was from ponds and 27–46% from lakes. The lateral export of dissolved carbon was negligible compared to the gaseous emissions due to the small volumes of runoff. The concentrations of dissolved inorganic carbon in the ponds were found to triple during freezeback, highlighting their importance for temporary carbon storage between the time of carbon production and its emission as CO2. If ponds are ignored the total summer emissions of CO2-C from water bodies of the islands within the entire Lena Delta (0.7–1.3 Tg) are underestimated by between 35 and 62%.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Temperate and boreal reservoirs excluding Lokka and Porttipahta
2. Site Description
 The study area is located within the continuous permafrost zone of northeastern Siberia (Figure 1a). This area has developed and preserved several hundreds of meters of deep, cold, permafrost as a result of the absence of extensive ice sheets during the Pleistocene and extremely cold climatic conditions (the present-day mean annual air temperature is about −14.9°C); it is believed to contain the world's largest carbon stocks in soils [Romanovskii et al., 2000; Tarnocai et al., 2009; Boike et al., 2012]. The Lena Delta, which covers about 29,000 km2, is the largest Arctic delta [Gupta, 2007] (Figure 1b). The study area on Samoylov Island (4.3 km2) is located in the central part of the Lena Delta (72°22′ N, 126°28′ E; Figure 1c). Large numbers of (polygonal) ponds and thermokarst lakes dominate the polygonized landscape of the island (Figures 1c–1e). The ponds typically range in diameter from about 10 to 30 m, with water depths of 0.5 to 1 m. In contrast, the larger lakes have surface areas of several thousands of square meters and have water depths up to 5 m; they developed as a result of thermokarst processes and thawing of the underlying permafrost. Some lakes have developed outflow channels that eventually reach the main channel of Lena River.
 The area acts as a carbon sink during July and August and becomes a CO2 source from the beginning of September until winter [Kutzbach et al., 2007]. Published methane emissions of 3.2 g CH4-C m−2 yr−1 in small ponds and 1.6 g CH4-C m−2 yr−1 in lakes here are relatively low [Zhang et al., 2012] compared to other areas in the Arctic [Walter Anthony et al., 2010], due to either limited nutrient availability or reduction of methane emissions as a result of oxidation by submerged brown mosses [Liebner et al., 2011].
3.1. Water Balance Measurements
 In order to cover a range of different freshwater systems on the island, we investigated selected ponds, thermokarst lakes, and island outflows (Figure 1c). The water balances for lakes and ponds were assessed for the period from April to September 2008 using the following equation:
where dS/dt is the change in water volume.
 Snow surveys were conducted within the study area prior to snowmelt. Rainfall was measured with recording rain gauges (Tipping bucket 52203, RM Young, ±2%) set at a height of 0.3 m above ground level. Water levels were monitored continuously at all of the investigated lakes and ponds using pressure transducers (SensorTechnics BTE6000, ±0.2%) attached to Campbell Scientific data loggers. The measurements were made every minute and averaged over 60-min intervals, and their reliability was checked against manual water table measurements. Discharge rates (in m3/s) from lake and floodplain outflows were monitored using gauging stations equipped with V notch weirs (Thomson-V 70°, ±5–10%) established at the start of the season. Water level recorders at these stations provided regular measurements (in cm) at 60 min intervals. The evaporation from water surfaces was evaluated using standard micrometeorological measurements: the applied methods are described in detail by Muster et al. .
3.2. Water Analyses
 Water samples from lakes, ponds, and outflows were collected at weekly intervals during the study period (i.e., 6 August to 21 September 2008). Samples for dissolved organic carbon (DOC) analysis were filtered using a PE (polyethylene) syringe and GF (glass microfiber) filters (0.7 μm pore diameter) at the time of sampling, collected in 30 ml HDPE plastic (High-density polyethylene) containers, and acidified to pH 2 by adding 2 M HCl. Unfiltered samples for the analysis of dissolved inorganic carbon (DIC) and dissolved nitrous oxide (N2O) concentrations were collected in headspace-free, sealed 20 ml glass vials. Samples in glass vials were stored in the dark at 4°C. Some of the glass vials were unfortunately damaged in transit, so that only the samples from September were available for the analysis of dissolved gases. Unfiltered samples for the determination of total organic carbon (TOC) concentrations were also collected in 30 ml HDPE plastic containers, and were kept frozen until laboratory analysis. Concentrations of DIC and dissolved N2O were analyzed using a gas chromatograph (Shimadzu GC-2014AF) equipped with an AOC-5000 autosampler, a 1 m × 1/8'' HayeSep Q 80/100-mesh column, an electron capture detector, and a flame ionization detector. Gas concentrations and total pressure of the headspace were analyzed after shaking the solutions at 90°C for 20 min on a rotary shaker. The gas concentrations in water were calculated from the headspace concentrations by applying Henry's law. Results were corrected for temperature, pressure-dependent residual gas concentrations, and pH-dependent carbonate equilibrium (only for DIC). We used the equations in Plummer and Busenberg  for temperature adjustment of equilibrium constants.
 The DOC and DIC stocks in lakes and ponds were estimated from the concentration (C) in mg/m3 and volume (V) in m3. The retention of carbon was calculated using the following equation:
 The CO2 emissions from lakes, ponds, and outflows were calculated according to Repo et al. :
where kgas is the gas exchange constant, Caq is the concentration of CO2 + H2CO3 (H2CO3*) in the water, and gassat is the H2CO3* concentration of the water in equilibrium with the atmosphere calculated using temperature-adjusted Henry's law constants. The kgas constant was estimated using the approaches of Cole and Caraco  and Crusius and Wanninkhof . The dissolved concentrations of CO2 (+ H2CO3) required for calculating the gaseous emissions were derived from DIC concentrations and temperature-corrected pH values [Ben-Yaakov, 1970], determined from the gas-tight vials in the laboratory, using temperature-adjusted dissociation constants [Plummer and Busenberg, 1982].
 The isotopic composition of water (δD, δ18O) was determined for the surface water samples and for water samples obtained from a frozen soil core by an equilibration technique [Meyer et al., 2000], using a mass-spectrometer (Finnigan MAT Delta-S).
3.3. CO2 Flux Measurements
 The net CO2 flux (net ecosystem exchange) for the terrestrial tundra was obtained from high frequency measurements of wind speeds and CO2 concentrations, using an eddy covariance system established within the polygonal tundra in the western part of Samoylov Island. The area around the eddy system features a comparably low density of ponds and lakes (Figure 1c). The three-dimensional wind vectors and sonic temperatures were measured with a sonic anemometer (CSAT3, Campbell Scientific Ltd., UK), at a height of 2.4 m above ground level. The CO2 concentrations were detected using an open path gas analyzer (LI-COR 7500). All measurements were conducted with sampling rate of 20 Hz and stored on a data logger (CR3000, Campbell Scientific Ltd., UK). A detailed technical description of the applied eddy system is provided in Langer et al. . The calculation of half hourly net ecosystem exchange (NEE) averages, including quality control on the stationarity and the integral turbulence characteristics, was performed using TK2 post-processing software [Mauder et al., 2006]. The flux source area for each NEE value was estimated on the basis of the footprint model after Schmid . Those NEE values with a greater than 4% probability of being affected by lakes or ponds within the flux source areas were discarded. The meteorological station was located in the same vicinity as the eddy covariance station (Figure 1c).
3.4. Up Scaling of CO2 Emissions
 Landscape-scale CO2 emissions were calculated as a linear combination of emissions from terrestrial tundra, lakes, and ponds, as follows:
where A is the dimensionless proportion of the total surface area that is made up of terrestrial tundra (terr.), lakes, and ponds, and E is the average CO2 emission from terrestrial tundra, lakes, and ponds. The surface area proportions of the three landscape units were inferred from a supervised classification of high-resolution aerial images of the study area [Muster et al., 2012].
4. Results and Discussion
4.1. Water Balance
 Our estimation of the seasonal water budget for the studied ponds in 2008 showed that losses through evaporation were offset by similar precipitation inputs prior to the freezeback, resulting in a general state of equilibrium (Table 2). On average, evaporation rates were about 2 mm d−1, with a maximum of 3 mm d−1. Lake and pond water levels varied less than 10 cm during the study period. Overall, the water balance from April to September 2008 was in equilibrium, i.e., the precipitation input (233 mm) was only slightly higher than the evapotranspiration output (190 mm). The average snow water equivalent (SWE) was about 65 mm (average SWE taken from transects over the island), of which about half evaporated during the month of May (15 mm). The remaining water was most likely stored in ponds and lakes, and no visible runoff was generated (automated camera pictures). The summer rain (June–September) totaled 163 mm, most of which fell during the month of June (60 mm). The total runoff from the island during the entire summer period amounted to only about 10% of the total precipitation, illustrating the dominance of vertical water fluxes.
Table 2. Water Balance Estimates for Samoylov Island in 2008
Table 3. Limnological and Chemical Characteristics as Well as Concentrations of Carbon and N2O, Together With Isotopic Compositions, for Ponds and Lakes of Samoylov Island During the Study Period From 6 August to 21 September 2008 (Mean Values, With Ranges in Parentheses)a
Thermokarst Lake 1
Thermokarst Lake 2
Freezeback values from 21 September 2008 are italicized. Note: n/a stands for not available data, an asterisk (*) indicates analytes sampled only from September 5 to September 21.
Surface area (m2)
Maximum depth (m)
 Aerial photographs indicate that lakes and ponds cover 13% and 12%, respectively, of the island's surface, with the remaining 75% being either wet or dry tundra [Muster et al., 2012] (Figure 1c). Assuming that the island's net CO2 emissions can be expressed as a linear combination of CO2 emissions from tundra, lakes, and ponds (equation (4)), we can calculate an average landscape-scale CO2 emission of 0.2–1.1 g C m−2 d−1 for September 2008 prior to freezeback, which is of the same order of magnitude as the NEE of −0.07 to 0.44 g CO2-C m−2 d−1 previously determined for September 2003 [Kutzbach et al., 2007]. The contribution to these emissions from lakes and ponds was between 74 and 81% (depending on the method used to calculate gas transfer coefficients), with roughly one half of the CO2 originating from ponds. During freezeback the landscape-scale CO2 emission equaled approximately 1.6 to 1.9 g C m−2 d−1, of which about 70% was derived from ponds and about 7% from lakes. Compared to these gaseous emissions of CO2 from the water surface, the export of carbon with flowing water during the study period was negligible because of the small volumes of runoff (Figure 3 and Table 2).
 A tentative carbon budget for lakes and ponds revealed that CO2 emissions to the atmosphere exceeded the retention of DIC and DOC as well as stocks of dissolved carbon in the water column (Figure 3). This then raises the question of where all the CO2 emitted to the atmosphere has come from. The required inputs of carbon could be explained by subsurface inflow, or by thawing of ground ice that is rich in dissolved carbon (average DOC concentration in ground ice: 155.07 ± 59.84 mg l−1 (J. Boike, unpublished data, 2011)), or by mineralization of organic carbon in the sediment of lakes and ponds. Similar absolute stable isotope ratios (δD and δ18O) and slopes of regression between δD and δ18O for thermokarst lake water (slope estimate: 6.20) and permafrost ground ice (slope estimate: 6.16) suggest that thawed ice is the main source of lake water with potential mixing with precipitation (Figure 4). The isotopic composition of pond waters is, however, determined by summer rain and evapotranspiration, as indicated by the slope of the regression line (Figure 4). The CO2 emissions from ponds are therefore unlikely to have been fuelled by carbon inputs from the thawing of ground ice or subsurface inflows, and sediment respiration is probably the main source of the CO2 in these shallow ponds [Kortelainen et al., 2006]. Furthermore, the oxidation of methane escaping from the sediment of ponds, for example in submerged layers of brown mosses [Liebner et al., 2011], could provide an additional source of CO2 in these water bodies.
 Ponds with surface areas of less than 1000 m2 cover 1577 km2 of the Lena Delta, with a further 3008 km2 covered by lakes that have areas of 3600 m2 or more [Muster et al., 2012]. By assuming an ice-free period of 100 days per year we can extrapolate the average CO2 emissions measured from lakes and ponds in September 2008 to an overall annual emission of approximately 0.7 to 1.3 Tg CO2-C from the entire area of islands within the Lena Delta, with the lower value calculated excluding freezeback data and transfer coefficient according to Cole and Caraco , and the upper value including freezeback data and transfer coefficient according to Crusius and Wanninkhof . These emissions are underestimated by 35–62% if emissions from water bodies are neglected. In addition to the lakes and ponds within the Lena Delta, the river itself is supersaturated with CO2 compared to the atmosphere by up to 1.5–2 fold in summer and up to 4–5 fold in winter [Semiletov et al., 2011]. Therefore direct emissions of CO2 from the river will increase the total CO2 emissions from surface waters of the delta in addition to the emissions we report here.
 Our results demonstrate that water bodies within polygonal tundra landscapes are hot spots for CO2 emission from late summer until the beginning of the period of frozen ground. The CO2 emissions from lakes and ponds are an order of magnitude higher than the net CO2 flux observed above vegetated tundra surfaces. The findings of this study are limited to only one month due to loss of samples during transportation (September, 2008), however these high CO2 emissions and their large areal extent indicate the importance of these water bodies in the carbon cycle of the polygonal tundra. This study has shown, for the first time, that not only lakes but also small ponds must be regarded as effective processors, transient stores, and conduits of carbon in permafrost landscapes, a conclusion that has wide-ranging implications for current evaluations of the Arctic carbon budget and its sensitivity to future changes in climate and landscape, since:
 I. The actual CO2 emissions from vast areas of the Arctic are commonly underestimated since ponds are commonly not included in the global land surface classifications employed in estimating the Arctic carbon budget.
 II. Small variations in, for example, precipitation, evaporation, or drainage have the potential to produce considerable changes in the CO2 emission potential of large Arctic regions as small changes in the water table level can have a major effect on the volumes and surface areas of ponds.
 III. Earth system modeling faces a formidable problem of scale since Arctic ponds are an important component in the carbon cycle of many Arctic regions but are far too small to be represented in land surface schemes used for global circulation models.
 We would like to thank Hanno Meyer for critical discussions of isotope data; Niko Bornemann and Maren Grüber for data collection; Gerhard Kattner, Boris Koch, Hanno Meyer, Antje Eulenburg, and Reimo Kindler for supporting laboratory analyses; Jonathan J. Cole for kindly checking the calculation of gaseous CO2 emissions; and the Russian–German research station on Samoylov for providing logistical and financial support. This research was supported financially by the Alfred Wegener Institute for Polar and Marine Research, a Helmholtz Young Investigator grant (VH-NG 203) awarded to Julia Boike, and by a DAAD Scholarship stipend awarded to Anna Abnizova.