Rates of atmospheric deposition are 5 – 8 kg N ha−1 y−1 in the Colorado Front Range, which represents 25 – 40% of estimated N mineralization rates in high-elevation soils [Fisk and Schmidt, 1995]. In the eastern U.S. and Europe, deposition rates are as high as 11 – 20 kg N ha−1 y−1 [Bergström and Jansson, 2006; Tørseth and Semb, 1998] but represent only 5 – 17% of estimated soil N mineralization rates [Groffman et al., 2006; Zak et al., 2006]. While deposition rates in Colorado are less than those of other regions, they constitute a relatively large fraction of N that is cycled. As a result, ecosystems in alpine and subalpine regions of Colorado may be more sensitive to ecological effects of atmospheric N. Indeed, soil N content and soil N mineralization rates have increased in old-growth forests and the C:N ratio of foliage has decreased compared to reference sites [Baron et al., 2000]. In Colorado lakes, changes in the composition and biomass of the diatom community, altered stoichiometric ratios of N to P in the water column, and shifted phytoplankton nutrient limitation have been associated with N deposition [Elser et al., 2009a]. In the present study, we found evidence that concentrations of dissolved N2O were greater in surface waters of lakes subject to elevated N deposition, while the reverse was true for dissolved CH4. Methane production by sediments and surface water CH4 concentrations were negatively related to NO3−, while N2O concentrations were positively related to NO3−. Our data suggest that atmospheric N loading affects greenhouse gas dynamics in lakes by increasing NO3− concentrations.
4.1. Carbon Mineralization in Sediments
 Contrary to our expectations, we found no evidence that atmospheric N deposition has enhanced C cycling rates in lake sediments. Rather, sediment CO2 production was greater in low-deposition lakes compared to high-deposition lakes. For all sampled lakes, C released by sediment CO2 production was two orders of magnitude greater than that released through CH4 production. Further, there was no difference in sediment OM and C content of bulk sediments between regions, but the concentration of DOC was greater in low deposition lakes compared to high-deposition lakes. The higher concentration of DOC in low-deposition lakes is consistent with greater sediment C mineralization rates here compared to the high-deposition region. However, we expected the opposite result because of the fertilization effect that N deposition could have on primary production in the catchments and water column. Indeed, previous fieldwork at many of the same lakes found that concentrations of chlorophyll and seston C were significantly greater in high-deposition lakes [Elser et al., 2009b]. Such differences between studies could reflect seasonal or year-to-year differences in catchment or lake productivity. It is also possible that growth of heterotrophic bacterioplankton in low deposition lakes are N limited, allowing for DOC to accumulate in the water column [Taylor and Townsend, 2010]. Last, at current rates of atmospheric N loading, catchment-specific properties for the sampled lakes might have a greater influence on lake DOC concentrations than does N deposition, as has been observed in other regions [Hessen et al., 2009].
 Methane may represent 20 – 60% of total C mineralization in lake sediments [Bastviken et al., 2008]. In our sediment slurry incubations, however, CH4 production represented only about 7% of measured C mineralization. In fact, CH4 production by sediments for the sampled lakes was two orders of magnitude less than that reported for a eutrophic lake and at the low end of that reported for boreal lakes [Algesten et al., 2005; Bastviken et al., 2008; Liikanen et al., 2002]. The low rate of CH4 production we observed in sediment could be partially explained by temperature, as we conducted our incubations at 4°C, which is colder than temperatures for other studies. Methane production in peat soils is strongly sensitive to temperature, with Q10 values of 5.3–16 [Dunfield et al., 1993] and similar results could occur in sediments. It is also possible that our slurry incubations were not completely anoxic. During denitrification, ratios of CO2:N2 production are 1.5–6 [Groffman et al., 2006], but for our incubations the average ratio of CO2:N2O production (as a proxy for CO2:N2) was significantly greater at 95 (±26 SE). This suggests that there was available oxygen for microbial respiration. Methanogenesis is sensitive to reduction-oxidation potential and the presence of strong oxidants, such as oxygen or NO3−, will suppress CH4 production [Le Mer and Roger, 2001]. We also observed a negative relationship between water NO3− concentrations and CH4 production (Table 3). This might explain why CH4 production was greater in low-deposition lakes compared to high-deposition lakes. Further, published data for CH4 production in sediments are often collected from eutrophic lakes and lakes surrounded by peatlands or bogs, which would likely have anoxic conditions in the sediments and, thus, be more favorable for CH4 production [Huttunen et al., 2003]. Last, the generally low CH4 fluxes we observed could also be result from relative poor quality OM in sediments of the sampled lakes.
4.2. Greenhouse Gas Emissions From Lakes
 Lake gas emissions not only depend on the concentration gradient between the lake and the atmosphere, but also on the gas exchange coefficient [Wanninkhof et al., 1987]. We did not measure gas exchange rates for the sampled lakes, so we cannot determine gas emissions. However, the surface waters were supersaturated with CO2, CH4, and N2O, suggesting that the lakes are net sources of these gases to the atmosphere. The surface water concentration of CO2 was three orders of magnitude greater than that of CH4 or N2O; thus, CO2 is likely the dominant greenhouse gas emitted from the sampled lakes. Also, while CO2 is 25 and 298 times less potent than CH4 and N2O, respectively, in terms of radiative forcing, we estimate that CO2 would represent about 80% of the global warming potential of total greenhouse gases emitted from the sampled lakes [Forster et al., 2007; Liikanen et al., 2002]. The dynamics of CO2 in the surface water do not appear to be influenced by atmospheric N inputs. We did observe that surface water concentrations of CO2 were positively related to DOC and NO3− (Table 3), although there was no significant relationship with either predictor variable when considered individually. This is interesting because we found NO3− to be negatively related to sediment CO2 production and because other studies have found strong positive relationships between CO2 concentrations and DOC [Sobek et al., 2003]. It is possible that respiration by bacterioplankton and sediment bacteria respond differently to N loading, thus, it would be useful to further investigate the extent to which bacteria in sediments and pelagic areas contribute to CO2 in surface water of the sampled lakes. Overall, however, our data are consistent with the well-documented finding that most lakes are heterotrophic and sources of CO2 to the atmosphere [Cole et al., 1994; Tranvik et al., 2009].
 There are two major pathways for CH4 emissions from lakes, diffusion and ebullition (bubble flux) of gas produced in the sediments. Ebullition may account for 20–70% of CH4 emissions from sediments, especially in shallow, eutrophic lakes [Bastviken et al., 2008; Juutinen et al., 2009]. Our approach for measuring CH4 concentrations only reflects diffusive sediment fluxes and is at the very low end of the range reported for lakes using similar methods [Bastviken et al., 2004]. While there is a general lack of CH4 data for high-elevation lakes, Smith and Lewis  sampled five lakes in the Colorado Front Range, including one of the lakes we visited (Long Lake). The dissolved CH4 concentration we observed for Long Lake is within the range they reported. Overall, evidence presented here suggests that N deposition has reduced sediment CH4 production and concentrations of dissolved CH4. Methanogenesis does not appear to be a significant source of greenhouse gas in the sampled lakes, independent of atmospheric N inputs. However, methanogenesis may be a significant C mineralization pathway if methane oxidation is important. Indeed, up to 80% of CH4 produced in sediments may be consumed in the water column [Bastviken et al., 2008].
 Our study lakes were generally supersaturated with N2O and concentrations were comparable to other lakes, although the number of data sets from other lakes is limited (Table 5). The data presented here suggest that atmospheric N deposition has increased concentrations of dissolved N2O and potentially emissions of N2O from lakes, consistent with studies of other aquatic ecosystems [Beaulieu et al., 2011; Liikanen et al., 2003; Seitzinger et al., 1984]. It is important to note that studies of N2O dynamics in aquatic ecosystems have largely been conducted in the Northern Hemisphere in temperate and boreal climates, and that data for areas such as the tropics are lacking. The lakes we sampled are located in the same region with similar elevations and climate. Thus, while they are comparable to those reported in the literature, they may not be representative of other regions or land uses, such as those in urban or agricultural areas. Our understanding of the factors influencing N2O dynamics, as well as spatial and temporal patterns of N2O emissions would greatly benefit from further experimental research in aquatic ecosystems.
Table 5. Comparison of Dissolved N2O and Sediment N2O Fluxes Among Different Studies
|Dissolved N2O, nmol N|
|26 lakes, CO, USA||26.4||11–58||112–208%||This study|
|Lacamas Lake, WA, USA||20.1||9–33||143–312%||Deemer et al. |
|Taihu Lake, China|| ||25–62||80–689%||Wang et al. |
|Lake Baldegg, Switzerland|| ||10–120|| ||Mengis et al. |
|15 lakes, Switzerland||43.9||14–152||99–798%||Mengis et al. |
|N2O Sediment Fluxes, μmol N m−2h−1|
|17 lakes, CO, USA||0.4||0.0–5.6|| ||This study|
|32 lakes, Norway||0.5||0.0–6.8|| ||McCrackin and Elser |
|Lake Kevätön, Finland|| ||0.4–7.1|| ||Liikanen et al. |
|Lake Kevätön, Finland|| ||0–1.3|| ||Liikanen et al. |
|Humber Estuary, UK|| ||0.2–25.2|| ||Barnes and Owens |
|Narragansett Bay, RI, USA|| ||0.0–0.9|| ||Seitzinger et al. |
 In lakes, N2O is produced by denitrification in sediments and by nitrification in sediments and the water column [Mengis et al., 1997; Wrage et al., 2001]. For the lakes we sampled, the microbial source of N2O in the sampled lakes is unclear. We expected a correlation between sediment N2O production and lake water concentrations of N2O. Denitrification-related N2O production, however, was only observed in sediments of four lakes. We do not know why there were no N2O fluxes for the majority of sediments we sampled. Similar assays conducted with sediments from lakes in Norway found significantly greater N2O fluxes in lakes that receive elevated levels of N deposition [McCrackin and Elser, 2010]. Incubation temperatures (4°C for Colorado and 15°C for Norway) could explain the differences between studies. If our assays were not completely anoxic, rates denitrification and related N2O production could have been repressed. Also, differences in N2O production may also be the result of incomplete denitrification in Norwegian lakes due to relatively higher NO3− concentrations. The sediments are still a possible source of N2O production because experiments conducted simultaneously found considerable denitrification capacity under NO3− enrichment [McCrackin and Elser, 2011]. Nitrification could produce N2O as has been observed in soils [Bateman and Baggs, 2005]. Assays conducted with sediments from Norwegian lakes in high-deposition regions showed significantly greater rates of nitrification potential (in response to non-limiting quantities of ammonium) than those in low-deposition regions (M. McCrackin, unpublished data, 2010). Also, nitrification in the water column is a potential significant source of N2O [Mengis et al., 1997; Twining et al., 2007]. Further investigation is required to determine the microbial sources of dissolved N2O observed in lakes. It is difficult to make generalizations of our data based on summer measurements, because we do not know how concentrations of dissolved CO2, CH4, and N2O in the sampled lakes vary seasonally. The alpine and subalpine lakes that we sampled are ice-covered for more than half of the year. Stream water NO3− concentrations in LVW of the Colorado Front Range tend to peak in lake May, decrease during June and July, and increase in the fall [Baron and Campbell, 1997]. Assuming lakes follow a similar pattern, our sampling likely missed peak NO3− concentrations. Because NO3− was an important predictor of gas concentrations and sediment gas fluxes, we might expect these variables to change seasonally. Gases that accumulate under ice during winter could be released during periods of thaw and mixing. Hence, concentrations of dissolved gases that we measured in summer months may be greater than average annual concentrations.
 While CH4 and N2O are more potent than CO2 in terms of radiative forcing, the low concentrations of these trace gases relative to CO2 indicate that lake greenhouse gas dynamics are dominated by CO2. The relatively small surface area of the sampled lakes suggests that they do not contribute disproportionately to such dynamics in the Colorado Rocky Mountains. Given that N2O plays a significant role in the depletion of stratospheric ozone, however, potential N2O emissions from lakes, and the effects of N deposition on them, deserve further consideration.
 The IPCC Guidelines for National Greenhouse Gas Inventories recognize that as N fertilizer leaches from agricultural soils and moves through groundwater and river networks, a fraction is converted to N2O. Such fluxes are considered indirect N2O emissions from N used in agriculture. Indirect emissions are estimated for groundwater, rivers, and estuaries using an emission factor (EF5) of 0.0075 kg N2O-N per kg N input, with an uncertainty range of 0.0005–0.025 kg N2O-N kg−1 N. The EF5 was originally based on the ratios of dissolved N2O-N to NO3− from groundwater and agricultural drainage water in published studies [Beaulieu et al., 2008; Mosier et al., 1998]. While lakes are not currently included in greenhouse gas inventories, we used the general approach for estimating emission factors as the ratio of N2O-N:NO3−. This ratio averaged 0.01 with a range of 0.001–0.07 for the lakes we sampled, somewhat larger than the default EF5 value of 0.0075 for aquatic ecosystems. Clarifying the relationship between N2O emissions and N loading rates is important, but the dynamics of N2O in lakes are not well documented. We surveyed the scientific literature and, where concentrations of N2O and NO3− were reported, determined the N2O-N:NO3− ratio to be between 0 and 0.12 (Table 6). This broad range indicates that a single N2O-N:NO3− ratio cannot be generalized across all lakes. Further, it is not known how the ratio of N2O-N:NO3− varies seasonally. For example, in regions where lakes develop ice cover during winter, allowing N2O to accumulate, ratios of N2O-N:NO3− at spring melt could be substantially greater than those in other seasons. Nonetheless, values for the high end of the range of N2O-N:NO3− ratios suggest there is potential for N2O emissions from lakes subject to elevated N loading. Indeed, Wang et al.  reported N2O production of 300 μmol N m−2 d−1 in pelagic areas of a hyper-eutrophic lake, which is an order of magnitude greater than maximum emissions reported for agricultural fields [Bouwman et al., 2002].
Table 6. Comparison of the Ratio of N2O-N:NO3− Among Different Studies
| ||Ratio of [N2O-N]:[NO3−]||Study|
|26 lakes, CO, USA||0.01||0.001–0.07||This study|
|Taihu Lake, China||0.002||0.0003–0.02||Wang et al. |
|Lacamas Lake, WA, USA||0.009||0.002–0.02||Deemer et al. |
|Greifensee, Switzerland||0.002||0–0.017||Mengis et al. |
|Lago di Lugano, Switzerland||0.0015||0–0.009||Mengis et al. |
|Lake Baldegg, Switzerland||0.016||0.001–0.12||Mengis et al. |
|Lake Huron, USA||0.0011||0.001–0.0012||Lemon and Lemon |
|Lake Ontario, USA||0.001||0.003–0.0006||Lemon and Lemon |
|Cayuga Lake, USA||0.0003||0.0002–0.0006||Lemon and Lemon |
| ||Default Value kg N2O-N per kg N Input||Uncertainty Range||Study|
|IPCC Indirect N2O Emission Factor (EF5)||0.0075||0.0005–0.025||IPCC |
4.3. Global N2O Emissions From Lakes Receiving N Deposition
 Our data suggest a possibility that lakes could be an underappreciated component of global N2O cycling, especially given elevated N inputs. Here we attempt to quantify this possibility. Globally, the largest source of anthropogenic N2O emissions is the enhanced conversion of N fertilizer by microorganisms in agricultural soils [Forster et al., 2007]. Emissions from aquatic ecosystems, however, are also significant. Globally, rivers may convert 0.68 Tg N y−1 of anthropogenic N to N2O [Beaulieu et al., 2011]. Kroeze et al.  estimated contemporary natural and anthropogenic N2O emissions by rivers and estuaries to be between 0.3 and 2.1 Tg N y−1, which represent approximately 2–12% of total anthropogenic N2O emissions (∼17 Tg N y−1) [Denman et al., 2007].
 Lakes may receive N from a variety of sources other than atmospheric deposition, such as groundwater, sewage, agricultural run-off, and natural sources. All of these sources may contribute to lake N2O emissions. Here, we will estimate global N2O emissions related to direct N deposition to the lake surface and N deposition that is leached from the surrounding catchment. First, we first estimated global N loading to lakes via atmospheric N deposition using published data sets of N deposition rates (for 1993 and 2050) and of lakes and reservoirs with surface area >1 km2 [Dentener, 2006; Lehner and Döll, 2004]. We quantified N deposited directly to the lake surface (N kg y−1) = atmospheric deposition kg N ha−1 y−1 * lake surface area (ha). Estimated N delivered directly to lake surfaces is 1.1 and 2.1 Tg N y−1 for 1993 (contemporary) and 2050 (modeled), respectively. Based on our analysis, over 90% of the global surface area of lakes (2.6 million km2) is subject to natural and/or anthropogenic atmospheric N deposition. We accounted for watershed inputs of N deposition as (N kg y−1) = N deposition rate kg ha−1 y−1 * watershed surface area (ha) * 30%, where we assumed that that 30% of all N deposited to the watershed is subsequently leached to the lake [Intergovernmental Panel on Climate Change (IPCC), 2006]. Watershed area data were only available for large lakes (>50 km2). Based on the relationship of watershed surface area to lake surface area for large lakes reported by Lehner and Döll , we assumed for small lakes that the watershed area was 12 times greater than the lake surface area. Inputs of atmospheric N from watershed to lakes were estimated as 13.7 Tg N y−1 for 1993 and 26.1 Tg N y−1 for 2050. The extent to which lakes are subject to N deposition is of interest not only because of the potential for increased N2O emissions but also because of the documented effects of N on lake stoichiometry and food web dynamics [Elser et al., 2009b; Hessen et al., 2009]. Even lakes that are not directly influenced by human activity are at risk from atmospherically delivered pollution, which is of particular concern because N deposition rates are expected to increase globally in the next few decades, driven by energy demands and agricultural activities [Dentener, 2006].
 We next estimated N2O production using the method of Kroeze et al. , which assumes that emissions from rivers and estuaries result from nitrification and denitrification of inorganic N inputs as, N2O-N kg y−1 = (Nitrification + Denitrification) * EF. Here, anthropogenic and natural inorganic N sources are not distinguished. Seitzinger and Kroeze  describe this approach in detail. Briefly, it is assumed that 50% of N inputs are denitrified and that the nitrification rate exceeds the denitrification rate by 20%. The emission factor (EF) is 0.3% of total denitrification and nitrification except where N loading rates exceed 10 kg N ha−1 y−1, when the EF is 3%. Using this method we estimated lake N2O emissions related to total atmospheric N deposition to be 0.33 Tg N y−1 in 1993 and 0.9 Tg N y−1 in 2050.
 We also estimated N2O emissions using IPCC methodology for indirect greenhouse emissions from aquatic ecosystems. This approach is intended to estimate anthropogenic N2O emissions resulting from N run-off from agricultural systems and does specifically address atmospheric deposition to lakes. In our opinion, however, this methodology is most appropriate under current IPCC guidelines. Here, N2O emissions are determined as N2O-N kg y−1 = N inputs to lake kg N y−1 * EF5. Nitrogen inputs to lake are defined as atmospheric N deposition, calculated as previously described, and EF5 is 0.0075 kg N2O-N per kg N input (uncertainty range of 0.0005–0.025 kg N2O-N per kg N input). The resulting estimate for 1993 is 0.6 Tg N y−1 (range 0.04–2 Tg N y−1). Emissions from Asia and North America contribute about 44% and 23% of total global emissions, respectively, because of the relative high densities of lakes and regional patterns of N deposition (Figure 3a). Estimated 2050 emissions increase to 1 Tg N y−1 (range 0.07–3.4 Tg N y−1), with nearly half of this increase occurring in Asia (Figure 3b). Our 1993 estimates represent 13–95% of N2O emitted from rivers and estuaries reported by Kroeze et al. , although our estimates are not strictly comparable because we do not consider all sources of N inputs to lakes.
Figure 3. Estimated annual global lake emissions of N2O (Gg N) by country based on (a) contemporary (1993) and (b) modeled 2050 atmospheric N deposition rates and the IPCC default EF5 of 0.0075 kg N2O-N per kg N input.
Download figure to PowerPoint
 Lake emissions calculated using the approach of Kroeze et al.  fall within the ranges determined using the IPCC methodology. It should be noted that the method used by Kroeze et al. includes all natural and anthropogenic sources of N2O, while the IPCC methodology is intended to quantify only anthropogenic emissions. Compared to modeled pre-industrial rates of atmospheric deposition (e.g., 1860) [Dentener, 2006], we estimate that about 75% and 88% of N deposition in 1993 and 2050, respectively, is attributable to human activity. Accordingly, the anthropogenic portion of emissions that we estimated using the IPCC methodology are ∼0.03–1.5 Tg N y−1 in 1993 and ∼0.6–3.0 in 2050. Contemporary anthropogenic N2O emission from rivers (0.68 Tg N y−1) [Beaulieu et al., 2011], are within the range of our estimates of deposition-related emissions from lakes. Again, these estimates are not strictly comparable because we do not consider all anthropogenic N inputs to lakes.