Respiration and methanogenesis respond more strongly to temperature change than photosynthesis6, 7, 8, 9. Consequently, warming has been shown to increase CO2 and CH4 emissions and reduce carbon sequestration in experimental ponds10, 11, 12, 13. However, these experiments have either been restricted to relatively short-term responses to warming (for example, 1 year or less)10, 11 or have not investigated how the effects of warming change over time13, 14. Therefore, a key unanswered question for understanding how greenhouse gas (GHG) dynamics in freshwater ponds will respond to global warming is: do the high temperature sensitivities of methanogenesis and respiration result in increased emissions of CH4 and CO2, and reduced carbon sequestration under warming that is sustained over the long term, and which could potentially accelerate the rate of climate change?
Unlike work in aquatic ecosystems, the long-term effects of experimental warming have been assessed in terrestrial ecosystems, and tend to show elevated emissions of CO2 in the short term, driven by the exponential temperature dependence of respiration, followed by a damped effect of warming later15, 16, 17. These diminishing long-term responses have been attributed to loss of labile carbon substrates15, physiological acclimation16, evolutionary adaptation18 or community turnover through ecological dynamics19. These findings are important because they imply carbon cycle responses to global warming should be more complex than the simple exponential effect of temperature on respiration rates alone17.
Small freshwater ponds contribute disproportionately to greenhouse gas emissions budgets from inland waters5. However, whether shifts in the fluxes of CH4 and CO2 observed in short-term warming experiments10, 11 are sustained in the long term (for example, >1 year to decadal timescales) in these ecosystems is unknown, which severely limits our ability to predict whether future changes in greenhouse gas fluxes4, 5 will contribute to accelerating or slowing global warming. We tackled this fundamental knowledge gap using an array of experimental mesocosms that were designed to mimic mid-latitude ponds, to investigate the effects of long-term warming on the ecosystem-level exchange of CO2 and CH4 with the atmosphere. We present a detailed analysis of the seasonal dynamics of the key metabolic and GHG fluxes in the carbon cycle in the sixth (2012) and seventh (2013) year of the experiment, which we contrast with our initial findings from the first year (2007) of warming10, 11 to explore whether, like terrestrial ecosystems15, 16, 17, the effects of warming on the carbon cycle in freshwater ponds are dampened in the long term.
CH4 emissions were elevated in the warmed treatments in both 2007 and 2013. However, the magnitude of the effect size increased over the seven years of the experiment (Fig. 1). Annual rates of CH4 emissions were 1.5-fold higher in the warmed treatments after one year of experimental warming11, but after seven years this effect size had increased to 2.5-fold (Fig. 1 and Tables 1 and 2). Consequently, a generalized additive mixed effects model (GAMM) that included a ‘treatment’ by ‘year’ interaction on the intercept provided the best fit to the data (Table 1), demonstrating that the effects of warming on CH4 emissions were larger in 2013 than 2007.
a, The seasonal variation in CH4 flux from 2007 and 2013 demonstrates that the effect of warming was larger in 2013 than 2007. b, Box–whisker plots of annual CH4 flux from each pond calculated by integrating the seasonal data over time. These data show that the effect size of warming increased from 1.5-fold in 2007 to 2.5-fold in 2013, highlighting that the effects of warming became amplified in the long term. The solid lines denote the fixed effects from the best fitting GAMM model (see Table 1 for model selection). Red circles and lines denote warmed treatments, while the ambient treatments are in black. Tops and bottoms of boxes in box–whisker plots correspond to the 25th and 75th percentiles, horizontal white lines correspond to medians, whisker extents correspond to 1.5× the interquartile range and blue points are outliers.
We also saw a comparable amplification of the effects of warming on total ecosystem-level carbon metabolism. The effect size of warming increased from 1.2- to 1.8-fold for gross primary production (GPP) and 1.4- to 2-fold for ecosystem respiration (Reco) between 2007 and 2012 (Fig. 2a–d and Table 1). Consequently, GAMMs fitted to the seasonal distributions of GPP and Reco, and including a ‘treatment’ by ‘year’ interaction on the intercept, provided the best fit to the data. Because rates of Reco increased more with warming than those of GPP, the Reco/GPP ratio was 1.15-fold higher in the warmed treatments, indicating reduced capacity for carbon sequestration in both 2007 and 2012 (Fig. 2e, f). High-frequency measurements of CO2 exchange between the ponds and the atmosphere in 2013 confirmed these findings, with annual net CO2 uptake reduced by 50% in the warmed mesocosms (Fig. 3 and Tables 1 and 2). Together these results demonstrate that the effects of warming on the key fluxes in the carbon cycle became amplified over the seven years of the experiment, in stark contrast to the damped effects of long-term warming reported for terrestrial systems15, 16, 17. So what mechanisms might be responsible for the amplified effects of warming in freshwater ponds?
a,c,e, Seasonal distributions of rates of gross primary production, GPP, (a), ecosystem respiration, Reco, (c), and the Reco/GPP ratio (e) were fitted to generalized additive mixed effects models (see Methods). For GPP and Reco the effects of warming on median rates of ecosystem metabolism were larger in 2012 than in 2007. The Reco/GPP ratio was higher in the warmed ponds and increased between 2007 and 2012. The solid lines denote the fixed effects from the best fitting GAMM model (see Table 2 for model selection). Red circles and lines denote warmed treatments, while the ambient treatments are in black. b,d,f, Box–whisker plots of annual rates of GPP (b), Reco (d) and Reco/GPP ratio (f) from each pond calculated by integrating the seasonal data over time. Tops and bottoms of boxes in box–whisker plots correspond to the 25th and 75th percentiles, horizontal white lines correspond to medians, whisker extents correspond to 1.5× the interquartile range and blue points are outliers.
a, Seasonal distribution of net CO2 flux data collected in 2013 reveals differences in the seasonality of net daily CO2 fluxes with maximal rates of net CO2 absorption peaking earlier in the year in the warmed treatments. The solid lines denote the fixed effects from the best fitting GAMM model. Red circles and lines denote warmed treatments, while the ambient treatments are in black. b, Box–whisker plot of the annual net CO2 fluxes calculated for each pond by integrating over the seasonal data reveal higher net fluxes in the warmed treatments (indicating lower CO2 absorption). Tops and bottoms of boxes in box–whisker plots correspond to the 25th and 75th percentiles, horizontal white lines correspond to medians, whisker extents correspond to 1.5× the interquartile range and blue points are outliers.
The experimental mesocosms were seeded in 2005 with organisms and organic matter (see Methods) and have since been on a trajectory of ecosystem development. Succession theory proposes that in the early stages of ecosystem development, as organic matter and biomass accumulate, rates of GPP exceed Reco and the ratio of Reco/GPP < 1 (ref. 20). As ecosystems develop towards later successional stages, energy fixed by GPP tends to be balanced by energy consumed through Reco (that is, the ratio of Reco/GPP ≈ 1) and biomass production is maximized20. Consistent with ecosystem succession theory, the Reco/GPP ratio and annual totals for GPP and Reco all increased substantially over the course of the experiment in both the warmed and ambient treatments, with increases in GPP and Reco much larger in the warmed mesocosms (Fig. 2a–d). In line with the data on total carbon metabolism, we also observed consistently higher biomass of macrophytes (Supplementary Fig. 5 and Supplementary Table 3), phytoplankton and zooplankton in the warmed treatments in the long term21. Together, these data show that warming enhanced rates of ecosystem development, community succession and biomass accumulation, amplifying the divergence between treatments in GHG emissions and metabolic fluxes. These results demonstrate that warming can fundamentally alter the energetic balance at the ecosystem level: firstly, because in the short term, rates of respiration rise more sharply with temperature than photosynthesis (increasing Reco/GPP); and secondly, over the long term, because higher rates of metabolism drive more rapid ecosystem development22, magnifying energetic imbalances and shifts in the carbon cycle. Natural ecosystems are typified by far-from-equilibrium dynamics23, and thus, because warming can act both as a stressor and a driver of physiology, understanding the long-term impacts of warming on ecosystem properties requires both an appreciation of the acute effects of temperature change on organism metabolism and subsequent impacts on the successional dynamics of ecosystems.
Focusing on the fluxes of CO2 and CH4 measured in 2013 at a high temporal resolution reveals that, in addition to driving shifts in the annual budgets, long-term warming also profoundly altered the seasonality of CH4 emissions (Fig. 1) and net daily exchange of CO2 (Fig. 3). Rates of net daily CO2 emission (that is, days where total CO2 emissions > absorption) peaked in October in the warmed treatments, whilst, on average, ambient ponds were net sinks for CO2 over the entire year. By contrast, rates of net daily CO2 absorption (that is, days where total CO2 absorption > emissions) peaked in the warmed treatments in June, while they peaked in July in the ambient ponds (Fig. 3). These respective peaks in net CO2 absorption coincided with peak CH4 emissions (Fig. 1), implying a strong coupling between CO2 drawdown by photosynthesis and substrate supply for methanogenesis, which is a well-known characteristic of many natural aquatic ecosystems24. Indeed, the most marked effects of warming on CH4 emissions occurred during the spring and early summer (Fig. 1). Later in the year, however, the effects were negligible (Fig. 1) when rates of respiration exceeded those of photosynthesis (for example, CO2 production > CO2 consumption leading to net CO2 emissions; see Figs 1 and 3) in the warmed treatments and rates of methanogenesis may have been limited by photosynthetically derived carbon. These results suggest that the effects of global warming could shift the seasonal timing of carbon fluxes, which, in turn, affect the supply and demand of substrates that support aquatic ecosystem productivity.
Overall, our findings provide the first experimental evidence that the annual balance of greenhouse gas fluxes from freshwater ponds remain profoundly altered at inter-annual timescales, with substantially elevated CH4 emissions and lower CO2 absorption. The extent to which these results are important for understanding how carbon fluxes from globally important pond ecosystems5 respond to warming depends on whether carbon dynamics in our experimental mesocosms are broadly representative of those in natural systems. One important distinction between the mesocosms and natural shallow lakes is that they are not embedded within a watershed and consequently receive little terrestrially derived organic carbon, which is often an important carbon flux in lakes and ponds25. In consequence, carbon cycle dynamics in the mesocosms are driven predominantly by autochthonous production, which could alter sediment characteristics, coupling between photosynthesis and respiration, and GHG emissions compared with natural ponds that receive allochthonous carbon subsidies. To investigate this and assess the relevance of our findings for natural ponds, we measured the carbon (C), nitrogen (N) and C:N ratios of the mesocosm sediments (Table 2) and compared them with values from natural lakes and ponds spanning the dystrophic to oligotrophic spectrum25. The sediment characteristics of the mesocosms are similar to those from natural oligotrophic lakes26. Mean annual rates of gross primary production and ecosystem respiration (Table 2) are also comparable to those from natural lakes5, 27, indicating that rates of total carbon metabolism reflect those of natural systems. Finally, recent work has shown that the size of lakes and ponds (in terms of surface area) are critical for determining their greenhouse gas (GHG) emissions, with CH4 flux per unit area increasing as a power function of decreasing lake surface area5, 28, 29. Our mesocosms, with a surface area of 3.14 × 10−4 ha, fall within the smallest category of ponds analysed in a global synthesis of GHG emissions5 and have average CH4 concentrations that are indistinguishable from natural ponds of a similar size (Supplementary Fig. 4). Taken together, this evidence demonstrates that results from our mesocosm experiment are of direct relevance for understanding carbon cycle responses to warming in freshwater ponds. Our results suggest that profound shifts in the carbon fluxes of small ponds should be expected over the long term in a warming world, which is of particular concern considering that we are only just beginning to appreciate the importance of such small water bodies in global budgets of GHG emissions from inland waters5, 28, 29.