SEARCH

SEARCH BY CITATION

Keywords:

  • climate change;
  • optimum temperature;
  • temperature acclimation;
  • temperature adaptation;
  • thermal optimality

Summary

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information
  • It is well established that individual organisms can acclimate and adapt to temperature to optimize their functioning. However, thermal optimization of ecosystems, as an assemblage of organisms, has not been examined at broad spatial and temporal scales.
  • Here, we compiled data from 169 globally distributed sites of eddy covariance and quantified the temperature response functions of net ecosystem exchange (NEE), an ecosystem-level property, to determine whether NEE shows thermal optimality and to explore the underlying mechanisms.
  • We found that the temperature response of NEE followed a peak curve, with the optimum temperature (corresponding to the maximum magnitude of NEE) being positively correlated with annual mean temperature over years and across sites. Shifts of the optimum temperature of NEE were mostly a result of temperature acclimation of gross primary productivity (upward shift of optimum temperature) rather than changes in the temperature sensitivity of ecosystem respiration.
  • Ecosystem-level thermal optimality is a newly revealed ecosystem property, presumably reflecting associated evolutionary adaptation of organisms within ecosystems, and has the potential to significantly regulate ecosystem–climate change feedbacks. The thermal optimality of NEE has implications for understanding fundamental properties of ecosystems in changing environments and benchmarking global models.

Introduction

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Ecosystems are represented in most Earth system models in order to simulate the responses and feedbacks of land processes to climate change (Chapin et al., 2002; Friedlingstein et al., 2006). Virtually all ecosystem carbon cycle models use temperature response functions that have been developed from our understanding of biochemical processes of enzymes and/or physiological processes of leaf photosynthesis and organic matter decomposition (Enquist et al., 2003; Friedlingstein et al., 2006) because we lack essential knowledge about ecosystem-level responses to temperature change.

The integrated response of ecosystem processes to temperature change is influenced by the responses of its constituents, including plants, animals, microbes, and their interactions. It is well documented that plants (Mooney et al., 1978; Berry & Bjorkman, 1980; Atkin & Tjoelker, 2003), animals (Parmesan, 2006), and microbes (Bradford et al., 2008) acclimate and/or adapt to prevailing environmental conditions in a way that can optimize their functioning under varying temperatures, which is collectively termed optimality (Parker & Maynard Smith, 1990). Thus, we hypothesize that the integrated response of an ecosystem, as an assemblage of interacting organisms, might also demonstrate thermal optimality under temperature change (Loreau, 2010). Nevertheless, thermal optimality at the ecosystem level has not yet been carefully examined.

Net ecosystem exchange (NEE) of carbon dioxides is an ecosystem process that reflects the balance of gross primary productivity (GPP) and ecosystem respiration (Re), and is often used to approximate net ecosystem productivity (NEP = −NEE)(Chapin et al., 2006). Previous studies have shown that plant photosynthesis and GPP can acclimate to temperature change via an increase in optimum temperature in a warmer environment (Mooney et al., 1978; Baldocchi et al., 2001; Niu et al., 2008). It has also been documented that Re responds exponentially to temperature as long as there is no soil water limitation (Law et al., 1999), and its temperature sensitivity (Q10) decreases in a warmer environment, a process also described as temperature acclimation (Lloyd & Taylor, 1994; Luo et al., 2001; Tjoelker et al., 2008; Piao et al., 2010). The temperature acclimation of either GPP or Re can lead to changes in the temperature response of NEP. As illustrated in the conceptual Fig. 1, if GPP acclimates to temperature by increasing its optimum temperature in a warmer environment while Re stays unchanged (Fig. 1a), the optimum temperature of NEP at the maximum differences between GPP and Re may also shift upward. Alternatively, if Re acclimates to temperature with decreasing Q10 at higher temperature, while GPP stays unchanged (Fig. 1b), the optimum temperature of NEE will also increase. If both GPP and Re acclimate, the optimum temperature of NEE will likely increase more than that in the previously described two conditions (Fig. 1c). A recent study suggests thermal adaptation of NEE based on a positive relationship between the optimum temperature of NEE and the average summer temperature across space (Yuan et al., 2011). Still, we do not know yet which of the these possible processes contributes to the shift of optimum temperature of NEE, and whether or not the optimum temperature of NEE shifts over time.

image

Figure 1. Conceptual figure for the shifts of optimum temperature of net ecosystem productivity (NEP; NEP = −NEE (net ecosystem exchange)) as a result of the changes in optimum temperature of gross primary productivity (GPP) or the temperature sensitivity of respiration. (a) Here it is assumed that in a warmer year or at a warmer site, the optimum temperature of NEP shifts higher owing to a shift of optimum temperature of GPP. In (b) it is assumed that the shifts in the optimum temperature of NEP are the result of the temperature acclimation of Re (decrease of Q10). In (c) it is assumed that the optimum temperature of NEP shifts higher owing to acclimation of both GPP and Re. The dashed curves represent the temperature response curve in a warmer year or at a warmer site. The vertical lines refer to the maximum NEP.

Download figure to PowerPoint

With the advent of eddy-covariance measurements of NEE, together with detailed meteorological variables (Baldocchi et al., 2001), it has become possible to characterize the ecosystem-level temperature response. In this study, we quantified temperature responses of NEE to determine whether NEE exhibits thermal optimality over time and space, and we further explored the underlying mechanisms. Here, we define acclimation as short-term reversible adjustments of NEE over time and adaptation as geographical changes which are the result of long-term genetic and species compositional changes in ecosystems and the physiological adjustment of those organisms to the prevailing environment. The thermal optimum of NEE and its acclimation and adaptation together are described as thermal optimality. Thermal optimality of whole ecosystem carbon exchange is here assumed to represent ensemble responses of multiple temperature optima of productivity, metabolic, and decomposition processes among organisms within the ecosystem.

Materials and Methods

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Site information and data analysis

Ecosystem carbon fluxes and meteorological data used in this analysis were taken from standardized files from the North America Carbon Program (NACP), AmeriFlux, CARBOEUROPE, and the FLUXNET-LaThuile datasets. These data have been quality-controlled and gap-filled using consistent methods (Papale et al., 2006; Moffat et al., 2007). We analyzed only those sites that have at least 1 yr of complete meteorological data because we needed to calculate mean annual temperature, precipitation and solar radiation. In total, 169 sites with 818 site-years were used to investigate the optimum temperature of NEE (inline image). The sites included eight major terrestrial biomes: deciduous broadleaf forests (DBF), mixed forests (MF), evergreen needleleaf forests (ENF), grasslands (GRA), evergreen broadleaf forests (EBF), wetlands (WET), savanna (SAV) and shrubs (SHR) (Supporting Information, Table S1, Fig. S1). The climatic zones of the sites included the polar tundra arctic, continental temperate, subtropical Mediterranean, and tropical areas. The quality-controlled and gap-filled database, including eddy-covariance fluxes of CO2 (NEE), GPP and Re, were used together with solar radiation, air temperature, and precipitation in this study.

We used daily accumulated NEE rather than half-hourly values to obtain the temperature response curve with the intention of avoiding diurnal variations caused by asynchrony between cyclic changes in photosynthesis and respiration. Daily values, which integrate diurnal cycles of photosynthesis and respiration, are more robust than hourly values in reflecting ecosystem responses to temperature over the season. Hourly data during a few h near midday were used to test the robustness of the temperature response functions of NEE as discussed in the supplementary materials.

For each site or year, temperatures were binned in 1°C daily temperature. The daily air temperature and NEE were averaged in each temperature bin to quantify the temperature response of NEE. In practice, the running mean of every three temperature bins was calculated, and the value inline image, at which the maximum carbon uptake was attained, was determined from the curve relating NEE with temperature. The robustness of the parabolic pattern of NEE response to temperature was analyzed as shown in the supplementary materials (Figs S2–S7). We used the same method of calculating inline image to estimate the optimum temperature of GPP (inline image).

To determine the temperature dependence of Re, we fitted the daily mean temperature and Re data to a modified Van’t Hoff equation: Re = RbebT (Lloyd & Taylor, 1994). The parameter b determines temperature sensitivity of respiration (Q10) and can be expressed in terms of the Q10 coefficient as b = ln(Q10)/10. Although there are other types of Re acclimation, for example, changes in absolute Re rates with a constant Q10 (type II), and functional switches in both absolute Re and Q10 (type III) (Atkin & Tjoelker, 2003; Bradford et al., 2008), we only assessed type I acclimation (changes in Q10) in this study as a result of the limitation of eddy flux data to address the other two types.

Uncertainty assessment

The bootstrapping method, which is a statistical technique based on building a sampling distribution by resampling from the dataset, was used to estimate the uncertainty of each optimum temperature (Banks et al., 2010). We constructed a number of resamplings of the observed dataset (of equal size to the observed dataset), each of which was obtained by random sampling with replacement from the original datasets. We estimated one optimal temperature from each resampled dataset. This was done 5000 times to obtain 5000 optimum temperatures, from which we computed a mean and standard deviation of inline image, as well as the distribution of statistics for each site or year (Fig. S8). The same methods were used to obtain the mean and standard deviation of inline image and Q10.

Statistical analysis

To examine the potential thermal acclimation of inline image and the underlying mechanisms, we analyzed the relationship of inline image with environmental factors and with inline image and Q10 across the years within each site and across all site-years at 12 sites with >10 yr of data. Because each site has relatively few yr of data, we used standardized residual analysis to remove outliers by using a magnitude of ± 2.0 to omit points from the regression. One outlier was removed from each of IT-Ren (Renon, Italy, 2007) and Ru-Fyo (Fyodorovskoye, Russia, 2006). To examine the potential thermal adaptation of inline image, we used linear regression to analyze the relationship of inline image with environmental factors and with inline image and Q10 across all the 169 sites. All the regression analyses were performed using SPSS 11.0 for Windows (SPSS Inc., Chicago, IL, USA).

Results

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

We used results from the five sites (Fig. 2) to illustrate the general pattern of peak-curve temperature responses of NEE with the optimum temperature shifting over different temperature zones (from cold to warm). NEE increased (i.e. became more negative) with temperature in the lower temperature range to reach a maximum, and then declined (became less negative) as temperature increased further (Fig. 2). The temperature responses of NEE were caused by the relative changes in GPP and Re. At low temperatures, GPP increased more than Re for a unit of temperature change. Nevertheless, at higher temperatures, Re increased further with temperature (without water limitation), while GPP leveled off or even decreased (Fig. S9). As a consequence, there was always a well-defined inline image, at which NEE attained the maximum value (Fig. 2).

image

Figure 2. The general pattern of peak-curve temperature response of net ecosystem exchange (NEE) at five representative sites over different temperature zones (mean annual temperature ranges from −1.17 to 12.5°C). At all five sites, net ecosystem CO2 uptake (negative NEE) increased with temperature in the low range, reached a maximum, and then declined in the high range.

Download figure to PowerPoint

Twelve sites with measurements longer than 10 yr in our database were examined to illustrate interannual shifts of inline image. In comparison with the values obtained in the average year, inline image shifted to a higher temperature in warmer years at each site (Fig. 3). Across all years, inline image tended to increase with annual mean air temperature at all sites, although this was statistically significant only at some sites (Fig. 4a). Across all years of the 12 sites, annual air temperature explained 35% of the changes in inline image. Solar radiation had no significant relationship with inline image across years within any site or across all site-years (> 0.05, Fig. 4b). Although there was no significant relationship of inline imagewith precipitation across the years within any site (= 0.11–0.97), we found that across all site-years, precipitation explained 23% of the changes in inline image (< 0.01, Fig. 4c). Multiple (stepwise) regression analysis showed that air temperature and precipitation together explained 39% of changes in inline image, while temperature and precipitation alone explained 31 and 8%, respectively, of the changes in inline image, suggesting that temperature was the main factor regulating the shifts of inline imageacross site-years.

image

Figure 3. Shift of the optimum temperature of net ecosystem exchange (NEE) to a higher value in a warmer year compared with the average of all studied years at the 12 sites with > 10 yr of data length. Black closed circles, average over all years; red closed circles, represent a warmer year (mean ± 1 SD). Ta is annual mean temperature. The numbers labeled in the figures are the optimum temperatures of NEE and their standard deviation.

Download figure to PowerPoint

image

Figure 4. The relationship between optimum temperature of net ecosystem exchange (NEE; mean ± SD) and annual mean air temperature (a), global solar radiation (b) and precipitation (c) at the 12 sites with > 10 yr of data. * and *** indicate that the relationship was significant at the < 0.1 and 0.001 levels, respectively. Site abbreviations and names are as follows: CA-Man, BOREAS NSA, Canada; CA-Oas, Sask.SSA Old Aspen, Canada; IT-Ren, Renon, Italy; Ru-Fyo, Fyodorovskoye, Russia; US-Ha1, Harvard Forest, USA; BE-Vie, Vielsalm, Belgium; IT-Col, Collelongo, Italy; DE-Tha, Anchor Station Tharandt, Germany; DK-Sor, Soroe, Denmark; FR-Hes, Hesse Forest, France; NL-Loo, Loobos, Netherlands; and BE-Bra, Brasschaat, Belgium.

Download figure to PowerPoint

Across all site-years, inline image was correlated positively with inline image, but negatively with Q10 (Fig. 5). Multiple regression analysis showed that inline imageand Q10 together explained 36% of the changes ininline image, but the contribution of inline imagewas greater (29%) than Q10 (7%). The optimum temperature of GPP (inline image) was positively related to mean annual air temperature across the site-years at the Canadian sites and at the other 10 sites (Fig. S10a). Solar radiation and precipitation contributed little to the shifts of inline image. Changes in Q10 correlated negatively with temperature and precipitation, but were not related to the changes in solar radiation (Fig. S10).

image

Figure 5. The relationship of optimum temperature of net ecosystem exchange (NEE) to optimum temperature of gross primary productivity (GPP) (a) and temperature sensitivity of respiration (Q10) (b) across the site-years at the 12 sites with > 10 yr of data.

Download figure to PowerPoint

Globally, inline imagewas higher at warmer than at cold sites, leading to a positive relationship between inline image and annual mean air temperature regardless of climate zone (Fig. 6a). Mean annual precipitation and solar radiation together only explained 5% of the changes in inline image based on multiple regression analysis. Across all sites, inline image had a positive linear relationship with inline image, but no significant relationship with Q10 (Fig. 6b,c).

image

Figure 6. The relationship of optimum temperature of net ecosystem exchange (NEE; means ± SD) to annual mean air temperature (a), optimum temperature of gross primary productivity (GPP) (b), and temperature sensitivity of respiration (Q10) (c) in tropical and subtropical areas (closed circles) and temperate and boreal areas (open circles) across all 169 sites.

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

We found an universal peak-curve pattern of NEE in response to temperature, a phenomenon which was first reported in a high-elevation subalpine forest (Huxman et al., 2003). The peak pattern of the temperature response curve could be explained by relative changes in GPP and Re. The increase in carbon uptake at low temperature resulted from a relatively larger increase in GPP than Re for a given change in temperature. The decrease of carbon uptake at higher temperatures resulted from the continuous increase in Re against the leveling-off or decline of GPP (Figs 1, S9).

Shifts of inline image across years were primarily caused by temperature (Fig. 4), suggesting thermal acclimation at interannual scales. The upward trend of inline image with increasing temperature was largely the result of the shifts of optimum temperature in photosynthesis (Fig. 5), which has been well documented at plant level (Mooney et al., 1978; Berry & Bjorkman, 1980; Niu et al., 2008; Gunderson et al., 2010) and primarily results from increased electron transport capacity and/or greater heat stability of Rubisco (Sage & Kubien, 2007). In addition, extended growing seasons, increased nitrogen mineralization, and enhanced root growth (Penuelas & Filella, 2001; Churkina et al., 2005; Luo et al., 2009) may also have contributed to the increased CO2 uptake under higher temperatures, leading to the upward shift in the optimum temperature of GPP in warmer years (Fig. S10a). For example, fine roots in boreal ecosystems are more active in warmer years, allowing greater access to water and nutrients (Jarvis & Linder, 2000; Pregitzer et al., 2000). Although respiration normally increases with temperature, water stress and respiratory acclimation during high temperatures can offset or reverse the direct temperature effect (Luo et al., 2001; Irvine et al., 2008; Scott et al., 2009). Increases in GPP overcompensated for any respiration increases at higher temperature, leading to the upward shift of inline image in warmer years. The relative roles of these processes may vary at different sites, leading to site-specific responses of inline image to temperature (Fig. 4). Thus, the mechanisms underlying thermal optimization of NEE across temporal scales most likely vary among sites and require further investigation.

Shifts of inline image with annual mean air temperature across space are consistent with previous studies on geographical shifts of optimum temperature of GPP and NEE with summer or growing season temperatures (Baldocchi et al., 2001; Yuan et al., 2011), also suggesting thermal adaptation of NEE. Such spatial shifts in inline image likely result from processes at several hierarchical scales and may involve adaptive changes of various organisms. At the ecosystem level, thermal adaptation of NEE likely reflects associated changes in the structure and function of plant, animal, and microbial communities, which are likely the result of both long-term genetic changes and physiological adjustment to the prevailing environment (Diaz et al., 2001; Janssens et al., 2001; Knohl et al., 2005; Larsen et al., 2007). At the community level, species competition may also be a very important determinant of the changes ininline image. The universal existence of inline image and its acclimation and adaptation to temporal and spatial temperature changes strongly suggest thermal optimality of NEE. Compared with previous studies (Baldocchi et al., 2001; Yuan et al., 2011), this study holistically reveals the thermal optimality of NEE via acclimation and adaptation and withstands more rigorous uncertainty analysis. A particularly novel point found here is the identification of the relative roles of photosynthesis and respiration. The shifts of inline imageare mostly attributable to changes in GPP rather than in Re. Although respiration has been documented to be the main determinant of intraregional forest carbon balance variation (Valentini et al., 2000), our study clearly showed that the temperature response of GPP overrides Re in determining thermal optimality of NEE. According to Atkin & Tjoelker (2003) and Bradford et al. (2008), changes in Q10 as used in this study only reflect type I acclimation. However, if the type II and type III acclimation occur, the upward shift of inline imagein a warmer environment can also happen. Owing to the limitations of eddy flux data, we can only address type I acclimation in this study. This might cause partial assessment on Re acclimation.

Thermal acclimation/adaptation of NEE, in conjunction with the adaptation of GPP (Baldocchi et al., 2001) and plant and soil respiration (Atkin & Tjoelker, 2003; Bradford et al., 2008), will influence the long-term responses and feedback of ecosystem carbon storage to global temperature changes. Shifts of inline image with interannual variations in temperature may dampen short-term temperature effects on ecosystem carbon balances and could contribute to a lower apparent sensitivity of global ecosystem carbon uptake to temperature than to rainfall (Tian et al., 1998; Angert et al., 2005). These shifting thermal optima with temperature over time and space may lower the sensitivity of the carbon cycle feedback to future climate warming (Frank et al., 2010; Mahecha et al., 2010).

The observed patterns of thermal optimality over time and space represent a useful benchmark for evaluating the capability of Earth system models (Randerson et al., 2009; Frank et al., 2010). Benchmarking against these observed sensitivities is not sufficient to validate models but represents an important first step in assessing model capabilities on centennial time scales(Randerson et al., 2009; Frank et al., 2010). Models that replicate short-term patterns of ecosystem optimization are not guaranteed to succeed on longer timescales. However, models that do not replicate these observed temperature sensitivities will almost certainly fail to simulate biosphere and carbon cycle responses to climate change accurately. Further research is needed to explore whether land process models can reproduce this thermally optimal pattern across space and time.

Acknowledgements

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

This work was financially supported by the Terrestrial Carbon Program at the Office of Science, US Department of Energy, grants DE-FG02-006ER64317 and the US National Science Foundation (NSF), grant DEB 0444518, DEB 0743778, DEB 0840964, DBI 0850290, and EPS 0919466 to Y.L. We used the eddy covariance data acquired by the FLUXNET community and, in particular, by the following networks: AmeriFlux (US Department of Energy, Biological and Environmental Research, Terrestrial Carbon Program (DE-FG02-04ER63917 and DE-FG02-04ER63911)), GHG-Europe, Fluxnet-Canada Research Network and Canadian Carbon Program (supported by CFCAS, NSERC, BIOCAP, Environment Canada, and NRCan), GreenGrass, KoFlux, LBA, NECC, OzFlux, TCOS-Siberia, and USCCC. We acknowledge the financial support to the eddy covariance data harmonization provided by the Office of Science US Department of Energy for AmeriFlux, CarboEuropeIP, FAP-GTOS-TCO, iLEAPS, NitroEurope, Max Planck Institute for Biogeochemistry, National Science Foundation, University of Tuscia, Universite Laval and Environment Canada, and database development and technical support from the Berkeley Water Center, Lawrence Berkeley National Laboratory, and Microsoft Research eScience.

References

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Supporting Information

  1. Top of page
  2. Summary
  3. Introduction
  4. Materials and Methods
  5. Results
  6. Discussion
  7. Acknowledgements
  8. References
  9. Supporting Information

Fig. S1 Global map of site distribution.

Figs S2-S7 Robustness test of the parabolic temperature response curves of NEE.

Fig. S8 Uncertainty test for ToptNEE using boostrapping methods.

Fig. S9 Examples of temperature response curves of NEE, GPP and Re in the average vs warmer years.

Fig. S10 Relationship of ToptGPP and Q10 with environmental factors.

Table S1 Site information

Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office.

FilenameFormatSizeDescription
NPH_4095_sm_TableS1_FigS1-S10.doc6862KSupporting info item