Volume 222, Issue 2 p. 768-784
Full paper
Free Access

Acclimation and adaptation components of the temperature dependence of plant photosynthesis at the global scale

Dushan P. Kumarathunge

Corresponding Author

Dushan P. Kumarathunge

Hawkesbury Institute for the Environment, Western Sydney University, Locked Bag 1797, Penrith, NSW, 2751 Australia

Plant Physiology Division, Coconut Research Institute of Sri Lanka, Lunuwila, 61150 Sri Lanka

Author for correspondence:

Dushan P. Kumarathunge

Tel: +614 78807875

Email: [email protected]

Search for more papers by this author
Belinda E. Medlyn

Belinda E. Medlyn

Hawkesbury Institute for the Environment, Western Sydney University, Locked Bag 1797, Penrith, NSW, 2751 Australia

Search for more papers by this author
John E. Drake

John E. Drake

Forest and Natural Resources Management, College of Environmental Science and Forestry, State University of New York, 1 Forestry Drive, Syracuse, NY, 13210 USA

Search for more papers by this author
Mark G. Tjoelker

Mark G. Tjoelker

Hawkesbury Institute for the Environment, Western Sydney University, Locked Bag 1797, Penrith, NSW, 2751 Australia

Search for more papers by this author
Michael J. Aspinwall

Michael J. Aspinwall

Department of Biology, University of North Florida, 1 UNF Drive, Jacksonville, FL, 32224 USA

Search for more papers by this author
Michael Battaglia

Michael Battaglia

CSIRO Agriculture and Food, Private Bag 12, Hobart, Tasmania, 7001 Australia

Search for more papers by this author
Francisco J. Cano

Francisco J. Cano

Hawkesbury Institute for the Environment, Western Sydney University, Locked Bag 1797, Penrith, NSW, 2751 Australia

Search for more papers by this author
Kelsey R. Carter

Kelsey R. Carter

School of Forest Resources & Environmental Science, Michigan Technological University, 1400 Townsend Dr., Houghton, MI, 49931 USA

Search for more papers by this author
Molly A. Cavaleri

Molly A. Cavaleri

School of Forest Resources & Environmental Science, Michigan Technological University, 1400 Townsend Dr., Houghton, MI, 49931 USA

Search for more papers by this author
Lucas A. Cernusak

Lucas A. Cernusak

College of Science and Engineering, James Cook University, Cairns, QLD, 4878 Australia

Search for more papers by this author
Jeffrey Q. Chambers

Jeffrey Q. Chambers

Department of Geography, University of California Berkeley, 507 McCone Hall #4740, Berkeley, CA, 94720 USA

Search for more papers by this author
Kristine Y. Crous

Kristine Y. Crous

Hawkesbury Institute for the Environment, Western Sydney University, Locked Bag 1797, Penrith, NSW, 2751 Australia

Search for more papers by this author
Martin G. De Kauwe

Martin G. De Kauwe

ARC Centre of Excellence for Climate Extremes, University of New South Wales, Sydney, NSW, 2052 Australia

Search for more papers by this author
Dylan N. Dillaway

Dylan N. Dillaway

Thomashow Learning Laboratories, Unity College, 90 Quaker Hill Road, Unity, ME, 04988 USA

Search for more papers by this author
Erwin Dreyer

Erwin Dreyer

Université de Lorraine, Inra, Silva, F54000 Nancy, France

Search for more papers by this author
David S. Ellsworth

David S. Ellsworth

Hawkesbury Institute for the Environment, Western Sydney University, Locked Bag 1797, Penrith, NSW, 2751 Australia

Search for more papers by this author
Oula Ghannoum

Oula Ghannoum

Hawkesbury Institute for the Environment, Western Sydney University, Locked Bag 1797, Penrith, NSW, 2751 Australia

Search for more papers by this author
Qingmin Han

Qingmin Han

Department of Plant Ecology, Forestry and Forest Products Research Institute (FFPRI), 1 Matsunosato, Tsukuba, Ibaraki, 305-8687 Japan

Search for more papers by this author
Kouki Hikosaka

Kouki Hikosaka

Graduate School of Life Sciences, Tohoku University, Aoba Sendai, 980-8578 Japan

Search for more papers by this author
Anna M. Jensen

Anna M. Jensen

Department of Forestry and Wood Technology, Linnaeus University, Växjö, Sweden

Search for more papers by this author
Jeff W. G. Kelly

Jeff W. G. Kelly

Center for Sustainable Forestry at Pack Forest, University of Washington, 9010 453rd Street E, Eatonville, WA, 98328 USA

Search for more papers by this author
Eric L. Kruger

Eric L. Kruger

Department of Atmospheric and Oceanic Sciences, University of Wisconsin-Madison, Madison, WI, 53706 USA

Search for more papers by this author
Lina M. Mercado

Lina M. Mercado

College of Life and Environmental Sciences, University of Exeter, Exeter, EX4 4PS UK

Centre for Ecology and Hydrology, Crowmarsh-Gifford, Wallingford, OX10 8BB UK

Search for more papers by this author
Yusuke Onoda

Yusuke Onoda

Graduate School of Agriculture, Kyoto University, Kyoto, 606-8502 Japan

Search for more papers by this author
Peter B. Reich

Peter B. Reich

Hawkesbury Institute for the Environment, Western Sydney University, Locked Bag 1797, Penrith, NSW, 2751 Australia

Department of Forest Resources, University of Minnesota, St Paul, MN, 55108 USA

Search for more papers by this author
Alistair Rogers

Alistair Rogers

Environmental and Climate Sciences Department, Brookhaven National Laboratory, Upton, NY, 11973-5000 USA

Search for more papers by this author
Martijn Slot

Martijn Slot

Smithsonian Tropical Research Institute, Apartado 0843-03092 Balboa, Ancón, Panama

Search for more papers by this author
Nicholas G. Smith

Nicholas G. Smith

Department of Biological Sciences, Texas Tech University, Lubbock, TX, USA

Search for more papers by this author
Lasse Tarvainen

Lasse Tarvainen

Department of Forest Ecology and Management, Swedish University of Agricultural Sciences (SLU), SE-901 83 Umeå, Sweden

Department of Biological and Environmental Sciences, University of Gothenburg, PO Box 461, Gothenburg, SE-405 30 Sweden

Search for more papers by this author
David T. Tissue

David T. Tissue

Hawkesbury Institute for the Environment, Western Sydney University, Locked Bag 1797, Penrith, NSW, 2751 Australia

Search for more papers by this author
Henrique F. Togashi

Henrique F. Togashi

Department of Biological Sciences, Macquarie University, North Ryde, NSW, 2109 Australia

Search for more papers by this author
Edgard S. Tribuzy

Edgard S. Tribuzy

Instituto de Biodiversidade e Florestas, Universidade Federal do Oeste do Pará (UFOPA), CEP 68035-110 Santarém, PA, Brazil

Search for more papers by this author
Johan Uddling

Johan Uddling

Department of Biological and Environmental Sciences, University of Gothenburg, PO Box 461, Gothenburg, SE-405 30 Sweden

Search for more papers by this author
Angelica Vårhammar

Angelica Vårhammar

Hawkesbury Institute for the Environment, Western Sydney University, Locked Bag 1797, Penrith, NSW, 2751 Australia

Search for more papers by this author
Göran Wallin

Göran Wallin

Department of Biological and Environmental Sciences, University of Gothenburg, PO Box 461, Gothenburg, SE-405 30 Sweden

Search for more papers by this author
Jeffrey M. Warren

Jeffrey M. Warren

Climate Change Science Institute and Environmental Sciences Division, Oak Ridge National Laboratory, Oak Ridge, TN, 37831 USA

Search for more papers by this author
Danielle A. Way

Danielle A. Way

Department of Biology, The University of Western Ontario, London, ON, Canada, N6A 5B6

Nicholas School of the Environment, Duke University, Box 90328 Durham, NC, 27708 USA

Search for more papers by this author
First published: 29 December 2018
Citations: 159

Summary

  • The temperature response of photosynthesis is one of the key factors determining predicted responses to warming in global vegetation models (GVMs). The response may vary geographically, owing to genetic adaptation to climate, and temporally, as a result of acclimation to changes in ambient temperature. Our goal was to develop a robust quantitative global model representing acclimation and adaptation of photosynthetic temperature responses.
  • We quantified and modelled key mechanisms responsible for photosynthetic temperature acclimation and adaptation using a global dataset of photosynthetic CO2 response curves, including data from 141 C3 species from tropical rainforest to Arctic tundra. We separated temperature acclimation and adaptation processes by considering seasonal and common-garden datasets, respectively.
  • The observed global variation in the temperature optimum of photosynthesis was primarily explained by biochemical limitations to photosynthesis, rather than stomatal conductance or respiration. We found acclimation to growth temperature to be a stronger driver of this variation than adaptation to temperature at climate of origin.
  • We developed a summary model to represent photosynthetic temperature responses and showed that it predicted the observed global variation in optimal temperatures with high accuracy. This novel algorithm should enable improved prediction of the function of global ecosystems in a warming climate.

Introduction

The capacity of species to cope with increasing growth temperature is one of the key determinants in range shifts and local extinction of species, because their distribution and range limits closely follow temperature isolines (Battisti et al., 2005). Evidence suggests that many species are adapted to their thermal environment of origin (Berry & Björkman, 1980) but also exhibit the capacity to adjust to temporal variations in the temperature of their environment (Rehfeldt et al., 2001; Valladares et al., 2014). However, the mechanisms that determine these responses are not well understood, making it challenging to predict the fate of plants in a changing climate.

Global vegetation models (GVMs) are one of the principal tools used to predict future terrestrial vegetation carbon (C) balance (Rogers et al., 2017a; Mercado et al., 2018). The temperature response of leaf-scale net photosynthesis (referred to as An-T response hereafter) is one of the key processes in these models. The effect of warming on modelled photosynthesis depends on the An-T response function used in the model and, in particular, the optimum temperature of photosynthesis (ToptA; Booth et al., 2012). Decades of empirical studies have shown that the An-T responses of plants vary geographically, suggesting genetic adaptation of species to their climate of origin (Fryer & Ledig, 1972; Slatyer, 1977, 1978; Berry & Björkman, 1980; Gunderson et al., 2009). Considerable evidence also shows that plants have the capacity to adjust the An-T response following temporal changes in ambient temperature, a response known as thermal acclimation (Way & Sage, 2008; Hall et al., 2013; Way & Yamori, 2014; Yamaguchi et al., 2016; Way et al., 2017). In a recent review, Yamori et al. (2014) reported inherent differences in the An-T response and its acclimation capacity among photosynthetic pathways (C3, C4 and CAM) and functional types (annual vs perennial, deciduous vs evergreen) that often differ in their climatic distributions. However, the current representations of An-T response in GVMs do not capture this empirical knowledge well (Smith & Dukes, 2013; Lombardozzi et al., 2015; Smith et al., 2016; Mercado et al., 2018). Most GVMs use either a single An-T response function for all species or represent broad geographical variation in the An-T response by using plant functional type (PFT)-specific functions without considering thermal acclimation. Robust representation of adaptation and acclimation of An-T response in GVMs is challenging, as we lack a quantitative assessment of acclimation and adaptation of photosynthetic temperature responses on a global scale (Stinziano et al., 2017).

Many GVMs incorporate the biochemical model of C3 photosynthesis (referred to as FvCB hereafter; Farquhar et al., 1980; Rogers et al., 2017a). Therefore, it is both tractable and valuable to encapsulate the mechanisms of photosynthetic temperature adaptation and acclimation in terms of parameters of the Farquhar model (Hikosaka et al., 1999; Dreyer et al., 2001; Medlyn et al., 2002b; Dillaway & Kruger, 2010). The model has two key parameters, for which the temperature response is particularly important: the maximum rate of ribulose-1,5-bisphosphate carboxylase-oxygenase (Rubisco) activity (Vcmax) and the potential electron transport rate (Jmax) (Farquhar et al., 1980). GVMs use two basic functional forms to characterize the instantaneous temperature response of the key FvCB model parameters, namely the standard and peaked Arrhenius functions (Medlyn et al., 2002a). Most empirical studies of the instantaneous temperature response of Vcmax and Jmax have used the peaked Arrhenius model, which has four key parameters: the basal rate of either Vcmax or Jmax at a standard temperature of 25°C (Vcmax25 or Jmax25), the activation energy (Ea), the deactivation energy (Hd) and the entropy term (∆S). The peaked Arrhenius model can also be used to calculate the optimum temperatures of Vcmax (ToptV) and Jmax (ToptJ). These parameters have now been documented for a wide range of species from different biomes and PFTs (Onoda et al., 2005; Rogers et al., 2017b; Slot & Winter, 2017). Evidence suggests that the Arrhenius model parameters vary significantly across plant taxa but also that these parameters have the capacity to acclimate to the growth temperature (Crous et al., 2013, 2018).

Several meta-analytic studies have attempted to characterize species variation in the model parameters. Medlyn et al. (2002a) compared the temperature response of key FvCB model parameters across different species but reported a poor relationship overall between the optimum temperature for photosynthesis and the temperature of the growing environment. They reported lower ToptV and ToptJ for plants grown in boreal vs temperate climates, but it was unclear whether this difference was a result of inherent genetic differences among the boreal and temperate species, or of acclimation to prevailing growth temperature. In an analysis of 23 species, Hikosaka et al. (2006) identified two important mechanisms of photosynthetic temperature acclimation, namely Ea of Vcmax (EaV) and Jmax (EaJ) and the ratio Jmax : Vcmax (JVr). The most comprehensive synthesis to date of the biochemically based plant photosynthetic temperature response is that of Kattge & Knorr (2007), who compared the instantaneous temperature response of Vcmax and Jmax across 36 species. This study found a lack of thermal acclimation of EaV and EaJ, but reported significant acclimation relationships for JVr and ∆S of Vcmax (∆SV) and Jmax (∆SJ). Importantly, Kattge & Knorr (2007) synthesized these relationships into a simple and generalizable form that enabled direct implementation into GVMs, thus providing a means to quantify the effect of thermal acclimation of photosynthesis on terrestrial C cycle predictions (Chen & Zhuang, 2013; Lombardozzi et al., 2015; Smith et al., 2016) as well as on biophysical consequences in future climates (Smith et al., 2017).

Despite the success of the Kattge & Knorr (2007) algorithms, the functions have several limitations. Firstly, the parameterization process did not consider potential interspecific differences in photosynthetic temperature response; all changes were attributed to differences in growth temperature. Hence, the response incorporates elements of both temperature adaptation and acclimation without resolving the extent of the contribution of the two processes. Given that acclimation can occur over days and that adaptation takes many generations, the importance of resolving the relative contribution of the two processes is critical. Recently, Mercado et al. (2018) showed that assuming the relationships represent both adaptation and acclimation, or adaptation only, leads to significantly different conclusions about the trajectory of future terrestrial C storage under warming. Their results further highlight the importance of separating photosynthetic thermal adaptation and acclimation when simulating current and future C storage. However, to date, few studies have separated species differences in temperature adaptation from temperature acclimation processes (Lin et al., 2013).

Second, the data used to derive the Kattge & Knorr (2007) functions came mainly from northern temperate and boreal trees and lacked globally important PFTs such as tropical forests and Arctic tundra. As a result, the growth temperature range varied only from 11 to 29°C (Kattge & Knorr, 2007), which is substantially narrower than growth temperatures simulated in GVMs. Therefore, the analysis of Kattge & Knorr (2007) could be improved with a broader global dataset directly addressing the relative roles of temperature acclimation and adaptation.

Third, the ability of the acclimation functions to capture the observed differences in temperature optima of light-saturated net photosynthesis (ToptA) has not been directly tested. It is not clear whether making adjustments to ToptV and ToptJ improves the ability of models to capture changes in ToptA; some studies have reported similar ToptA values even with significantly different ToptJ among species (Vårhammar et al., 2015). Moreover, the photosynthetic temperature response is controlled not only by the photosynthetic biochemistry, but also by stomatal and respiratory processes. Sensitivity analysis suggests that all three component processes are equally important in determining the ToptA at leaf scale (Lin et al., 2012) as well as at canopy scale (Tan et al., 2017) but none of the previous review studies addressed how the latter two components affected ToptA.

Given the need for robust representation of photosynthetic temperature acclimation and adaptation in GVMs, and its importance in predicting future global C budget (Lombardozzi et al., 2015; Smith et al., 2016; Mercado et al., 2018) and climate (Smith et al., 2017), we quantified and modelled the mechanisms that underlie the observed differences in ToptA among species and growth temperatures. We hypothesized that ToptA would be strongly driven by adaptation to the climate of origin, whereas temperature acclimation would further modify the temperature optimum in response to seasonal changes in temperature of the growth environment. To test these hypotheses, we compiled a global database of photosynthetic CO2 response curves measured at multiple leaf temperatures to simultaneously resolve the temperature optima of Anet, Vcmax and Jmax. The data comprised a total of 141 species from tropical rainforests to Arctic tundra. Included in this database were datasets from common-garden studies, which were used to quantify effects of adaptation alone on ToptA, and comprising time course studies that measured plants under contrasting prevailing ambient temperatures, which are used to quantify effects of temperature acclimation alone. We combined the identified effects of climate adaptation and temperature acclimation to derive a general global model of temperature responses, which is then tested against a third, independent, biogeographic dataset measured on mature plants growing in their native environments across the globe.

Materials and Methods

Data sources

We compiled a global database of datasets consisting of leaf photosynthetic CO2 response measurements (referred to as ACi curves hereafter) measured at multiple leaf temperatures and saturating irradiance intensities. The database covers 141 species from 38 experiments conducted around the world (Supporting Information Fig. S1; Table S1). Site latitude ranged from 42°48′S to 71°16′N, and mean annual growing season temperature (long-term average temperature of months where mean monthly temperature is > 0°C) ranged from 3 to 30°C.

The method of data collection was consistent across all datasets. In most datasets, measurements were started at ambient CO2 concentrations (360–400 ppm; depending on the year of data collection) and changed stepwise through a series of subambient (40–400 ppm) to superambient saturating CO2 concentrations (400–2000 ppm). The same measurement protocol was repeated on the same leaf at different leaf temperatures. Measurements were made at saturating irradiance (Table S1) using a portable photosynthesis system with standard leaf chambers, in most cases the Li-Cor 6400 (Li-Cor Biosciences, Lincoln, NE, USA), although some measurements were made with the Walz-CMS system (Walz, Effeltrich, Germany). We visually inspected every ACi curve in the dataset for possible outliers and erroneous data points (e.g. negative intercellular CO2 concentrations). We used criteria based on De Kauwe et al. (2016) to screen individual ACi curves for the analysis performed in this paper. Curves were excluded from the analysis if the fitted function (see later) had an r2 value < 0.99 (however, if the number of replicates available for a given occasion was limited, the threshold r2 was reduced to 0.90; c. 9% of the total number of ACi curves included in the analysis). After screening, the dataset contained a total of 3498 ACi curves measured at leaf temperatures ranging from 1 to 50°C.

Estimating temperature optimum for leaf net photosynthesis (ToptA)

Ambient leaf net photosynthesis (Anet) at each temperature was either obtained from the initial direct measurements at ambient CO2 concentrations or extracted from the ACi curves. For curves where the first point was not measured at ambient CO2 concentration, we extracted the Anet value at the measured sample CO2 concentration falling between 300 and 400 ppm. We estimated the temperature optimum for Anet, ToptA, by fitting a widely used model of instantaneous photosynthetic temperature response (Gunderson et al., 2009; Crous et al., 2013; Sendall et al., 2015; Vårhammar et al., 2015) 1) to the net photosynthesis measurements. The model is a quadratic equation, expressed as:
urn:x-wiley:0028646X:media:nph15668:nph15668-math-0001(Eqn 1)
where Anet is the net photosynthetic rate (μmol m−2 s−1) at a given leaf temperature, T (°C), ToptA is the temperature optimum for photosynthesis (°C), Aopt is the net photosynthetic rate at ToptA, and the parameter b (unitless) describes the degree of curvature of the relationship.

Parameterizing biochemical component processes of photosynthesis

We used the FvCB model to characterize photosynthetic biochemical component processes. The model represents leaf net photosynthesis rate as the minimum of three rates: the Rubisco carboxylation limited photosynthetic rate (Wc), the RuBP-regeneration limited photosynthetic rate (Wj), and the triose phosphate utilization limited rate (Wp). The widely used formulation and parameterization of the FvCB model is of the form Eqn 2-Eqn 6.

urn:x-wiley:0028646X:media:nph15668:nph15668-math-0002(Eqn 2)
urn:x-wiley:0028646X:media:nph15668:nph15668-math-0003(Eqn 3)
urn:x-wiley:0028646X:media:nph15668:nph15668-math-0004(Eqn 4)
urn:x-wiley:0028646X:media:nph15668:nph15668-math-0005(Eqn 5)
where Vcmax is the maximum rate of Rubisco activity, Ci and Oi (μmol mol−1) are intercellular CO2 and O2 concentrations, respectively, Kc and Ko (μmol mol−1) are Michaelis–Menten coefficients of Rubisco activity for CO2 and O2 respectively, Γ* (μmol mol−1) is the CO2 compensation point in the absence of photorespiration, TPU (μmol m−2 s−1) is the rate of triose phosphate export from the chloroplast, RL (μmol m−2 s−1) is the nonphotorespiratory CO2 evolution in the light, and J (μmol m−2 s−1) is the rate of electron transport at a given light intensity. J is related to incident photosynthetically active photon flux density (Q, μmol m−2 s−1) by
urn:x-wiley:0028646X:media:nph15668:nph15668-math-0006(Eqn 6)
where Jmax (μmol m−2 s−1) is the potential rate of electron transport, α (μmol mol−1) is the quantum yield of electron transport, and θ (dimensionless) is the curvature of the light response curve (Farquhar et al., 1980; Medlyn et al., 2002a,b; Kattge & Knorr, 2007; Sharkey et al., 2007).

We parameterized Eqn 3-Eqn 6 using the fitacis function within the plantecophys package (Duursma, 2015) in R version 3.3.2 (R Development Core Team, 2012). We assumed the Bernacchi et al. (2001) kinetic constants for the temperature response of Kc, Ko and Γ* as given in Medlyn et al. (2002a). We used measurement Q in Eqn 6 whenever available (see Table S1); otherwise we assumed a fixed value of 1800 μmol m−2 s−1. We assumed constant values of α (0.24 μmol mol−1) and θ (0.85 (unitless)) for all datasets (Medlyn et al., 2007); these parameter values have a relatively minor effect on the magnitude of estimated Jmax (Medlyn et al., 2002a). The estimated parameters, Vcmax and Jmax, are apparent values as we assumed infinite mesophyll conductance (gm). The significance of gm for Vcmax and Jmax estimates and their temperature response has been discussed elsewhere (Crous et al., 2013; Bahar et al., 2018), Here, there are insufficient data to quantify gm and hence it would have been inappropriate to include in our analysis (see Rogers et al., 2017a).

We tested two ACi curve-fitting routines, one with and one without TPU limitation (Eqn 5). Accounting for TPU limitation in the FvCB model did not affect the estimated photosynthetic capacities, apparent Vcmax and Jmax (Fig. S2), suggesting that at ambient CO2 concentrations, net photosynthesis was rarely limited by TPU (results not shown). Hence, we focused on the temperature responses of apparent Vcmax and Jmax as the principal biochemical components affecting the ToptA.

The temperature responses of Vcmax and Jmax were fitted using the peaked Arrhenius function:
urn:x-wiley:0028646X:media:nph15668:nph15668-math-0007(Eqn 7)
where urn:x-wiley:0028646X:media:nph15668:nph15668-math-0008 is the process rate (i.e. Vcmax or Jmax; μmol m−2 s−1) at a given temperature, Tk (K), k25 is the process rate at 25°C, R is the universal gas constant (8.314 J mol−1 K−1), and Ea (kJ mol−1) is the activation energy term that describes the exponential increase in enzyme activity with the increase in temperature, Hd (kJ mol−1) is the deactivation energy term that describes, for example, decline in enzyme activity at higher temperature as a result of denaturation of enzymes, and ΔS (J mol−1 K−1) is the entropy term which characterizes the changes in reaction rate caused by substrate concentration (Johnson et al., 1942). To avoid overparameterization, we assumed a fixed value of 200 kJ mol−1 for Hd in Eqn 7 for all species (Dreyer et al., 2001; Medlyn et al., 2002a).
The optimum temperature for urn:x-wiley:0028646X:media:nph15668:nph15668-math-0009 is given by:
urn:x-wiley:0028646X:media:nph15668:nph15668-math-0010(Eqn 8)

Assessing the contribution of stomatal and respiratory processes

The optimum temperature for photosynthesis is determined by stomatal and respiratory processes as well as biochemical processes (Medlyn et al., 2002a; Lin et al., 2012). Stomatal conductance values are potentially affected by the measurement protocol used in ACi curve measurements which rarely replicates the ambient conditions. Therefore, to assess the relative contribution of stomatal processes to ToptA, we calculated the net photosynthesis rate at a fixed Ci of 275 μmol mol−1 from each ACi curve, interpolating the curve using the FvCB model with parameters fitted to that curve. A fixed Ci of 275 μmol mol−1 was chosen, as it roughly corresponds to 70% of ambient [CO2]. When the photosynthetic rate is scaled to a common Ci, it eliminates the effect of variation in stomatal conductance on photosynthesis, isolating the temperature effects on photosynthetic biochemistry. Similar to net photosynthesis, the temperature optimum for photosynthesis at a fixed Ci (ToptA275) was estimated for each species by fitting Eqn 1. We compared ToptA275 with ToptA to estimate the effect of variation in stomatal conductance on the temperature optimum for photosynthesis.

We fit the standard Arrhenius function (Eqn 9) to RL values obtained from ACi curves to assess the effect of respiratory component processes on ToptA. We estimated two parameters, RL25 (RL at 25°C) and activation energy of RL (Ea). Similar to Jmax and Vcmax, linear regression was used to test for temperature adaptation and acclimation of RL.

urn:x-wiley:0028646X:media:nph15668:nph15668-math-0011(Eqn 9)
where urn:x-wiley:0028646X:media:nph15668:nph15668-math-0012 is the rate of respiration in light at 25°C.

Test for local adaptation and seasonal temperature acclimation of ToptA

We divided the database into three subsets: mature plants growing in their native environments, common-garden datasets, and datasets with seasonal photosynthetic measurements. We used a subset of the data collected in mature plants to identify the patterns in photosynthetic temperature responses of plants in native environments and for model evaluation. Temperature responses in this subset include the effects of both adaptation to the native environment and acclimation to the prevailing temperature. We used the common-garden and seasonal measurements subsets to estimate the relative contributions of adaptation and acclimation, respectively, in determining the observed trends with temperature for plants in native environments.

For plants growing in native environments, we derived relationships between photosynthetic parameters and the prevailing temperature of the growing environment, defined as the mean air temperature for the 30 d before gas exchange measurements (Kattge & Knorr, 2007) (Tgrowth), to identify the temporal trends in photosynthetic temperature responses. We derived Tgrowth using on-site measured real-time daily air temperature for most of the datasets, but for three datasets (Hinoki cypress, Japan; Mongolian oak, Japan; and Scots pine, Finland; Table S1), we extracted Tgrowth values from the original publications, as on-site temperature measurements were not available. We used a general linear model to parameterize the observed responses in the mature plants dataset (Eqn 10)
urn:x-wiley:0028646X:media:nph15668:nph15668-math-0013(Eqn 10)
where a and b are the intercept and slope, respectively.

Seasonal datasets provide the opportunity to test the acclimation capacity of different species to temporal changes in the ambient temperature of the growing environment. Here, we correlated photosynthetic parameters with growth temperature, Tgrowth, defined as the mean air temperature for the 30 d before gas exchange measurements. Similar to the mature plants dataset, we derived Tgrowth using on-site measured daily air temperature for most of the datasets. For datasets where real-time meteorological data were not available, we extracted Tgrowth values from the original publications.

Common gardens provide an opportunity to test for adaptation, as species with different climates of origin are grown at a common growth temperature. The common-garden datasets included field trials and experiments in controlled environmental conditions, which included two or more species or provenances with contrasting climates of origin. We located the seed source of each species or provenance (latitude and longitude) using published information (Table S1). We used 30″ resolution WorldClim climatology data (WorldClim 1.4; Hijmans et al., 2005) to estimate long-term average (1960–1990) air temperature at seed source. With reference to the species selection criteria used in several common-garden studies (Lin et al., 2013; Vårhammar et al., 2015), we defined mean maximum air temperature of the warmest month at species’ seed source as the species’ home temperature (Thome) and derived relationships between photosynthetic parameters and Thome to test for adaptation of species’ An-T response to climate of origin. We repeated the same analysis with two other forms of species’ home temperature – mean growing season air temperature and mean temperature of the warmest quarter – to test whether our results were altered depending on the definition of climate of origin.

For both common-garden and seasonal subsets, we used linear regression against Thome and Tgrowth Eqn 11-Eqn 12 to test for temperature adaptation and acclimation, respectively, of ToptA, ToptA275, the photosynthetic biochemical parameters (Vcmax, and Jmax), and their temperature response parameters (see Eqn 7-Eqn 8). To test the effect of different biochemical parameters on temperature optimum for photosynthesis, we used linear regression between ToptA275 and temperature response parameters of Vcmax and Jmax.

Representing acclimation and adaptation in vegetation models

We derived functions to represent photosynthetic temperature acclimation and adaptation in GVMs. If a given parameter showed only acclimation to growth temperature, the function used was:
urn:x-wiley:0028646X:media:nph15668:nph15668-math-0014(Eqn 11)
where Aac is the parameter value when Tgrowth = 0 and αac is the acclimation coefficient (°C−1).
If a parameter showed only adaptation to climate of origin, the function was:
urn:x-wiley:0028646X:media:nph15668:nph15668-math-0015(Eqn 12)
We combined Eqn 11 and Eqn 12 to represent both acclimation and adaptation, defined as
urn:x-wiley:0028646X:media:nph15668:nph15668-math-0016(Eqn 13)
where, δac is the acclimation coefficient corresponding to a unit deviation in Tgrowth from the species’ Thome (°C−1). We parameterized Eqn 11 and Eqn 12 independently using data from seasonal photosynthetic response studies (Eqn 11) and common-garden experiments (Eqn 12). Eqn 13 was parameterized using combined seasonal and common-garden datasets. We implemented the modified functions into the FvCB model (see Duursma, 2015) to simulate photosynthetic temperature response curves at a constant Ci of 275 μmol mol−1 and tested how well the leaf-scale photosynthesis model captured the observed temperature optimum of photosynthesis in the mature plants dataset. This provided an independent comparison, as the mature plants dataset was not used to parameterize the temperature acclimation and adaptation functions (Eqn 11-Eqn 13).

Statistical analysis

Parameters of Eqns Eqn 1-Eqn 9 were estimated in a nonlinear mixed model framework (Zuur et al., 2009) using the nlme function within the nlme package in R v.3.3.2 (R Development Core Team, 2012). Replicate trees and/or leaves of the same species were included as random effects. However, when datasets contained measurements of multiple species (e.g. Brazilian rainforests, Australian rainforests and Australian semiarid woodland datasets; Table S1), individual species were considered as a random variable in the model. Similarly, Eqn 11-Eqn 13 were parameterized in a linear mixed model framework using the inverse of the standard error (SE) of each parameter of Eqns Eqn 1-Eqn 9 as the weighting scale to account for parameter uncertainty (Zuur et al., 2009; Lin et al., 2015). We tested whether the model parameters (Eqns Eqn 11-Eqn 13) differed significantly among datasets (and/or species) by fitting linear mixed models with and without random slopes and intercepts for each dataset (and or species). These models were then compared using a likelihood ratio test (Zuur et al., 2009) to determine whether the acclimation and adaptation coefficients differed among species. We used standard model validation tools (normal quantile plots and residual plots) to test the underlying assumptions in linear mixed models, and marginal and conditional r2 values to evaluate the goodness of fit (Nakagawa & Schielzeth, 2013). The complete database used for this analysis is available as a public data product (Kumarathunge et al., 2018) and the code used for the entire analysis is publicly available through https://bitbucket.orgKumarathungephotom.

Results

Temperature optimum for net photosynthesis at saturating irradiance (ToptA)

The temperature optimum for leaf-level net photosynthesis at saturating irradiance (ToptA) of mature plants in their natural habitats was strongly correlated with the temperature of the growth environment (Tgrowth; mean air temperature of preceding 30 d; Fig. 1a; Table 1). Values of ToptA ranged from 16.3 to 32.4°C, where the minimum and maximum values were observed for Arctic vegetation and tropical evergreen trees, respectively. The rate of increase in ToptA was 0.62 ± 0.07°C per °C increase in Tgrowth.

Details are in the caption following the image
Temperature optimum for leaf net photosynthesis (ToptA) (a–c) and net photosynthesis at an intercellular CO2 concentration of 275 μmol mol−1 (ToptA275) (d–f) of mature plants growing in their native environments (a, d), species in the field (grown at ambient growth temperatures) measured in at least two or more seasons (b, e) and species or provenances from contrasting climates of origin grown in common growth temperatures (common gardens or controlled environments) (c, f). Tgrowth, mean air temperature of the preceding 30 d; Thome, long-term (1960–1990) mean maximum temperature of the warmest month at species’ seed origin. Different colours in (a, d) depict plant functional types: orange, tropical evergreen angiosperms (EA-Tr); light blue, arctic tundra; red, temperate deciduous angiosperms (DA-Te); blue, temperate evergreen angiosperms (EA-Te); green, boreal evergreen gymnosperms (EG-Br); purple, temperate evergreen gymnosperms (EG-Te); those in (b,c,e,f) depict different datasets. The thick black lines are: (a, d) least-squares linear regression fits; (b, c, e, f) linear mixed-effect model fits with random intercepts for each dataset. The thin lines in respective colours are the fitted random intercept models for individual datasets. Error bars represent ± 1 SE.
Table 1. Results of the linear regression analysis of the parameters of Eqns 1, Eqn 7-Eqn 9
Mature plants in native environment (Eqn 10) Seasonal dataset (Eqn 11) Common-garden dataset (Eqn 12)
Parameter a b r 2 P-value A ac α ac r2 (marginal) r2 (conditional) P-value A ad α ad r2 (marginal) r2 (conditional) P-value
T optA 12.5 (1.4) 0.62 (0.1) 0.80 < 0.001 18.2 (1.1) 0.34 (0.05) 0.27 0.87 < 0.001 24.8 (2.1) 0.07 (0.1) 0.01 0.71 0.309
T optA275 14.9 (1.5) 0.63 (0.1) 0.84 < 0.001 20.5 (1.2) 0.24 (0.05) 0.16 0.85 < 0.001 26.8 (2.3) 0.07 (0.1) 0.03 0.30 0.400
Biochemical parameters
V cmax25 85.3 (16.7) −1.84 (0.8) 0.19 0.404 58.2 (12.0) 0.50 (0.4) 0.01 0.94 0.252 33.4 (28.0) 1.62 (0.9) 0.07 0.91 0.096
J max25 194.7 (24.1) 5.13 (1.2) 0.53 < 0.001 141.3(18.8) −1.35 (0.7) 0.03 0.95 0.053 92.7 (47.2) 1.63 (1.6) 0.02 0.95 0.312
E aV 48.7 (7.8) 0.82 (0.4) 0.14 0.067 39.7 (6.2) 1.14 (0.3) 0.32 0.91 < 0.001 79.4 (13.1) −0.37 (0.5) 0.14 0.14 0.450
E aJ 43.5 (9.8) −0.19 (0.5) 0.05 0.7143 27.2 (5.0) 0.26 (0.3) 0.04 0.82 0.325 51.5 (8.7) −0.38 (0.3) 0.20 0.20 0.247
SV 662.0 (8.7) 1.31 (0.5) 0.30 0.011 645.1 (4.6) −0.38 (0.2) 0.09 0.82 0.089 647.9 (9.5) −0.36 (0.3) 0.08 0.66 0.302
SJ 667.3 (7.8) 1.34 (0.4) 0.36 0.005 653.9 (4.6) 0.85 (0.2) 0.22 0.94 < 0.001 662.3 (7.5) 0.99 (0.3) 0.49 0.84 < 0.001
T optV 24.3 (3.8) 0.71 (0.2) 0.40 0.002 30.3 (1.9) 0.36 (0.1) 0.23 0.77 < 0.001 34.3 (3.3) 0.12 (0.1) 0.05 0.36 0.335
T optJ 19.9 (2.9) 0.63 (0.2) 0.52 < 0.001 27.6 (1.8) 0.31 (0.1) 0.13 0.91 < 0.001 24.8 (3.4) 0.42 (0.1) 0.42 0.60 < 0.001
JV r 2.9 (0.2) 0.06 (0.01) 0.66 < 0.001 2.3 (0.2) 0.03 (0.01) 0.07 0.17 < 0.001 2.5 (0.3) 0.03 (0.01) 0.13 0.64 0.005
Respiratory parameters
 R L25 2.8 (0.5) −0.09 (0.03) 0.38 0.0037 1.54 (0.42) −0.01 (0.02) 0.01 0.25 0.502 1.16 (0.45) 0.01 (0.01) 0.01 0.61 0.583
Ea −20.7 (14.3) 1.18 (0.78) 0.07 0.1508 −9.17 (11.49) 0.42 (0.61) 0.02 0.83 0.485 −4.25 (43.38) 0.12 (1.57) 0.01 0.93 0.937
RL25 :  Vcmax25 0.036 (0.01) −0.001 (0.0003) 0.22 0.033 0.03 (0.01) 0.001 (0.0003) 0.04 0.60 0.043 0.03 (0.01) −0.0005 (0.0004) 0.06 0.53 0.149
  • For common-garden and seasonal datasets, linear mixed models were fitted accounting for between-dataset variations of a given parameter (see the Materials and Methods section for details). For mature plants in native environments, parameter values were derived by fitting simple linear regression models (Eqn 10). Values in parentheses are standard errors of estimates. Bold values are the significant parameters at α = 0.05. Vcmax25, maximum rate of Rubisco activity (Vcmax) at a standard temperature of 25°C; Jmax25, potential rate of electron transport (Jmax) at a standard temperature of 25°C; JVr, ratio Jmax25 : Vcmax25; EaV, activation energy of Vcmax; EaJ, activation energy of Jmax; ∆Sv, entropy of Vcmax; ∆SJ, entropy of Jmax; ToptV, optimum temperature of Vcmax; ToptJ, optimum temperature of Jmax. RL25, non-photorespiratory CO2 evolution in the light at a standard temperature of 25°C; Ea, activation energy of RL.

In the seasonal dataset (Fig. 1b), we found strong evidence for acclimation of ToptA to the prevailing growth temperature. ToptA showed a significant increasing trend with Tgrowth. The mean rate of increase in ToptA was 0.34 ± 0.05°C per unit increase in Tgrowth (Table 1). By contrast, no trend was observed with climate of origin in common-garden studies (Table 1). Here, we tested for a relationship between ToptA and the Thome (1960–1990 mean maximum air temperature of the warmest month at species’ seed source) and we did not find any significant relationship for ToptA with Thome (Fig. 1c; Table 1). The results were similar for the two alternative definitions of the climate of origin (Table S2). The lack of a significant relationship with the species’ home temperature in the common-garden datasets suggests that the variation in ToptA of mature plants across ecosystems (Fig. 1a) is more strongly driven by acclimation to growth temperatures (Fig. 1b) than by local adaptation to climate of origin (Fig. 1c).

Temperature optimum for photosynthesis at a common Ci (ToptA275)

Similar to ToptA, ToptA275 showed a strong correlation with Tgrowth in mature plants across ecosystems (Fig. 1d; Table 1). We found no significant differences in either intercept or slope of the linear regression between ToptA and ToptA275 vs Tgrowth (Table 1), in both the mature (Fig. 1a,d) and seasonal (Fig. 1b,e) datasets, strongly suggesting that the observed variation in ToptA among ecosystems is not a result of variation in the stomatal limitation of ToptA. This result also suggests that the observed seasonal pattern of ToptA (Fig. 1b) was not driven by stomatal processes but rather by the effects of photosynthetic biochemical processes. Similar to ToptA, species in common-garden studies did not show significant trends for ToptA275 with Thome (Fig. 1f).

Temperature dependence of biochemical capacities, Jmax and Vcmax

Similar to ToptA, we found a strong increase in both ToptV and ToptJ with Tgrowth in the mature plants dataset (Fig. 2a,d). The slopes of the linear regression with Tgrowth were similar for ToptV and ToptJ (0.71 ± 0.20 and 0.63 ± 0.15°C °C−1, respectively). These sensitivities are similar in magnitude to the sensitivity of ToptA and ToptA275 to Tgrowth in the mature plants dataset. For Vcmax, the trend in Topt was caused by an increase (P ≈ 0.06) in EaV with increasing Tgrowth, and a strong decline in ∆SV (Fig. 2b,c). For Jmax, however, there was no change in EaJ, only a decline in ∆SJ with increasing Tgrowth (Fig. 2e,f).

Details are in the caption following the image
Biochemical temperature response parameters for the mature plants dataset in relation to mean air temperature of the preceding 30 d (Tgrowth). Different colours represent plant functional types as in Fig. 1(a, d). Solid and dotted lines in each panel are the least-squares linear regression fits (this study; coefficients and r2-values given in Table 1) and the linear models proposed by Kattge & Knorr (2007), respectively. Error bars represent ±1 SE. Legend follows Fig. 1(a, d). EaV, activation energy of the maximum rate of Rubisco activity (Vcmax); EaJ, activation energy of the maximum potential electron transport rate (Jmax); ToptV, optimum temperature of Vcmax; ToptJ, optimum temperature of Jmax; ∆SV, entropy of Vcmax; ∆SJ, entropy of Jmax.

We decomposed the observed trends across biomes shown in Fig. 2 by looking at seasonal datasets (Fig. 3) and common-garden studies (Fig. 4) independently to identify the effect of seasonal acclimation and local adaptation of photosynthetic biochemical component processes. We found a strong increase in ToptV and ToptJ with Tgrowth (Fig. 3a,d). The rate of increase in ToptJ per unit increase in Tgrowth was slightly higher than that in ToptV (Table 1), but the difference was not significant. Further, these sensitivities were found to be similar to the sensitivity of both ToptA and ToptA275 to Tgrowth. Similar to the mature plants dataset, we found a significant positive trend for EaV and a decreasing trend (P ≈ 0.08) for ∆SV with increasing Tgrowth (Fig. 3b,c). For Jmax, however, there was no change in EaJ, only a strong decline in ∆SJ with increasing Tgrowth (Fig. 3e,f).

Details are in the caption following the image
Biochemical temperature response parameters for the seasonal dataset in relation to mean air temperature of the preceding 30 d (Tgrowth). Data were measured on field-grown plants (including whole-tree chamber experiments) in two or more seasons. Solid and dotted lines in each panel are the linear mixed-effect model fits (this study; coefficients and r2-values are given in Table 1) and the linear models proposed by Kattge & Knorr (2007), respectively. Error bars represent ± 1 SE. EaV, activation energy of the maximum rate of Rubisco activity (Vcmax); EaJ, activation energy of the potential rate of electron transport (Jmax); ToptV, optimum temperature of Vcmax; ToptJ, optimum temperature of Jmax; ∆SV, entropy of Vcmax; ∆SJ, entropy of Jmax. Legend follows Fig. 1(b, e).
Details are in the caption following the image
Biochemical temperature response parameters for the common-garden dataset in relation to the long-term (1960–1990) mean maximum temperature of the warmest month at species’ seed origin (Thome). Data were measured in species or provenances from contrasting climates of origin grown at common growth temperatures (common gardens and controlled environments). Solid lines in each panel are the linear mixed-effect model fits (this study; coefficients and r2-values are given in Table 1). Error bars represent ± 1 SE. EaV, activation energy of the maximum rate of Rubisco activity (Vcmax); EaJ, activation energy of the potential rate of electron transport (Jmax); ToptV, optimum temperature of Vcmax; ToptJ, optimum temperature of Jmax; ∆SV, entropy of Vcmax; ∆SJ, entropy of Jmax. Legend follows Fig. 1(c, f).

We found no evidence to support adaptation of ToptV, EaV and ∆SV to climate of origin as there were no significant trends observed with temperature at species’ seed source (i.e. Thome) in the common-garden dataset (Fig. 4a–c). These observations were consistent with the lack of significant trends for ToptA in the common-garden dataset. However, ToptJ and ∆SJ showed significant trends with Thome (Fig. 4d–f; Table 1), suggesting the adaptation of both parameters to climate of origin. The results were similar for the two alternative definitions of the climate of origin (Table S2).

The balance between Jmax and Vcmax

We found no detectable correlation between Tgrowth and the basal rate of Vcmax at a standard temperature 25°C for mature plants in their natural habitats, but the basal rate of Jmax showed a strong decrease (Fig. 5a,b). The ratio of Jmax : Vcmax at 25°C (JVr) showed a significant decrease with increasing Tgrowth (Fig. 5c, Table 1). We excluded the Scots pine, Finland dataset when fitting linear regression, as the JVr value departed significantly from the general trend and was therefore identified as an outlier (black circle in Fig. 5c).

Details are in the caption following the image
Maximum rate of Rubisco activity (Vcmax), potential rate of electron transport (Jmax) and Jmax : Vcmax ratio (JVr) at a standard leaf temperature (25°C) of: (a–c) mature plants growing in their native environments; (d–f) field-grown plants measured in two or more seasons; and (g–i) species or provenances from contrasting climates of origin grown in common growth temperatures (common gardens or controlled environments). Tgrowth, mean air temperature of preceding 30 d; Thome, long-term (1960–1990) mean maximum temperature of the warmest month at species’ seed origin, respectively. Solid lines in each panel are the least-squares linear regression fits (b, c), and linear mixed-effect model fits with random intercepts for each dataset (f, i). One outlier is circled in (c) (see the Results section). Error bars represent ± 1 SE. Legend follows Fig. 1.

Basal rates of Vcmax and Jmax did not show significant trends with Tgrowth, but JVr responded negatively to Tgrowth in the seasonal dataset (Fig. 5d,f). We found no evidence to support adaptation of basal rates of Vcmax and Jmax to climate of origin; no parameters showed any significant trend with Thome in the common-garden dataset (Fig. 5g,h; Table 1). However, there was evidence of adaptation of JVr to climate of origin, as JVr showed a significant decrease with Thome in the common-garden dataset (Fig. 5I; Table 1).

Assessing the role of day respiration

We found no detectable trends (Fig. S3; Table 1) for either RL25 or Ea of mature plants in native environments. Similar results were found for common-garden studies and no seasonal trends were observed for either RL25 or Ea in the seasonal dataset. However, the data showed a slight negative trend for RL25 : Vcmax25 ratio with increasing Tgrowth (of mature plants in native environments) and Tgrowth (of seasonal datasets) (Fig. S4). We also observed negative Ea values in all three datasets (Fig. S4).

Model to represent acclimation and adaptation in vegetation models

Our results provide evidence that changes in the temperature response of photosynthesis among datasets are principally driven by acclimation of photosynthetic biochemistry to growth temperature. Both EaV and JVr showed strong acclimation to growth temperature with significant (albeit weak) acclimation of ∆SV. We found little evidence to support local adaptation of photosynthetic biochemistry to climate of origin. Only JVr and ∆SJ showed statistically significant, but weak, signals of local adaptation. We further tested whether variation in EaV and JVr can explain the seasonal acclimation of temperature optimum of photosynthesis observed in the seasonal dataset using linear regression analysis (JVr and EaV vs ToptA275). We found a strong negative trend for the relationship between JVr and ToptA275 (Fig. 6a). ToptA275 increased by c. 6°C for a unit decrease in JVr. Also, we found a significant trend between EaV and ToptA275; ToptA275 increased by c. 0.2°C for a unit increase in EaV (Fig. 6b). Therefore, the observed trends in ToptA of mature plants in native habitats (Fig. 1a) can be explained by the effect of growth temperature on EaV, ∆SV and JVr and the effects of both growth temperature and climate of origin on ∆SJ and JVr. Hence, photosynthetic temperature acclimation and adaptation can be implemented in GVMs using these parameters. Therefore, we modified the baseline peaked Arrhenius functions (Eqn 8) to represent temporal variability of EaV and ∆SV using Eqn Eqn 12, and geographical and temporal variation of the JVr ratio at 25°C and ∆SJ using Eqn 13. The full final model is given in Table 2.

Details are in the caption following the image
Relationship between potential rate of electron transport (Jmax) : maximum rate of Rubisco activity (Vcmax) ratio (JVr) at a standard leaf temperature (25°C) and temperature optimum for photosynthesis at a fixed intercellular CO2 concentration of 275 μmol mol−1 (ToptA275) (a) and relationship between activation energy of Vcmax (EaV) and ToptA275 (b). Data were measured on field-grown plants (including whole-tree chamber experiments) in two or more seasons. Lines in each panel are the linear mixed effect regression model fits (a, ToptA275 = 35.78–5.93 × JVr, R2 = 0.36; b, ToptA275 = 13.11 + 0.20 × EaV, R2 = 0.49. Error bars represents ± 1 SE.
Table 2. Parameters of the temperature acclimation and adaptation functions developed in this study
Parameter Model representation Value Units
V cmax25 PFT specific DA-Te 39.0 μmol m−2 s−1
EA-Te 82.9
EG-Te 42.8
EG-Br 80.4
EA-Tr 39.4
Arctic tundra 78.3
J max25 Acclimation + adaptation Vcmax25 × JVr μmol m−2 s−1
JV r Acclimation + adaptation 2.56 − 0.0375 Thome − 0.0202 (Tgrowth − Thome) Unitless
E aV Acclimation 42.6 + 1.14 Tgrowth kJ mol−1
E aJ Global mean 40.71 kJ mol−1
Sv Acclimation 645.13 − 0.38Tgrowth J mol−1 K−1
SJ Acclimation + adaptation 658.77 − 0.84Thome − 0.52 (Tgrowth − Thome) J mol−1 K−1
  • Thome, long-term (1960–1990) mean maximum temperature of the warmest month; Tgrowth, mean air temperature of preceding 30 d. Plant functional types (PFTs): DA-Te, temperate deciduous angiosperms; EA-Te, temperate evergreen angiosperms; EG-Te, temperate evergreen gymnosperms; EG-Br, boreal evergreen gymnosperms; EA-Tr, tropical evergreen angiosperms; Arctic tundra, Arctic spp.; Vcmax25, maximum rate of Rubisco activity (Vcmax) at a standard temperature of 25°C; Jmax25, potential rate of electron transport electron transport rate (Jmax) at a standard temperature of 25°C; JVr, ratio Jmax25 : Vcmax25; EaV, activation energy of Vcmax; EaJ, activation energy of Jmax; ∆Sv, entropy of Vcmax; ∆SJ, entropy of Jmax.

We found that the new temperature response functions were able to predict the temperature optima of photosynthesis observed in field-grown mature plants with a high degree of accuracy (r2 = 0.80). The slope (1.09 ± 0.15) and intercept (−2.20 ± 4.10) of the linear regression between the predicted and observed ToptA were not significantly different from unity and zero, respectively (Fig. 7a; Table S3). Our new model outperformed the Kattge & Knorr (2007) algorithms, which tend to underpredict ToptA (Fig. 7b; Table S3). Further, the use of PFT-specific values of Vcmax, together with a standard unacclimated photosynthetic temperature responses (Leuning, 2002), was not able to predict the observed variability in ToptA, as it predicts a ToptA ≈ 25°C for all datasets (Fig. 7a). Note that the mature plant dataset was not included in fitting Eqn 11-Eqn 13, so that the predicted ToptA275 in Fig. 7(a) was independent of the data used to derived the model parameters.

Details are in the caption following the image
Observed and predicted temperature optimum for photosynthesis at a fixed intercellular CO2 concentration (Ci) of 275 μmol mol−1 using model parameterizations given in Table 2. (a) With acclimation and adaptation functions developed in this study (y = 1.09 x − 2.20, r2 = 0.80); (b) Kattge & Knorr (2007) acclimation function (y = 1.58x − 13.82, r2 = 0.83). The crossed circle in the x-axis of (a) depicts the predicted temperature optimum for photosynthesis at a fixed intercellular CO2 concentration of 275 μmol mol−1 (ToptA275) with a fixed set of parameters without acclimation and adaptation (Leuning, 2002). Colours reflect plant functional types as in Fig. 1. Thin lines, 1 : 1 relationship; thick lines, least-squares regression fit. In (a), the intercepts and the slope of the linear regression were not significantly different from zero and unity respectively (Supporting Information Table S3). Error bars represent ± 1 SE.

Discussion

We developed new mathematical functions to represent the photosynthetic temperature response in vegetation models to account for both acclimation to growth temperature and adaptation to climate of origin using a global database that contains > 140 species. We found acclimation to growth temperature to be the principal driver of the photosynthetic temperature response, and observed only a few modest effects of adaptation to temperature at the climate of origin. The observed variation of temperature optimum for leaf net photosynthesis was primarily explained by the photosynthetic biochemical component processes rather than stomatal or respiratory processes. The new temperature response functions presented here capture the observed ToptA across biomes with higher degree of accuracy than previously proposed algorithms.

Adaptation of ToptA to climate of origin

Despite a significant range in long-term mean temperature at species’ seed sources, we found no predictable relationship for ToptA with climate of origin when species were grown in common gardens. Therefore, our results do not support the hypothesis that ToptA is adapted to species’ climate of origin (hypothesis 1). Our results contrast with previous studies which found that ToptA is related to species climate of origin (Fryer & Ledig, 1972; Slatyer, 1977, 1978; Robakowski et al., 2012), but there are a number of studies which compare the temperature response of photosynthesis and report a lack of local adaptation of ToptA (Ledig & Korbobo, 1983; Gunderson et al., 2000). We propose two hypotheses to explain the lack of local adaptation of ToptA: there is a lack of specialization in photosynthetic biochemistry in relation to climate of origin; and the capacity of species to adjust their ToptA to temporal variations in local thermal environment could mask ecotypic thermal adaptation of ToptA (Robakowski et al., 2012).

With respect to the first hypothesis, Rubisco activity is one of the key photosynthetic biochemical determinants and one of the most temperature-responsive physiological process (Galmés et al., 2015). Several lines of evidence suggest that Rubisco catalytic properties, including the relative specificity for CO2/O2 (Sc/o), the Michaelis–Menten constants for CO2 (Kc) and O2 (Ko), and the maximum turnover of carboxylation (kc), differ among species that have evolved under different thermal environments (Andersson & Backlund, 2008; Galmes et al., 2014). However, it is not clear whether these differential responses are a result of genetic adaptation of Rubisco kinetics to climate of origin or to the temporal effects of growth temperature. Galmés et al. (2015) argued that closely related species could be less adapted to their current thermal environment as a result of past strategies that limit adaptation of Rubisco to new thermal regimes (Lambers et al., 2008). This hypothesis was further supported by Savir et al. (2010), who suggested that point mutations may not cause a significant improvement in Rubisco activity owing to its close optimality in the net photosynthetic rate (Tcherkez et al., 2006). As a result, the adaptive evolution of Rubisco to novel thermal environments may be rare, as adaptation to a local environment will be working against the selective pressure to cope with seasonal and annual temperature variations and would reduce species fitness and expansion into new niches with different thermal environments. Other than the parameters ∆SJ and JVr, our results do not show evidence of thermal adaptation of photosynthetic biochemical parameters. Thus we suggest that the lack of local adaptation of ToptA may be partially explained by the lack of specialization in photosynthetic biochemistry, particularly Rubisco kinetic properties to species climate of origin.

Regarding the second hypothesis, we suggest that the capacity of Rubisco kinetic properties to adjust to temporal variations in growth temperature could potentially mask the species’ preadaptive responses to their original thermal environments. Here, we show strong evidence for the acclimation of ToptA to species Tgrowth, which is primarily a result of the variations in photosynthetic biochemical component processes JVr, EaV, ∆SV and ∆SJ in relation to the seasonal temperature dynamics. Potential mechanisms by which the kinetic properties of Rubisco could be altered in response to changes in temperature include structural changes in the Rubisco enzyme itself (Huner & Macdowall, 1979; Huner, 1985; Yamori et al., 2006); changes in the concentration of other photosynthetic enzymes such as Rubisco activase (Yamori et al., 2005, 2011); expression of cold/heat-stable isozymes (Yamori et al., 2006); and alterations in membrane fluidity (Falcone et al., 2004). A number of previous studies have demonstrated short-term acclimation of Rubisco kinetics to growth temperature (Medlyn et al., 2002b; Yamori et al., 2006; Kattge & Knorr, 2007; Lin et al., 2013; Yamaguchi et al., 2016; Smith & Dukes, 2017; Crous et al., 2018), although the sensitivities of the responses varied. In addition, studies that have compared the acclimation capacity of multiple species in common growth temperatures have shown similar direction and magnitude of short-term temperature acclimation of ToptA (Berry & Björkman, 1980; Sendall et al., 2015) and Rubisco kinetics (Lin et al., 2013; Smith & Dukes, 2017) across species irrespective of their climate of origin. Therefore, we argue that the capacity of species to adjust their photosynthetic biochemistry to temporal variations in growth temperature provides a fitness advantage over that of local climatic adaptation of ToptA and its related mechanisms, by enabling species to optimize C balance in their current habitat (Hikosaka et al., 2006).

The lack of a temperature adaptation response in this study contrasts with the results of a previous meta-analysis which found both evolutionary changes and an acclimation effect on ToptA (Yamori et al., 2014). Our common-garden studies compared closely related species (or provenances of the same species) in most cases. The most climatically divergent sets of species included in this study were those of Vårhammar et al. (2015) (lowland and montane tropical species) and Dillaway & Kruger (2010) (North American boreal and temperate deciduous species; see Table S1). By contrast, Yamori et al. (2014) compared temperature responses of C3, C4 and CAM plants and found evidence of evolutionary shifts among these functional groups. Other common-garden studies with taxonomically diverse species have also provided evidence of evolutionary changes in ToptA in relation to climate of origin (Cunningham & Read, 2002; Reich et al., 2015).

Acclimation of ToptA to growth temperature

Our observations of seasonal photosynthetic temperature response datasets suggest that the seasonal plasticity in ToptA is principally driven by the adjustment of the temperature response of the Rubisco-limited photosynthetic rate and the balance between Rubisco-limited and electron transport-limited photosynthetic rates. These two mechanisms control the seasonal shifts in ToptA as follows. First, at biologically relevant leaf temperatures, the light-saturated net photosynthetic rate is mostly limited by Rubisco activity (Rogers & Humphries, 2000; De Kauwe et al., 2016; Yamaguchi et al., 2016). An increase in EaV along with a decrease in ∆SV increases the Rubisco-limited photosynthetic rate with temperature, and thus affects the shape of the photosynthetic temperature response. The rate of increase in EaV with Tgrowth in this study (1.14 kJ mol−1 C−1) aligns closely with previous reports (1.01 kJ mol−1 C−1 in Hikosaka et al., 2006). A number of potential causes have been suggested for variations in EaV across species, including mesophyll conductance to CO2 diffusion (Bernacchi et al., 2002; Warren et al., 2007; Walker et al., 2013; von Caemmerer & Evans, 2015), kinetic properties of Rubisco (Yamori et al., 2006), distribution of leaf nitrogen among photosynthetic proteins (Yin et al., 2018) and the influence of other enzymes that affect the in vivo activity of Rubisco (Onoda et al., 2005). Furthermore, the Rubisco activation status could also be a significant factor contributing to the observed trends in EaV with Tgrowth., as evidence suggested that plants have the capacity to maintain high Rubisco activation status through an increase in Rubisco activase concentration and expression of heat-stable Rubisco activase isoforms (Crafts-Brandner & Salvucci, 2000; Sage et al., 2008; Yamori et al., 2014). However, not all authors find a change in EaV with growth temperature. Kattge & Knorr (2007) did not find any temperature acclimation in EaV. They argued that the choice of a standard, rather than peaked, Arrhenius model to fit the temperature response for Vcmax without considering the deactivation energy would be a possible reason for the observed acclimation responses of EaV in previous studies (e.g. Hikosaka et al., 2006). However, here we used the peaked Arrhenius model, and thus the acclimation of EaV that we observed is not an artifact of model choice.

The second important mechanism for acclimation was a change in the magnitude of JVr, as has also been observed by others (Kattge & Knorr, 2007; Crous et al., 2013, 2018; Lin et al., 2013). The ratio determines the transition between the two limiting steps, Wc and Wj. As the temperature responses of Wc and Wj are different from each other with different optimum temperatures (Topt of Wc < Topt of Wj), ToptA is potentially determined by the limiting step (von Caemmerer & Farquhar, 1981; Hikosaka, 1997). At higher JVr, the photosynthetic rate is mostly limited by RuBP carboxylation, and therefore ToptA tends to be of a lower value, and vice versa.

The acclimation capacity of ∆SV observed in this study (−0.38 J mol−1 K−1) was lower than the −1.07 J mol−1 K−1 C−1 reported in Kattge & Knorr (2007). The higher sensitivity observed in Kattge & Knorr (2007) could potentially be explained by the lack of variation in EaV. Both EaV and ∆SV are correlated; a high sensitivity of EaV to Tgrowth could potentially cause ∆SV to be less sensitive, and vice versa.

We observed changes in JVr with temperature in all three datasets (Fig. 5), but only the mature plant dataset showed a change in either of the two terms contributing to this ratio. In this dataset, the reduction in JVr is driven by a reduction in Jmax25, whereas in the other two datasets, there is no overall effect on either Vcmax25 or Jmax25. Some previous studies have observed changes in Vcmax25 with growth temperature in more limited datasets (Way & Oren, 2010; Lin et al., 2013; Ali et al., 2015; Scafaro et al., 2017; Crous et al., 2018; Smith & Dukes, 2018), but here we did not find any consistent pattern in Vcmax25. It appears that JVr responded strongly and consistently to growth temperature, but whether this is achieved by increasing Vcmax, decreasing Jmax, or both, is highly variable. We speculate that the global pattern in Jmax observed in Fig. 5(b) may be a response to increasing light availability in the tropics, following the colimitation hypothesis, as proposed by Dong et al. (2017), rather than a response to growth temperature.

Improved temperature response functions for photosynthetic capacity

We demonstrate acclimation to growth temperature to be the principal driver, and only a few modest effects of adaptation, in photosynthetic temperature responses at global scale. Our results highlight the limitation of using a fixed set of parameters to determine ToptA, and challenge the use of PFT-specific Vcmax25 and Jmax25 with a fixed set of temperature response parameters without accounting for temperature acclimation and adaptation (Leuning, 2002) in GVMs (Harper et al., 2016; Rogers et al., 2017a). We also demonstrate that the current representation of photosynthetic temperature acclimation (Kattge & Knorr, 2007) that has been implemented in some vegetation models (Smith & Dukes, 2013; Lombardozzi et al., 2015; Smith et al., 2016) was not able to predict the observed patterns in ToptA across biomes.

We proposed new algorithms for temperature response that are based on a broad range of data, account for both geographical and temporal variability in photosynthetic biochemical component processes, and are able to capture observed variation of ToptA across biomes with a high degree of accuracy. The temperature response functions that we propose have a broad temperature domain (c. 3–37°C), which should enable their use in GVMs without outer domain uncertainties (Stinziano et al., 2017), a limitation of the algorithms proposed previously (Kattge & Knorr, 2007) that are widely implemented in GVMs (BETHY, CLM4.5, Orchidee). As a result of these advantages, our new photosynthetic temperature algorithms provide an improved representation of geographical and temporal variability in ToptA and should ultimately improve the accuracy of predicted future C cycle in GVMs.

Acknowledgements

This research was supported by a Western Sydney University PhD scholarship to DPK. AR was supported by the Next Generation Ecosystem Experiments (NGEE Arctic) project, which is supported by the Office of Biological and Environmental Research in the United States Department of Energy (DOE), Office of Science, and through the United States Department of Energy contract no. DE-SC0012704 to Brookhaven National Laboratory. KYC was supported by an Australian Research Council DECRA (DE160101484). DAW acknowledges an NSERC Discovery grant and funding from the Hawkesbury Institute Research Exchange Program. JU, LT and GW were supported by the Swedish strategic research area BECC (Biodiversity and Ecosystem Services in a Changing Climate; www.becc.lu.se). JQC was supported by the NGEE-Tropics, United States DOE. MDK was supported by Australian Research Council Centre of Excellence for Climate Extremes (CE170100023). MS was supported by an Earl S Tupper postdoctoral fellowship. AMJ and JMW were supported by the Biological and Environmental Research Program in the Office of Science, United States DOE under contract DEAC05-00OR22725. MAC was supported by United States DOE grant DE-SC-0011806 and USDA Forest Service 13-JV-11120101-03. Several of the Eucalyptus datasets included in this study were supported by the Australian Commonwealth Department of the Environment or Department of Agriculture, and the Australian Research Council (including DP140103415). We are grateful to Jens Kattge, Yan Shih-Lin, Alida C. Mau and Remko Duursma for useful discussions.

    Author contributions

    The project was conceived by BEM. The analyses were designed and carried out by DPK with guidance from BEM, JED, MGT, and with contributions from MGDK. Manuscript writing was led by DPK, BEM and JED. All other co-authors (MJA, MB, FJC, KRC, MAC, LAC, JQC, KYC, DND, ED, DSE, OG, QH, KH, AMJ, JWGK, ELK, LMM, YO, PBR, AR, MS, NGS, LT, DTT, HFT, EST, JU, AV, GW, JMW, DAW) contributed data and ideas, and edited the manuscript.