Leaf Trait Acclimation Amplifies Simulated Climate Warming in Response to Elevated Carbon Dioxide
Abstract
Vegetation modifies Earth's climate by controlling the fluxes of energy, carbon, and water. Of critical importance is a better understanding of how vegetation responses to climate change will feedback on climate. Observations show that plant traits respond to elevated carbon dioxide concentrations. These plant trait acclimations can alter leaf area and, thus, productivity and surface energy fluxes. Yet the climate impacts of plant structural trait acclimations remain to be tested and quantified. Here we show that one leaf trait acclimation in response to elevated carbon dioxide—a one-third increase in leaf mass per area—significantly impacts climate and carbon cycling in Earth system model experiments. Global net primary productivity decreases (−5.8 PgC/year, 95% confidence interval [CI95%] −5.5 to −6.0), representing a decreased carbon dioxide sink of similar magnitude to current annual fossil fuel emissions (8 PgC/year). Additional anomalous terrestrial warming (+0.3 °C globally, CI95% 0.2 to 0.4), especially of the northern extratropics (+0.4 °C, CI95% 0.2 to 0.5), results from reduced evapotranspiration and enhanced absorption of solar radiation at the surface. Leaf trait acclimation drives declines in productivity and evapotranspiration by reducing leaf area growth in response to elevated carbon dioxide, as a one-third increase in leaf mass per area raises the cost of building leaf area and productivity fails to fully compensate. Our results suggest that plant trait acclimations, such as changing leaf mass per area, should be considered in climate projections and provide additional motivation for ecological and physiological experiments that determine plant responses to environment.
Plain Language Summary
Plants have been observed to change their traits, such as the thickness of leaves, in response to future environmental conditions, but the implications of these changes for climate have not yet been quantified. We show that changes in plant traits could have large-scale climate impacts, including higher temperatures and relative decreases in plant photosynthesis which have not been previously accounted for. Our findings suggest an urgent need for observations of how plant traits will respond to future environmental conditions as well as a need for a better understanding of the underlying mechanisms so that they can be included in climate projections.
1 Introduction
Feedbacks between vegetation and climate change are of critical importance to future climate projections but remain highly uncertain (Arora et al., 2013; Friedlingstein et al., 2014; Lovenduski & Bonan, 2017; Pu & Dickinson, 2012). Vegetation strongly influences the Earth's climate by controlling the fluxes of carbon, water, and energy between the land surface and the atmosphere (Bonan, 2008). Changes in biologically mediated carbon fluxes, such as productivity and respiration, can alter the concentration of carbon dioxide (CO2) in the atmosphere, leading to warming of the Earth due to the radiative effects of CO2. Given that these radiative effects are driven by biological sources of carbon, we refer to the associated temperature increase as biogeochemical warming. Since the start of the industrial era, Earth's vegetation has removed about 30% of anthropogenic CO2 emissions from the atmosphere (Ciais et al., 2013). Changes in vegetation can also induce warming by altering water and energy fluxes through their influence on Earth surface properties such as evapotranspiration, albedo, and roughness. We refer to increases in temperature due to alterations of the surface energy balance as biogeophysical warming. Transpiration, the biologically controlled flux of water from soil through plants into the atmosphere, makes up an estimated 60% of current terrestrial water fluxes (Wei et al., 2017), which physically cool the land surface. Rising CO2 concentrations are expected to have profound and wide reaching effects on vegetation functioning and growth, with important implications for global carbon uptake and evapotranspirative cooling. Yet large uncertainty exists in the magnitude, and even the sign, of vegetation feedbacks on climate change (Arora et al., 2013; Friedlingstein et al., 2014; Lovenduski & Bonan, 2017; Pu & Dickinson, 2012). This uncertainty stems in large part from the challenge of representing complex and diverse life-forms at the global scale in the Earth system models used to project future climate (Lovenduski & Bonan, 2017). Key biological processes must be missing or poorly constrained, but we lack a clear understanding of which processes are essential for predicting climate and carbon cycling changes.
Incorporating observations of plant trait distributions and their responses to environmental drivers into Earth system models is proposed as a way to improve predictions of ecosystem functioning (Butler et al., 2017; Fisher et al., 2015; Kattge et al., 2011; Kattge & Knorr, 2007; Pavlick et al., 2013; Reich et al., 2014; Reichstein et al., 2014; Scheiter et al., 2013; Van Bodegom et al., 2012; Verheijen et al., 2015, 2013; Wright et al., 2004). Trait databases and studies that aggregate observations across species are beginning to make it possible to characterize current plant trait distributions and their responses to environmental drivers at the global scale (e.g., Kattge et al., 2011; Kattge & Knorr, 2007; Niinemets, 2001; Van Bodegom et al., 2012; Verheijen et al., 2013; Wright et al., 2004). However, the biogeographic relationship between traits and climate across ecosystems, caused primarily by environmental filtering, does not tell us about short-term responses to changes in climate within an ecosystem, caused by acclimation (Van Bodegom et al., 2012; Verheijen et al., 2013). The climate impacts of these two distinct responses, environmental filtering and acclimation, have been tested in previous work.
Studies focused on environmental filtering have shown that allowing traits to vary temporally based on observed spatial relationships between these traits and environmental drivers (i.e., space-for-time substitution) has carbon uptake and climate implications (Verheijen et al., 2015, 2013). This approach estimates the integrated outcome of numerous biological responses to climate (e.g., adaptation, changes in species distribution, and acclimation; Van Bodegom et al., 2012; Verheijen et al., 2015). However, it does not separate the impacts of individual biological responses (e.g., acclimation, adaptation, and species turnover) from one another and therefore cannot mechanistically explain the underlying causes of trait variation (Verheijen et al., 2013). Further, it is uncertain if space-for-time relationships used in the environmental filtering approach will hold under future climate in part because acclimation of traits may alter these trait-environment relationships (Fisher et al., 2015; Verheijen et al., 2015). Acclimation responses can differ in magnitude and even direction from trait responses to environmental filtering (e.g., Poorter et al., 2009; Verheijen et al., 2013).
Other studies have directly investigated the influence of some trait acclimations to temperature and elevated CO2 (e.g., photosynthetic and stomatal conductance rates) and found profound effects on large-scale climate and carbon cycling (Betts et al., 1997; Cao et al., 2010; Lombardozzi et al., 2015; Pu & Dickinson, 2012; Sellers et al., 1996; Smith et al., 2017). Acclimation occurs within the same individual plant and on short-time scales (e.g., a growing season), making it immediately relevant for 21st century climate. Prior studies have focused on rate traits and have not considered the potential climate feedbacks of plant structural traits. Trait responses to climate change that alter plant structure could feedback on climate and carbon cycling by modifying the surface areas (e.g., leaf area) over which the rates of photosynthesis and stomatal conductance are summed.
Among the most widely observed plant structural trait responses to elevated CO2 is an increase in leaf mass per area (g leaf carbon/m2 leaf area). Leaf mass per area represents the carbon cost of building leaf area and is a quantity commonly used in Earth system models to convert from carbon available for leaf growth to leaf area. Field and greenhouse manipulation experiments show that leaf mass per area increases by as much as one third in response to elevated CO2 in a wide range of C3 plants, including trees, shrubs, and crops, across a variety of ecosystems on many continents (Ainsworth & Long, 2005; Medlyn et al., 1999, 2015; Poorter et al., 2009). Acclimation to warming temperatures could potentially offset leaf mass per area increases due to elevated CO2 but is limited to cold regions such as the boreal and arctic (Poorter et al., 2009). Most Earth system models project increases in leaf area in response to CO2 over the 21st century (Mahowald et al., 2016; Swann et al., 2016), which are expected to negatively feedback on climate change by promoting carbon uptake from the atmosphere and evapotranspirative cooling over land (Betts et al., 1997; Bounoua et al., 2010; Pu & Dickinson, 2012). However, few models capture the decreased sensitivity of leaf area index to increases in leaf biomass at elevated CO2 because they fail to represent the concomitant increase in leaf mass per area (De Kauwe et al., 2014; Medlyn et al., 2015).
Here we quantify the potential extent of climate and carbon cycling impacts of leaf mass per area acclimation to rising CO2 using a series of Community Earth System Model coupled atmosphere-land-carbon cycle simulations (supporting information Table S1). In the model, vegetation responds to climate by changing carbon assimilation, stomatal conductance, biomass, and leaf area. These vegetation responses can induce biogeophysical warming through feedbacks on the surface energy balance and atmosphere via changes in albedo, evapotranspiration, and surface roughness. We quantify the additional climate impacts, beyond those of elevated CO2, of leaf mass per area acclimation to CO2 as the difference between a leaf acclimation experiment and a climate change control simulation (CCLMA-CC). As atmospheric CO2 concentration is held invariant over time in all simulations, biogeochemical warming is estimated from the difference in net primary productivity. The level of leaf acclimation, a one-third increase in leaf mass per area in C3 plants, was estimated from the upper bound of acclimation to a doubling of CO2 (355 to 710 ppm) from Poorter et al.'s (2009) meta-analysis of approximately 200 studies, which provides the most plant-type-specific CO2 acclimation relationships for leaf mass per area currently available. The control simulation (CTRL) provides a reference for whether the effects of leaf acclimation at elevated CO2 (CCLMA-CC) moderate (e.g., reduce the increase in leaf area) or enhance (e.g., further increase leaf area) changes due to elevated CO2 alone (CC-CTRL). We also estimate the effects of leaf mass per area acclimation to temperature (TCCLMA-CC) and the historical influence of changing leaf mass per area (LMA-CTRL). Maximum photosynthetic rates (e.g., Vcmax25 and Jmax25) are the same across these simulations (CCLMA, CC, CTRL, TCCLMA, and LMA) before acclimating to temperature following Kattge and Knorr (2007). We test the sensitivity of our results to increasing maximum photosynthetic rates concurrently with leaf mass per area (CCLMAPS).
2 Materials and Methods
This study used the Community Earth System Model version 1.3beta11 with interactive land and biogeochemistry (CLM4.5-BGC; Oleson et al., 2013), atmosphere (CAM5; Neale et al., 2012), mixed-layer ocean (Neale et al., 2012), and sea ice (CICE4; Hunke & Lipscomb, 2010) models. Simulations that couple the land and atmosphere, such as performed here, are required to quantify the climate impacts of changes in the land surface, as they capture the atmospheric response and land-atmosphere feedbacks. To allow for ocean heat transport and atmosphere-ocean interaction while retaining computational economy, we used a mixed-layer ocean model with prescribed lateral heat fluxes rather than a more computationally expensive full dynamical ocean model. We ran the simulations with a spatial resolution of approximately 1.9° by 2.5° gridcells. The biogeochemistry model represents a full terrestrial carbon cycle with growth, mortality, and decay—and hence leaf area and carbon storage in aboveground and belowground pools. The distribution of 15 plant functional types was prescribed by a map of present day vegetation and held invariable; however, under unsuitable growing conditions, plants diminish to a minimum leaf area.
The climate change control simulation (CC; 2×CO2, no leaf acclimation) represents the mean climate state when atmospheric CO2 is fixed at 710 ppm. The CO2 leaf acclimation experiment (CCLMA; 2×CO2, +1/3 leaf mass per area) is identical to the climate change control simulation (CC) except that it includes a plausible extent of leaf mass per area acclimation to CO2 in all C3 plants (Poorter et al., 2009). (See Text S1.2 for details.) A second experiment (TCCLMA; 2×CO2, no change in leaf mass per area in boreal and arctic biomes, +1/3 leaf mass per area in all other C3 plants) tests the impact of leaf acclimation to both CO2 and temperature (Poorter et al., 2009). (See Texts S1.3 and S2.1 for further details.) Leaf mass per area acclimation to CO2 and temperature were estimated using the most plant-type-specific acclimation relationships currently available (Poorter et al., 2009). A third experiment (CCLMAPS; 2×CO2, +1/3 leaf mass per area, +1/3 maximum photosynthetic rates) tests the sensitivity of our results to increasing maximum photosynthetic rates and quantifies the increase in maximum photosynthetic rates required to offset the biogeophysical warming due CO2 acclimation of leaf mass per area. All elevated CO2 simulations (CC, CCLMA, TCCLMA, and CCLMAPS) include the effects of CO2 radiative forcing, CO2 fertilization, and gains in water use efficiency. A fourth experiment (LMA; 1×CO2, +1/3 leaf mass per area) tests the sensitivity of historical climate to increased leaf mass per area. A separate control simulation (CTRL; 1×CO2, no leaf acclimation) represents the equilibrium climate state when CO2 concentration is fixed at 355 ppm, a common baseline for Earth system model simulations.
We held maximum photosynthetic rates (Vcmax25, Jmax25, and Tp25) constant, so that they did not differ between the control (CC and CTRL) and CCLMA, TCCLMA, and LMA simulations prior to temperature acclimation. As the default model calculates maximum photosynthetic rates from leaf mass per area, we modified this relationship so that these rates did not differ (except CCLMAPS). In our simulations a decrease in leaf nitrogen concentration, which can also be thought of as an increase in leaf carbon-to-nitrogen ratio (gC/gN) and a reduction in leaf nutrition, is coupled to the increase in leaf mass per area (except CCLMAPS) to maintain maximum photosynthetic rates at control (CTRL and CC) levels. (See Text S1.2 for details.) This represents a conservative estimate of acclimation of maximum photosynthetic rates to CO2, as evidence supports a decrease in these rates in response to elevated CO2 (Ainsworth & Long, 2005; Leakey et al., 2012; Rogers et al., 2017; Smith & Dukes, 2013). The decrease in leaf nitrogen concentration with elevated CO2 is also supported by observations (reviewed in Ainsworth & Long, 2005; Leakey et al., 2012; Way et al., 2015). All simulations include temperature acclimation of maximum photosynthetic rates (Kattge & Knorr, 2007; Oleson et al., 2013). The maximum photosynthetic rate values of all simulations were within the observed range used to generate the empirical temperature acclimation function, and acclimation was not allowed outside of the range of temperature values used to generate the empirical function.
All simulations were integrated for 85 years, except the CCLMAPS experiment was integrated for 44 years. All experiment simulations were initiated by branching from the beginning of year 56 of the control run (CTRL). Temperature, leaf area index, net and gross primary productivity, evapotranspiration, and live carbon pools (leaf, live stem, live root, and fine root) reached equilibrium before year 30 in each simulation. The first 30 years of each simulation were discarded to allow for spin up. The remaining years were used in our analysis and represent many samples of the equilibrium state. Model results are available through the University of Washington Libraries ResearchWorks digital repository. The URL for the data in the ResearchWorks system is https://digital.lib.washington.edu/researchworks/handle/1773/41856.
We use annual mean changes in biogeophysical warming and net primary productivity to quantify the upper bound of the potential climate and carbon cycling influences of leaf mass per area acclimation. We tested for differences between simulations in the annual mean at the global, latitude band, zonal mean (average for a given latitude), and gridcell scales using bootstrap methods (n = 50,000; Text S1.4) with model years as the unit of replication. Spatial relationships between variables at the gridcell scale were tested using simple, multiple, and stepwise linear regression methods on annual mean values. Differences and relationships were considered significant at the 95% level. (See Text S1.4 for details.) Latitude bands were defined as southern extratropics (60°S to 20°S), tropics (20°S to 20°N), northern extratropics (20°N to 65°N), and northern high latitudes (65°N to 90°N).
Biogeochemical warming was calculated by converting the change in net primary productivity to a change in atmospheric CO2 level (2 PgC to 1 ppm). After accounting for compensatory carbon uptake by the ocean of 60–85% (Archer et al., 2009; Broecker et al., 1979), we converted the change in atmospheric CO2 concentration to a radiative forcing in watt per square meter following the methods of Hansen et al. (1998) and Myhre et al. (1998). The resulting global temperature change was then estimated from the forcing using a range of climate sensitivities (temperature change due to a doubling of CO2) from 1.5 to 4.5 °C.
3 Results
3.1 Biogeophysical Warming
Acclimation of leaf mass per area to elevated CO2 induced significant biogeophysical warming in addition to the warming caused by the radiative effects of a doubling of CO2 in Earth system model experiments. The change in temperature from the direct effects of a doubling of CO2 (from 355 to 710 ppm) in our model (CC-CTRL) was 5.0 °C (95% confidence interval [CI95%] 5.0 to 5.1), with a higher mean warming over land of 6.1 °C (CI95% 6.0 to 6.1). The influence of doubling CO2 included plant responses such as carbon fertilization (Oleson et al., 2013) and increased water use efficiency (+27% for CC-CTRL, CI95% 27 to 28) but did not account for acclimation of leaf mass per area. Consideration of leaf mass per area acclimation to CO2 (CCLMA-CC) increased annual mean temperature over land by an additional +0.3 °C (CI95% 0.2 to 0.4, Figure 1a and Tables 1 and S2) and + 0.2 °C (CI95% 0.1 to 0.2) globally on top of the direct effects of CO2. This acclimation-driven warming was especially pronounced over land in the northern extratropics (+0.4 °C, CI95% 0.2 to 0.5) due to above average warming over Eurasia (Figures 1a and 2a and Table 1). The influence of temperature acclimation of leaf mass per area (TCCLMA-CC) was limited to cold biomes and did not significantly alter the amount of additional warming over land and globally due to CO2 acclimation (Text S2.1; Figure S1). The influence of leaf mass per area changes at historical CO2 levels (LMA-CTRL) was also small (Text S2.2).
| Global | S. extratropics | Tropics | N. extratropics | |
|---|---|---|---|---|
| Biogeophysical warming (°C) | 0.3 (0.1%) | 0.3 (0.1%) | 0.3 (0.1%) | 0.4 (0.1%) |
| Net primary productivity (PgC/year) | −5.8 (−6.4%) | −0.8 (−9.1%) | −2.5 (−6.1%) | −2.1 (−6.2%) |
| Leaf area index (m2/m2) | −0.9 (−26.0%) | −0.8 (−24.0%) | −1.0 (−24.3%) | −1.0 (−27.4%) |
| Evapotranspiration (W/m2) | −0.7 (−1.5%) | −0.9 (−1.6%) | −1.2 (−1.6%) | −0.4 (−1.1%) |
| Transpiration (W/m2) | −1.4 (−5.8%) | −1.9 (−7.2%) | −1.7 (−4.6%) | −1.1 (−6.7%) |
| Leaf evaporation (W/m2) | −0.8 (−8.6%) | −0.7 (−8.5%) | −1.3 (−8.3%) | −0.5 (−9.0%) |
| Soil evaporation (W/m2) | 1.4 (9.5%) | 1.6 (7.0%) | 1.9 (10.6%) | 1.3 (9.9%) |
| Absorbed solar radiation (W/m2) | 0.6 (0.4%) | 0.8 (0.5%) | 0.6 (0.4%) | 0.6 (0.4%) |
- Note. All changes significant at the 95% level. Percent change ((CCLMA-CC)/CC) in parentheses. Confidence intervals reported in Table S2.
Leaf trait acclimation enhanced biogeophysical warming over land under future CO2 levels by offsetting the CO2-induced increase in leaf area index (m2 leaf area/m2 ground). Doubling of CO2 (CC-CTRL) increased the annual mean leaf area index by 1.2 m2/m2 (CI95% 1.2 to 1.2) in our simulations. This magnitude of change is at the high end of Coupled Model Intercomparison Project Phase 5 model leaf area responses to RCP8.5 over the 21st century (Mahowald et al., 2016). Inclusion of leaf mass per area acclimation strongly limited the increase in leaf area index to 0.3 m2/m2 (CI95% 0.2 to 0.3) over the ambient CO2 simulation (CCLMA-CTRL). This attenuation of leaf area growth occurred in almost all vegetated areas (Figures 1b and 2b and Table 1). However, leaf area index decreased more in response to leaf acclimation in places with high initial leaf areas, as shown by the negative spatial relationship (r = −0.91, R2 = 0.83, Figure S2a) between leaf area index in the control climate change case (CC) and the change in leaf area index in response to leaf acclimation (CCLMA-CC).
The reduced increase in leaf area in response to leaf trait acclimation (CCLMA-CC) induced biogeophysical warming over land by shifting the balance between surface energy budget terms. Near-surface temperature warmed in response to a moderation of the increase in evapotranspirative cooling and an increase in solar radiation absorbed at the Earth's surface (Figures 2 and 3c and Tables 1 and S2). These two factors shifted additional energy to sensible heat, the term in the surface energy balance that directly drives surface temperature changes. In the tropics, warming was primarily the result of reduced evapotranspiration, followed by greater solar radiation absorbed at the surface (Figures 2c and 2d and Tables 1 and S2). In the extratropics, increased absorption of solar radiation and reduced evapotranspiration induced warming in more equal proportion (Figures 2b and 2c and Tables 1 and S2). The strong influence on the surface energy budget of evapotranspiration in the tropics and the combination of evapotranspiration and solar radiation in the midlatitudes is consistent with previous studies (Bonan, 2008).
Evapotranspiration is the combination of several contributing water fluxes. Reduced transpiration (CCLMA-CC) represented the largest contribution to evapotranspiration declines in all regions, followed by lower evaporation from leaf surfaces (Tables 1 and S2). However, greater soil evaporation partially offset the decline from transpiration and leaf evaporation. The reduced increase in leaf area index in response to leaf acclimation drove the reduction in evapotranspiration (Figure 2), aided by a slight increase in water use efficiency (CCLMA-CC; +0.5%, CI95% 0.2 to 0.8). Reductions in evapotranspiration were spatially positively related to changes in leaf area (CCLMA-CC; r = 0.57, R2 = 0.32; Figure S2b). As leaf area provides the surface area over which transpiration and leaf evaporation occur, the acclimation-induced reduction of leaf area index diminished evapotranspiration to drive biogeophysical warming.
More solar radiation reached land when leaf mass per area acclimation was included (Figures 2d and 3c and Table 1) due to reduced low cloud cover over the tropics and northern extratropics (Figure S3a). Acclimation-driven warming decreased the relative humidity of the lower atmosphere in these regions (Figure S3b), making it less likely for water vapor to saturate the air and condense to form clouds. Relative humidity decreased because warming of the atmosphere (Figure S3c) raised the saturation vapor pressure, outcompeting the influence of greater absolute amounts of water vapor (i.e., specific humidity) in some areas (Figure S3d). The overall increase in solar radiation at the surface demonstrates that the effect of reduced cloud cover overwhelmed the opposing influence of a small surface albedo increase. Albedo increased because the reduced increase in leaf area index (CCLMA-CC) allowed more radiation to reach and reflect away from bare ground which is brighter than vegetation (Bonan, 2008; Oleson et al., 2013). Albedo changes (Figure S4) were measured by comparing the difference in solar radiation absorbed at the surface under clear-sky conditions (a model calculation that ignores the influence of clouds).
3.2 Carbon Cycle and Biogeochemical Warming
In addition to biogeophysical warming, acclimation of leaf mass per area reduced carbon uptake by the biosphere (Figures 1c and 3c), which would induce further warming by increasing atmospheric CO2 levels. Net primary productivity increased 51% (+30.1 PgC/year, CI95% 29.8 to 30.4) in response to a doubling of CO2 (CC-CTRL). Acclimation of leaf mass per area strongly moderated the positive effect of carbon fertilization on net primary productivity in response to elevated CO2, reducing the gain in productivity by −5.8 PgC/year (CCLMA-CC; CI95% −5.5 to −6.0, Tables 1 and S2). This decrease in net primary productivity in response to leaf acclimation was driven by declines in the tropics, followed by the northern extratropics (Tables 1 and S2).
Smaller increases in leaf area and higher temperatures in response to leaf acclimation both contributed to the reduced gains in productivity relative to the climate change control. Decreases in gross primary productivity (CCLMA-CC) were best described by a multiple regression using both changes (CCLMA-CC) in temperature and leaf area as predictors (multiple regression R2 = 0.32; Figure S2d). Changes in net primary productivity were weakly but best related to temperature change (r = −0.49, R2 = 0.24; Figure S2c).
From the reduced gains in carbon uptake in response to leaf mass per area acclimation, we estimate a change in global mean temperature. Our simulations did not directly account for this biogeochemical warming, as atmospheric CO2 levels within each simulation were held fixed in time. Instead, we estimate biogeochemical warming (see section 2) associated with the net change in carbon storage from the difference in carbon uptake by vegetation, as measured by net primary productivity, when leaf acclimation is considered (CCLMA-CC). The −5.5 to −6.0-PgC/year reduction in net primary productivity gains would increase global atmospheric CO2 concentration by +0.4 to +1.2 ppm/year when considering the effect of oceanic buffering. We estimate that this additional atmospheric CO2 induces biogeochemical warming of +0.1 to +1.0 °C over 100 years, the approximate average time scale for a doubling of CO2 from 355 to 710 ppm under the Intergovernmental Panel on Climate Change RCP8.5 and RCP6 emissions scenarios (Cubasch et al., 2013). The sum of this biogeochemical warming and the biogeophysical warming reported above brings the total additional warming over land due to leaf mass per area acclimation (CCLMA-CC) to +0.3 to +1.4 °C greater than the warming due to a doubling of CO2 in the control climate change simulation.
4 Discussion
We find that leaf trait responses could have significant large-scale climate implications. Increased leaf mass per area enhances warming beyond the direct effects of elevated CO2 by moderating evapotranspiration and enhancing absorption of solar radiation and by lessening the rise in leaf area which lowers net primary productivity gains (Figure 3).
The surface temperature change in response to leaf trait acclimation is of comparable magnitude to the climate response to other important climate forcings (Figure 4). For example, the enhanced warming in our experiment (+0.3 to +1.4 °C) is smaller but of the same order of magnitude as the change in temperature in response to a doubling of CO2 estimated by the Intergovernmental Panel on Climate Change (+1.5 to +4.5 °C) from observed 20th century climate change, paleoclimate, feedback analysis, and climate models (Ciais et al., 2013). While these comparisons are not exact, as the methods and measures of uncertainty differ, they provide an order of magnitude comparison for our results. Enhanced warming in our experiment is also of greater or comparable magnitude to the temperature response to large-scale land cover change (Figure 4d), such as anthropogenic land cover change over the 20th century (−0.04 °C physical, +0.27 chemical, +0.22 total, over land; Pongratz et al., 2010) and theoretical global deforestation (−1.1 °C biogeophysical over land; Davin & de Noblet-Ducoudré, 2010).
Furthermore, our results show that the surface temperature change in response to leaf trait acclimation can exceed or match several well-studied plant physiological feedbacks to elevated CO2 that are included in most climate projections (Figure 4c). These include the vegetation carbon-concentration feedback (0 to −1.0 °C; estimated from the change in CO2 implemented in this study of 355 to 710 ppm and the Coupled Model Intercomparison Project Phase 5 model range for land carbon-concentration feedback parameter from Arora et al., 2013); stomatal conductance response to elevated CO2 (+0.2 to +0.5 °C biogeophysical over land; Betts et al., 1997, 2007; Boucher et al., 2009; Cao et al., 2010; Cox et al., 1999; Pu & Dickinson, 2012; Sellers et al., 1996); photosynthetic down-regulation (−0.1 to +0.3 °C biogeophysical over land; Bounoua et al., 2010; Pu & Dickinson, 2012); and increased leaf area index (+30 to 60%) due to CO2 fertilization and increased water use efficiency under elevated CO2 (−0.1 to −0.4 °C biogeophysical over land; Betts et al., 1997; Bounoua et al., 2010; Pu & Dickinson, 2012).
The reduced increase in terrestrial productivity in response to leaf mass per area acclimation is on the order of other large-scale carbon cycle perturbations and moderates the effect of CO2 fertilization on plant growth and carbon uptake from the atmosphere. The −5.8-PgC/year (CI95% −5.5 to −6.0) reduction in net primary productivity in response to leaf mass per area acclimation in our simulations (CCLMA-CC) is a reduced carbon sink comparable in magnitude to current global fossil fuel emissions (8 PgC/year; Ciais et al., 2013). It is larger than the total current terrestrial biosphere uptake of CO2 from the atmosphere (3 PgC/year; Le Quéré et al., 2016).
Leaf mass per area acclimation to CO2 represents a shift in the relationship between two key ecosystem properties—productivity and leaf area. As such, this acclimation will remain important for climate and carbon cycling if other trait responses further modify estimates of productivity. Notably, the magnitude of maximum photosynthetic rate (e.g., Vcmax25 and Jmax25) acclimation to CO2 remains uncertain and difficult to represent at the global scale (Rogers et al., 2017; Smith & Dukes, 2013). While most estimates suggest that maximum photosynthetic rates will decrease in response to CO2 (Ainsworth & Long, 2005; Leakey et al., 2012; Rogers et al., 2017; Smith & Dukes, 2013), which would amplify our results, we conservatively do not change these rates in our primary experiment (CCLMA-CC). We note that our results should be considered in relation to our treatment of maximum photosynthetic rates, which were equivalent across experiments prior to temperature acclimation in all simulations except CCLMAPS. The CCLMAPS experiment tests the sensitivity of our results to increasing maximum photosynthetic rates. Using this experiment (CCLMAPS-CC), we estimate that maximum photosynthetic rates would need to increase (opposite direction of expected CO2 acclimation) by one third to bolster net primary productivity enough to offset the biogeophysical warming over land due to leaf acclimation in our experiments (Text S2.3; CCLMAPS-CC). This altered balance between productivity (biogeochemical warming) and leaf area (biogeophysical warming) demonstrates the importance of including leaf mass per area acclimation to CO2.
In addition to leaf mass per area, other changes in coordinated leaf traits could be expected to occur under climate change and further influence biogeophysical and biogeochemical warming. Longer leaf lifespans are correlated with higher leaf mass per area across species (Wright et al., 2004) and could be expected to offset the climate influence of leaf mass per area by enhancing productivity beyond current estimates. However, this correlation observed across species does not necessarily hold for trait changes within a species, such as in response to acclimation (Anderegg et al., 2018; Fisher et al., 2015; Lusk et al., 2008). Observations of leaf lifespan acclimation to elevated CO2 indicate that the response is highly variable in magnitude and sign and inconsistently associated with higher leaf mass per area (e.g., Norby et al., 2003, 2010; Taylor et al., 2008, and references therein). As the observational evidence does not support an increase in leaf lifespan in coordination with leaf mass per area acclimation to CO2, we chose not to impose this change in our simulations. However, we do include changes in leaf area duration due to phenological responses to warming temperature and soil moisture in all simulations (Oleson et al., 2013). Litter decomposition has also been hypothesized to slow with leaf responses to elevated CO2 with implications for carbon cycling (Strain & Bazzaz, 1983). However, a meta-analysis of observations found that the effect of elevated CO2 on leaf decomposition processes was not significant, despite changes in leaf litter traits (Norby et al., 2001). We therefore do not test changes in litter decomposition here. Lastly, changes in leaf nitrogen concentration and anatomy in response to climate change could alter albedo through their influence on leaf reflectance and transmittance (e.g., Ollinger, 2011), a possible avenue for future research. Leaf acclimation in our simulations was allowed to influence albedo indirectly by altering leaf area index but did not alter leaf optical characteristics because the influence of individual leaf traits (e.g., leaf mass per area) on these properties remains highly uncertain especially under future conditions (Ollinger, 2011).
Several environmental drivers of leaf mass per area acclimation—CO2, temperature, and nutrient limitation—will likely be modified by climate change. We estimate that the influence of temperature acclimation of leaf mass per area globally is secondary to CO2 (Text S2.1 and Figure S1). The effect of temperature warming on leaf mass per area occurs under cold conditions; thus, the acclimation is limited to high latitude boreal regions (Figure S5). Nutrient limitation is expected to increase with CO2 fertilization of plant growth (Norby et al., 2010; Wieder et al., 2015) and has been found to enhance leaf mass per area in manipulation experiments (Poorter et al., 2009), which could further amplify the impacts of leaf acclimation to elevated CO2. The magnitude of leaf mass per area acclimation in response to climate change may ultimately depend upon the combined influence, including potential interaction effects, of multiple climate drivers.
Accounting for leaf acclimation in climate projections will require the ability to represent the functional relationship between leaf mass per area and its climate drivers, especially CO2, by biome at the global scale. This remains challenging (Medlyn et al., 2015). Poorter et al.'s (2009) empirical relationship, used herein, shows that on average leaf mass per area increases with CO2 in C3 species. However, the proportion of variance in the magnitude of acclimation explained by this relationship is relatively low (Poorter et al., 2009), suggesting that other key drivers, such as plant type, still need to be incorporated. A mechanistic model of leaf mass per area acclimation also remains elusive. The leading hypothesis for why elevated CO2 increases leaf mass per area is that the abundance of carbon causes nonstructural carbohydrates to accumulate in leaves (Poorter et al., 1997, 2009; Pritchard et al., 1999; Roumet et al., 1999). One possible advantage for plants of increasing leaf mass per area under elevated CO2 is that it maintains a high level of leaf nitrogen per leaf area (g N/m2 leaf area), an essential component of photosynthetic machinery, by counteracting a decrease in leaf nitrogen concentration (g N/g leaf) driven by larger pools of nonstructural carbohydrates (N per area = N per mass × leaf mass per area; Ishizaki et al., 2003; Luo et al., 1994; Peterson et al., 1999; Poorter et al., 1997; Stitt & Krapp, 1999). However, this process operates differently across environments, plant species, and even genotypes (Körner et al., 1997; Luo et al., 1994; Peterson et al., 1999; Poorter et al., 1997, 2009; Pritchard et al., 1999; Roumet et al., 1999; Stitt & Krapp, 1999). Further research into the underlying mechanism, influences of multiple environmental drivers, and differences in acclimation between plant types is needed to develop a representation of leaf mass per area acclimation suitable for use in Earth system models.
The climate implications of increased leaf mass per area reveal an urgent need for observational constraints on the magnitude and mechanism of leaf trait acclimation to future climate conditions. Other structural trait acclimations that influence leaf area may have similar climate implications that require testing. Our findings suggest that the uncertainty in vegetation-climate feedbacks, and therefore climate change projections, is even larger than previously thought.
Acknowledgments
We thank M. Laguë and E. Garcia for help with model setup. We acknowledge support from the National Science Foundation AGS-1321745 and AGS-1553715 to the University of Washington. High-performance computing support from Yellowstone (ark:/85065/d7wd3xhc) was provided by NCAR's Computational and Information Systems Laboratory, sponsored by the National Science Foundation. M. K. thanks the UW Program on Climate Change Graduate Fellowship for support. Model results are available through the University of Washington Libraries ResearchWorks digital repository. The URL for the data in the ResearchWorks system is https://digital.lib.washington.edu/researchworks/handle/1773/41856.
