Stomatal conductance influences interannual variability and long-term changes in regional cumulative plant uptake of ozone

Ambient ozone uptake by plant stomata degrades ecosystem and crop health and alters local-to-global carbon and water cycling. Metrics for ozone plant damage are often based solely on ambient ozone concentrations, overlooking the role of variations in stomatal activity. A better metric is the cumulative stomatal uptake of ozone (CUO), which indicates the amount of ozone entering the leaf over time available to cause physiological damage. Here we apply the NOAA GFDL global earth system model to assess the importance of capturing interannual variations and 21st century changes in surface ozone versus stomatal conductance for regional mean CUO using 20-year time-slice simulations at the 2010s and 2090s for a high-warming climate and emissions scenario. The GFDL model includes chemistry-climate interactions and couples atmospheric and land components through not only carbon, water, and energy exchanges, but also reactive trace gases—in particular, ozone dry deposition simulated by the land influences surface ozone concentrations. Our 20-year time slice simulations hold anthropogenic precursor emissions, well-mixed greenhouse gases, and land use distributions fixed at either 2010 or 2090 values. We find that CUO responds much more strongly to interannual and daily variability in stomatal conductance than in ozone. On the other hand, long-term changes in ozone explain 44%–90% of the annual CUO change in regions with decreases, largely driven by the impact of 21st century anthropogenic NOx emission trends on summer surface ozone. In some regions, increases in stomatal conductance from the 2010s to 2090s counteract the influence of lower ozone on CUO. We also find that summertime stomatal closure under high carbon dioxide levels can offset the impacts of higher springtime leaf area (e.g. earlier leaf out) and associated stomatal conductance on CUO. Our findings underscore the importance of considering plant physiology in assessing ozone vegetation damage, particularly in quantifying year-to-year changes.


Introduction
Plant stomata control the uptake of carbon dioxide for photosynthesis and release of water vapor into the atmosphere through transpiration. Ambient ozone diffuses through open stomata and reacts quickly with fluids and tissues once inside the leaf (Laisk et al 1989, Wang et al 1995. Stomatal uptake of ozone serves as an important removal pathway of tropospheric ozone (Wesely and Hicks 2000, Fowler et al 2009, Clifton et al 2020a, which is a potent greenhouse gas, air pollutant, and a strong lever on the atmospheric oxidation capacity. Oxidation inside the leaf following stomatal ozone uptake causes cell death and decreases carbon fixation, leading to necrosis, reduced ecosystem productivity and carbon storage over time (Fiscus et al 2005, Ainsworth et al 2012, and lost crop yields (Mauzerall and Wang 2001, Morgan et al 2003, Feng et al 2008, Tai et al 2014, McGrath et al 2015. By changing local-to-global carbon cycling as well as altering energy and water exchanges, stomatal ozone uptake influences meteorology, climate, and air quality (Sitch et al 2007, Lombardozzi et al 2015Super et al 2015, Li et al 2016, Sadiq et al 2017, Zhou et al 2018.
Changes in ecosystem functioning and landatmosphere exchanges due to ozone plant damage depend on the cumulative stomatal uptake of ozone (CUO) (e.g. Musselman et al 2006). While the argument for including CUO in ozone plant damage metrics is decades old (Reich 1987, Musselman and Massman 1999, Matyssek et al 2004, Paoletti and Manning 2007, damage or risk is often evaluated based solely on metrics of ambient ozone concentrations (McLaughlin et al 2007, Hollaway et al 2012, Sun et al 2012, Tai et al 2014, Lapina et al 2014, 2016, Mills et al 2018 given the paucity of observational constraints on CUO. Recent work, leveraging a gridded surface ozone observational product (Schnell et  Ozone damage to plants depends not only on CUO but also on the plants' ability to detoxify and respond to ozone (Musselman and Massman 1999, Massman et al 2000, Musselman et al 2006, Paoletti and Manning 2007, Matyssek et al 2008. For example, plants scavenge some of the ozone inside the leaf such that a certain amount of ozone does not pose a risk to the plant other than by depleting detoxification reserves. Detoxification has been shown to depend on environmental conditions and species (e.g. Musselman et al 2006), and recently on the ratio of dry leaf mass to leaf area in woody plants (Feng et al 2018), but is highly uncertain, especially at regional scales (Karnosky et al 2007, Lombardozzi et al 2015, Jolivet et al 2016. Parameterizations that include plant detoxification and responses to ozone in regional-to-global models are used to assess the impacts of CUO on crop yields, carbon and water cycling, climate, and air quality (Sitch et al 2007, Lombardozzi et al 2015, Li et al 2016, Sadiq et al 2017, Oliver et al 2018, Arnold et al 2018, Unger et al 2020, Lei et al 2020. Most studies employing damage parameterizations in large-scale models probe either the carbon and water cycling impacts of 'turning on' schemes, or the impacts of changes in either surface ozone or stomatal conductance (g s ). While some of these studies examine the impacts of changes in both ozone and g s , they do not separate how changes in ozone versus g s drive the changes in impacts.
Because surface ozone and g s both influence CUO but respond to meteorology and the land surface differently, there may be individual changes in ozone and g s that counteract and limit, or compound and amplify, changes in CUO. Indeed Ronan et al (2020) use the Ducker et al (2018) CUO dataset to illustrate that recent reductions in ozone air pollution at sites in the US and Europe due to NO x (= NO + NO 2 ) emission controls do not benefit plants due to offsetting increases in g s . Other work modeling CUO suggests counteracting changes in g s and surface ozone on CUO at present day (Anav et al 2019) and by the end of the 21st century (Klingberg et al 2011) over Europe. These studies use uncoupled modeling frameworks, where meteorology from a regional climate model is fed into a regional chemical transport model and g s used for CUO is inconsistent with g s used for determining ambient ozone through dry deposition as well as with g s used for energy and water exchanges. Here, we use a new version of a global earth system model with chemistry-climate interactions and self-consistent land-atmosphere exchanges of carbon, water, heat, and reactive gases including ozone (Paulot et al 2018, Clifton et al 2020b to explore the roles of surface ozone versus g s in driving interannual and long-term variability in CUO. In particular, we show a critical role for interannual variations and 21st century changes in g s on regional mean CUO.

Methods
We use the NOAA GFDL global chemistryclimate model AM3, which includes stratospheretroposphere gas-phase and aerosol chemistry (Donner et al 2011, Naik et al 2013. AM3 is the atmospheric component of the fully coupled atmosphereocean general circulation model CM3, which was used and evaluated extensively in the 5th phase of the Coupled Model Intercomparison Project (CMIP5). The underlying land surface model of AM3/CM3 is LM3 (Shevliakova et al 2009, Milly et al 2014, which includes water, energy, and carbon cycling, vegetation dynamics and land use and management, and is coupled to atmospheric dynamics and radiation via surface albedo, surface roughness, and exchanges of water, energy, and momentum. We use a new version of AM3 called AM3DD where the land and tropospheric chemistry are coupled through dry deposition of reactive gases like ozone (Paulot et al 2018, Clifton et al 2020b. Because AM3 and LM3 are fully coupled, we refer to the GFDL model as an earth system model (note that we reduce computational expense by forcing sea surface temperatures and sea ice).
We examine time-slice AM3DD simulations of RCP8.5 at the 2010s and 2090s. RCP8.5 is an emissions and climate scenario designed by CMIP5 for the IPCC Fifth Assessment Report ( Central to the land-atmosphere exchanges of water, energy, and reactive trace gases in AM3DD is the stomatal resistance (R s ) simulated by LM3 (note that a resistance is the inverse of a conductance). The prognostic variable R s for water vapor (m s −1 ) is calculated from net photosynthesis (A net ) via the Leuning (1995) model: The parameter R is the universal gas constant (J mol air −1 K −1 ); T leaf is leaf temperature (K); p s is surface pressure (Pa); m is an empirical constant (unitless); d s is the vapor pressure deficit (kg H 2 O kg air −1 ); d 0 is an empirical constant (kg H 2 O kg air −1 ); c i is carbon dioxide concentration internal to the leaf (mol CO 2 mol air −1 ); Γ is carbon dioxide compensation point of assimilation in the presence of dark respiration (mol CO 2 mol air −1 ); LAI is leaf area index (m 2 m −2 ). A net (mol CO 2 m −2 s −1 ) is calculated following Farquhar et al (1980) and Collatz et al (1991Collatz et al ( , 1992. A net is only calculated when LAI and photosynthetically active radiation at the canopy top are greater than zero. The variable g s is scaled by a fractional parameter that balances the water supply from the roots with demand when supply is less than demand (Milly et al 2014). The minimum value of g s is 0.01 mol m −2 s −1 , and the maximum is 0.25 mol m −2 s −1 (both applied before conversion to m s −1 ). g s of ozone is estimated by scaling g s by the ratio of the diffusivity of ozone by the diffusivity of water vapor.
CUO requires concurrent estimates of the effective stomatal conductance (eg s ) and ambient ozone concentrations. eg s is the contribution of stomatal uptake to the ozone deposition velocity (v d ), a measure of the efficiency of the total ozone depositional sink irrespective of surface ozone concentration, in velocity units. Hereinafter, we will use the term g s to represent the conductance for ozone diffusion through stomata, whereas eg s to represent the strength of the removal of ozone by stomata. The variable v d (m s −1 ) is given by equation (2) in the dry deposition parameterization in AM3DD: This parameterization is based on a resistance network analogous to the treatment of resistances in Ohm's law for electrical circuits. The variable R a is the resistance to turbulent transport of ozone from the bottom of the atmospheric model to canopy height. In our big-leaf parameterization, all leaves are at canopy height. The variable R b,veg is the resistance to transport through the quasilaminar boundary layer around vegetation, R m is the resistance to ozone reactions inside the leaf, R cut is the resistance to ozone uptake by leaf cuticles, R stem is the resistance to ozone uptake by stems, R ac is the resistance to turbulent transport through the canopy to the soil, R b,soil is the resistance to transport through the quasi-laminar boundary layer around soil, and R soil is the resistance to ozone uptake by soil. Descriptions of R m , R cut , R b,veg , R stem , R ac , and R b,soil can be found in Clifton et al (2020b). CUO (mmol O 3 m −2 ) should be estimated at a frequency that captures surface ozone and eg s diel cycles. We calculate CUO for the 2010s and 2090s as the cumulative sum of hourly stomatal ozone fluxes (F stom, O3 ; mmol m −2 h −1 ) over a year. F stom, O3 is calculated by multiplying hourly fields of ozone in mmol m −3 and eg s in m h −1 . F stom, O3 follows Fick's law and assumes no ozone internal to the leaf given ozone's high reactivity with internal fluids and tissues (Laisk et al 1989, Wang et al 1995, Omasa et al 2000, Sun et al 2016. We do not employ a detoxification threshold for ozone damage here. A threshold is primarily used to Figure 1. Regional mean impact of interannual and daily variability of either surface ozone or effective stomatal conductance (egs) on annual CUO. Each symbol represents the percentage difference in annual CUO for each year. In particular, year-specific annual CUO is subtracted from annual CUO calculated with year-specific hourly-varying egs but multiyear monthly mean diel cycles of ozone (∆CUO O3 ) or annual CUO calculated with year-specific hourly-varying ozone but multiyear monthly mean diel cycles of egs (∆CUO egs ). Only grid cells with >50% land area are included. account for the plant's ability to detoxify ozone after it enters stomata and pertains more to the estimation of plant damage from CUO than the amount of ozone actually entering the leaf. The focus of our paper is not to quantify plant damage, but instead to quantify how variability in eg s and ambient ozone concentrations affect CUO.
We quantify the influence of daily and interannual variations in surface ozone versus eg s on CUO by calculating CUO from hourly archived fields of eg s and ozone from AM3DD (Clifton 2020). We identify the impact of variations in ozone on CUO by subtracting year-specific annual CUO from annual CUO calculated with year-specific hourly-varying eg s but multiyear monthly mean diel cycles of ozone. In other words, for each year (y), the difference in where the overbar is the multiyear monthly mean diel cycle transposed into an hourly array for all hours in a year, and h is hour. To identify the impact of variations in eg s on CUO, we subtract year-specific annual CUO from annual CUO calculated with year-specific hourlyvarying ozone but multiyear monthly mean diel To identify how changes in eg s versus surface ozone alter CUO over the 21st century, we calculate CUO in two ways: (i) with multiyear monthly mean diel cycles of ozone from the 2010s, but multiyear monthly mean diel cycles of eg s from the 2090s (CUO O3,2010 ) and (ii) with multiyear monthly mean diel cycles of eg s from the 2010s, but multiyear monthly mean diel cycles of ozone from the 2090s (CUO egs,2010 ). Inferring the role of changes in eg s versus ozone with our offline calculation fails for any grid cell where eg s is zero during the 2010s but nonzero at the end of the century. This happens for <1.4% of the grid-cell-hours for most regions examined (except 3.3% of the grid-cell-hours in the Midwest US and 4.2% in east Asia). However, we find that eg s values that are nonzero at the 2090s but zero at the 2010s are too small or infrequent to impact CUO.

Large role for interannual variability in stomatal uptake on CUO
CUO varies strongly from year to year, with the 2010s annual CUO relative interannual spread (coefficient of variation) ranging from 3.7% to 21.4% across regions. Meteorological variability influences both surface ozone and g s . For example, there is a strong correlation between ozone and temperature on daily and interannual timescales largely from the influence of transport patterns (Vukovich 1995, Barnes and Fiore 2013, Porter and Heald 2019, Kerr et al 2019. Variations in ecosystem-scale evapotranspiration and gross primary productivity, observable quantities related to g s , are influenced by meteorology on hourly-to-interannual timescales and by phenology and soil moisture, which vary more slowly, on seasonal and interannual timescales (Wilson and Baldocchi 2000, Katul et al 2001, Stoy et al 2005, Chen et al 2009, Baldocchi et al 2018. The influence of interannual variations in eg s on annual CUO is substantially larger than the influence of interannual variations in surface ozone for most Figure 2. Regional multiyear mean yearly progression of CUO for the 2010s, 2090s, and sensitivity calculations. CUOeg s ,2010 is calculated with 2010s multiyear monthly mean diel cycles of effective stomatal conductance, but 2090s multiyear monthly mean diel cycles of ozone while CUOO 3 ,2010 is calculated with 2090s multiyear monthly mean diel cycles of effective stomatal conductance, but 2010s multiyear monthly mean diel cycles of ozone. Only grid cells with >50% land area are included. regions (figure 1). Variations in eg s are critical for a given year's CUO relative to variations in ozone. Neglecting the role of eg s variations yields over-or underestimates in annual CUO by up to 6%-58% across regions.
Only in east Asia is there a comparatively large role for variability in surface ozone. The relative interannual spread in annual eg s in east Asia is weak relative to the other regions (4.1% versus 11.3%-22.8%) while the relative interannual spread in annual ozone is more within the range of other regions (1.4% versus 1.4%-2.6%) at the 2010s, suggesting that low eg s variability leads to the larger relative role for ozone variability there. Low eg s variability follows little hydroclimate variability-east Asia has high simulated summer rainfall and low relative interannual variation in rainfall relative to other regions.
While eg s interannual and daily variability is still more important for CUO than surface ozone interannual and daily variability at the 2090s, the absolute impact of eg s variability lessens for several regions at the 2090s (figure 1). The smaller role of eg s variability at the 2090s may be due to stomatal closure under high carbon dioxide and thus a weaker plant sensitivity to environmental stress such as drought (e.g. Field et al 1995, Swann et al 2016). Indeed, the model projects increases in regional summer mean water use efficiency (gross primary productivity divided Figure 3. Regional multiyear mean daily changes from the 2010s to the 2090s in effective stomatal conductance (egs) versus surface ozone. The color of symbols is the regional multiyear mean daily egs at the 2010s. Only grid cells with >50% land area are included.
by transpiration) of 40%-100% depending on the region. Large decreases in annual CUO tend to occur in regions with large decreases in summer surface ozone (compare figures 2 and S1). Summer ozone decreases under RCP8.5 from the 2010s to the 2090s in all of the regions examined here (figure S1). Decreases in summer surface ozone follow regional reductions in anthropogenic NO  (Clifton et al 2020b). Differences in 2010 regional NO x emissions, local ambient chemistry, and dry deposition, as well as background ozone contribute to regionally varying responses to changes in regional NO x emissions. While springtime CUO is higher for all regions by the 2090s, summertime CUO is lower for many regions because summertime eg s is lower by the 2090s (figures 2 and S3). Lower summertime CUO from changes in eg s counteracts higher springtime CUO from changes in eg s for all regions except the IMW and southwest US (compare 2010s and ozone_2010s CUO in figure 2). Similar summertime eg s at the 2010s and 2090s in the IMW and southwest US (figure S3) is likely due to offsetting between the expansion of vegetation coverage in these regions (Clifton et al 2020b) and the short-term impacts of high carbon dioxide on eg s (i.e. stomatal closure) (e.g. Field et al 1995, Betts et al 1997, Ainsworth and Rogers 2007. Lower 2090s summer eg s in the other regions (figure S3) likely follows stomatal closure due to high carbon dioxide.

Twenty-first century changes in CUO under RCP8.5
The 21st century eg s changes sometimes counteract or amplify the influence of surface ozone changes on CUO (figure 2). For the IMW US, slightly higher eg s for most of the year increases CUO and lower ozone decreases CUO, yielding little 21st century CUO change by the end of the year. For most other regions, lower eg s during nonwinter months (figure S3) leads to larger reductions in annual CUO by the 2090s, relative to the CUO reductions due to changes in ozone alone. For the SW US, reductions in annual CUO mostly stem from ozone reductions. While offsetting by temporally opposing changes in eg s and/or ozone imply that 21st century changes in annual CUO in some regions may be relatively small, temporal differences in the plant sensitivity to ozone

Conclusion
Here we probe the cumulative ozone uptake by stomata, a metric that accounts for the amount of ozone entering the leaf that can cause physiological injury. We examine the relative importance of temporal changes in surface ozone versus stomatal uptake using a new version of the GFDL global earth system model where the atmosphere and land are coupled through exchanges of carbon, water, and energy as well as dry deposition of reactive gases including ozone. We find that accurate estimates of the cumulative stomatal ozone uptake require considering interannual variations in stomatal functioning, supporting observational and modeling evidence that recent changes in cumulative stomatal ozone uptake cannot be explained by ozone changes alone (e.g. Anav et al 2019, Ronan et al 2020). We emphasize that our study is a sensitivity analysis of the influence of ozone versus stomatal conductance on the cumulative stomatal ozone uptake-an assessment of changes in ozone damage requires advanced understanding of plant detoxification ability and responses to ozone at regional scales. Decreases in water use efficiency from ozone plant damage (Lombardozzi et al 2015, Hoshika et al 2015 may increase the effect of water stress on plants and thus alter interannual variability in stomatal activity, implying a need to better understand how variability in stomatal ozone uptake feeds back on itself. Nonetheless, our results suggest that, without substantial changes in NO x emissions from year to year, the highest ozone damage may occur in highly productive (i.e. high stomatal conductance) years, rather than high-ozone years. The important role for interannual variability and 21st century changes in stomatal conductance highlighted here challenges the validity of widely used approaches employing only ambient ozone concentrations to assess ozone plant damage and protect vegetation. model with ozone dry deposition was supported by NOAA's Climate Program Office's Atmospheric Chemistry, Carbon Cycle, and Climate program Grant NA14OAR4310133.

Data availability statement
The data that support the findings of this study are openly available at the following DOI: https://doi.org/ 10.5065/wkdj-2s62