Introduction

The complex interplay between air pollution and climate change presents a challenge for their prediction and for the assessment of their total risk in the future [1, 2]. Changes in climate conditions will alter the emission, transport, chemical evolution, and removal of air pollutants and their precursors [1, 3, 4]. In turn, ozone (O3) and fine particulate matter (PM2.5), the two major air pollutants, act as near-term climate forcers (NTCFs) that perturb the Earth’s radiative energy budgets [1, 3, 5]. Moreover, many air pollutants and their precursors share common anthropogenic sources with major greenhouse gases (GHGs) and are thus concertedly tied to the trajectory of future emission scenarios and climate mitigation strategies [6, 7].

In increasing levels of complexity and human involvement, anthropogenic climate warming, which is driven by increased levels of GHGs, can impact air quality in the following ways:

  1. (a)

    Through changes in physical climate conditions affecting atmospheric processes, such as temperature, precipitation, humidity, circulation, and actinic flux

  2. (b)

    Through climate-induced changes in natural emissions of pollutants and their precursors, such as emissions from vegetation, soil, lightning, and wildfires

  3. (c)

    Through pollutant produced by increasing atmospheric GHGs, such as the ozone produced by methane

  4. (d)

    Through climate-induced changes in the land surface and the ecosystem, such as the aridification of land, and the changes in types and densities of vegetation

  5. (e)

    Through changes in climate-sensitive anthropogenic emissions in the absence of changes in anthropogenic activities, such as the enhanced volatilization of volatile organic compounds (VOCs) from fuels and solvents

  6. (f)

    Through changes in anthropogenic pollutant emissions due to adaptation or mitigation policies against climate change, such as the increased emissions due to additional energy usage for cooling, the reduced emissions by actively mitigating NTCFs, or the co-reductions in anthropogenic pollutant emissions as a result of GHG mitigation

An important, quantitative diagnostic of the impacts of future climate on air quality is the “climate change penalty,” which is the deterioration of air quality due to a warming climate, in the absence of changes in anthropogenic polluting activities [1, 3]. This is often defined as the combined impacts of pathways (a) to (d) above, although the impacts of pathway (e) are sometimes included as well (e.g., [8]). The flip side of the climate change penalty is the “climate control benefit,” which is the amelioration in air pollution when climate warming is slowed in the absence of changes in anthropogenic polluting activities [9]. In addition, improvements in air quality can be gained from the co-reduction of anthropogenic pollutant emissions as a result of GHG and NTCF mitigation policies (pathway (f) above), often referred to as the “climate policy co-benefit” to air quality [6, 7, 10]. The sum of the climate control benefit and the climate policy co-benefit represent the total improvement in air quality as a result of climate mitigation, relative to unconstrained climate warming [9]. Many recent studies have shown that capping the global average temperature increase by 2100 to within 1.5 °C or 2 °C of the pre-industrial average can restrain the deterioration of air quality, such that the climate policy co-benefits on regional and global air quality may win out [11••, 12,13,14,15]. The associated economic gain in terms of public health and agriculture can offset, potentially completely, the cost of GHG mitigation for both individual nations and the world (e.g., [11••, 12, 13,14,15,16]). An integrated mitigation strategy for both air quality and climate is therefore necessary but highlights the need to better quantify the impacts of climate change on air quality.

In particular, the climate change penalty on ozone air quality involves many complex meteorological, chemical, and biological processes and feedbacks, many of which are still not well understood [1, 3]. A large number of recent studies have examined in depth the statistical relationships between ozone and meteorology, as well as the changes in surface ozone in response to climate-sensitive processes in the atmosphere and the ecosystem. In this way, many recent studies have uncovered critical processes that materially altered our understandings of the climate change penalty to surface ozone.

Aims and Scope

In this brief review, we examine the latest understanding on the climate change penalty to surface ozone, with emphases on the underlying processes and knowledge gaps. Table 1 provides an updated summary of the climate-sensitive processes reviewed here, their effects on ozone air quality, and the associate uncertainties. We build on several previous reviews on the subject with regional and global perspectives [1, 3, 4, 17,18,19] and report mainly on the new findings since 2015. The climate change penalties on PM2.5 air quality also involve many of the same processes examined here but are further complicated by opposing influences on the various PM2.5 components, as previously reviewed (e.g., [1, 20]). In addition, the atmospheric concentrations and life cycles of many other atmospheric hazardous substances, such as persistent organic pollutants, mercury, and many aeroallergens, are also known to be climate sensitive [17, 21,22,23]. There may be considerable climate mitigation benefits associated with these hazardous substances that warrant in-depth assessment.

Table 1 Impacts of climate-sensitive processes on surface ozone over land, updated from Jacob and Winner [17] and Fiore et al. [3] to reflect recent findings
Full size table

Climate-Sensitive Processes That May Impact Future Ozone Air Quality

Background Tropospheric Chemistry and Ozone Transport from the Stratosphere

Ozone in the troposphere is mainly produced by the oxidation of methane, CO, and non-methane VOCs in the presence of solar radiation and nitrogen oxides (NOx≡NO + NO2), with some transport from the stratosphere. Its removal from the troposphere involves photolysis followed by reaction with water vapor, reactions with HO2 and OH, and dry deposition to the Earth’s surface, resulting in a global mean lifetime of approximately 23 days but with considerable seasonal and spatial variation [5].

Previous studies generally predicted a decline in baseline surface ozone with climate warming, as a result of greater water vapor abundance accelerating ozone chemical loss in the background lower troposphere [1, 3, 24]. However, the rapid increase of global methane levels under the less controlled climate scenarios may raise 2100 baseline surface ozone by as much as 8 ppb, annihilating both the water vapor effect and the efforts of reduced anthropogenic precursor emissions on surface ozone over some regions [5, 25, 26••, 27, 28•, 29•]. In addition, transport of ozone from the stratosphere will likely increase as a consequence of both climate change and the recovery of stratospheric ozone as levels of ozone depleting substances decline [30•]. Under the RCP8.5 scenario, the annual average difference in stratospheric origin ozone concentration at sea level between 2000 and 2100 may exceed 5 ppb in the Southern Hemisphere, over mountainous areas, and over certain regions such as the Mediterranean Basin [30•].

Climate-induced changes in lightning NOx emissions may affect baseline surface ozone concentrations in complex ways [31•, 32]. On a global scale, sustained enhancement of lightning NOx emissions will initially increase background ozone, but the associated increase in OH will decrease global methane concentrations, which in turn ultimately reduce background ozone production on decadal timescales [31•]. However, the direction of change of lightning activities in the future remains highly uncertain. A recent study found that the simulated lightning flash densities based on relatively simple parameterizations of cloud top height or cold cloud depth agreed well with the observed spatial distributions of lightning [33]. Predictions based on those two parameterizations projected moderate to large increases in global lightning for future climate conditions. However, another study used a new parameterization that took into account the key effects of cloud ice fluxes and projected a decrease in global lightning under climate warming [34].

Surface Sources/Sinks of Ozone and Its Natural Precursors

Atmosphere–Vegetation Exchange of Ozone and Biogenic VOCs

Over vegetated continental areas, surface ozone production is often dominated by the photo-oxidation of biogenic VOCs (BVOCs), which include not only most importantly isoprene (C5H8) but also other alkenes, terpenes, and oxygenated VOCs. The emission of BVOCs from plants is highly variable across species and non-linearly dependent on temperature, sunlight, soil moisture, leaf physiology, and other environmental variables [35]. As such, an important way by which climate warming can affect surface ozone is by altering the fluxes of BVOCs. Early studies mostly scaled up leaf-level temperature and solar radiation sensitivities, predicting a large increase in BVOC emissions as a result of climate warming on local to global scales (e.g., [35, 36]).

A growing number of new studies are contesting this paradigm of much higher BVOC emissions in the future when the complex interactions between the biosphere, climate, and surface ozone are taken into account. For example, while future elevated CO2 concentrations may enhance the global net primary productivity and thus isoprene emissions, they also may “inhibit” isoprene emissions by uncoupling them from photosynthesis, an effect that has been shown to substantially offset or even reverse the sign of future BVOC emission changes on regional and global scales [37, 38•, 39•, 40]. Intra- and interspecies competitions within the ecosystem responding to climate change may also reduce the abundance of high BVOC-emitting species, nullifying the effect of warming on BVOC emissions [39•, 41].

Ozone itself interacts with vegetation to affect the net surface fluxes of BVOCs and ozone. Cumulative stomatal uptake of ozone damages photosynthesis, leading to reduction in leaf area index (LAI), gross primary productivity (GPP), crop yields, and transpiration, ultimately constituting a negative feedback via BVOC emissions [42, 43••, 44•, 45••, 46•, 47••, 48]. At the same time, reduction in LAI slows the dry deposition velocity of ozone via stomatal uptake, constituting a positive feedback [42, 47••, 49]. A model study by Zhou et al. [47••] showed that, on a global scale, the positive feedback of ozone via its dry deposition out wins the negative feedback via BVOC emissions, particularly over the tropical forests. However, several studies showed that the dry deposition of ozone over many vegetated surfaces may be driven mainly by non-stomatal mechanisms and not by stomatal uptake [50,51,52], or the stomatal uptake maybe strongly modulated by climate conditions [48]. As such, the positive feedback on ozone may not be as pronounced as modeled by Zhou et al. [47••]. Other feedbacks recently explored included the modified vegetation transpiration feeding back into local climate [44•] and the difference in ozone tolerance leading to competition within the ecosystem [53]. Overall, it has become clear that the response of vegetation to climate change and ozone stress is too uncertain to reliably predict the change in future BVOC emissions and ozone deposition.

Soil Emission of NOx

In addition to the emissions of BVOCs, the chemical and microbial emissions of NOx from soils are also highly sensitive to meteorological variables [54]. The impacts of the soil NOx climate sensitivity on the climate change penalty of surface ozone have so far largely been overlooked, but they may become increasingly important as anthropogenic NOx emissions continue to decline over many areas. An example of this was over the Southeast USA, where one analysis showed that the ozone enhancements on hotter days were mainly driven by changes in local chemistry, with an integrated ozone–temperature sensitivity of 2.3 ppb K−1 [55•]. Nearly half of that ozone–temperature sensitivity was attributable to temperature-enhanced soil NOx emissions of 0.23 ng N m−2 s−1 K−1, most likely by soil microbes [55•]. Most current models still use an empirical parameterization developed in the 1990s to compute soil NOx emission as a linear function of temperature and a step function of precipitation [56]. However, the soil NOx fluxes thus computed were significantly lower than those derived from satellite- and ground-based measurements and misrepresented the observed spatiotemporal patterns [57]. Hudman et al. [54] developed a more mechanistic parameterization for soil NOx emission that is non-linearly dependent on soil temperature, soil moisture, soil wetting by precipitation after a drying period, as well as the soil nitrogen content from deposition and fertilizer application [54]. They found that this new parameterization better reproduced the magnitude and the spatiotemporal of tropospheric NOx columns observed from space [54]. However, the impacts of this new parameterization on the climate change penalty of surface ozone have not yet been explicitly evaluated.

Wildfire Emissions of Ozone Precursors

Wildfires are strong sources of ozone precursors [46•, 58, 59•], and climate change alone will likely increase the global risks of wildfires in the future [58, 60••, 61]. Changes in climate conditions alter the abundance, distribution, and aridity of vegetation biomass available for burning, as well as the natural ignition, spread, and intensity of wildfires [58, 62]. Many analyses of satellite and ground-based fire observations have found marked increases in regional wildfire activities in recent decades attributable to climate warming, including in particular over the western USA [63, 64], where wildfire emissions currently contribute 2–8 ppb of summertime monthly mean ozone at individual sites [59•]. In contrast, the historic emission inventory recently compiled for the Coupled Model Intercomparison Project Phase 6 (CMIP6) showed relative stable global biomass burning emissions between 1750 and 2015 despite pronounced global warming [65], but the lack of robust fire data during the early periods may have confounded that estimate. A recent study diagnosed the fire weather index (FWI) in 17 simulations of historical climates and future climates under the RCP8.5 scenario [60••]. Relative to historical conditions (1861 to 2005), that study showed a distinct increase in the number of high FWI days over 22% of the world’s burnable land area by 2019, including over the Amazon. By the mid-twenty-first century, the number of high FWI days was found to increase over 62% of the world’s burnable lands, including much of North America, Europe, and South China. Increases in the daily maximum temperature and decreases in relative humidity were found to be the dominant driver for the future increase in fire risks [60••].

More uncertainties arise when the dynamic interactions between fire, climate, air quality, and the terrestrial ecosystem are considered [58, 62]. Fires modify the structure and distribution of the terrestrial ecosystem and feedback to the atmosphere via changed albedo, ecosystem properties, transpiration, and emissions of BVOC, GHG, and other trace species [58, 62, 66, 67]. In addition, fire-induced surface ozone pollution may significantly contribute to the ozone damage of global terrestrial productivity, potentially forming a negative feedback [46•]. Investigations of these complex air quality–climate–ecosystem interactions and their impacts on wildfires and ozone have only just begun [46•, 62].

Photochemistry of Ozone Formation

Even if the emissions of biogenic VOCs do increase with climate warming, the associated response of surface ozone over BVOC-rich areas is still quantitatively uncertain [68]. A canonical example of this is the Southeast USA, where model studies disagreed on even the sign of surface ozone change due to warming-enhanced isoprene emissions [1, 17]. This lack of consensus has been mainly attributed to the different model assumptions on the yields of isoprene nitrates and their subsequent NOx-recycling ratios [3, 17, 68], as the formation of non-NOx-recycling isoprene nitrates terminally removes NOx from the photochemical reactions leading to ozone formation. Over the last decade, great progress has been made in laboratory and field experiments regarding the photochemistry of isoprene [69, 70]. The yields of isoprene nitrates are now better constrained at about 12–16% and their subsequent NOx-recycling ratios at about 50% [69, 70], but not all models have been updated to reflect these changes [68, 71]. In addition, it has been shown that the effects of these photochemical parameters in models are resolution-dependent, as they partially compensate for the artificial mixing between NOx and isoprene in coarser resolution models [72, 73]. As such, the laboratory-derived values for these parameters may not be directly applicable in models. One recent study pointed out that the interannual variation of August temperature over the Southeast USA between 1988 and 2011 manifested a continental-scale temperature oscillation, which provided an opportunity to diagnose the response of surface ozone to large-scale warming [74••]. That study concluded that surface ozone does increase in response to the enhanced isoprene emissions driven by large-scale warming and that the observed ozone–isoprene sensitivity was consistent with a 12% first-generation yield and a 55% subsequent NOx-recycling ratio for isoprene nitrate for their model at 2.5° longitude × 2° latitude resolution.

Synoptic- to Seasonal-Scale Meteorology

Over polluted areas, high levels of surface ozone have often been shown to correlate with individual local meteorological variables, such as warm temperature, strong solar radiation, low relative humidity, and low precipitation, on synoptic and seasonal timescales (e.g., [75••, 76••, 77•, 78•, 79, 80]). These correlations form some evidence that changes in regional meteorology due to climate warming will likely incur changes in future surface ozone. However, as these local, short-term correlations do not indicate actual causal processes, they cannot be directly extrapolated to estimate surface ozone in future climate. Instead, much effort has been in identifying the synoptic weather patterns that drive surface ozone variability and the observed correlations between ozone and meteorological variables, with the goal of quantifying the changes of those weather events in future climate to predict surface ozone change (e.g., [81,82,83,84,85]).

Many such synoptic and seasonal-scale weather patterns have been identified as partial drivers for high surface ozone, and the potential changes in some of those weather patterns in future climate have been explored. Over eastern North America, the variability of summertime surface ozone depends strongly on the north–south shift of the polar jet and the westward extension of the Bermuda High [86•, 87•]. Severe drought conditions have also been shown to enhance summertime US surface ozone by 3.5 ppb [79], and an increase of 1–6% for surface ozone has been estimated by 2100 compared to the 2000s due to increasing drought alone [79]. Over Europe and the UK, anti-cyclonic conditions, easterlies, blocking, and subtropical ridges have all been shown to significantly enhance surface ozone [81, 83, 84]. Over eastern China, ozone pollution events are often linked to the anti-cyclonic conditions or atmospheric subsidence associated with the Pacific Subtropical High [82]. Over coastal southern China, high ozone events are often associated with the subsidence ahead of an approaching tropical cyclone [85]. As a result, changes in tropical cyclone numbers and trajectories in the future will potentially both impact local ozone air quality, although those metrics are difficult to predict in climate models [85].

A condition commonly coupled to many of the weather events above is regional air stagnation, which has been widely accepted as a large driver for the observed positive ozone–temperature sensitivity [3, 78•, 88•, 89, 90]. Regional stagnation often occurs under the influence of anti-cyclonic air masses and subsidence, and in that way, stagnation may be correlated with high temperature, shallower mixing layer, less cloud cover, less precipitation, and strong radiation. Mechanistically, stagnation suppresses the horizontal and vertical ventilation of ozone and its precursors from the boundary layer and can thus build up surface ozone [71]. A number of studies have characterized future stagnation occurrences using the Air Stagnation Index (ASI), which is a Boolean index defined by light winds near the surface and in the mid-troposphere and no precipitation [91, 92]. These studies generally found more frequent stagnation occurrences in the future and implied a potential worsening of ozone air quality [91, 92]. However, recent analyses found that the observed summertime surface ozone and high ozone events were only weakly correlated against either the ASI or the number of stagnant days over the USA on both daily and seasonal scales [75••, 93•, 94••, 95]. In addition, no positive correlation was found between daily temperature and the ASI [95]. As such, the use of ASI as a predictor for present or future levels of surface ozone appears not to be sound. It is yet unclear why the ASI is not significantly correlated with surface ozone and temperature. It may be that the meteorological variables that define the ASI (light wind and no precipitation) do not capture the actual “stagnant” conditions leading to high surface ozone [95]. Alternatively, it is possible that suppressed ventilation is simply not a key driver for high ozone, nor is it a key driver for the observed ozone-temprature sensitivity.

Early model studies showed that climate warming can lead to preferential increases in the high-end of the ozone probability distribution, which has been widely attributed to more frequent occurrences of extreme high temperature and heat waves in the future [3, 17]. Much attention has thus been directed at the association of extreme ozone pollution with extreme warmth or heat wave occurrences, as the combined stress may produce disproportionally serious public health risks [77••, 88•, 93•, 96, 97, 98••]. Some observational analyses over the USA showed that the ozone–temperature sensitivities are larger at higher ozone quantiles [76••], while the probability of summertime severe ozone pollution when there are heat waves could be up to seven times of the average probability [93•]. A model simulation for present-day climate also showed enhancements of surface ozone exceeding 10 ppb over eastern USA, Europe, and China on heat wave days relative to non-heat wave days [99]. Schnell and Prather [88•] showed that extreme events of ozone and temperature over North America do generally overlap, but their occurrence shows consistent offsets in space and in time, possibly reflecting the response scale and time of natural precursor emissions and pollutant advection [88•].

However, other recent observations and model analyses showed muted sensitivities of ozone to temperature on the higher end of their probability distributions [100] and no preferential increase in extreme ozone probability associated with heat wave days on the interannual timescale [94••]. One model projection indicated slight decreases in ozone over the Southeast USA on heat wave days in the future, despite that model showing large ozone enhancements over the same region on heat wave days for the present climate [99]. Thus, the connection between extreme ozone and extreme temperature may not be as robust as previously believed. Possible explanations may involve chemical and biophysical feedbacks, such as the diminished role of nitrogen oxide sequestration by peroxyacetyl nitrates and the reduced biogenic isoprene emissions at high temperatures [101]. Interestingly, two recent studies, based on observations and simulations, respectively, found larger ozone enhancements associated with the co-occurrence of heat wave and stagnation events than when either of those events occur in isolation [93•, 98••]. This may provide an explanation as to why the ASI or the heat wave occurrence individually is not robust predictors of high surface ozone.

Another newly proposed way by which climate change affect surface ozone is by lengthening the local ozone season [102••]. Zhang and Wang [102••] showed that the high ozone observed over the Southeast USA in October 2010 was driven by the abnormally warm and dry conditions that fall, which enhanced biogenic isoprene emissions, photochemical production, air stagnation, and fire emissions. They pointed out that the further warming and drying of fall climate over the Southeast USA in the future, as projected by climate models, will likely deteriorate seasonal ozone air quality [102••]. This seasonal effect of climate change on ozone can presumably also happen over other vegetated areas.

Interannual and Interdecadal Climate Variability

Several modes of natural climate variability on interannual, decadal, and multidecadal timescales have been shown to influence regional surface ozone, possibly through their impacts on regional meteorology [74••, 103•], wildfires [104], long-range transport [105], and ozone transport from the stratosphere [106, 107]. While these internal variability modes may not be GHG forced, their modulation of surface ozone concentrations in the coming years and decades is extremely relevant for near-term air quality–climate policy-making [108••, 109••]. Over the USA, the warmer, drier, and more stagnant weather, as well as the anomalous circulation patterns during the warm phase of the Atlantic Multidecadal Oscillation (AMO), has been shown to enhance observed summertime surface ozone by 1–4 ppb relative to the cold phase [103•]. Observed surface ozone over northeastern USA has also been shown to correlate with the Arctic Oscillation (AO), the Pacific Decadal Oscillation (PDO), and the Quasi-Biennial Oscillation (QBO) [75••]. Over Southeast Asia, the hot and dry conditions during strong El Niño years can lead to intense fires, which in turn can enhance annual mean surface ozone by 50 ppb near fire sources [104]. Analyses of satellite observations showed that the interannual variability of the tropospheric column ozone (TCO) over the tropics and many mid-latitude areas can be largely explained by El Niño–Southern Oscillation (ENSO) [107], and some of that ENSO influences on the TCO are manifested in surface ozone [106, 110].

Chemistry–Climate Model Projections of the Climate Change Penalties on Ozone Air Quality

Finally, Fig. 1 shows the projections of the climate change penalties on the ozone air quality in polluted regions reported by chemistry–climate model studies published since 2015 (further details are summarized in Table S1). The purpose of this summary is twofold: (1) to compare with previous reviews [3, 17, 120] to examine whether there have been material changes in regional model projections and (2) to identify regions with large changes or uncertainties to direct future research. Almost all recent studies summarized in Fig. 1 focused on North America or Europe; the exceptions were one study over East Asia [117•], one study over India [118•], one study for the globe [24], and one study that analyzed the results from four global models over North America, Europe, and East Asia [119••].

Fig. 1
figure 1

Model projections of changes in summertime or annual mean ozone air quality due to climate change only, reported in recent studies. This figure is adapted and updated from Jacob and Winner [1] and Fiore et al. [2]. Individual studies are indicated by alphabets: A [111•], B [11••], C [112•], D [113•], E [114•], F [115•], G [116•], H [28•], I [117•], J [118•], K [8], L [26••], M [119••], and N [24]. Climate scenarios are indicated by color: POL3.7 (blue); RCP4.5 and POL4.5 (green); RCP6 and A1B (orange); RCP8.5, A1Fi, and A2 (red). Most studies reported changes (symbols) or range of changes (bars) in mean ozone concentrations or maximum daily 8-h average ozone concentrations averaged over June–August (circles and solid line bars) or averaged over the full year (triangles and dashed line bars). One study [26••] reported changes in the sum of daily mean ozone concentration over 35 ppb (SOMO35). Further details of these recent studies are summarized in Table S1

Full size image

An important implication for the ozone modulation by climate variability is that interpretation of observations and simulations to diagnose ozone trends forced by changes anthropogenic climate warming must be done with caution [108••, 109••, 121]. Early model assessments of the climate change penalty on surface ozone did so by comparing ozone differences between a future time slice against the present-day time slice, averaging the simulations over 3–5 years to represent each time slice [3, 17]. Recent studies showed that at least 15-year averaging windows are required smooth out the “noise” due to natural climate variabilities and to distinguish signals forced by anthropogenic climate warming [8, 109••]. As a result, most studies reported in Fig. 1, although not all of them, used 10- to 30-year simulations to represent the ozone air quality of a particular time slice. In addition, Barnes et al. [108••] showed that simulated trends of surface ozone robustly attributable to climate warming alone may not emerge above the noise driven by internal variability until after 2050. Also, chemistry–climate models need to be properly initialized for multidecadal variabilities (e.g., AMO) or use initial condition ensembles to represent the uncertainties associated natural variabilities [74••, 109••].

Over North America, a previous summary [17] concluded that surface ozone will generally increase by 1–10 ppb in response to climate warming, particularly over present-day high-ozone areas such as Northeastern USA. Significant declines in future ozone were only predicted by some models over the Southeastern USA, where the assumptions associated with isoprene nitrate chemistry have great impacts on the simulated response of ozone to warming-enhanced isoprene emissions, as discussed above [17]. Most of the recent studies reported in Fig. 1 also projected rising future surface ozone concentrations, with the larger amounts of ozone increases associated with more significant warming. However, the projected ozone increases up to 7 ppb were more moderate compared to those projected by earlier studies [11, 111•, 112•, 113•, 114•, 115•]. The reason for the smaller ozone changes in recent studies may be that the uses of longer averaging windows smooth out the ozone fluctuations driven by internal climate variability. In addition, recent studies mostly predicted largest ozone increases occurring over the Midwest [11, 111•, 112•, 113•, 114•, 115•], although the reason for this is yet unclear. Notable exceptions were the projections by Rieder et al. [28•, 116•] and Schnell et al. [119••]. Rieder et al. [28•, 116•] found that future ozone will increase over Northeastern USA but will decrease significantly over Southeastern USA and Western USA. However, their model did not include temperature-sensitive BVOC emissions. Three of the four global models in Schnell et al. [119••] also did not include temperature-sensitive BVOC emissions and predicted varying directions of ozone change; the one model that did include temperature-sensitive BVOC predicted large ozone increase throughout the USA. Interestingly, Val Martin et al. [113•] argued that the CO2 inhibition of isoprene emissions will likely substantially reduced the role of isoprene in the climate change penalty on ozone. As a result, they did not consider temperature-sensitive BVOC emissions in their simulations but still predicted increased future ozone throughout USA due to changes in meteorology alone [113•].

Over East Asia, Lee et al. [117•] simulated the impacts of climate warming on surface ozone but ignored the temperature sensitivity of BVOC emissions. They found a large decrease in annual mean ozone concentrations, particularly over Northeastern Asia, due to projected increase in precipitation and cloud coverage. Over India, Pommier et al. [118•] projected a significant increase in surface ozone of + 2 ppb over Northern India and a decrease of − 1.4 ppb over Southern India. They attributed this regional contrast to climate-induced changes in stomatal conductance in their model, which in turn led to changes in the dry deposition velocity of ozone.

Over Europe, a previous synthesis of 25 model projections found that climate warming will likely bear a penalty on surface ozone over most parts of continental Europe while slightly reducing ozone in Scandinavia in the future [120]. A similar latitudinal dependence of future ozone change in response to climate warming was also projected in two recent studies [8, 26••]. This latitudinal dependence of ozone sensitivity was likely driven by changes in BVOC emissions, as only the one global model with temperature-sensitive BVOC emissions in Schnell et al. [119•] showed a similar pattern. That global model also found dramatic delays in the seasonal timing of peak ozone by 18 to 60 days over Southern Europe, Eastern USA, and Southern California, possibly driven by enhanced isoprene emissions due to warmer and drier conditions in future falls [102••].

Finally, the two recent global studies both found general increases in surface ozone in a warming climate over mid-latitude continental areas if temperature-sensitive BVOC emissions were taken into account, but neither studies included the CO2 inhibition on BVOC emissions [24, 119••]. Schnell et al. [119••] found an increase in wintertime surface ozone over the Northern mid-latitude oceans, presumably due to enhanced ozone transport from the stratosphere. In summer, they found consistent decreases in surface ozone over the Northern hemispheric oceans as a result of increased water vapor. In contrast, Glotfelty et al. [24] found an increase in global surface ozone (global average 0.6 ppb) due to climate warming, including over most parts of the global ocean. Significant reduction in surface ozone due to water vapor increases was only found over the tropical ocean and some tropical land areas [24].

The examination of Fig. 1 and comparison against earlier studies indicate that current models generally projected deteriorating surface ozone air quality in the future, but the amount of deterioration attributable to climate change may be smaller than previous assessments due to the noise from climate variability. In addition, the climate change penalty on ozone is very much dependent on the sensitivity of BVOC emissions to climate change. It is important to note that none of the more complex atmosphere–ecosystem feedbacks reviewed in this paper, such as the CO2 inhibition of isoprene emissions, the ozone damage of leaf and photosynthesis, and the dynamical changes of plant species, in response to climate, ozone, and fire stress, were taken into account in the projections reported in Fig. 1.

Conclusions

This review examined recent studies on the impacts of climate warming on ozone air quality and identified critically uncertain processes that impede the robust quantification of the climate change penalty on surface ozone. The complex ecosystem feedbacks in response to climate warming, rising CO2 levels, ozone pollution, and other environmental stresses may materially alter natural precursor emissions in the future, but the directions of changes cannot be robustly determined at this time. Previously accepted associations between high ozone events with stagnation and heat waves require further examination. In addition, natural climate variability may drive large fluctuations in surface ozone and may obscure the warming-forced signal until the latter half of the twenty-first century. Recent literatures generally project a climate change penalty on ozone air quality, although the magnitudes are smaller than those projected by earlier studies. However, many of the known processes and feedbacks have not been taken into account in the recent projections. These key knowledge gaps require further study and representation in models to better quantify the climate change penalty on surface ozone to better inform air quality and climate policy-making.