Uncovering the key role of stratospheric oZOne in Northern Hemispheric climate variability and change (UZONH)
1. Proposal Summary
While the influence of stratospheric ozone on tropospheric climate in the Southern Hemisphere (SH) has been extensively assessed, large uncertainties remain for its effects on the Northern Hemisphere (NH) climate. Arctic ozone exhibits small long-term trends, but large interannual variability, and has been commonly viewed as a passive constituent that responds to stratospheric polar vortex variability. This view has been challenged by recent studies, suggesting that ozone in the NH high latitudes can also feed back into the circulation, playing an active role in surface climate. However, there is controversy about the causality in the ozone-circulation relationship. This calls for a more conclusive assessment of the impact of ozone on stratosphere-troposphere coupling, and its role in predictability. Stratospheric ozone influences in the NH may also extend to longer time-scales. Over the 21st century, Arctic ozone is projected to exceed historical levels, and this long-term trend could also affect future climate projections through both dynamical and radiative processes. Moreover, recent studies have documented that the inclusion of an interactive ozone chemistry in climate change simulations can reduce (by up to 20%) the projected surface warming: this is largely due to a feedback involving stratospheric water vapor. However, the magnitude of the feedback is model-dependent, and the source of this uncertainty is unclear. In addition, there is considerable spread in the projections of ozone recovery, whose dynamical effects on NH surface climate has not been studied so far. In summary, the impact of ozone on the NH climate remains largely unknown, and is subject to large uncertainties due to complex coupling between ozone, the circulation and climate: these interactions are not well represented in state-of-the-art climate models.
This project will provide answers to these questions by combining observational data, and a hierarchy of models, thus offering the unprecedented opportunity of assessing the role of ozone in NH climate on all time-scales. Using two models that allow coupling and de-coupling of stratospheric ozone chemistry, we will assess the influence of ozone on the variability of the Arctic polar vortex, stratosphere-troposphere coupling and surface climate, including its influence on the predictability of the North Atlantic Oscillation (NAO), the largest mode of climate variability in Europe. We will quantify the impact of ozone on longer time-scale, such as future ozone recovery and its effects on the mid-latitude jet in the North Atlantic, an important proxy for storminess over Europe, and link the inter-model spread in ozone recovery with uncertainty in NH circulation trends. Finally, we will explore the role of ozone as a radiative feedback in tropospheric climate, via its coupling with stratospheric water vapor, by using offline radiation codes and radiative-convective-equilibrium models.
This project will provide novel insights into the role of stratospheric ozone in the Earth system, through a process-based understanding of the ozone feedbacks. It will assess the benefit from an accurate representation of ozone chemistry for seasonal to multi-decadal predictions, whilst proposing ways to improve the representation of this coupling in the current generation of models. Finally, it will contribute to narrowing the uncertainty on long-term climate projections. Output from this project could be used to guide environmental policies in the future, and thus be valuable for policy-makers. In this sense, this project will help the Ambizione program in its mission to provide impactful research for the benefit of the society.
2. Research Plan
2.1 Current state of research in the field
Due to its high efficiency in absorbing solar radiation and emitting at longer wavelengths, ozone largely determines the radiative budget of the stratosphere (Ramanathan and Dickinson, 1979). The stratospheric ozone distribution can be altered by changes in dynamics and chemical composition (Pyle et al.,, 2005), and the resulting ozone changes can induce radiative forcing (WMO 1998) and circulation changes in the troposphere (WMO 2014). Thus, the coupling between ozone, radiation and the atmospheric circulation is an important element of the climate system, and a proper simulation of these interactions is essential to capture key aspects of climate variability and change. Yet, due to computational constraints the majority of the climate models to date simply prescribe the ozone concentrations, and hence do not have interactive ozone. The impact of this coupling on the NH climate is the subject of this proposal.
2.1.1 The role of ozone in Northern Hemisphere climate variability
Springtime ozone depletion in the Antarctic stratosphere has been deemed as one of the key drivers of recent circulation changes in the SH (Previdi and Polvani, 2014). While the impact of stratospheric ozone in the SH are well understood, the role of ozone in NH climate has received less attention. This is due to the small size of Arctic ozone trends, from which negligible effects on surface climate are expected (WMO,2007;2010;2014). However, Arctic ozone concentrations exhibit large interannual variability. An example is the ozone minimum of March 2011 (Manney et al, 2011), comparable in some altitudes (18-20 km) to the Antarctic ozone hole, which preceded anomalous patterns in the troposphere, such as a poleward shift of the mid-latitude westerly jet, and positive phase of the NAO. These ozone variations are mostly determined by the strength of the stratospheric polar vortex, which controls dynamical ozone supply and chemical ozone loss (Tegtmeier et al., 2008; Strahan et al., 2016). However, low levels of ozone of Arctic ozone cause the polar vortex to further cool and strengthen due to reduced absorption of shortwave (SW) radiation, providing a positive feedback between chemistry and dynamics. In the presence of present-day stratospheric chlorine loadings, cold winters/springs become more frequent for years with very low concentrations in the Arctic (Dameris et al., 2014). This raises the question whether Arctic ozone actively contributes to the circulation anomalies, which lead to NH surface climate anomalies.
In this context, a number of model studies have investigated the impact of Arctic ozone on surface climate, by forcing Atmosphere-only General Circulation Models (AGCMs) with zonal-mean ozone anomalies for the year 2011 (Karpechko et al., 2014) or mimicking inter-annual variability in observations (Smith and Polvani, 2014). In order to obtain a robust signal in these studies, ozone anomalies larger than observed (Smith and Polvani, 2014), or in combination with observed sea surface temperature anomalies for the year 2011 (Karpechko et al., 2014) were imposed. Similarly, another study found no surface response to Arctic zonal-mean ozone anomalies of that year (Cheung et al., 2014). . The failure of some AGCMs in reproducing this relationship could possibly be due to their use of zonal- and monthly-mean ozone forcing, as by doing so, they do not capture the interactions between ozone and planetary waves, which are manifest as longitudinal ozone asymmetries (Albers et al., 2012). Neglecting these asymmetries has been shown to lead to a weaker tropospheric circulation response (Gillet et al., 2009; Waugh et al., 2009a), but their role for observed Arctic ozone variability has not been explored.
A more recent study based on a chemistry-climate-model (CCM) found a robust relationship between Arctic ozone and surface climate (Calvo et al., 2015). Most importantly, observational evidence of a lagged correlation between Arctic ozone anomalies in March and surface temperature in April was documented in Ivy et al. (2017), lending support to the hypothesis of an influence of Arctic ozone on NH surface climate Ozone in CCMs can both respond to and affect the circulation: these effects are non-linear, making the detection of an ozone signal in NH surface climate difficult. The separation of cause and effect is also complicated because changes in stratospheric ozone and the polar vortex have similar surface signatures. For example, Sudden Stratospheric Warmings (SSWs) lead to surface patterns resembling the negative NAO phase (Baldwin and Dunkerton, 2001; Thompson et al., 2002; Kidston et al, 2015; Palmeiro et al., 2017): this is also observed after positive ozone anomalies (Ivy et al., 2017). A possible explanation is that ozone modulates polar vortex variability and its effects on surface climate. Stratospheric Final Warmings (SFWs) generally occur in early-spring, when Arctic ozone can exert a feedback on the circulation, due to increasing solar radiation, and the resulting heating from ozone absorption. A proper simulation of SFWs can lead to improved predictive skill of the NAO in April (Hardiman et al., 2011). Accordingly, by influencing the stratospheric zonal mean flow, ozone could play an important role in the dynamical coupling between the stratosphere and the NAO, and provide a potential source of predictability.
An active area of research concerns the impact of a well-resolved stratosphere in forecasting systems (Domeisen et al., 2015; Butler et al., 2016). Similar studies should be performed for the coupling between ozone and the circulation, in order to quantify the predictive skill that can be gained from including interactive ozone chemistry. To this aim, more work is needed to quantify the impact of ozone chemistry on the polar vortex and NH surface climate on sub-seasonal time-scales. This should be done by separating the effects of ozone from the circulation, which is unfeasible in CCMs. New modeling studies allowing coupling and de-coupling of ozone from the circulation are needed. This represents a novel and yet undocumented pathway whereby the stratosphere can influence the occurrence of surface climate anomalies and extremes. Similarly, we need new methods to improve the representation of ozone and its variability in the current generation of models without interactive chemistry: this can be achieved by reconstructing ozone asymmetries based on the dynamical state of the polar vortex.
2.1.2 The impact of ozone recovery on future climate projections in the Northern Hemisphere
Due to decreasing emissions of Ozone Depleting Substances (ODS) after the implementation of the Montreal Protocol in 1989, ozone is already recovering and will reach historical levels (1955–1975) by the latter half of this century (WMO 2007; 2010; 2014). In the SH, ozone recovery trends are expected to drive a weakening and equatorward shift of the westerly jet (Previdi and Polvani, 2014), opposing the effect of greenhouse-gases (GHG). The degree of cancellation depends on the GHG emission scenario, and will determine the sign of SH circulation trends (Eyring et al., 2013).
Future stratospheric ozone evolution is sensitive to both ODS and GHG. The influence of ODS maximizes in the SH polar cap (Solomon et al., 2015), whereas GHG play a larger role in driving ozone recovery in the NH (Eyring et al., 2013). Relative to the 1955-1975 period, the projected recovery trends for the Arctic and Antarctic are comparable in magnitude (Butler et al., 2016). In extreme GHG emission scenarios, ozone concentrations at high latitudes are projected to surpass historical levels, yielding a so-called “super-recovery”. Arctic ozone increases in such scenarios can be as large as 80 DU, ~50% of the SH recovery. As such, Arctic ozone recovery could lead to substantial NH circulation changes, similar to what has been documented for the SH, such as a warmer stratospheric polar vortex, an equatorward shift of the mid-latitude jet, and a negative NAO, with consequences for storminess over Europe. Yet, while the effects of future ozone recovery on SH circulation trends are well documented, those on NH have not been quantified at all.
Climate models without interactive chemistry prescribe an ozone forcing that is independent of the GHG-scenario (Eyring et al., 2013). In so doing, they miss the ozone response to GHG, thereby underestimating the Arctic ozone recovery, with possible consequences for future projections. It has been shown that ozone responses to GHGs can act as a negative feedback, leading to a reduction in the circulation response to carbon dioxide (CO2) in the SH (Chiodo and Polvani, 2016b). Yet, the impact of this feedback in the NH still needs to be assessed. In this context, more work is needed to accurately quantify the impact of ozone recovery on NH surface projections, as well as the feedbacks between ozone and GHG. In addition to the scenario-related uncertainty, ozone recovery is largely model-dependent. A spread of 15-20 DU is seen in both low and high latitudes for the RCP8.5 scenario (Butler et al., 2016), and quadrupling of CO2 can produce an inter-model spread in ozone that is as large as the response itself (Chiodo et al., 2017, sub.). Spread in the recovery rates can affect polar stratospheric temperature and wind, which have been shown to contribute substantially to the inter-model spread in Arctic Sea Level Pressure (SLP) (Manzini et al., 2014). The identification of sources of uncertainty in the circulation response to anthropogenic forcings is one of the key challenges in climate change assessments (Shepherd, 2014). More work is needed to ascertain if ozone recovery, through its effects on stratospheric circulation, represents an undocumented source of uncertainty in future climate projections for the NH.
2.1.3 The impact of ozone chemistry on climate feedbacks
It has been recently suggested that interactive ozone chemistry can lead to a sizable (~20%) reduction in equilibrium climate sensitivity, quantified as the surface warming response to quadrupling of CO2 concentrations (Nowack et al., 2014). A key contribution to the radiative feedback comes from stratospheric water vapor (SWV). In response to increased CO2 levels, ozone in the tropical lower stratosphere (TLS) decreases by up to 50%; inducing a localized cooling of 3-4 K. This cooling dehydrates air parcels entering into the stratosphere, reducing the SWV concentrations (Stuber et al., 2001). A drying of the stratosphere exerts a negative long-wave (LW) radiative forcing on the troposphere/surface system, reducing the projected surface warming. This feedback is seen in all recent studies on interactive chemistry (Muthers et al, 2014; Dietmuller et al., 2014; Nowack et al., 2014; Marsh et al., 2016), but the magnitude varies greatly among models. For example, the reduction in SWV concentrations due to ozone chemistry ranges from 15% in the WACCM model (Marsh et al., 2016) to over 40% in HadGEM3 (Nowack et al., 2014) and SOCOL (Muthers 2017, pers.comm.). The reasons of this spread, as well as their implications for radiative forcing are largely unknown. As a consequence, the influence of interactive ozone on surface temperature in these models is also uncertain, ranging from 7-8% (Dietmüller et al., 2014; Muthers et al., 2014) to nothing at all (Marsh et al., 2016; Chiodo et al., 2017). Understanding the origin of this spread is of crucial importance to determine whether ozone chemistry feedbacks can contribute to inter-model spread in climate sensitivity.
One of the possible sources of inter-model spread is the cold-point temperature (CPT), as it largely controls the dehydration rates and thus the water vapor entering into the stratosphere (Dessler et al., 2013). Further work is needed to isolate the impact of the spread in ozone and SWV on the CPT, and ultimately, their radiative effects on surface climate in the NH. An improved quantification of the ozone chemistry feedback is a crucial task for progress towards understanding the role of ozone chemistry as a source of the inter-model spread in climate sensitivity, reported in the models involved in the IPCC-AR5.
2.1.4 Scientific questions to be addressed
To summarize, the questions to be addressed are:
1) Does ozone influence climate variability in the Northern Hemisphere?
2) What is the impact of ozone recovery on climate change projections?
3) What is the impact of ozone chemistry as a radiative feedback?
2.1.5 Current projects in Switzerland and abroad
Several projects on topics related to ozone and climate variability are underway in the framework of the World Climate Research Program (Geneva, Switzerland), under the umbrella of Stratospheric Processes And their Role in Climate (SPARC). A core project on chemistry-climate is represented by the IGAC/SPARC Chemistry-Climate Model Initiative (CCMI), established to coordinate evaluation and associated activities involving CCMs. SPARC also includes additional programs on Atmospheric Dynamics and Predictability, such as Dynamical Variability (DynVar) and the Stratospheric Network for the Assessment of Predictability (SNAP). Recently, the SNAP initiative also led the creation of the “Subseasonal to Seasonal Prediction project database” (S2S), aiming to improve forecast skill (Vitart et al., 2017). The proposal’s activities on ozone-circulation-climate interactions fit well in these projects, and would help bridging research interests among each of the aforementioned programs. Close synergies with these activities through data sharing and active participation are foreseen, as detailed in 2.3.3.
2.2 Current state of personal research
During my post-doc at Columbia University, I developed a new line of research on the role of interactive ozone chemistry in the climate response to external forcings. Preliminary results motivating the project are outlined below.
2.2.1 Impact of ozone on stratospheric variability
Arctic ozone is both influenced by and responds to stratospheric variability, but the magnitude of this feedback is difficult to assess in CCMs and observations. To address this issue, I recently performed 200-year long simulations with the WACCM model with current ODS and GHG levels, contrasting different configurations of ozone chemistry, one using interactive stratospheric chemistry, and one using the “specified chemistry” (Smith et al., 2014), with a fixed climatological ozone, obtained from the interactive chemistry run. Thus, the only difference between the runs lies in the ability of ozone to vary on inter-annual time-scales, and thus affect stratospheric temperatures.
Consistent with their identical ozone climatology, there are no differences in the long-term average temperature between both runs. However, the inter-annual temperature variability in April and May is larger in the interactive ozone run. Cold extremes coincide with decreased Arctic ozone, and are several degrees colder in the interactive, compared to the fixed ozone runs (Figure 1). A similar effect is also seen for warm events, albeit of smaller magnitude. These results suggest that interactive Arctic ozone, via its year-to-year variability, enhances the frequency and magnitude of stratospheric temperature extremes in late spring, through changes in absorption of SW radiation. By thermal wind relationship, warm (cold) extremes imply a weak (strong) stratospheric polar vortex, which can possibly influence surface climate via downward propagation of westerly flow anomalies (Baldwin et al, 2003). These results support the hypothesis formulated in this proposal regarding an influence of ozone on NH variability. However, the extent to which this feedback can modulate stratosphere-troposphere coupling and surface climate is still unclear. These WACCM integrations could be used to address this issue, as detailed in section 2.3.
2.2.2 The role of ozone in climate change projections
Stratospheric ozone in the NH is projected to change in response to GHG and ODS trends; can this exert a dynamical influence on the troposphere? Useful insights can be obtained from recent studies under idealized GHG forcings, such as 4xCO2 (Chiodo and Polvani, 2016b). When imposing an ozone climatology, the WACCM model employed in this study simulates a poleward shift of the SH westerly jet in response to 4xCO2, in agreement with CMIP5 models (Grise and Polvani, 2014). However, interactive chemistry reduces this response, due to the ozone response to CO2. Increased CO2 causes stratospheric ozone to decrease in the tropics, and increase at high latitudes. These ozone anomalies modify the meridional temperature gradient at the tropopause, leading to an equatorward shift of the tropospheric jet, opposing the effects of CO2. This implies that the bulk of CMIP5 models (without interactive chemistry) may neglect an important dynamical feedback between ozone and the circulation, possibly overestimating the poleward shift in the SH westerly jet.
A preliminary inspection for the NH in these same runs performed by myself reveals that, similar to what occurs in the SH, ozone increases at high latitudes, leading to a warming of the Arctic stratosphere in boreal winter and spring. In spring, the ozone-induced weakening of the polar vortex strongly projects onto surface circulation patterns, resembling a negative Arctic Oscillation (AO; see the SLP anomaly induced by interactive ozone shown on Figure 2). The CO2-induced Arctic ozone increase in these integrations is ~50 DU (Figure 3), similar to the projected recovery in RCP4.5 scenario by 2100 (Eyring et al., 2013). Thus, it is likely that ozone recovery in the NH has a sizable and yet undocumented effect on the circulation response to GHG in the NH, with potential consequences for the long-term trends. It is important to note, however, that these preliminary results are based on 4xCO2 forcing, which is a highly idealized forcing, and does not consider future reductions in ODS. This motivates further work on elucidating the role of ozone recovery in NH climate, in the context of the IPCC climate change scenarios.
2.2.3 The role of ozone as a radiative feedback
In addition to its effects on the circulation (2.2.2), ozone can also play an important role in the radiative forcing of climate, through its feedback with SWV. This is supported by the role of ozone under different external forcings (Chiodo and Polvani, 2016a). In this study, we quantified the climate response to a solar irradiance increase equivalent to the transition from the Maunder Minimum (1600) to present. Most importantly, the UV-induced increase of stratospheric ozone concentrations leads to increased absorption of clear-sky SW radiation, reducingthe incident SW at the surface and thus, the surface temperature response by approx. 30%. Hence, ozone induces a negative feedback to solar forcing.
Similar to what occurs in the case of solar forcing, the feedback induced by ozone in the case of a CO2 forcing is negative, and is likely mediated by longwave (LW) (Nowack et al., 2014). Increased CO2 levels induce cooling in the upper stratosphere, and a strengthening of the Brewer Dobson, the global-stratosphere meridional circulation (Butchart et al., 2006; 2010). As a consequence, CO2 enhances ozone production in the upper stratosphere (Haigh and Pyle, 1982). In the TLS, enhanced upwelling causes ozone to decrease and cooling this atmospheric region, which in turn leads to a negative LW forcing. However, the effect of this feedback on surface temperatures is model dependent (Marsh et al., 2016). To investigate the role of ozone in creating the spread, Chiodo et al. (2017, sub.) compared four different CCMs. The stratospheric ozone response to CO2 in these models is robust, but the magnitude is remarkably different, with even opposite signed changes in total column ozone (Figure 3). Possible causes of inter-model spread in the ozone response are differences in tropical upwelling (Oman et al., 2010), or in gas-phase chemistry (Haigh and Pyle, 1982); although their exact contribution is still unclear. The spread in the tropical ozone response results in a range of radiative forcing estimates. For example, the ozone responses in the low latitudes (30S-30N) lead to a stratosphere-adjusted LW tropopause forcing of -0.15 W/m2 and -0.5 W/m2 for WACCM and SOCOL, respectively. These differences in the LW forcing are considerable and warrant further research on the sources of inter-model spread in ozone. In addition to ozone, SWV changes resulting from interactive stratospheric chemistry are largely uncertain, ranging from 5 ppmv in WACCM to over 10 ppmv in SOCOL. A larger TLS ozone decrease in SOCOL possibly leads to stronger cooling and hence larger stratospheric drying than in WACCM. This points at SWV as a peace-maker in the ozone-CO2 feedback. However, the exact role of ozone and SWV in the CPT and surface climate cannot be tested in simulations with interactive chemistry, as ozone and SWV are coupled. This project will prove answers to these issues, by performing idealized modeling studies, which isolate the influence of ozone and SWV on radiative forcing and tropospheric temperature (section 2.3).
2.3 Detailed research plan
In this project, we will tackle the following items:
1) Stratospheric ozone as source of intra-seasonal climate variability
2) Stratospheric ozone as a driver in long-term projections
3) Stratospheric ozone as a radiative feedback
These items constitute the three main Working Packages (henceforth “WP”) of the proposal, as shown in Figure 4.
Figure 4.Schematic of the three main objectives of the project. Despite its primary role in the stratospheric radiative budget, ozone in the NH is generally seen as passive constituent, which is affected by halogen chemistry and polar vortex variability. The conventional view is represented by the thick blue and violet lines. We propose three new pathways whereby ozone can play an active role in the NH climate; they are represented by the dashed lines. The project will quantify the dynamical influence of stratospheric ozone on troposphere/stratosphere coupling and surface climate on sub-seasonal (WP1, red line) and multi-decadal (WP2, violet) time-scales, as well as the radiative feedback induced via its coupling with stratospheric water vapor (WP3, green).
Conceptually, the three goals and corresponding WPs are closely intertwined. First, WP1 provides the basis for our understanding of the impact of stratospheric ozone on intra-seasonal time-scales, exploring the dynamical mechanisms whereby ozone can affect the circulation in the troposphere. Second, WP2 will test whether the same dynamical processes explored in WP1 can affect projections on longer time-scales. Third, WP3 will quantify the radiative influence exerted by long-term changes in ozone. The methods employed to accomplish the WPs are outlined in section 2.3.1. Each WP will be divided into three activities, which are described in section 2.3.2.
2.3.1 Methodology: models and observations
To accomplish the three main goals of this proposal, we will combine a hierarchy of models ranging from fully coupled Earth system models (a), to simplified models (b), as well as observational data (c) and other available model data (d).
a) Fully coupled models:
The Whole Atmosphere Community Climate Model (WACCM)
WACCM is the tool upon which preliminary results motivating the proposal are based. It is a comprehensive CCM, which accurately simulates stratosphere-troposphere coupling (Marsh et al., 2013). In its standard configuration, it has interactive stratospheric ozone chemistry, and will be herein referred to as WACCM_CHEM. The model can also be run without interactive stratospheric chemistry, as “Specified Chemistry” (Smith et al.,2014), which will be denoted as WACCM_NOCHEM. In this configuration, ozone concentrations are specified, while retaining the same physical parameterizations, so that differences between both versions can be attributed to interactive ozone.
The SOlar Climate Ozone Links model (SOCOL)
The SOCOL model has been developed in Switzerland (ETH and PMOD/WRC). In its standard configuration, SOCOL includes a comprehensive chemistry scheme (henceforth denoted as “SOCOL_CHEM”), solving for gas-phase and heterogeneous reactions, and advection of chemical species (Stenke et al., 2013). SOCOL_CHEM performs well in many respects, such as stratospheric variability, stratosphere-troposphere coupling and tracer distribution (Stenke et al., 2013; Muthers et al., 2014). It can also be run without interactive chemistry (henceforth denoted as SOCOL_NOCHEM), where ozone is prescribed.
b) Models of reduced complexity:
The Parallel Offline Radiative Transfer (PORT)
The Parallel Offline Radiative Transfer (PORT) model is well suited for calculations of the stratosphere-adjusted radiative forcing at the tropopause (Conley et al. 2013). As input, we will use ozone and SWV from WACCM and SOCOL. PORT employs a fixed dynamical heating assumption above the tropopause, thereby allowing stratospheric temperatures to adjust radiatively. PORT agrees with reference calculations made with line-by-line models (Conley et al.,2013), and thus provides a valid tool for the purposes of the project.
The Community Atmosphere Model – Single column mode (SCAM)
To explore the radiative effect of ozone and SWV on the tropospheric temperature profile and the CPT (both are held fixed in PORT) and isolate them from the dynamics, we will employ a Radiative-Convective Equilibrium (RCE) approximation, as it was done in pioneering studies on ozone radiative forcing (e.g. Ramanathan and Dickinson, 1979). This can be done in the “Single Column” configuration of the Community Atmosphere Model (SCAM), which enables an atmospheric column to be run with full physics, prescribing ozone and SWV as input.
c) Observations: reanalysis and satellite composites
To evaluate the interannual variability of ozone from 1979 until present, we will combine reanalysis and satellite data, including (i) the 3-D gridded data for ozone and water vapor from the Modern-Era Retrospective analysis for Research and Applications (MERRA; Rieckner et al., 2011); and (ii) the new global satellite observational ozone product developed at ETH (Ball et al., 2017 sub.), which combines the Stratospheric Water and Ozone Satellite Homogenized (SWOOSH; Davis et al., 2016) and the Global OZone Chemistry And Related trace gas Data records for the Stratosphere (GOZCARDS; Froidevaux et al., 2015), correcting their biases and artifacts.
d) Other available modeling datasets
As a reference case for many analyses of future projections, we will use available model output data from the ensemble of CCMI simulations (Morgenstern et al., 2017). Aside from CCMI, we will also use available simulations from NASA’s GEOSCCM and the Max-Planck-Institute model ESM-MPI, as detailed in 2.3.2.
SOCOL and WACCM are the appropriate tools to investigate the role of ozone-climate interactions in the Earth System, since they are able to quantify the effects of interactive ozone in a coupled system. The use of two independent models with different dynamics and physics will be useful to test the robustness of the results. The combination of two coupled models (a), and simplified codes (b) will allow us to carefully separate the role of ozone as a dynamical and radiative feedback, and explore physical processes. Finally, the different observational datasets (c) provide a unique opportunity to study ozone variability and extremes, allowing for validation of WACCM and SOCOL, and construction of a new ozone forcing dataset for NOCHEM models.
2.3.2 Tasks to be performed within the project
WP 1. Stratospheric ozone as source of surface climate variability
The objective is to understand the role of stratospheric ozone in NH sub-seasonal variability, with focus on Europe and North America. We will study whether interactive ozone can influence stratosphere-troposphere coupling (WP1.1), the predictability of the NAO (WP1.2), and the mechanisms whereby this occurs (through ozone asymmetries, WP1.3).
WP 1.1: Role of stratospheric ozone in stratosphere-troposphere coupling
My preliminary results show a role of interactive ozone in enhancing stratospheric variability during springtime (2.2.1). In this WP, we will explore a follow-up question: does enhanced variability in the stratosphere affect the coupling with the troposphere? We will address this question by quantifying the effects of interactive ozone on polar vortex variability, represented by the frequency of SSWs and strong vortex events. In particular, we will assess the role of ozone in the final warmings, and timing of the vortex breakup in spring. First, we will compare the WACCM and SOCOL ozone variability with satellite data. Then, we will quantify the impact of interactive ozone on Northern Annular Mode (Wallace, 2000), including its magnitude and persistence, which are useful metrics for the characterization of stratosphere-troposphere coupling (e.g. Baldwin et al., 2003). Finally, the impact of interactive ozone on the NAO and its extremes will be explored. In practice, we will contrast two pairs of ocean-coupled 200-year long control experiments (named “CTRL”): one with CHEM, and one with NOCHEM in WACCM and SOCOL, using present-day ODS and GHG forcings. Using long runs and constant forcings will enable us to obtain robust statistics. Collaborations with Dr W. Ball are foreseen for obtaining and analyzing observational data for ozone. More specifically, we will validate the ozone variability in the CHEM version of both models against observational data recently developed by him (Ball et al., 2017 sub.). In addition, collaborations with D. Domeisen (ETH) will be useful to compare the data generated in WACCM and SOCOL with MPI-ESM simulations, which were used by her in predictability studies.
WP 1.2: Stratospheric ozone and the predictability of the NAO
This activity aims to assess the role of interactive ozone on the predictability of the NAO, based on the effects on stratosphere-troposphere coupling explored in WP1.1. To achieve this, we will employ a probabilistic approach, quantifying the impact of ozone on the NAO and on the predictive skill for the NH. To quantify the predictive skill introduced by ozone, we will hindcast experiments using CHEM and NOCHEM for January-June of the observed years with extreme) March Arctic ozone anomalies, documented in Ivy et al. (2017). Observed Sea Surface Temperatures (SSTs) will be used as boundary conditions, as they are precursors of polar vortex perturbations (Polvani and Waugh, 2004). A set of 20 initial conditions will be produced using the “large-ensemble” approach (Kay et al., 2015): employing 20 members will allow us to optimize the ensemble skill score in the NAO (Scaife et al., 2014). Ensembles that are 6 months long ensures capturing the full evolution, from the onset of large ozone anomalies in March to the tropospheric responses in April-May. The CHEM experiments will, by construction, employ interactive ozone, while NOCHEM will be forced with fixed (climatological) ozone. The prediction skill will be calculated based on the correlation between modeled and observed fields, as it was done in Butler et al. (2016). Active collaboration with Prof. D. Domeisen (ETH) is foreseen for this activity. More specifically, she will provide us with data from the MPI-ESM model, employed in previous predictability studies (Domeisen et al., 2015), for a multi-model comparison of the skill correlation for winter and spring NAO. In addition, we will collaborate with Prof. S. Solomon on identifying ozone extremes in MERRA data, based on the analysis of Ivy et al 2017 (in which she was a co-author).
WP 1.3: Role of ozone asymmetries and dynamically consistent ozone forcing
Previous AGCM studies on the tropospheric response to observed ozone anomalies all imposed zonal-mean profiles, which are inconsistent with observations (e.g. Karpechko et al., 2014). For example, the Arctic ozone depletion event in 2011 was longitudinally asymmetric. Moreover, the ozone forcing data below 100 hPa in MERRA is inaccurate: in the Arctic, this is well in the stratosphere, and could thus affect the results in other studies using AGCMs. This WP activity aims to investigate the impact of longitudinal ozone asymmetries on the circulation, using a more accurate observational data to construct ozone forcings for NOCHEM models, and a new method to derive a dynamically consistent ozone.
First, we will use the ozone observational datasets of Ball et al. (2017, sub.) to construct a zonal-mean (2-D) and 3-D ozone forcing for (14) EXTR years in Ivy et al. (2017). Then, we will simulate the January-June period of each year, by using observed SSTs and comparing two ensembles of 20 members each; one imposing 3-D, and one using 2-D ozone anomalies from the observational product developed at ETH.
One could question whether an interactive ozone chemistry, with all its high computational costs, is actually really needed to capture the observed ozone asymmetries. We propose a groundbreaking new method to calculate ozone asymmetries in NOCHEM models, based on a PV-mapping technique (Randall et al., 2005). By scaling ozone with PV at each model time-step, a standard 2-D climatology (which is commonly imposed in NOCHEM models) can be made “dynamically” consistent with the model dynamics. This can be important for situations in which the polar vortex is strongly displaced with respect to the geographical pole, such as winter/spring of 2011. This method provides a way to simulate the coupling between ozone and the circulation, without the need to include a comprehensive chemistry, which results in a large reduction in the computational resources needed. This technique will be tested in two sets of 30-years long experiments with a specified SSTs climatology; one with PV scaled ozone (PV), and one with simple 2-D ozone.
The activity of WP1.3 is well suited to be carried out by the PhD student as part of a dissertation. The analysis and construction of a 3-D ozone forcing will be done using the observational dataset developed at ETH (Dr. W.Ball). For the implementation and testing phase in the SOCOL model, collaborations with Dr. E. Rozanov, an ETH contributor to the development of SOCOL, are foreseen. Moreover, data from NASA’s GEOSCCM, provided by Dr. D. Waugh, will be used to compare the longitudinal ozone asymmetries calculated in WACCM and SOCOL.
WP 2. Stratospheric ozone as a driver in long-term projections in the NH circulation
The main objective of this WP is to understand the importance of ozone and its feedbacks on the circulation response to climate change, focusing on dynamical aspects that are key for the surface climate of the North Atlantic, such as the NAO. The following questions will be addressed: Does future ozone recovery affect surface projections in the NH? Does ozone contribute to uncertainty in regional-scale climate change? First, we will explore the impact of the ozone recovery on future circulation trends, by assessing the effects of the recovery due to Montreal Protocol on future circulation trends, as well as the role of GHG in the ozone response (WP2.1). Finally, we will explore the impact of inter-model spread in ozone recovery on surface projections (WP2.2).
WP 2.1: The impact of ozone recovery on future projections
Future ozone recovery will occur largely as a consequence of successful regulation of ODS emissions in accordance with the Montreal Protocol, as well as due to the influence of GHG. This activity will quantify the effects of the Montreal Protocol, and GHG on ozone trends, and the impact of these ozone trends on NH surface climate.
First, we will compare climate change scenarios for the 21st century performed with and without future reductions in ODS. To do so, we will analyze the existing multi-model ensemble performed within the CCMI project (Morgenstern et al., 2017). As a reference, we will use the “REF-C2” ensemble, forced with the A1 scenario for ODS and the RCP 6.0 for GHGs. This ensemble will be compared, for each of the participating 20 CCMI models, with the “SEN-C2-fODS2000”. This forcing scenario is identical to “REF-C2”, except for ODS surface-mixing ratios, which are fixed at year 2000. As both ensemble were run with interactive chemistry, their comparison will quantify the impact of ODS-induced ozone recovery on the projected NH circulation trends. Moreover, the availability of an extensive number of models (20) will allow us to detect robust responses, and assess inter-model uncertainty in the ozone recovery.
Second, we will isolate the effect of GHG on ozone, and their influence on the circulation by contrasting future projections with interactive chemistry against runs imposing the IGAC/SPARC data-set (Cionni et al., 2011). This data-set has only been produced for the A1B scenario (Eyring et al., 2013), and will thus be inconsistent with high GHG scenarios, such as RCP8.5. Thus, the difference between runs with interactive chemistry and these integrations will quantify the ozone response to GHG, and its role in NH projections. We will use interactive chemistry runs for the SEN-C2-RCP8.5 scenario done with SOCOL_CHEM and WACCM_CHEM, which are available in CCMI.. These will be compared with two ensembles of 3-members, using RCP8.5 forcing in SOCOL_NOCHEM and WACCM_NOCHEM, performed imposing the IGAC/SPARC data-set. Collaborations with Dr. E.Rozanov are sought for this activity. More specifically, he will provide us with data from an older version of SOCOL (v1), which will be useful for comparisons with SOCOL and WACCM.
WP 2.2: Role of ozone as source of uncertainty in future projections
Ozone recovery rates in simulations with interactive chemistry are model-dependent. In this WP, we will quantify the current degree of uncertainty and its implications in future projections. We will explore this question, by using interactive chemistry runs from the CCMI database for the RCP8.5 scenario. First, the analysis performed for the 20 models included in CCMI in the WP 2.1 will provide us with a range of ozone recovery projections for the 21st century. Then, the largest and smallest ozone recovery rates will be used to drive a set a 3 members ensembles using SOCOL_NOCHEM and WACCM_NOCHEM. The comparison of the future projections obtained for the lower (denoted as SMALL) and upper-(LARGE) ends of ozone recovery will allow us to estimate the degree of uncertainty in the projections that can be ascribed to that in ozone recovery. Collaborations with Prof. D.Waugh are foreseen for this WP activity. More specifically, he will provide simulated data of ozone recovery using different GHG forcings in the NASA’s GEOSCCM model (Waugh et al., 2009b), which will be compared with SOCOL and WACCM. This will be useful to further quantify model uncertainties in recovery rates.
Table 2:As Table 1, for WP2
WP 3. Stratospheric ozone as a radiative feedback
Once we establish the dynamical impact of ozone (WP1 and WP2), we address the radiative effects of the ozone response to a large GHG forcing, such as 4xCO2, including the causes of inter-model differences in the ozone response (WP3.1), the coupling with stratospheric water vapor (WP3.2), and the radiative effects of stratospheric ozone and water vapor on tropospheric temperatures (WP3.3).
WP 3.1: Understanding sources of uncertainty in ozone
This activity aims to quantify the contribution of dynamical (i.e. tropical upwelling) and gas-phase chemistry to the inter-model spread in the ozone response to GHG (see Fig.3). This will be done by combining “Free Running” (FR) and “Specified Dynamics” (SD) runs (e.g. Kunz et al., 2011) of SOCOL and WACCM, forced with RCP8.5. The FR runs will identify the full (chemical and dynamical) response, and are already available in the CCMI database. The SD integrations will consist in 100-year long RCP8.5 runs with the meteorological fields relaxed towards an identical wind profile, obtained by averaging over the dynamical fields in the FR runs of both models. Hence, SD will only account for the role of chemistry in producing differences in the ozone response between SOCOL and WACCM.
WP 3.2: On the role of water vapor in the radiative feedback
The role of stratospheric water vapor in the chemistry feedback will be explored in this activity, by assessing the radiative forcing due to the SWV and ozone changes, as well as its effects on tropospheric temperature. To do so, we will first provide an observational estimate of the ozone-SWV coupling by studying their co-variability in the TLS in the observational data developed at ETH. This will also be used as a benchmark to validate WACCM and SOCOL, and understand the differences in the forced response. Collaborations with Dr. W. Ball is foreseen for this activity, as his satellite data for ozone, in combination with MERRA-2, will be exploited for the analysis of inter-annual variability in ozone and SWV in the TLS region.
To quantify the SWV changes that are induced by interactive ozone, we will compute the difference in SWV between 4xCO2 experiments with CHEM and NOCHEM. These changes in SWV, in combination with ozone, will be imposed in PORT to derive the adjusted radiative forcing for both models (PORT_WACCM, PORT_SOCOL), to provide a quantitative estimate of the radiative feedback. Second, we will explore the response of tropospheric temperatures to the radiative perturbation induced by the SWV and ozone perturbations. This information cannot be extracted in models or observations, as the ozone, stratospheric upwelling and SWV are tightly coupled, hampering identification of a clear causality. To overcome these limitations, we propose using the RCE approximation in SCAM. In practice, this will be done by performing 4 months long SCAM experiments, forced with ozone and SWV changes induced by interactive chemistry (i.e. the CHEM vs NOCHEM difference) in SOCOL and WACCM. The effects of ozone and water vapor on the CPT will be evaluated, by simulating the response of tropospheric temperature in tropical atmospheric conditions. Then, these calculations will be repeated for atmospheric columns in the NH, to assess the extent to which the ozone radiative feedback can (radiatively) affect NH surface climate.
Table 3:As Table 1, for WP3
2.3.3 High-Performance-Computing (HPC) resources and data management plan
The simulations designed for the project are summarized on the Table within each of the WPs. In total, we will run 1690 model years using WACCM, and 2090 years using SOCOL, in a total of 9 and 11 simulations for the two models, respectively. Based on the computational cost of WACCM (2 degrees resolution) on NCAR and other high-performance-computing (HPC) facilities across Europe, we estimate that a total of ~170 days of computation will be required for all WACCM integrations. As for SOCOL, using a T42 horizontal resolution as in Muthers et al., (2014) would require ~190 days of computation. Both estimates can be reduced by 50% when running ensembles in parallel, thus reducing the total computation time to ~170 days, which is ~15% of the overall project duration. The construction of the boundary conditions, including the ozone forcing for NOCHEM, SSTs and initial conditions will take 10 months, corresponding to 1 month for each pair of SOCOL and WACCM simulations (note that the same forcing files will be imposed in both models), which equates to a total of 15-20% of the total project lifetime time. We will also exploit existing datasets, such as CCMI, and other runs that have been previously done by myself. The remaining time in the project (65-70%) will be devoted to analysis and publication of the results. Thus, a timely execution of all the proposed tasks is feasible.
All the experiments will be performed at the Swiss National Supercomputing Center (“Centro Svizzero di Calcolo Scientifico” or CSCS – see http://www.cscs.ch/). The “Piz-Daint” cluster located at the CSCS is one of the most powerful HPC facilities in the world, and is similar to the National Center for Atmospheric Research facility located in Boulder (USA). It grants access to institutions across Switzerland, including ETH, and provides HPC allocations for projects on Earth Sciences. Most importantly, both WACCM and SOCOL have already been previously ported to this cluster and are thus operative, which allows us to perform the experiments without any delays. The raw output from all the model simulations is estimated to occupy 61 Tb, which will be reduced to 10 Tb after post-processing. This data will be saved in a storage system, to be purchased with the project’s funds at ETH. The most relevant output data, including projections for the 21st century, will be made available to researchers across Swizerland and Europe by using FTP servers offered at IAC/ETH, in keeping with the data compliance policy at ETH. By doing so, the data can be used by the scientific community, increasing the visibility and impact of this project. In particular, the simulations data from the project’s activities will be of interest to the community involved in the SPARC initiative of the World Climate Research Program (WCRP). Data from the project on the NAO predictability could be useful for the S2S and SNAP initiatives.
2.3.4 Risks involved in the project, and alternative strategies
The risks associated with the proposal’s activities are considered to be very low. This is due to the availability of the satellite data for ozone (developed at ETH), which allows to avoid logistical and technical challenges of collecting observational data. Moreover, the modeling activities will be carried out using scientific codes (i.e., WACCM and SOCOL) that have been scientifically validated (Marsh et al., 2013; Stenke et al., 2013). These codes are already operative of the SCSC cluster, reducing the risk of delays in the production of the model simulations. The models of reduced complexity (i.e. SCAM and PORT) are part of the a code distribution, which includes WACCM. Thus, having an operative WACCM version on SCSC ensures the possibility of using these codes. Finally, CCMI data are located on the British Atmospheric Data Center located in the UK, to which I have already gained access: this ensures immediate availability of these data.
2.3.5 Role of each member of the team, and collaborations
The team will be composed by myself as Principal Investigator (PI), one PhD student, one group leader (Dr. T.Peter, ETH) and five collaborators: Dr. E.Rozanov (ETH), Prof. D.Domeisen (ETH), Dr. W.Ball (ETH), Prof. S.Solomon (MIT) and Prof. D.Waugh (J.Hopkins University). As PI of the project, I will ensure a timely execution of each of the project deliverables. The PhD project (WP1.3) represents a major advancement in our understanding of the interactions between the polar vortex and ozone, via ozone asymmetries. The dissertation’s goals are sufficiently distinct to be carried out independently from the main proposal. Moreover, the modeling tasks do not rely on input data from other WPs. Thus, the feasibility of the project, along with my supervision will ensure a successful and timely completion of the PhD thesis.
The progression and execution of the whole project will be overseen by Dr. T. Peter (ETH) as a group leader. He is the former head of the Department of Environmental Systems Science at ETH Zurich, and has been a member of the scientific steering group of the World Climate Research Programme SPARC since 1997. His research interests are aerosol-chemistry and chemistry-climate interactions, and he has published influential papers in these areas. Close synergies with the other research teams at ETH are foreseen. For example, Prof. D. Domeisen (ETH) and her ETH group on atmospheric predictability will collaborate on the activities focused on ozone and the NAO. She will provide useful insights due to her experience in stratosphere-troposphere coupling and predictability. Close collaboration with Dr. W.Ball (ETH) will be useful for the analysis of the observational ozone data-set developed by himself. Finally, collaborations with Dr. E. Rozanov (ETH) will be useful for SOCOL modeling tasks. Heplayed an active role in several ozone assessments and inter-comparison projects. Dr. Rozanov has plenty of expertise testing parameterizations in SOCOL, and will collaborate on the implementation of the PV scaling technique.
Part of the project will be carried out with external collaborators, such as Prof. S. Solomon (MIT) and D. Waugh (J Hopkins). Prof. S.Solomon was the lead author of IPCC-4AR and has co-authored recent papers motivating this project, such as Calvo et al.,(2015) and Ivy et al. (2017). Prof. D.Waugh.has been one of the key contributors to CCM intercomparisons, such as CCMVal (2010). Collaborations with these scientists will be a valuable asset in the project, due to their willingness of sharing data from model simulations, and their insights in ozone-climate interactions.
2.4 Schedule and milestones
The schedule for each of the WP activities over the course of the 48 months (16 Quarters) is shown in Table 2. WP1 will take a total of 6 Quarters (Q), and is projected to be completed by the third Quarter (Q2) of Year 2. The main responsibility of the PhD student will be WP1.3. His/her work will be initially aligned with the other WP1 activities, due to the topical overlap (ozone and predictability). WP1 will take 1.5 years for completion, in large part due to the computational resources needed for the project on predictability of the NAO (WP1.2). WP2 activity will start in Q3 of Year 2, and will take one full year to be completed (because of the availability of CCMI data). The time of completion of WP2 is aligned with the sixth IPCC report. Note that the RCP scenarios imposed in the WP2 experiments are compatible with the scenarios imposed in CMIP6 models (Eyring et al., 2016): this ensures that results of WP2 can be extrapolated to models participating in the IPCC-AR6. Results on the projections for the NH will become available while the IPCC report is still in the drafting phase, and are thus going to be available to contribute to the final report (which is foreseen for 2021). WP3 will start in Q4 of year 3, and finish on Q4 of Year 4. Like WP2, it is expected to be completed within one single year, due to data availability and low complexity of the idealized calculations needed.
2.5 Relevance and impact
The project will provide new insights into the role of stratospheric ozone in NH climate, via its coupling with atmospheric dynamics and radiation. The current generation of atmospheric models neglect these coupling processes, so the implications for Europe, and more generally NH climate, are unclear. The project will assess the benefit from an accurate representation of the ozone-climate interactions for seasonal predictions of rainfall and wind patterns in the NH. We also propose novel ways to improve the representation of this coupling, based on simple scaling of an ozone climatology, without the need for a comprehensive atmospheric chemistry scheme. By elucidating the role of ozone in NH surface climate, we also propose mechanisms whereby the stratosphere can affect surface climate, which is one of the key purposes of the WCRP/SPARC activity. Finally, by assessing the future ozone recovery and its dynamical and radiative effects, this project will contribute towards improved understanding of sources of uncertainty in long-term climate projections. . As dynamical aspects of climate change exert a strong control on regional climate, the project is also a step forward in narrowing the current uncertainties of regional-scale projections in the NH, with potential implications for planning of extreme weather patterns.
The results from this project will be of relevance to the modeling communities involved in SPARC. For this reason, publication of manuscripts corresponding to each of the WP activities is foreseen. They will be submitted to peer-reviewed journals in Earth Sciences, such as Proceedings of the National Academy of Sciences, Atmospheric Chemistry and Physics, Journal of Climate, and Geophysical Research Letters. This project addresses cutting-edge questions in the frontier of Earth System Sciences, and hence high-impact results can be expected. Thus, their publication in top journals, such as Science and Nature, is foreseen. A final review paper – to be submitted to Nature Climate Change – will summarize the results of the project, and other relevant recent studies on the role of ozone in NH climate. Active participation in the production of the sixth WMO ozone assessment, as well as IPCC-AR6, which are foreseen for the years 2020 and 2021, respectively. Each of the WP activities will be presented at major conferences in Europe and North America, including the annual meetings of the European Geophysical Union, and American Geophysical Union, as well as topical workshops, such as CCMI, SPARC-Dynvar, and SPARC general assembly. To increase visibility within the Swiss science network, we also seek participation in the Swiss Global Change day meeting. Finally, we will create a website exclusively dedicated to outreach for the results obtained within the project (e.g. www.uzonh-project.com)
Climate change uncertainty poses a major conceptual challenge to policy-makers in the decision-making process, as they can no longer have confidence in single projections of the future. Part of the uncertainty stems from processes that are poorly constrained in models, such as ozone-climate interactions. Results from this project will thus contribute towards improved seasonal climate predictions and future climate projections at the regional level. We will provide new constraints on the uncertainty in seasonal predictions of Arctic ozone. Using information from the hindcast experiments and PV scaling technique, we can also build a statistical-dynamical model for seasonal prediction of ozone over Europe and the Arctic: this can offer a direct application of the science behind this project for the general public. The identification of sources of uncertainty in ozone recovery can be useful to monitor Arctic and tropical ozone, and to guide future environmental policies aimed at safe-guarding the ozone layer.
 For MINIMUM years, these are 1990,1993,1995,1996,1997,2000,2011
 For MAXIMUM years, these are 1979,1980,1984,1985,1989,1999,2010