The Spillover of Tropospheric Ozone Increases Has Hidden the Extent of Stratospheric Ozone Depletion by Halogens

Stratospheric ozone depletion from halocarbons is partly countered by pollution‐driven increases in tropospheric ozone, with transport connecting the two. While recognizing this connection, the ozone assessment's evaluation of observations and processes have often split the chapters at the tropopause boundary. Using a chemistry‐transport model we find that air‐pollution ozone enhancements in the troposphere spill over into the stratosphere at significant rates, that is, 13%–34% of the excess tropospheric burden appears in the lowermost extra‐tropical stratosphere. As we track the anticipated recovery of the observed ozone depletion, we should recognize that two tenths of that recovery may come from the transport of increasing tropospheric ozone into the stratosphere.


Prologue
The Climatic Impact Assessment Program (CIAP, 1974) in the early 1970s completed an integrated U.S. assessment of the ozone (O 3 ) layer as being under future threat from a proposed fleet of high-flying Concorde-like supersonic transport aircraft (SSTs).The threat of depletion of the ozone layer was based on the catalytic ability of oxides of nitrogen (NO x ) emitted from the engines to destroy stratospheric ozone (Johnston, 1971;McElroy et al., 1974).The hypothetical threat of SSTs was quickly replaced by real ozone depletion caused by chlorofluorocarbons (Molina & Rowland, 1974;Stolarski & Cicerone, 1974;Wofsy & McElroy, 1974).Ozone assessments became international and remained focused on halogen-driven ozone depletion (e.g., WMO, 1981WMO, , 1985WMO, , 2022)).In 1988 the NASA High Speed Research Program revived the dream of a commercial SST fleet and initiated the Atmospheric Effects of Stratospheric Aircraft (AESA) to re-assess SST impacts based on the maturity of stratospheric science.In the first AESA report, a figure appeared (Schmeltekopf, 1992; reproduced here as Figure 1) that predicted increases in lowermost stratospheric O 3 from the aircraft NO x due to "smog reactions" that include methane (CH 4 ).This result was surprising at first to those focusing on O 3 depletion, and it may be the first assessment to show increased lower stratospheric O 3 production associated with human pollution.

Introduction
The community has continued to study pollution of the lower stratosphere whereby ozone precursors or ozonedepleting substances are transported from the troposphere into the stratosphere, thereby altering the chemical production and loss of O 3 .Some examples include the convective overshooting at the tropical tropopause (Oman et al., 2016;Wales et al., 2018), the South Asian Monsoon (Dubé et al., 2022;Fu et al., 2006;Lelieveld et al., 2018;Randel et al., 2010), and extensive wildfires (Hirsch & Koren, 2021;Santee et al., 2022;Solomon et al., 2022), all of which can lead to increases or decreases of stratospheric O 3 .What is less studied is the direct impact of enhanced tropospheric O 3 on the stratosphere.This spillover effect, whereby tropospheric O 3 pollutes the stratosphere directly rather than through precursors, is not readily diagnosed because separating the chemistry and transport of O 3 across the tropopause region is impossible with just O 3 observations and remains a difficult task in most models.
In climate forcing, increased tropospheric O 3 dominates stratospheric depletion.Gauss et al. (2006) in their multimodel assessment recognized that "tropospheric ozone increase since preindustrial times has moderated lower stratospheric ozone depletion," but they did not have the scenarios or diagnostics to evaluate it.Other model studies (Reader et al., 2013;Shindell et al., 2013) derived the amount of tropospheric "pollution" O 3 in the stratosphere but their pollution scenarios included also the increase in CH 4 .Given the very large impact CH 4 has on stratospheric O 3 (see Figures 1 and 14 of Zeng et al., 2022) this obfuscates their results.In the AerChemMIP study (Collins et al., 2016) a scenario singling out the O 3 -precursor emissions without changing CH 4 provided an opportunity for Zeng et al. (2022) to identify direct O 3 transport, but, even with this scenario, O 3 precursors are emitted directly into the lower stratosphere (e.g., 18%-44% of aviation NO x , Gettelman & Baughcum, 1999) or transported there.
In this paper, a set of carefully designed chemistry-transport model (CTM) simulations is used to assess the impact of direct O 3 influx from the troposphere into the stratosphere.These simulations follow from the use of direct O 3 emissions to derive the lifetime of tropospheric O 3 (Prather & Zhu, 2024).Given a 20th century tropospheric column O 3 (trpcolO 3 ) increase of about 9 Dobson Units (DU) (Gauss et al., 2006;Griffiths et al., 2021;Young et al., 2013), this work identifies a 2-3 DU increase in stratospheric column O 3 (strcolO 3 ) without including any stratospheric increase in O 3 precursors such as NO x and CH 4 .Failure to recognize trends of this level can notably alter the projected year-of-recovery for stratospheric ozone depletion in response to policy options (see Figures ES-1 and 3-24 of WMO, 2022) and also the projected net integrated ozone depletion (Pyle et al., 2022).While this trpcolO 3 spillover is inherently included in current models, it is likely misdiagnosed in both models and observations as reduced levels of strcolO 3 depletion and earlier recovery.Modeling is covered in Section 2; analysis of past and future ozone changes are in Section 3; and a summary discussion is in Section 4.

The UCI Chemistry Modules
The UC Irvine CTM, like many early CTMs, developed separate chemistry modules for stratosphere (Linoz: Hsu & Prather, 2009, 2010;McLinden et al., 2000) and troposphere (ASAD: Wild & Prather, 2000;Wild et al., 2003).The current stratospheric chemistry module (Linoz version 3) calculates the chemical production and loss of O 3 , NO y (the family of nitrogen oxides except N 2 O), N 2 O, and CH 4 as a first-order Taylor expansions about 3D climatological means (monthly, latitude by pressure) of these species.The expansion also allows for independent variations in temperature and overhead column O 3 .Local stratospheric H 2 O is inferred from CH 4 decay through conservation of H and a tropopause boundary condition of 3.65 ppm (micromol mol 1 ) H 2 O.For the present-day (circa year 2000) simulations shown here the CTM uses lower tropospheric boundary conditions (LBCs) of 30 ppb (nmol mol 1 ) for O 3 , 0.02 ppb for NO y , 316 ppb for N 2 O, and 1,800 ppb for CH 4 .Rapid O 3 loss driven by heterogeneous chemistry on polar stratospheric clouds uses the Cariolle et al. (1990) parameterization, which produces an excellent simulation of the mean and interannual variability of the Antarctic ozone hole (e.g., Figure 8 of Ruiz & Prather, 2022).
To calculate the chemical rates and the Taylor expansions, a stratospheric chemistry box model (Pratmo: e.g., Dubé et al., 2022) is linearized about the climatological mean values of the independent species scaled to the LBCs for N 2 O and CH 4 , plus stratospheric climatologies for the nitrogen (NO y ), chlorine (Cl y ), and bromine (Br y ) families of species.The Pratmo calculation forces all the individual species in each of the three families to be in 24-hr photo-stationary state.During a CTM simulation, stratospheric CH 4 , N 2 O, and NO y may evolve in response to changing LBCs, but the Cl y and Br y remain fixed per the initial calculation of the Linoz tables, that is, 3.24 ppb peak Cl y and 16.8 ppt peak Br y for year 2000.The tropospheric chemistry module ASAD includes about 30 species with a reduced set of hydrocarbon reactions, but no halogen chemistry.Many ASAD species are set to decay slowly in the stratosphere; but the NO y species are reconciled to the Linoz production and loss of NO y by adjusting NO, NO 2 , and HNO 3 to ensure the observed relative influx of these species into the troposphere generated from the N 2 O loss.
The choice of which chemistry to apply in each air parcel is triggered by the e90 surface tracer (Prather et al., 2011) using a threshold of 90 ppb.Thus, a full-chemistry CTM simulation includes both a standard O 3 tracer that experiences Linoz chemistry when in the stratosphere and ASAD chemistry when in the troposphere.It also includes a tagged tracer O 3 S that experiences only Linoz chemistry plus the lower boundary condition (LBC), which is invoked by the Linoz module.A recent example of the full-chemistry simulations are shown in Prather and Zhu (2024, see their Table 1).Present-day tropospheric O 3 budgets are similar to recent model intercomparison projects (Griffiths et al., 2021;Young et al., 2018).

Identifying the Spillover Effect
The work here was triggered by the belated discovery that the O 3 and O 3 S tracers in the UCI CTM differed, not just in the troposphere as expected (where O 3 has ASAD chemistry and O 3 S has only a LBC reset), but also in the lower stratosphere (where both are subject only to the same Linoz chemistry).
To study this effect we set up some diagnostic experiments.DU) and the seasonal cycle (minimum in DJF) found by Ziemke and Chandra (2012).The annual mean strcolO 3 for the full chemistry run is 280 DU.As expected, the O 3 S profiles with 30 ppb LBC, either alone (b) or with tropospheric chemistry (c), are essentially identical and not shown.
The profiles in Figure 2b are for stratospheric air only and show the percent difference of O 3 (full)-O 3 S (30 ppb) (blue) and O 3 S (10 ppb)-O 3 S (30 ppb) (red).The annual mean strcolO 3 difference in these two pairs is +3.8 DU and 1.6 DU, respectively.For both cases the relative differences in stratospheric O 3 are largest (5%-20%) below 16 km and negligible above 22 km.Absolute differences in the lowermost stratosphere below 19 km for these two cases are nearly constant at about +30 ppb and 15 ppb, respectively (not shown, but see Figure 3).Both trace species O 3 and O 3 S experience only Linoz chemical tendencies in the lower stratosphere.These tendencies drive the O 3 species toward the O 3 S profile, the preferred chemical state in the stratosphere, and thus damp any injection of tropospheric O 3 .Any flux of pollution precursors (e.g., ASAD's NO, CO, H 2 O 2 ) into the stratosphere will be ignored by Linoz.Thus we are certain that the stratospheric enhancements observed here are due entirely to the transport of tropospheric O 3 into the stratosphere.
Figure 3 provides a latitude by pressure-altitude color map of the absolute increase in stratospheric O 3 attributable to excess tropospheric O 3 , calculated as O 3 minus O 3 S with a 30 ppb LBC.We calculate that 99% of the increased mass occurs below 21 km.As in Figure 2, stratospheric values from troughs and folds appear well below the mean tropopause.Based on this map, the enhanced tropospheric O 3 finds its way into the stratosphere through isentropic flow across the jet streams into the extratropical lowermost stratosphere and also partly through ascent across the tropical tropopause.The stratospheric increases are asymmetric favoring the northern hemisphere (NH) because the tropospheric O 3 changes (O 3 minus O 3 S) in the full chemistry model are driven by predominantly NH emissions of ozone precursors.For our full chemistry run, the NH and southern hemisphere (SH) trpcolO 3 are 35.2 and 28.7 DU, respectively.For O 3 S (30 ppb LBC), these values are 21.2 and 20.6 DU.Thus, the increase in trpcolO 3 is about twice as large in the NH as evidenced in Figure 3.The small NH-SH difference in tropospheric O 3 S reflects the nearly balanced STE O 3 flux in the UCI CTM (Ruiz & Prather, 2022), even though the SH flux has been reduced by about 14% due to the Antarctic ozone hole.
Given our ability to measure trpcolO 3 from satellites (e.g., Ziemke and Chandra (2012)) and the ready diagnostic of trpcolO 3 from the model intercomparison projects (MIPs: Young et al., 2013;Griffiths et al., 2021;Zeng et al., 2022), it would be great if can derive a spillover factor that relates tropospheric column changes to the induced stratospheric column changes.For this study, we have available two independent cases: (a) the two Linoz runs with different LBCs, and (b) the single modern full chemistry run that contrasts O 3 with O 3 S.For these cases, the change in trpcolO 3 is an extremely poor predictor of strcolO 3 : (a) spillover factor of 0.13 DU per DU (1.6/ 12.4), and (b) 0.34 DU per DU (3.8/11.0).Searching for a better predictor and considering Figure 3, we identified the O 3 abundance (mole fraction) in the upper tropical troposphere, specifically the two model levels from 88 to 123 hPa.The predictor now is consistent for the two cases, 0.083 DU per ppb, but it is difficult, if not impossible, to observe the troposphere-only mole fraction of O 3 between 88 and 123 hPa except with sondes.Likewise, such a diagnostic for the MIPs would not work.

Analysis of Past and Future Ozone Changes
How can we use our limited set of CTM simulations to estimate the magnitude of the spillover effect from preindustrial (PI) to present day (PD)?The change in trpcolO 3 from 1900 to 2000 based on multi-model assessments give values of about +9 DU (Gauss et al., 2006;Griffiths et al., 2021;Young et al., 2013;Zeng et al., 2022), consistent with observed trends over the past two decades (Gaudel et al., 2018;Tarasick et al., 2019).We do not have the resources for a full PI to PD CTM simulation here and, further, lack the necessary PI meteorological data to drive the transport.We propose that our Linoz 30 ppb simulation is similar to PI because: (a) trpcolO 3 is 11 DU less than our PD simulation with full chemistry; (b) tropospheric O 3 S is remarkably north-south symmetric; and (c) in most models tropospheric pollution-driven changes in O 3 are largest in the upper troposphere (e.g., Figure 8 of Young et al., 2013, converting mass to abundance).Thus, assuming that the PI to PD profile of tropospheric O 3 has changed as in our O 3 S to O 3 simulation, we scale our spillover (3.8 DU) with trpcolO 3 (from 11 to 9 DU) and get 3.1 DU.Thus, our best estimate of the increase in strcolO 3 from PI to PD due simply to the increase in tropospheric O 3 is 3 DU.It is difficult to provide a formal error analysis, but our estimate would be ±1.5 DU, due primarily to abundance uncertainty in the upper tropical troposphere.
How important is 3 DU spillover into strcolO 3 ?The difficulty, but necessity, of separating trends in total column O 3 (totcolO 3 ) into pollution-driven tropospheric and halogen-driven stratospheric is discussed in Box 3-3 of WMO ( 2022), but a reconciliation was not made.A global mean loss of about 8.0 DU in totcolO 3 is observed between 1980 and 2020 with most occurring before 1995 (see Figures 3-6 of WMO, 2022).The return to 1980 O 3 levels is the established measure of "recovery" (Prather & Watson, 1990).The observed trpcolO 3 trend gives an increase of 3.6 DU for these four decades, and thus the true loss in strcolO 3 since 1980 is 11.6 DU.The tropospheric spillover based on the increase in trpcolO 3 means that this inferred halogen-driven loss is further underestimated by 0.5-1.2DU (The upper and lower range is based on trpcolO 3 spillover factors ranging from 0.34 (full chemistry O 3 -minus-O 3 S) to 0.13 (Linoz 30 ppb-minus-10 ppb cases)).
It is useful to examine the MIP-based attribution study of Zeng et al. (2022), which presented PI-PD changes in strcolO 3 and trpcolO 3 .Here we use year 2000 as PD for reading numbers from their figures.The PI-PD total change in trpcolO 3 is about +9.0 DU (consistent with previous studies); it is comprised of +4.0 DU from CH 4 increases and +7.5 DU from increased emissions of other O 3 precursors (e.g., NO x , CO) minus the decreases driven by halocarbons and N 2 O.The total change in strcolO 3 is predominantly negative ( 14.5 DU) but with some large positive changes from CH 4 (+6.5 DU) and the other O 3 precursors (+2.5 DU) that parallel the tropcolO 3 changes.The large direct role of CH 4 in stratospheric chemistry is well known and expected.What is unclear from the Zeng et al. analysis is whether the +2.5 DU from non-CH 4 precursors is a direct O 3 spillover effect from the +7.5 DU in trpcolO 3 or if it results from increased stratospheric chemical production of O 3 from the tropospheric NO x entering the stratosphere or direct stratospheric emissions of aviation NO x as described by Schmeltekopf (1992).
For the future, the multi-model assessments project changes in trpcolO 3 over the 21st century that vary by scenario from 5.6 DU (RCP2.6) to +5.3 DU (RCP8.5) to +7.3 DU (SSP3-7.0)(Archibald et al., 2020).This wide range is due primarily to changes in the O 3 precursor emissions and secondarily to recovery of halogendriven depletion combined with overall climate change (e.g., warmer wetter troposphere, colder stratosphere).The spillover effect on strcolO 3 with the large factor (0.34 DU per DU) then ranges from 1.9 to +2.4 DU; but because these future changes likely preserve the current O 3 profile shape and hence have a smaller factor (0.13), the spillover is likely much smaller, 0.7 to +0.9 DU.For perspective, the totcolO 3 loss from 1900 to 2000 attributed to ozone depleting substances is about 20 DU, and the loss from all emissions is 14 DU (Zeng et al., 2022).In terms of the observed totcolO 3 loss since 1980, we are looking for recovery of about 8 DU.Thus the scenario range of projected changes in trpcolO 3 is comparable in size to the recovery ( 70% to +90%), and the spillover effect onto strcolO 3 will still be an important fraction (∼25%).With fully diagnosed models, we should be able recognize and correct for this, but with observations, the spillover effect is hidden within the recovery and cannot be readily diagnosed.

Discussion
The spillover effect will vary depending on what is driving tropospheric O 3 changes and how that is manifest in the upper tropical troposphere.It is necessary for other models to quantify this effect, and to examine a range of scenarios.In simulating the spillover effect with other models, one should avoid the use of tagged tracers such as labeling tropospheric O 3 and watching its buildup in the stratosphere.Such experiments give unreliable results because tropospheric O 3 molecules entering the stratosphere are not differentiable from stratospheric O 3 molecules.There are chemical feedbacks on O 3 and its lifetime (in parallel with tropospheric CH 4 ), and thus addition of O 3 changes the whole of stratospheric O 3 chemistry, a process that cannot be simply simulated with a tagged tracer (see the problems with O 3 S as a measure of stratospheric influence in Prather & Zhu, 2024).It is important that other model studies are able to differentiate between the transport of O 3 (e.g., 10%-30% spillover) and that of chemical precursors (NO x , CH 4 ), and thus warrant such diagnostics in future ozone assessments.
The reverse of this process is well known, that is, strcolO 3 changes propagate into the troposphere through STE fluxes and alter trpcolO 3 .Recent results allow us to quantify this.The Antarctic ozone hole is shown to reduce the SH STE ozone flux by on average 30 Tg-O 3 yr 1 with interannual variations of ±20 Tg-O 3 yr 1 (Ruiz & Prather, 2022).Using the 24-day perturbation lifetime for tropospheric O 3 from STE sources, Prather and Zhu (2024) calculate that the total STE perturbation is only 8%, and hence the SH mean reduction in trpcolO 3 from the current ozone hole is only 0.4 DU.These relatively large ozone fluxes have much less impact in the troposphere because the photochemistry is much more reactive (shorter lifetime) there than in the lower stratosphere.
Unfortunately, this spillover effect exacerbates the conflicts inherent in environmental mitigation strategy.For example, the RCP2.6 scenario greatly improves air quality by reducing tropospheric O 3 , but this leads to reductions in stratospheric O 3 .Even the integrated O 3 depletion metric envisaged by Pyle et al. (2022) needs to be calculated with a full CTM that includes spillover effects when emission scenarios include O 3 precursors.

•
Our world has seen damaging levels of ozone depletion over the past several decades, and these will continue into this century • Currently, ozone column changes are a combination of halogen-driven stratospheric decreases counteracted by pollution-driven tropospheric increases • Tropospheric increases spillover into the stratosphere, stratospheric depletion can thus be underestimated, and the observed recovery of assumed halogen-driven loss is overestimated Supporting Information: Supporting Information may be found in the online version of this article.

Figure 2 .
Figure 2. (a) Mean profiles of tropospheric O 3 (ppb, nmol mol 1 ) versus pressure altitude (z*, km).Only tropospheric air as defined by e90 is included, and results are area-averaged over 60°S-60°N.Three profiles are shown along with their global mean tropospheric column (trpcolO 3 in Dobson Units = 10 3 cm-amagat): (green) No tropospheric chemistry plus Linoz with 10 ppb LBC; (red) Ditto with 30 ppb lower boundary condition (LBC); and (blue) Full tropospheric chemistry plus Linoz with 30 ppb LBC.Each curve is also labeled with mean tropospheric abundance (ppb) for two model layers from 88 to 123 hPa, which lie in the tropical upper troposphere, see text.(b) Relative change in stratospheric O 3 (%) versus pressure altitude (km).Only stratospheric air as defined by the e90 tracer is included.O 3 is area-averaged over 60°S-60°N.The positive profile (blue) shows difference O 3 (full tropospheric plus stratospheric ozone chemistry) minus O 3 S (Linoz stratospheric chemistry only with 30 ppb LBC).The negative profile (red) contrasts two Linoz-only runs: 10 ppb LBC minus 30 ppb LBC.

Figure 1 .
Figure 1.Ozone formation from the smog reactions based on methane and nitrogen oxides (45°latitude, spring).The primary perturbation shown here is that of added NO x from a fleet of SSTs. Figure 1, p. 166 from Schmeltekopf (1992).

Figure 3 .
Figure 3. Latitude-by-altitude color map of the spillover effect, that is, the stratospheric O 3 excess (ppb) due to injection tropospheric O 3 .Values shown here are for stratospheric air parcels only (based on e90).The color map is calculated as the difference between full chemistry simulation (O 3 ) and the stratospheric tagged tracer O 3 S calculated using only stratospheric chemistry (Linoz) with a lower boundary condition of 30 ppb, see text.The plot shows the annual, zonal mean taken from the first of each month for year 2001 and includes stratospheric air parcels in folds or troughs below the climatological mean tropopause.For perspective, horizontal white lines mark pressure altitudes of 12 and 16 km.