Skip to main content
NCBI home page
ACS AuthorChoice logoLink to ACS AuthorChoice
. 2024 Aug 2;58(32):14306–14317. doi: 10.1021/acs.est.4c01289

Anthropogenic Fingerprint Detectable in Upper Tropospheric Ozone Trends Retrieved from Satellite

Xinyuan Yu †,*, Arlene M Fiore †, Benjamin D Santer ‡,§, Gustavo P Correa ∥, Jean-François Lamarque ⊥, Jerald R Ziemke #,∇, Sebastian D Eastham ○,◆, Qindan Zhu †
PMCID: PMC11325641  PMID: 39092829

Abstract

graphic file with name es4c01289_0005.jpg

Tropospheric ozone (O3) is a strong greenhouse gas, particularly in the upper troposphere (UT). Limited observations point to a continuous increase in UT O3 in recent decades, but the attribution of UT O3 changes is complicated by large internal climate variability. We show that the anthropogenic signal (“fingerprint”) in the patterns of UT O3 increases is distinguishable from the background noise of internal variability. The time-invariant fingerprint of human-caused UT O3 changes is derived from a 16-member initial-condition ensemble performed with a chemistry-climate model (CESM2-WACCM6). The fingerprint is largest between 30°S and 40°N, especially near 30°N. In contrast, the noise pattern in UT O3 is mainly associated with the El Niño–Southern Oscillation (ENSO). The UT O3 fingerprint pattern can be discerned with high confidence within only 13 years of the 2005 start of the OMI/MLS satellite record. Unlike the UT O3 fingerprint, the lower tropospheric (LT) O3 fingerprint varies significantly over time and space in response to large-scale changes in anthropogenic precursor emissions, with the highest signal-to-noise ratios near 40°N in Asia and Europe. Our analysis reveals a significant human effect on Earth’s atmospheric chemistry in the UT and indicates promise for identifying fingerprints of specific sources of ozone precursors.

Keywords: climate change, climate variability, detection and attribution, chemistry-climate modeling, atmospheric composition

Short abstract

A clear signal of human influence on upper tropospheric ozone trends is identifiable with high statistical confidence in a 17-year satellite dataset.

Introduction

Tropospheric ozone (O3) is a known powerful oxidant and an important component of air pollution with adverse effects on human health1,2 and terrestrial vegetation.3 After the anthropogenic greenhouse gases carbon dioxide and methane, tropospheric ozone is estimated to have contributed the next largest radiative forcing since preindustrial times, 0.4 ± 0.2 Wm–2.4 Increases in O3 in the upper troposphere (UT), where the temperature difference with the surface is largest and ozone is more effective at trapping thermal infrared radiation, exert a stronger radiative effect than at lower altitudes,5 particularly in the tropical tropopause region.6−8 Here we explore UT O3 trends derived from model simulations and retrieved from the Aura Ozone Monitoring Instrument/Microwave Limb Sounder (OMI/MLS) satellite instrument and disentangle the role of external forcings from internal climate variability.

Ground stations, ozonesondes, aircraft measurements, and satellites have observed overall increases in O3 in recent decades throughout much of the troposphere.9−14 For example, In-Service Aircraft for a Global Observing System (IAGOS) data indicate increases in tropospheric column O3 of 3.6–18.5% per decade (varying by region) over 1994–2016.13 Terms in the tropospheric O3 budget include in situ photochemical production during the oxidation of volatile organic compounds (VOCs, especially methane) and carbon monoxide (CO) in the presence of nitrogen oxides (NOx), as well as chemical loss, dry deposition, and stratosphere-troposphere exchange (STE). In the UT, O3 is most directly influenced by STE and in situ chemical production from aircraft and lightning NOx. The O3 production efficiency is generally higher aloft than in the polluted continental boundary layer where NOx mostly originates from combustion processes including in vehicle engines and electricity generation.15,16 Relative to O3 in the lower troposphere (LT), the lifetime of O3 is also longer in the UT (weeks-to-months versus days in the continental boundary layer), increasing the importance of transport processes over emissions in shaping the spatial distribution of UT O3.5,10 From 1995 to 2014, model simulations and observations indicate that UT and LT O3 trends differ both in magnitude and sign in certain regions.17,18

Observed tropospheric O3 trends are imprinted by both the “signal” arising from external forcings and the “noise” of natural internal climate variability.19 External forcings are driven by a combination of human factors (e.g., changes in anthropogenic emissions of ozone precursors and greenhouse gases (GHGs)) and natural factors (changes in volcanic eruptions and solar irradiance). The natural internal climate variability of O3 consists of both direct effects (e.g., through natural fluctuations in atmospheric circulation) and indirect effects (e.g., via the production of NOx by lightning). Our goal here is to explore the feasibility of using a pattern-based “fingerprint” method to identify the signal of human influences in the relatively short (17-year) observed (satellite-derived) record of UT O3 trends. A previous study (which did not apply fingerprinting) has suggested that an observational record length of two decades may be insufficient for signal detection due to the large amplitude and spatially complex internal variability of tropospheric O3 variations, which can produce regional trends of opposite sign to the forced changes in O3.17 Neglecting to account for the uncertainty introduced by internal climate variability may lead to inaccurate assessment of the causes of observed O3 changes and misallocation of resources in mitigation efforts.20

Internal variability arises from the chaotic nature of the coupled climate system.21 It includes components that have some level of short-term predictability, such as the El Niño-Southern Oscillation (ENSO) which is the dominant mode of interannual climate variations observed in the troposphere. The response of tropospheric column ozone (TCO) to ENSO has been found in many observational and modeling studies.22−25 For example, Ziemke et al. used TCO measurements to show that ENSO modulates interannual variability in tropospheric O3 over the western and eastern Pacific.26 Changes in convection induced by ENSO affect vertical mixing and NOx produced during lightning, both of which influence UT O3. The longer lifetime of UT versus LT O3 makes the former more susceptible to internal climate variability through large-scale atmospheric circulation patterns. For example, the Quasi-Biennial Oscillation (QBO) affects the magnitude of STE, which in turn affects UT O3.23

Due to the presence of large-amplitude climate noise containing the imprint of ENSO and other modes of internal variability, assessing the causes and statistical significance of observed O3 trends is challenging. Previous studies have used different approaches to separate signal and noise in observed O3 trends, including local analysis at a single mountaintop site in the Pacific that samples free tropospheric O3,27 regional analysis of surface O3 changes over the western U.S.A.,28 and global analysis under different climate forcing scenarios.29 Using the same simulations we analyze here, Fiore et al. found that the continued global-scale increases in tropospheric O3 during recent decades were largely in response to human activities, while internal variability may strongly affected some regional trends.17

Here, we exploit differences between the spatial structure of signal and noise patterns over 50°S to 50°N to examine whether an externally forced signal in UT O3 is detectable in available satellite observations and attributable to human influence. We apply a standard method that relies on pattern information to search for anthropogenic climate change “fingerprints” embedded in the background noise of internal variability.30−35 This method has been successfully applied to a wide variety of observations, including atmospheric and ocean temperatures, rainfall, surface pressure, sea level, ocean acidity, and climate extremes.36 To date, however, pattern-based fingerprint methods have not been used to investigate the causes of changes in tropospheric O3. We demonstrate below that the use of this formal detection and attribution (D&A) method allows us to identify an anthropogenic fingerprint of UT O3 changes in both a satellite-derived product and in model simulations of recent climate change. We contrast our D&A results for UT O3 with those obtained for LT O3, where different physical and chemical processes govern the O3 changes.

Our study relies on tropospheric O3 observations during 2005–2021 from the OMI/MLS satellite product, from which a partial tropospheric column is retrieved in the UT, and on historical simulations (1950–2014) from a 16-member initial-condition chemistry-climate model (CESM2-WACCM6) ensemble. These ensemble members differ from each other only in their initial climate state, providing 16 possible realizations of the combined response of O3 to external forcings and internal climate variability. Details of all data sets and analysis methods are given in Materials and Methods and the SI Appendix. The data and code that support the findings of this study are openly available at the following URL/DOI: https://doi.org/10.5281/zenodo.11895904.

Materials and Methods

Chemistry-Climate Initial-Condition Ensemble Simulations

We use monthly O3 fields archived from a 16-member initial-condition ensemble generated with version 2 of the Community Earth System Model, which incorporates version 6 of the Whole Atmosphere Community Climate Model (CESM2-WACCM6). Simulated fields from 1950 to 2014 have previously been described and evaluated in Fiore et al.17 The model includes a complex representation of tropospheric and stratospheric chemistry, and the atmosphere is coupled to ocean, sea ice, and land models.37−40 The global atmospheric domain has a horizontal resolution of 0.95° (latitude) × 1.25° (longitude). Emissions were taken from the Coupled Model Intercomparison Project Phase 6 (CMIP6), with historical emissions from fossil fuel and biofuel combustion from Hoesly et al.41 and biomass burning emissions from van Marle et al.42 Emissions of biogenic volatile organic compounds, lightning NO, dust, and sea salt are calculated online (i.e., they respond to meteorology; details in Gettelman et al.39) and are therefore sensitive to climate variations. Stratosphere-troposphere exchange (STE) of ozone is also affected by the initial climate conditions, which impact the atmospheric circulation and climate variability (e.g., in response to the quasi-biennial oscillation (QBO) simulated in this model39). We extracted O3 trends in the lower (>690 hPa) and upper (300 hPa to tropopause) troposphere using the thermal lapse rate tropopause (2 K km–1), for consistency with the tropospheric column O3 (TCO) retrieved using the OMI and MLS satellite instruments. Except for Figure 1, we analyze O3 partial column concentrations in Dobson Units (DU). We restrict the domain for analysis to 50°S–50°N because the tropopause pressure can be close to (or even below) 300 hPa in winter poleward of 50°.

Figure 1.

Figure 1

Open in a new tab

Annual-mean ozone burden trends from 1950 to 2014 in the upper (300 hPa to tropopause; blue) and lower troposphere (pressures >690 hPa; pink) in the CESM2-WACCM6 historical simulations. The ozone burden is summed over 50°S–50°N. Thin lines show the 16 individual CESM2-WACCM6 ensemble members and thick lines show the ensemble-mean value in each year.

The procedure for initializing the 16-member ensemble of CESM2-WACCM6 historical simulations was previously described in Fiore et al.17 and follows the methods of Deser et al.19 and Kay et al.43 The ensemble includes three members available from the National Center for Atmospheric Research (NCAR) contribution to CMIP6; they were launched from a long preindustrial control simulation and then follow the historical emissions and forcing trajectories from 1850 until 2014.37 In the 13 other ensemble members, the initial conditions (ocean, sea ice, atmosphere, land) on January 1, 1950, from one of these three CMIP6 ensemble members were applied and a slight perturbation was imposed on the initial atmospheric temperature field (magnitudes of the perturbation ranged from 1 × 10–14 to 5 × 10–14 K). All emissions of tropospheric ozone precursors (except those tied to meteorology in the model) and climate forcings follow the CMIP6 historical protocol (e.g., refs (41, 42, 44, and 45)) and are identical across all ensemble members. These 16 realizations provide different plausible realizations of historical O3 changes, each with a unique sequence of internal variability (“noise”) superimposed on the underlying response to anthropogenic and natural external forces (“signal”). We rely on this ensemble to quantify the contributions of external forcing and internal climate variability to tropospheric O3 trends.

Prior work shows that the climate simulated by CESM2-WACCM6 captures key features of observed 20th century climate and variability,37−39 as well as observed tropospheric O3 trends available from measurements at long-term surface observatories, IAGOS commercial aircraft, and space-based satellite instruments.17 Fiore et al.,17 show that the free tropospheric column ozone trends (700–250 hPa) derived from this initial-condition ensemble bracket the IAGOS aircraft measurements in 8 out of 11 world regions with long-term sampling by aircraft. Most pertinent to our upper tropospheric trend analysis is our comparison with IAGOS-derived UT O3 trends over 1994–2013 from Cohen et al. (see their Figure 10h),46 which shows that the observed UT O3 trends for all 7 northern hemispheric regions fall within the range of our CESM2-WACCM6 initial-condition ensemble (Figure S1).

Satellite Products

Annual mean upper tropospheric column O3 data over 50°S–50°N are available from 2005 to 2021 with a horizontal resolution of 5° × 5° from the Aura Ozone Monitoring Instrument/Microwave Limb Sounder (OMI/MLS) satellite record.14 UT column O3 is obtained by differencing two OMI/MLS products that are routinely retrieved and publicly available. Specifically, we subtract the ground-to-300 hPa product from the full tropospheric column ozone (TCO). The latter is retrieved to the World Meteorological Organization (WMO) definition of the thermal tropopause, with the tropopause diagnosed from the National Centers for Environmental Prediction (NCEP) reanalysis.14 The TCO product was estimated by subtracting the stratospheric column O3 derived from MLS from the OMI total column O3 at each grid point. A recent global uniform drift adjustment of −0.6 ± 0.38 DU/decade has been applied equivalently to both the TCO and ground-to-300 hPa retrievals,47 therefore the UT column O3 trend is not affected by the recent correction. The corrected dataset is available at https://acd-ext.gsfc.nasa.gov/anonftp/acd/atmos/ziemke/.

In Fiore et al.,17 the TCO trends estimated from the OMI/MLS record were spatially averaged into 10° latitude bands and compared to trends over 2005–2014 derived from the CESM2-WACCM6 initial-condition ensemble. The OMI/MLS TCO trends fall within the inter-ensemble range of the simulated trends, and the recent drift adjustment is sufficiently small such that the corrected observed trends still fall within the inter-ensemble range. The UT O3 trends derived from OMI/MLS lie within the CESM2-WACCM6 inter-ensemble range in the majority of grid cells (Figure S2). Furthermore, the OMI/MLS data show spatial similarity to the model fingerprint (defined in the following section), most notably in terms of the pronounced increase in O3 in the ∼30°N Asian area.

A comparison of annual mean UT O3 data from OMI/MLS with ozonesonde data from the Southern Hemisphere ADditional OZonesondes (SHADOZ) network is shown in Figure S3, indicating consistent latitudinal gradients and annual mean UT O3 with a tendency for the satellite product to overestimate relative to the ozonesondes. A comparison of MLS with CAM-chem and WACCM6 revealed that climatological UT O3 values generally fall within 10–20% of each other and the positive tropical trends (the 147 and 215 hPa pressure levels) are also similar, though the WACCM6 trend is slightly smaller.48

Pattern-Based Fingerprint Method

We apply a standard pattern-based “fingerprint” method to tropospheric O3 between 50°N and 50°S to determine whether the model fingerprint in response to external forcing is statistically identifiable in each model realization of historical climate change and in the satellite record.30 This method has been successfully employed to identify anthropogenic fingerprints in different independently monitored aspects of climate change.31,32,34,49 We seek to determine whether the pattern similarity between the model fingerprint and satellite data (or individual model realizations) increases with time and whether such an increase is significant relative to random changes in the similarity between the fingerprint and patterns of natural internal variability. The statistical methodology follows Santer et al.;33 full details are provided in the SI Appendix. A brief description is given below.

The model fingerprint F(x) represents the signal in response to the applied external forcings, where x refers to the spatial index. Here, F(x) is defined as the leading empirical orthogonal function (EOF) of the area-weighted annual-mean ensemble-mean UT O3 anomalies over 1950 to 2014 derived from the CESM2-WACCM6 simulations. We compare F(x) with patterns of area-weighted annual-mean UT O3 anomalies from (1) O(x,t), the UT O3 derived from the OMI/MLS satellite products, where t = time (years); (2) S(i,x,t), the individual model historical realizations, where i = 1,16 ensemble members; and (3) C(x,t), the model-based estimates of internal variability (“climate noise”), which are defined as the residual UT O3 after removing the ensemble-mean O3 (“signal”) from each of the 16 model historical realizations. Concatenation of the 16 “signal removed” historical realizations in C (i,x,t) yields C(x,t), which consists of 16 × 65 = 1040 years of noise data.

By calculating the uncentered spatial covariance between F(x) and O(x,t) or S(i,x,t), we obtain “signal” time series (Zo(t) or Zs(i,t), respectively). The spatial covariance statistic (Zo(t) or Zs(i,t)) preserves overall differences between the spatial variances in the fingerprint and in the time-evolving UT O3 fields. In the same manner, we obtain a “noise” time series N(t) by computing the uncentered spatial covariance between F(x) and C(x,t). The “signal” time series shows a clear increase over time (see the Results section), indicating that a forced signal is emerging more strongly with time. We compare “signal trends” (trends in Zo(t) or ZS(i,t), starting from the first year of the time series) of varying year lengths (L) with the distribution of noise trends of lengths L. The noise trend distributions are constructed from non-overlapping L-year trends of N(t) (see details in SI Appendix). The distributions are Gaussian and centered on zero.

For each trend length L (with L = 10, 11,···, Nt years, where Nt = 65 for individual model realizations and Nt = 17 for satellite data), we define timescale-dependent signal-to-noise ratios (S/N) as the signal trend divided by the standard deviation of the noise trend distribution. The 1% significance threshold is derived from the one-tailed Student’s t-table and used to determine if the signal trend is significantly larger than the trends arising by chance due to internal climate variability. The detection time td is defined as the final year of the L-year analysis period for which the S/N first exceeds the 1% significance threshold, and then remains above that threshold for all trend lengths L > td. We calculate S/N and td for the OMI/MLS (satellite) UT O3 and from each individual model realization. To better understand the differences in the upper and lower troposphere, we also calculate the “model only” S/N for LT O3 (results are shown in SI Appendix), noting that it is not currently possible to derive partial LT O3 columns from space, and sparse in situ observations prevent us from determining an observational value of td for LT O3 changes.

Results

Ozone Trends in the Upper and Lower Troposphere

Long-term trends in annual-mean O3 burdens over 50°S–50°N increase in both the upper and lower troposphere in every one of the 16 ensemble members (thin lines in Figure 1). Each ensemble member provides a unique realization of the noise of natural internal variability overlaid on the underlying externally forced signal, where the signal is the response of O3 to historical changes in GHGs, O3 precursor emissions, aerosols, land use, and natural forcings. Because the noise is uncorrelated (except by chance) from realization to realization, averaging over the 16 ensemble members reduces the noise amplitude and provides the best estimate of the signal (thick lines in Figure 1). The climate noise is defined here as the deviation of each realization from the ensemble mean.19,43,50,51

In the model simulations for the period 1950 to 2014, the UT (300 hPa to tropopause) contributed 21 Tg of the total rise of 75 Tg in the ensemble-mean column burden of tropospheric O3. This increase is primarily driven by rising anthropogenic tropospheric O3 precursor emissions, including NOx, CO, and methane.52 UT O3 rises more steeply than LT (>690 hPa) O3 after roughly 1980, likely due to an equatorward shift in anthropogenic NOx emissions.13,53 This shift toward low latitudes leads to enhanced convective lofting of O3 and precursors to the free troposphere, where both the ozone production efficiency and lifetime are longer. Increasing aircraft NOx emissions may also contribute to the positive trend in UT O3.54

Some of the short-term downturns in the tropospheric O3 changes reflect temporary decreases in global anthropogenic NOx and CO emissions (e.g., in the 1970s). Other short-term O3 decreases that are identical across all ensemble members are driven by sporadic volcanic eruptions and biomass burning (BB) events. For example, the 1–2 year decreases in tropospheric O3 in 1963, 1982, and 1991 coincide with enhanced O3 depletion in the lower stratosphere following major volcanic eruptions, which decreased STE O3 transport.55,56 The upturn in both UT and LT O3 in 1997–1998 reflects the large El Niño event that triggered extreme BB emissions in Indonesia.57 This highlights the fact that in the real world, there is a strong influence of internal climate variability on the timing of some BB events. However, since our simulations follow the Coupled Model Intercomparison Project Phase 6 (CMIP6) protocol, the same time history of observed BB emissions is applied in each ensemble member, yielding synchronicity of BB effects on O3 across all 16 ensemble members.

In both the UT and LT, the signal of the long-term annual mean O3 increase clearly exceeds the amplitude of between-realization internal climate variability, particularly in the LT where the majority of direct anthropogenic emissions occur (Figure 1). While year-to-year variations in near-global mean UT O3 caused by internal climate variability can be appreciable (up to 9 Tg), the spatial patterns of long-term UT O3 trends are remarkably similar in each ensemble member (Figure S4A). When we examine UT O3 trends over a shorter 20-year period (1995–2014), however, results are more sensitive to the noise of internal variability, and the sign and magnitude of O3 trends can show noticeable regional differences among the individual ensemble members (Figure S4B). The larger inter-ensemble spread of UT O3 relative to LT O3 (Figure 1) reflects the stronger influence of internal climate variability on UT O3, presumably due to variations in atmospheric circulation and lightning NOx, combined with a longer O3 lifetime in the UT. Below, we examine the implications of this larger climate variability in the UT for detecting anthropogenic influence on O3 trends. We compare UT and LT signal detection properties using both a local signal-to-noise ratio approach and a pattern-based fingerprinting method that considers global-scale O3 changes.

Local Signal-to-Noise Ratios

Analysis of local signal-to-noise ratios considers forced and unforced changes at individual model grid points,19,35,58,59 which can help identify regions with strong anthropogenic change signals relative to internal variability. The signal S is the local trend over a selected period and the noise N is the local standard deviation of the trend across the 16 ensemble members. Figure 2 shows the local S, N, and S/N for both UT and LT O3.

Figure 2.

Figure 2

Open in a new tab

Local signal (S), noise (N), and signal-to-noise ratios (S/N) of annual-mean ozone trends in the 16-member CESM2-WACCM6 ensemble in the upper (A, C, and E) and lower (B, D, and F) troposphere. The signals (A and B) are the ensemble-mean least-squares linear trends in annual-mean tropospheric ozone over 1950 to 2014. The noise (C and D) is the standard deviation of the ozone trends in the 16 ensemble members. E and F show the S/N.

The UT O3 signal is characterized by strong latitudinal gradients, with little meridional variation in trends (Figure 2A). In contrast, the LT signal has more meridional structure and land-ocean contrast, reflecting a shorter O3 lifetime and slower zonal mixing in the LT (Figure 2B). The increasing UT O3 trend maximizes at northern midlatitudes, where the largest trends occur over South and East Asia, likely due at least in part to the uplift of surface air pollutants to the UT by the Asian monsoon.60,61 In contrast, the largest LT O3 trends are confined to areas with high anthropogenic precursor emissions, including India and East China, and areas susceptible to BB events (such as central Brazil).

Consistent with Figure 1, the between-realization variability of O3 trends in the UT is substantially larger than in the LT (compare Figure 2C, D) and is greatest over the Pacific Ocean between 30°S and 30°N, which is dominated by ENSO.24 The larger noise in the UT O3 trend also reflects STE, which is known to influence O3 across North America, Europe, the southern Pacific, and the southern Indian Ocean.62,63 The spatial average S/N is smaller for UT than for LT O3 changes (Figure 2E, F), implying that the forced signal in CESM2-WACCM6 may be harder to detect in the UT than in the LT, except in a few specific regions (such as Indonesia). The largest S/N for UT O3 changes is displaced toward the tropics relative to the LT.

Temporal variations of the local signal, noise, and S/N are shown in Figure S5 for UT O3 for three different 20-year periods in the CESM2-WACCM6 simulations: 1950–1969, 1970–1989, and 1995–2014. The local S/N generally increases over these three periods, indicating that the UT O3 signal should be easiest to detect in the most recent period when the OMI/MLS satellite record starts. The local signal of UT O3 during 1950–1969 increased between 30°S and 50°N but decreased south of 30°S, suggesting that the lower STE under stratospheric O3 depletion outweighed any increases from the uplift of surface O3 and precursor pollutants as surface emissions rose over this period. The influence of stratospheric O3 depletion has weakened in the 21st century,64 and we find that during 1995–2014, the UT O3 signal increased at virtually all locations. This significant increase of UT O3 has previously been reported for most regions sampled by a routine commercial aircraft for the period 1994–2013.46

Comparing Figures S5 and S6 shows strikingly different trend patterns in UT and LT O3, which arise because different processes control UT versus LT O3 distributions. The increasing signal trend in LT O3 is largest in the first two decades, in accord with the time evolution of the overall trends in global anthropogenic precursor emissions,17 with an equatorward redistribution of NOx emissions over time as US and European emissions decrease but Asian emissions increase. Since 1970, Asia has dominated the signal of LT O3 trends (Figure S6). Declining LT O3 trends occurred after 1990 in the U.S. and Europe as emission controls were implemented to reduce acid rain and ground-level O3. Similar decreases in LT O3 trends occur in South America where BB emissions decreased, producing the negative local S/N in Figure S6I. Regional NOx trends are more consistent with O3 trends in the LT compared to the UT, reflecting the greater direct impacts of in situ production on LT O3 (compare Figure S7 versus S5 and Figure S8 versus S6). For example, the UT NOx in the eastern US decreased during 1995–2014 while the UT O3 increased (Figures S5C and S7C). Observations indicate that transport of O3 from Asia impacts O3 trends in other regions; this transport is estimated to have offset 43% of the expected reduction in free tropospheric O3 over the western US from 2005 to 2013.65 Our analysis indicates that differences in the UT and LT O3 trends mainly reflect differences in the vertical structure and spatial coherence of the ozone response to external forcings rather than to internal climate variability.

Features of the Pattern-Based Fingerprint

Pattern-based fingerprinting utilizes the signal and noise properties of entire spatial fields.31−33,35 Our pattern-based method relies on F(x), the “fingerprint” of the UT O3 response to external forcing, defined here as the first empirical orthogonal function (signal EOF1, hereafter referred to as SEOF1) of ensemble-mean changes in annual-mean UT O3 in the CESM2-WACCM6 simulations. It is calculated over a domain extending between 50°S–50°N using simulated O3 from 1950 to 2014. SEOF1, shown in Figure 3A, explains 96.5% of the total variance in the ensemble-mean UT O3 changes and is dominated by the response to anthropogenic external forcings.

Figure 3.

Figure 3

Open in a new tab

Leading modes of the UT O3 response to external forcing versus natural internal climate variability. Results are (A) fingerprint of changes in UT O3 in CESM2-WACCM6. The fingerprint is the leading EOF of annual-mean ensemble-mean UT O3 anomalies from 1950 to 2014 between 50°S and 50°N; anomalies are calculated relative to the climatological annual-mean O3 over this 65-year period; (B) first principal component (PC) of ensemble-mean UT O3; (C) first EOF of the internal variability in UT O3, estimated from the concatenated between-realization variations of the 16 CESM2-WACCM6 ensemble members; (D) First PC of natural internal variability. Each colored line represents a different ensemble member. The total variance explained by each EOF is listed in A and C.

The fingerprint captures the zonally coherent changes in UT O3 and is spatially similar to the local signal trend pattern (compare Figures 2A and 3A). The hemispheric asymmetry of F(x), with larger loadings at northern midlatitudes, reflects the larger externally forced UT O3 increases in the Northern Hemisphere (NH) that are driven by greater population density and (in consequence) greater anthropogenic precursor emissions. The leading principal component (SPC1, Figure 3B) is similar to the time series of global-mean ensemble-mean changes in UT O3 (Figure 1), with short-term excursions due to natural volcanic forcings superimposed on an overall positive trend driven by anthropogenic external forcings.

The implicit assumption in our pattern-based fingerprint method is that F(x) does not change markedly over time. We examine the validity of this assumption for both UT and LT O3 changes by defining the fingerprint F(x) over three different 20-year periods (Figure S9). The UT O3 fingerprint pattern is relatively robust over these time periods, with the correlation coefficients between each of the three possible pairs of different fingerprint patterns varying between 0.66 and 0.86. In contrast, the spatial pattern of the LT O3 fingerprint changes markedly over time, with three correlation coefficients ranging from 0.11 to 0.61 (Figure S9B, D, and F). The relative temporal stability of the fingerprint in the upper troposphere reflects the longer O3 lifetime combined with stronger zonal mixing at midlatitudes relative to the lower troposphere.

Characterizing UT O3 Noise

Fingerprint studies require estimates of internal climate variability to determine whether the fingerprint is detectable.31,32 As in previous work,35 we rely here on internal variability (noise) estimates calculated with the between-realization variability of the CESM2-WACCM6 historical runs after removal of the ensemble-mean signal from each realization (details in Materials and Methods). The underlying assumptions here are that our ensemble size (16) is sufficient for obtaining a reliable estimate of the externally forced signal (the fingerprint F(x)) and that external forcing does not significantly modulate internal variability.35 The estimated model noise allows us to evaluate the statistical significance of the time-increasing similarity between F(x) and the satellite-derived tropospheric O3 changes (Figure 3B).

An EOF analysis of the UT O3 internal variability information yields estimates of the dominant noise patterns (Figure 3C, E). The first and second noise EOFs (NEOF1 and 2) explain 28.8 and 10.2% of the total variance (respectively); the variance explained by NEOFs 3 and higher are each ≤3.2%. Note that there is some overlap in SEOF1 and NEOF1 in the extratropical Pacific Ocean and Indian Ocean. One interpretation of this overlap is that 16 ensemble members may be insufficient to dampen the noise of internal variability superimposed on the externally forced UT O3 signal. Alternatively, this overlap could indicate a physical link between signal and noise associated with anthropogenically forced modulation of internal variability.66 NEOF1 has the highest loadings in the tropical Pacific, where there is anticorrelated UT O3 variability between O3 fluctuations at low latitudes and at 30°N and 30°S. This distinctive pattern is reminiscent of ENSO. The amplitude of noise PC1 (NPC1) grows with time, possibly reflecting an intensification of ENSO-driven variability in tropospheric temperature, humidity, and atmospheric circulation under global warming.66 An alternative (and not mutually exclusive) explanation for the increase in amplitude of NPC1 with time is that the growth of tropospheric O3 precursor emissions in the historical simulations enhances the variability of O3. Such an effect would not be removed by the subtraction of the ensemble-mean O3 signal from each historical realization.

The UT O3 NPC1 time series (Figure 3D) correlates strongly (r = 0.93) with the Niño 3.4 sea surface temperature (SST) anomalies calculated using the concatenated SST residuals after subtracting the ensemble-mean from each of the 16 ensemble members, implying that ENSO plays an important role in modulating UT O3 in the absence of external forcing. Earlier analysis of tropospheric column O3 (TCO)24,26 and UT O323 supports this interpretation. The high correlation coefficient (r = 0.94) between the LT O3 NPC1 (Figure S10D) and Niño 3.4 SST anomalies confirms that the ENSO-O3 covariation extends throughout the troposphere.

Both chemical and dynamical processes operate during ENSO to produce this pattern of UT O3 variations (Figure 3C). Since tropospheric O3 is generally low in marine air, especially in the tropical Pacific,67,68 observed UT O3 trends will be easily influenced by shifts in the patterns and strength of deep convection driven by internal climate variability. During El Niño, the region of active convection shifts eastward to the central Pacific, leading to smaller upper tropospheric O3 concentrations there, while positive tropospheric ozone anomalies occur over the Indonesian warm pool region, associated with enhanced downward flow.23,24 During La Niña, the situation reverses, although the strengthening of convection in the western tropical Pacific also enhances the production of NOx from lightning.69,70

Pattern-Based Signal-to-Noise Ratios

As noted above, we search for the model fingerprint F(x) of UT O3 (Figure 3A) in each individual model realization and in the satellite record. Specifically, we test for a statistically significant trend in the pattern similarity between F(x) and the time-varying model realizations or observations.35

To perform this test, we first project UT O3 anomalies from the individual CESM2-WACCM6 ensemble members and observations onto F(x) to obtain the signal time series S(t). Each signal time series is a measure of the spatial covariance between F(x) and the time-varying model or satellite UT O3 fields. We then fit trends of increasing length of years (L) to these signal time series using ordinary least-squares regression. The assumed start date for estimating trends in pattern similarity is 1950 for the model realizations and 2005 for the OMI/MLS satellite data (which commence in late 2004).

Figure 4A shows the signal trends as a function of the time scale L in the 16 model realizations of historical climate change. For the shortest time scale (L = 10 years), the spread across the different ensemble members is appreciable because internal variability is large relative to the signal over 1950 to 1959. This ensemble spread in S decreases as L increases. There are two reasons for this reduction. First, the spatial patterns of UT O3 changes in individual ensemble members are becoming increasingly similar to F(x) as the anthropogenic forcing increases. This is evident not only in the spatial covariance in Figure 4A but also in the time evolution of the spatial covariance between F(x) and OMI/MLS-derived UT O3 trends shown in Figure 4B as well as that of the spatial correlation in Figure S11. Second, the amplitude of natural internal variability superimposed on the signal S(t) decreases as L increases (Figure 4C).

Figure 4.

Figure 4

Open in a new tab

Signal, noise, and S/N from a pattern-based fingerprint analysis of annual-mean upper tropospheric ozone changes in the CESM2-WACCM6 ensemble and in satellite data within 50°S–50°N. Results are shown for each of 16 realizations (A, C, and E) and satellite observations (B, D, and F) and are a function of the trend length L. Detailed definitions of the signal trends (A and B), the standard deviation of the noise trend distribution (C and D), and the time-dependent S/N (E and F) are provided in the Materials and Methods. In brief, the signal represents trends in the pattern similarity between the time-invariant model fingerprint F(x) and the time-varying UT O3 simulated in each historical realization, or between F(x) and the time-varying satellite data. Trends are a function of the time scale L (in years), with L varying from 10, 11, 12··· Nt years, where Nt = 65 years for the historical realizations and Nt = 17 years for the satellite data. The last year of the L-year analysis period is shown in red in the upper x-axis. Time-dependent trends in the noise are defined analogously, and are obtained by comparing the fingerprint with the time-varying between-realization internal variability estimated from the CESM2-WACCM6 ensemble.

Because the start date of the satellite data is over 50 years later than in the model simulations, the signal strength of the UT O3 fingerprint in observations is shown separately in Figure 4B. The signal strength in observations is invariably positive as a function of L and shows appreciable variability as L increases from 10 to 17 years (the maximum length of the satellite record). The latter behavior is consistent with the large year-to-year variability of model signal trends for L values less than 20 years (Figure 4A).

To assess the likelihood that the time-increasing similarity between the model or satellite UT O3 data and F(x) could have arisen randomly, we require null distributions where we know a priori that any trend in the similarity between F(x) and time-varying patterns of UT O3 changes arose through internal variability alone. Here, we generate suitable null distributions by projecting the UT O3 noise dataset onto F(x) (see Materials and Methods). For each value of L, the noise trends are normally distributed about a mean of zero. As the trend-fitting period L increases, there is a decrease in σN(L), the standard deviation of the distribution of L-year noise trends (Figure 4C, D). This reduction occurs because with increasing L, it becomes more difficult for internal variability to generate large positive or negative trends in pattern similarity.

The UT O3 S/N is simply the ratio between the time-dependent signal and the time-dependent noise. S/N increases with longer trend-fitting periods (Figure 4E, F), primarily driven by the decay of σN(L). In each of the 16 model realizations of historical climate change and in the OMI/MLS satellite record, the model UT O3 fingerprint is identifiable with high statistical confidence (α = 0.01). This indicates that natural internal variability is highly unlikely to explain the time-increasing similarity between the model fingerprint and satellite-retrieved patterns of UT O3 change (or between F(x) and individual model realizations of UT O3 changes). The small changes in the UT O3 fingerprint F(x) over time do not hamper the detection of the UT O3 fingerprint in satellite observations (see Figure S12).

We also estimate td, the time required to detect the forced UT O3 fingerprint pattern. As shown in previous work, td is defined as the last year of the L-year analysis period in which S/N first rises above (and then remains above) a stipulated 1% significance level, assuming a Gaussian distribution of noise trends (see Materials and Methods). For a start date of 1950 and an initial trend length of 10 years, the detection time td in the 16 model realizations varies from 12 to 18 years–i.e., the earliest detection time is in 1961 and the latest detection time is in 1967 (Figure 4E). For the 17-year satellite record, in which the start date of the analysis is at the beginning of the satellite O3 record (in 2005), the detection time is 13 years (Figure 4F).

An interesting feature shown in Figure 4E is that S/N exhibits small “jumps” as a function of L and that there is a strong correlation of S/N(L) across the 16 ensemble members. This behavior is due to two factors: the large amplitude of S relative to N, and the slight fluctuations in σN(L) arising from the relatively small sample sizes used in estimating σN(L) (compare Figure 4A, C).

For trends in LT O3 over the full 1950–2014 period, the assumption of a time-invariant F(x) pattern is not satisfied but the time-varying spatial patterns are nevertheless key signatures of anthropogenic influence. We find that irrespective of the choice of the analysis period used for defining the LT O3 fingerprint, F(x) is robustly identifiable in the individual model historical realizations (discussed in SI Appendix).

Discussion

The continuous increase of UT O3 in recent decades and particularly its difference from trends in LT O3 since roughly 198017,18 raises the question of whether the UT O3 trend is forced by human influence on atmospheric composition. A 16-member CESM2-WACCM6 ensemble of chemistry-climate simulations, with each member commencing from different initial conditions, allows us to isolate the “signal” (driven by combined anthropogenic and natural external forcings) from the “noise” (arising from internal climate variability). The UT O3 signal varies by latitude, with maximum values occurring between 30°N and 30°S. In contrast, LT O3 trends vary strongly with location and over time and are closely linked to spatiotemporal changes in precursor emissions (Figures 2, S5, and S6). The UT O3 noise is primarily associated with ENSO, showing features similar to those previously identified in tropospheric column O3 (TCO).24,26 The small-scale UT O3 noise patterns are strikingly different from the large-scale, spatially coherent signals in UT O3, which is a favorable situation for signal detection.

We performed two types of S/N analysis - a local analysis at individual model grid points and an analysis of global patterns of tropospheric O3 change. Locally, the S/N for O3 trends is almost always higher in the LT than in the UT. The local analysis also reveals that south and east Asia are regions where anthropogenic signals in tropospheric composition can be detected most readily in both LT and UT O3. This finding can help inform future observing systems targeting rapid detection of anthropogenic impacts on tropospheric O3.

We applied a standard pattern-based climate fingerprint method to historical O3 changes in satellite observations and individual model realizations (Figure 3). Although the satellite UT O3 record began in 2005 and is only 17 years in length, the model-predicted fingerprint in response to combined anthropogenic and natural external forcing is identifiable with high statistical confidence after only 13 years of monitoring. Robust detection of the model UT O3 fingerprint also occurs in each of the 16 CESM2-WACCM6 realizations of historical climate change after 12 to 18 years.

Our study is the first attempt to apply standard climate detection and attribution methods to observed and modeled changes in tropospheric chemistry. Despite relying on only 16 ensemble members, which is a relatively small size compared with the large ensembles of 30–100 members created by the physical climate modeling community,50 our findings illustrate the potential to gain an understanding of climate change causes and processes by applying fingerprint methods to atmospheric chemical composition.

Future work could build upon our findings in a number of ways. A multimodel assessment would reveal whether S/N estimates obtained from CESM2-WACCM6 are robust to model structural uncertainty. Extending the ensemble simulations considered here (which end in 2014) would facilitate a direct comparison of the model and observational S/N results for the full period of the OMI/MLS satellite record (2005 to 2022). Additional sets of single-forcing initial-condition ensemble simulations that isolate the influence of biomass burning or other emission sectors would help separate the role of different anthropogenic emission changes, thereby providing valuable information to guide future mitigation plans.

Acknowledgments

This work was funded by the National Aeronautics and Space Administration (NASA) grant (80NSSC23K0925). X.Y. was also supported by the Thomas C. Desmond Fellowship at the Massachusetts Institute of Technology (MIT) and the MathWorks Science Fellowship. B.D.S. was supported by the Francis E. Fowler IV Center for Ocean and Climate at Woods Hole Oceanographic Institution. We would like to acknowledge high-performance computing support from Cheyenne (doi: https://doi.org/10.5065/D6RX99HX) provided by NCAR’s Computational and Information Systems Laboratory, sponsored by the National Science Foundation. We are grateful to all members of the Community Earth System Model (CESM) working groups, who develop and continually improve the model. We thank the Harmonization and Evaluation of Ground-Based Instruments for Free Tropospheric Ozone Measurements (HEGIFTOM) Focus Working Group for providing the ozonesonde data (http://hegiftom.meteo.be/datasets) to compare with the satellite observations. We also acknowledge helpful discussions with Susan Solomon.

Supporting Information Available

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.est.4c01289.

  • Additional details for the pattern-based fingerprint method; additional analysis on the “model-only” fingerprint analysis for lower tropospheric O3; and supplementary figures for model and satellite data evaluation, trend analysis, and fingerprint detection (PDF)

The authors declare no competing financial interest.

Supplementary Material

es4c01289_si_001.pdf (5.7MB, pdf)

References

  1. Jerrett M.; Burnett R. T.; Pope C. A.; Ito K.; Thurston G.; Krewski D.; Shi Y.; Calle E.; Thun M. Long-Term Ozone Exposure and Mortality. N Engl J. Med. 2009, 360 (11), 1085–1095. 10.1056/NEJMoa0803894. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Turner M. C.; Jerrett M.; Pope C. A.; Krewski D.; Gapstur S. M.; Diver W. R.; Beckerman B. S.; Marshall J. D.; Su J.; Crouse D. L.; Burnett R. T. Long-Term Ozone Exposure and Mortality in a Large Prospective Study. Am. J. Respir Crit Care Med. 2016, 193 (10), 1134–1142. 10.1164/rccm.201508-1633OC. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Emberson L. Effects of Ozone on Agriculture, Forests and Grasslands. Philos. Trans. R. Soc. A 2020, 378 (2183), 20190327. 10.1098/rsta.2019.0327. [DOI] [PMC free article] [PubMed] [Google Scholar]
  4. Stevenson D. S.; Young P. J.; Naik V.; Lamarque J.-F.; Shindell D. T.; Voulgarakis A.; Skeie R. B.; Dalsoren S. B.; Myhre G.; Berntsen T. K.; Folberth G. A.; Rumbold S. T.; Collins W. J.; MacKenzie I. A.; Doherty R. M.; Zeng G.; Van Noije T. P. C.; Strunk A.; Bergmann D.; Cameron-Smith P.; Plummer D. A.; Strode S. A.; Horowitz L.; Lee Y. H.; Szopa S.; Sudo K.; Nagashima T.; Josse B.; Cionni I.; Righi M.; Eyring V.; Conley A.; Bowman K. W.; Wild O.; Archibald A. Tropospheric Ozone Changes, Radiative Forcing and Attribution to Emissions in the Atmospheric Chemistry and Climate Model Intercomparison Project (ACCMIP). Atmos. Chem. Phys. 2013, 13 (6), 3063–3085. 10.5194/acp-13-3063-2013. [DOI] [Google Scholar]
  5. Monks P. S.; Archibald A. T.; Colette A.; Cooper O.; Coyle M.; Derwent R.; Fowler D.; Granier C.; Law K. S.; Mills G. E.; Stevenson D. S.; Tarasova O.; Thouret V.; Von Schneidemesser E.; Sommariva R.; Wild O.; Williams M. L. Tropospheric Ozone and Its Precursors from the Urban to the Global Scale from Air Quality to Short-Lived Climate Forcer. Atmos. Chem. Phys. 2015, 15 (15), 8889–8973. 10.5194/acp-15-8889-2015. [DOI] [Google Scholar]
  6. Forster P. M.; Bodeker G.; Schofield R.; Solomon S.; Thompson D. Effects of Ozone Cooling in the Tropical Lower Stratosphere and Upper Troposphere. Geophys. Res. Lett. 2007, 34 (23), L23813. 10.1029/2007GL031994. [DOI] [Google Scholar]
  7. Iglesias-Suarez F.; Kinnison D. E.; Rap A.; Maycock A. C.; Wild O.; Young P. J. Key Drivers of Ozone Change and Its Radiative Forcing over the 21st Century. Atmos. Chem. Phys. 2018, 18 (9), 6121–6139. 10.5194/acp-18-6121-2018. [DOI] [Google Scholar]
  8. Lacis A. A.; Wuebbles D. J.; Logan J. A. Radiative Forcing of Climate by Changes in the Vertical Distribution of Ozone. J. Geophys. Res. 1990, 95 (D7), 9971–9981. 10.1029/JD095iD07p09971. [DOI] [Google Scholar]
  9. Archibald A. T.; Neu J. L.; Elshorbany Y. F.; Cooper O. R.; Young P. J.; Akiyoshi H.; Cox R. A.; Coyle M.; Derwent R. G.; Deushi M.; Finco A.; Frost G. J.; Galbally I. E.; Gerosa G.; Granier C.; Griffiths P. T.; Hossaini R.; Hu L.; Jöckel P.; Josse B.; Lin M. Y.; Mertens M.; Morgenstern O.; Naja M.; Naik V.; Oltmans S.; Plummer D. A.; Revell L. E.; Saiz-Lopez A.; Saxena P.; Shin Y. M.; Shahid I.; Shallcross D.; Tilmes S.; Trickl T.; Wallington T. J.; Wang T.; Worden H. M.; Zeng G. Tropospheric Ozone Assessment Report. Elementa: Science of the Anthropocene 2020, 8 (1), 034. 10.1525/elementa.2020.034. [DOI] [Google Scholar]
  10. Christiansen A.; Mickley L. J.; Liu J.; Oman L. D.; Hu L. Multidecadal Increases in Global Tropospheric Ozone Derived from Ozonesonde and Surface Site Observations: Can Models Reproduce Ozone Trends?. Atmos. Chem. Phys. 2022, 22 (22), 14751–14782. 10.5194/acp-22-14751-2022. [DOI] [Google Scholar]
  11. Cooper O. R.; Schultz M. G.; Schröder S.; Chang K.-L.; Gaudel A.; Benítez G. C.; Cuevas E.; Fröhlich M.; Galbally I. E.; Molloy S.; Kubistin D.; Lu X.; McClure-Begley A.; Nédélec P.; O’Brien J.; Oltmans S. J.; Petropavlovskikh I.; Ries L.; Senik I.; Sjöberg K.; Solberg S.; Spain G. T.; Spangl W.; Steinbacher M.; Tarasick D.; Thouret V.; Xu X. Multi-Decadal Surface Ozone Trends at Globally Distributed Remote Locations. Elementa: Science of the Anthropocene 2020, 8, 23. 10.1525/elementa.420. [DOI] [Google Scholar]
  12. Gaudel A.; Cooper O. R.; Ancellet G.; Barret B.; Boynard A.; Burrows J. P.; Clerbaux C.; Coheur P.-F.; Cuesta J.; Cuevas E.; Doniki S.; Dufour G.; Ebojie F.; Foret G.; Garcia O.; Granados-Muñoz M. J.; Hannigan J. W.; Hase F.; Hassler B.; Huang G.; Hurtmans D.; Jaffe D.; Jones N.; Kalabokas P.; Kerridge B.; Kulawik S.; Latter B.; Leblanc T.; Le Flochmoën E.; Lin W.; Liu J.; Liu X.; Mahieu E.; McClure-Begley A.; Neu J. L.; Osman M.; Palm M.; Petetin H.; Petropavlovskikh I.; Querel R.; Rahpoe N.; Rozanov A.; Schultz M. G.; Schwab J.; Siddans R.; Smale D.; Steinbacher M.; Tanimoto H.; Tarasick D. W.; Thouret V.; Thompson A. M.; Trickl T.; Weatherhead E.; Wespes C.; Worden H. M.; Vigouroux C.; Xu X.; Zeng G.; Ziemke J. Tropospheric Ozone Assessment Report: Present-Day Distribution and Trends of Tropospheric Ozone Relevant to Climate and Global Atmospheric Chemistry Model Evaluation. Elementa: Science of the Anthropocene 2018, 6, 39. 10.1525/elementa.291. [DOI] [Google Scholar]
  13. Gaudel A.; Cooper O. R.; Chang K.-L.; Bourgeois I.; Ziemke J. R.; Strode S. A.; Oman L. D.; Sellitto P.; Nédélec P.; Blot R.; Thouret V.; Granier C. Aircraft Observations since the 1990s Reveal Increases of Tropospheric Ozone at Multiple Locations across the Northern Hemisphere. Sci. Adv. 2020, 6 (34), eaba8272. 10.1126/sciadv.aba8272. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Ziemke J. R.; Oman L. D.; Strode S. A.; Douglass A. R.; Olsen M. A.; McPeters R. D.; Bhartia P. K.; Froidevaux L.; Labow G. J.; Witte J. C.; Thompson A. M.; Haffner D. P.; Kramarova N. A.; Frith S. M.; Huang L.-K.; Jaross G. R.; Seftor C. J.; Deland M. T.; Taylor S. L. Trends in Global Tropospheric Ozone Inferred from a Composite Record of TOMS/OMI/MLS/OMPS Satellite Measurements and the MERRA-2 GMI Simulation. Atmos. Chem. Phys. 2019, 19 (5), 3257–3269. 10.5194/acp-19-3257-2019. [DOI] [Google Scholar]
  15. Granier C.; Brasseur G. P.. The Impact of Road Traffic on Global Tropospheric Ozone Geophys. Res. Lett. 2003, 30( (2), ), 10.1029/2002GL015972. [DOI]
  16. Zhang R.; Lei W.; Tie X.; Hess P. Industrial Emissions Cause Extreme Urban Ozone Diurnal Variability. Proc. Natl. Acad. Sci. U.S.A. 2004, 101 (17), 6346–6350. 10.1073/pnas.0401484101. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Fiore A. M.; Hancock S. E.; Lamarque J.-F.; Correa G. P.; Chang K.-L.; Ru M.; Cooper O.; Gaudel A.; Polvani L. M.; Sauvage B.; Ziemke J. R. Understanding Recent Tropospheric Ozone Trends in the Context of Large Internal Variability: A New Perspective from Chemistry-Climate Model Ensembles. Environ. Res.: Climate 2022, 1 (2), 025008 10.1088/2752-5295/ac9cc2. [DOI] [Google Scholar]
  18. Gulev S. K.; Thorne P.W.; Ahn J.; Dentener F.J.; Domingues C.M.; Gerland S.; Gong D.; Kaufman D.S.; Nnamchi H.C.; Quaas J.; Rivera J.A.; Sathyendranath S.; Smith S.L.; Trewin B.; von Schuckmann K.; Vose R.S.. 2021: Changing State of the Climate System. In Climate Change 2021: The Physical Science Basis. Contribution of Working Group I to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change; Masson-Delmotte V., Zhai P., Pirani A., Connors S.L.; Péan C., Berger S., Caud N., Chen Y., Goldfarb L., Gomis M.I., Huang M., Leitzell K., Lonnoy E., Matthews J.B.R., Maycock T.K., Waterfield T., Yelekçi O., Yu R., Zhou B., Eds.; Cambridge University Press: Cambridge, United Kingdom and New York, NY, USA; pp 287–422. [Google Scholar]
  19. Deser C.; Phillips A.; Bourdette V.; Teng H. Uncertainty in Climate Change Projections: The Role of Internal Variability. Clim Dyn 2012, 38 (3–4), 527–546. 10.1007/s00382-010-0977-x. [DOI] [Google Scholar]
  20. Schwarzwald K.; Lenssen N. The Importance of Internal Climate Variability in Climate Impact Projections. Proc. Natl. Acad. Sci. U.S.A. 2022, 119 (42), e2208095119 10.1073/pnas.2208095119. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Hawkins E.; Smith R. S.; Gregory J. M.; Stainforth D. A. Irreducible Uncertainty in Near-Term Climate Projections. Clim Dyn 2016, 46 (11–12), 3807–3819. 10.1007/s00382-015-2806-8. [DOI] [Google Scholar]
  22. Hsu J.; Prather M. J. Stratospheric Variability and Tropospheric Ozone. J. Geophys. Res. 2009, 114 (D6), D06102. 10.1029/2008JD010942. [DOI] [Google Scholar]
  23. Neu J. L.; Flury T.; Manney G. L.; Santee M. L.; Livesey N. J.; Worden J. Tropospheric Ozone Variations Governed by Changes in Stratospheric Circulation. Nature Geosci 2014, 7 (5), 340–344. 10.1038/ngeo2138. [DOI] [Google Scholar]
  24. Oman L. D.; Douglass A. R.; Ziemke J. R.; Rodriguez J. M.; Waugh D. W.; Nielsen J. E. The Ozone Response to ENSO in Aura Satellite Measurements and a Chemistry-Climate Simulation: OZONE RESPONSE TO ENSO. J. Geophys. Res. Atmos. 2013, 118 (2), 965–976. 10.1029/2012JD018546. [DOI] [Google Scholar]
  25. Voulgarakis A.; Hadjinicolaou P.; Pyle J. A. Increases in Global Tropospheric Ozone Following an El Niño Event: Examining Stratospheric Ozone Variability as a Potential Driver. Atmosph. Sci. Lett. 2011, 12 (2), 228–232. 10.1002/asl.318. [DOI] [Google Scholar]
  26. Ziemke J. R.; Chandra S.; Oman L. D.; Bhartia P. K. A New ENSO Index Derived from Satellite Measurements of Column Ozone. Atmos. Chem. Phys. 2010, 10, 3711–3721. 10.5194/acp-10-3711-2010. [DOI] [Google Scholar]
  27. Lin M.; Horowitz L. W.; Oltmans S. J.; Fiore A. M.; Fan S. Tropospheric Ozone Trends at Mauna Loa Observatory Tied to Decadal Climate Variability. Nature Geosci 2014, 7 (2), 136–143. 10.1038/ngeo2066. [DOI] [Google Scholar]
  28. Lin M.; Fiore A. M.; Horowitz L. W.; Langford A. O.; Oltmans S. J.; Tarasick D.; Rieder H. E. Climate Variability Modulates Western US Ozone Air Quality in Spring via Deep Stratospheric Intrusions. Nat. Commun. 2015, 6 (1), 7105. 10.1038/ncomms8105. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Barnes E. A.; Fiore A. M.; Horowitz L. W. Detection of Trends in Surface Ozone in the Presence of Climate Variability. JGR Atmospheres 2016, 121 (10), 6112–6129. 10.1002/2015JD024397. [DOI] [Google Scholar]
  30. Hasselmann K.On the signal-to-noise problem in atmospheric response studies; Royal Meteorological Society: London, 1979; pp 251–259. [Google Scholar]
  31. Hegerl G. C.; von Storch H.; Hasselmann K.; Santer B. D.; Cubasch U.; Jones P. D. Detecting Greenhouse-Gas-Induced Climate Change with an Optimal Fingerprint Method. Journal of Climate 1996, 9 (10), 2281–2306. . [DOI] [Google Scholar]
  32. Santer B. D.; Mikolajewicz U.; Brüggemann W.; Cubasch U.; Hasselmann K.; Höck H.; Maier-Reimer E.; Wigley T. M. L. Ocean Variability and Its Influence on the Detectability of Greenhouse Warming Signals. J. Geophys. Res. 1995, 100 (C6), 10693. 10.1029/95JC00683. [DOI] [Google Scholar]
  33. Santer B. D.; Po-Chedley S.; Zelinka M. D.; Cvijanovic I.; Bonfils C.; Durack P. J.; Fu Q.; Kiehl J.; Mears C.; Painter J.; Pallotta G.; Solomon S.; Wentz F. J.; Zou C.-Z. Human Influence on the Seasonal Cycle of Tropospheric Temperature. Science 2018, 361 (6399), eaas8806. 10.1126/science.aas8806. [DOI] [PubMed] [Google Scholar]
  34. Santer B. D.; Po-Chedley S.; Zhao L.; Zou C.-Z.; Fu Q.; Solomon S.; Thompson D. W. J.; Mears C.; Taylor K. E. Exceptional Stratospheric Contribution to Human Fingerprints on Atmospheric Temperature. Proc. Natl. Acad. Sci. U.S.A. 2023, 120 (20), e2300758120 10.1073/pnas.2300758120. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Santer B. D.; Fyfe J. C.; Solomon S.; Painter J. F.; Bonfils C.; Pallotta G.; Zelinka M. D. Quantifying Stochastic Uncertainty in Detection Time of Human-Caused Climate Signals. Proc. Natl. Acad. Sci. U.S.A. 2019, 116 (40), 19821–19827. 10.1073/pnas.1904586116. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Bindoff N. L.; Stott P. A.; AchutaRao K. M.; Allen M. R.; Gillett N.; Gutzler D.; Hansingo K.; Hegerl G.; Hu Y.; Jain S.; Mokhov I. I.; Overland J.; Perlwitz J.; Sebbari R.; Zhang X.. Detection and Attribution of Climate Change: From Global to Regional. In Climate Change 2013 – The Physical Science Basis; Intergovernmental Panel On Climate Change, Ed.; Cambridge University Press, 2014; pp 867–952. [Google Scholar]
  37. Danabasoglu G.; Lamarque J.-F.; Bacmeister J.; Bailey D. A.; DuVivier A. K.; Edwards J.; Emmons L. K.; Fasullo J.; Garcia R.; Gettelman A.; Hannay C.; Holland M. M.; Large W. G.; Lauritzen P. H.; Lawrence D. M.; Lenaerts J. T. M.; Lindsay K.; Lipscomb W. H.; Mills M. J.; Neale R.; Oleson K. W.; Otto-Bliesner B.; Phillips A. S.; Sacks W.; Tilmes S.; Kampenhout L.; Vertenstein M.; Bertini A.; Dennis J.; Deser C.; Fischer C.; Fox-Kemper B.; Kay J. E.; Kinnison D.; Kushner P. J.; Larson V. E.; Long M. C.; Mickelson S.; Moore J. K.; Nienhouse E.; Polvani L.; Rasch P. J.; Strand W. G. The Community Earth System Model Version 2 (CESM2). J. Adv. Model. Earth Syst. 2020, 12 (2), e2019MS001916 10.1029/2019MS001916. [DOI] [Google Scholar]
  38. Emmons L. K.; Schwantes R. H.; Orlando J. J.; Tyndall G.; Kinnison D.; Lamarque J.; Marsh D.; Mills M. J.; Tilmes S.; Bardeen C.; Buchholz R. R.; Conley A.; Gettelman A.; Garcia R.; Simpson I.; Blake D. R.; Meinardi S.; Pétron G. The Chemistry Mechanism in the Community Earth System Model Version 2 (CESM2). J. Adv. Model. Earth Syst. 2020, 12 (4), e2019MS001882 10.1029/2019MS001882. [DOI] [Google Scholar]
  39. Gettelman A.; Mills M. J.; Kinnison D. E.; Garcia R. R.; Smith A. K.; Marsh D. R.; Tilmes S.; Vitt F.; Bardeen C. G.; McInerny J.; Liu H.-L.; Solomon S. C.; Polvani L. M.; Emmons L. K.; Lamarque J.-F.; Richter J. H.; Glanville A. S.; Bacmeister J. T.; Phillips A. S.; Neale R. B.; Simpson I. R.; DuVivier A. K.; Hodzic A.; Randel W. J. The Whole Atmosphere Community Climate Model Version 6 (WACCM6). JGR Atmospheres 2019, 124 (23), 12380–12403. 10.1029/2019JD030943. [DOI] [Google Scholar]
  40. Tilmes S.; Hodzic A.; Emmons L. K.; Mills M. J.; Gettelman A.; Kinnison D. E.; Park M.; Lamarque J.-F.; Vitt F.; Shrivastava M.; Campuzano-Jost P.; Jimenez J. L.; Liu X. Climate Forcing and Trends of Organic Aerosols in the Community Earth System Model (CESM2). J. Adv. Model. Earth Syst. 2019, 11 (12), 4323–4351. 10.1029/2019MS001827. [DOI] [Google Scholar]
  41. Hoesly R. M.; Smith S. J.; Feng L.; Klimont Z.; Janssens-Maenhout G.; Pitkanen T.; Seibert J. J.; Vu L.; Andres R. J.; Bolt R. M.; Bond T. C.; Dawidowski L.; Kholod N.; Kurokawa J.; Li M.; Liu L.; Lu Z.; Moura M. C. P.; O’Rourke P. R.; Zhang Q. Historical (1750–2014) Anthropogenic Emissions of Reactive Gases and Aerosols from the Community Emissions Data System (CEDS). Geosci. Model Dev. 2018, 11 (1), 369–408. 10.5194/gmd-11-369-2018. [DOI] [Google Scholar]
  42. van Marle M. J. E.; Kloster S.; Magi B. I.; Marlon J. R.; Daniau A.-L.; Field R. D.; Arneth A.; Forrest M.; Hantson S.; Kehrwald N. M.; Knorr W.; Lasslop G.; Li F.; Mangeon S.; Yue C.; Kaiser J. W.; van der Werf G. R. Historic Global Biomass Burning Emissions for CMIP6 (BB4CMIP) Based on Merging Satellite Observations with Proxies and Fire Models (1750–2015). Geosci. Model Dev. 2017, 10 (9), 3329–3357. 10.5194/gmd-10-3329-2017. [DOI] [Google Scholar]
  43. Kay J. E.; Deser C.; Phillips A.; Mai A.; Hannay C.; Strand G.; Arblaster J. M.; Bates S. C.; Danabasoglu G.; Edwards J.; Holland M.; Kushner P.; Lamarque J.-F.; Lawrence D.; Lindsay K.; Middleton A.; Munoz E.; Neale R.; Oleson K.; Polvani L.; Vertenstein M. The Community Earth System Model (CESM) Large Ensemble Project: A Community Resource for Studying Climate Change in the Presence of Internal Climate Variability. Bulletin of the American Meteorological Society 2015, 96 (8), 1333–1349. 10.1175/BAMS-D-13-00255.1. [DOI] [Google Scholar]
  44. Eyring V.; Bony S.; Meehl G. A.; Senior C. A.; Stevens B.; Stouffer R. J.; Taylor K. E. Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) Experimental Design and Organization. Geosci. Model Dev. 2016, 9 (5), 1937–1958. 10.5194/gmd-9-1937-2016. [DOI] [Google Scholar]
  45. Meinshausen M.; Vogel E.; Nauels A.; Lorbacher K.; Meinshausen N.; Etheridge D. M.; Fraser P. J.; Montzka S. A.; Rayner P. J.; Trudinger C. M.; Krummel P. B.; Beyerle U.; Canadell J. G.; Daniel J. S.; Enting I. G.; Law R. M.; Lunder C. R.; O’Doherty S.; Prinn R. G.; Reimann S.; Rubino M.; Velders G. J. M.; Vollmer M. K.; Wang R. H. J.; Weiss R. Historical Greenhouse Gas Concentrations for Climate Modelling (CMIP6). Geosci. Model Dev. 2017, 10 (5), 2057–2116. 10.5194/gmd-10-2057-2017. [DOI] [Google Scholar]
  46. Cohen Y.; Petetin H.; Thouret V.; Marécal V.; Josse B.; Clark H.; Sauvage B.; Fontaine A.; Athier G.; Blot R.; Boulanger D.; Cousin J.-M.; Nédélec P. Climatology and Long-Term Evolution of Ozone and Carbon Monoxide in the Upper Troposphere–Lower Stratosphere (UTLS) at Northern Midlatitudes, as Seen by IAGOS from 1995 to 2013. Atmos. Chem. Phys. 2018, 18 (8), 5415–5453. 10.5194/acp-18-5415-2018. [DOI] [Google Scholar]
  47. Gaudel A.; Bourgeois I.; Li M.; Chang K.-L.; Ziemke J.; Sauvage B.; Stauffer R. M.; Thompson A. M.; Kollonige D. E.; Smith N.; Hubert D.; Keppens A.; Cuesta J.; Heue K.-P.; Veefkind P.; Aikin K.; Peischl J.; Thompson C. R.; Ryerson T. B.; Frost G. J.; McDonald B. C.; Cooper O. R. Tropical Tropospheric Ozone Distribution and Trends from in Situ and Satellite Data. EGUsphere 2024, 2024, 1–51. 10.5194/egusphere-2023-3095. [DOI] [Google Scholar]
  48. Froidevaux L.; Kinnison D. E.; Gaubert B.; Schwartz M. J.; Livesey N. J.; Read W. G.; Bardeen C. G.; Ziemke J. R.; Fuller R. A.. Tropical Upper Tropospheric Trends in Ozone and Carbon Monoxide (2005–2020): Observational and Model Results, 2024. 10.5194/egusphere-2024-525. [DOI]
  49. Sippel S.; Meinshausen N.; Székely E.; Fischer E.; Pendergrass A. G.; Lehner F.; Knutti R. Robust Detection of Forced Warming in the Presence of Potentially Large Climate Variability. Sci. Adv. 2021, 7 (43), eabh4429. 10.1126/sciadv.abh4429. [DOI] [PMC free article] [PubMed] [Google Scholar]
  50. Deser C.; Phillips A. S.; Simpson I. R.; Rosenbloom N.; Coleman D.; Lehner F.; Pendergrass A. G.; DiNezio P.; Stevenson S. Isolating the Evolving Contributions of Anthropogenic Aerosols and Greenhouse Gases: A New CESM1 Large Ensemble Community Resource. Journal of Climate 2020, 33 (18), 7835–7858. 10.1175/JCLI-D-20-0123.1. [DOI] [Google Scholar]
  51. Huang X.; Zhou T.; Turner A.; Dai A.; Chen X.; Clark R.; Jiang J.; Man W.; Murphy J.; Rostron J.; Wu B.; Zhang L.; Zhang W.; Zou L. The Recent Decline and Recovery of Indian Summer Monsoon Rainfall: Relative Roles of External Forcing and Internal Variability. Journal of Climate 2020, 33 (12), 5035–5060. 10.1175/JCLI-D-19-0833.1. [DOI] [Google Scholar]
  52. Turnock S. T.; Allen R. J.; Andrews M.; Bauer S. E.; Deushi M.; Emmons L.; Good P.; Horowitz L.; John J. G.; Michou M.; Nabat P.; Naik V.; Neubauer D.; O’Connor F. M.; Olivié D.; Oshima N.; Schulz M.; Sellar A.; Shim S.; Takemura T.; Tilmes S.; Tsigaridis K.; Wu T.; Zhang J. Historical and Future Changes in Air Pollutants from CMIP6Models. Atmos. Chem. Phys. 2020, 20 (23), 14547–14579. 10.5194/acp-20-14547-2020. [DOI] [Google Scholar]
  53. Zhang Y.; Cooper O. R.; Gaudel A.; Thompson A. M.; Nédélec P.; Ogino S.-Y.; West J. J. Tropospheric Ozone Change from 1980 to 2010 Dominated by Equatorward Redistribution of Emissions. Nature Geosci 2016, 9 (12), 875–879. 10.1038/ngeo2827. [DOI] [PMC free article] [PubMed] [Google Scholar]
  54. Terrenoire E.; Hauglustaine D. A.; Cohen Y.; Cozic A.; Valorso R.; Lefèvre F.; Matthes S. Impact of Present and Future Aircraft NOx and Aerosol Emissions on Atmospheric Composition and Associated Direct Radiative Forcing of Climate. Atmos. Chem. Phys. 2022, 22 (18), 11987–12023. 10.5194/acp-22-11987-2022. [DOI] [Google Scholar]
  55. Brenna H.; Kutterolf S.; Krüger K. Global Ozone Depletion and Increase of UV Radiation Caused by Pre-Industrial Tropical Volcanic Eruptions. Sci. Rep 2019, 9 (1), 9435. 10.1038/s41598-019-45630-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  56. Harris N. R. P.; Ancellet G.; Bishop L.; Hofmann D. J.; Kerr J. B.; McPeters R. D.; Prendez M.; Randel W. J.; Staehelin J.; Subbaraya B. H.; Volz-Thomas A.; Zawodny J.; Zerefos C. S. Trends in Stratospheric and Free Tropospheric Ozone. J. Geophys. Res. 1997, 102 (D1), 1571–1590. 10.1029/96JD02440. [DOI] [Google Scholar]
  57. Thompson A. M.; Witte J. C.; Hudson R. D.; Guo H.; Herman J. R.; Fujiwara M. Tropical Tropospheric Ozone and Biomass Burning. Science 2001, 291 (5511), 2128–2132. 10.1126/science.291.5511.2128. [DOI] [PubMed] [Google Scholar]
  58. Hawkins E.; Sutton R.. Time of Emergence of Climate Signals Geophys. Res. Lett. 2012, 39( (1), ), 10.1029/2011GL050087. [DOI]
  59. Mahlstein I.; Hegerl G.; Solomon S.. Emerging Local Warming Signals in Observational Data Geophys. Res. Lett. 2012, 39( (21), ), 10.1029/2012GL053952. [DOI]
  60. Dubé K.; Randel W.; Bourassa A.; Degenstein D. Tropopause-Level NOx in the Asian Summer Monsoon. Geophys. Res. Lett. 2022, 49 (18), e2022GL099848 10.1029/2022GL099848. [DOI] [PMC free article] [PubMed] [Google Scholar]
  61. Tomsche L.; Pozzer A.; Ojha N.; Parchatka U.; Lelieveld J.; Fischer H. Upper Tropospheric CH4 and CO Affected by the South Asian Summer Monsoon during the Oxidation Mechanism Observations Mission. Atmos. Chem. Phys. 2019, 19 (3), 1915–1939. 10.5194/acp-19-1915-2019. [DOI] [Google Scholar]
  62. Liu J.; Rodriguez J. M.; Oman L. D.; Douglass A. R.; Olsen M. A.; Hu L. Stratospheric Impact on the Northern Hemisphere Winter and Spring Ozone Interannual Variability in the Troposphere. Atmos. Chem. Phys. 2020, 20 (11), 6417–6433. 10.5194/acp-20-6417-2020. [DOI] [Google Scholar]
  63. Williams R. S.; Hegglin M. I.; Kerridge B. J.; Jöckel P.; Latter B. G.; Plummer D. A. Characterising the Seasonal and Geographical Variability in Tropospheric Ozone, Stratospheric Influence and Recent Changes. Atmos. Chem. Phys. 2019, 19 (6), 3589–3620. 10.5194/acp-19-3589-2019. [DOI] [Google Scholar]
  64. Solomon S.; Ivy D. J.; Kinnison D.; Mills M. J.; Neely R. R.; Schmidt A. Emergence of Healing in the Antarctic Ozone Layer. Science 2016, 353 (6296), 269–274. 10.1126/science.aae0061. [DOI] [PubMed] [Google Scholar]
  65. Verstraeten W. W.; Neu J. L.; Williams J. E.; Bowman K. W.; Worden J. R.; Boersma K. F. Rapid Increases in Tropospheric Ozone Production and Export from China. Nature Geosci 2015, 8 (9), 690–695. 10.1038/ngeo2493. [DOI] [Google Scholar]
  66. Hu K.; Huang G.; Huang P.; Kosaka Y.; Xie S.-P. Intensification of El Niño-Induced Atmospheric Anomalies under Greenhouse Warming. Nat. Geosci. 2021, 14 (6), 377–382. 10.1038/s41561-021-00730-3. [DOI] [Google Scholar]
  67. Kley D.; Crutzen P. J.; Smit H. G. J.; Vömel H.; Oltmans S. J.; Grassl H.; Ramanathan V. Observations of Near-Zero Ozone Concentrations Over the Convective Pacific: Effects on Air Chemistry. Science 1996, 274 (5285), 230–233. 10.1126/science.274.5285.230. [DOI] [Google Scholar]
  68. Newton R.; Vaughan G.; Hintsa E.; Filus M. T.; Pan L. L.; Honomichl S.; Atlas E.; Andrews S. J.; Carpenter L. J. Observations of Ozone-Poor Air in the Tropical Tropopause Layer. Atmos. Chem. Phys. 2018, 18 (7), 5157–5171. 10.5194/acp-18-5157-2018. [DOI] [Google Scholar]
  69. Nussbaumer C. M.; Fischer H.; Lelieveld J.; Pozzer A. What Controls Ozone Sensitivity in the Upper Tropical Troposphere?. Atmos. Chem. Phys. 2023, 23 (19), 12651–12669. 10.5194/acp-23-12651-2023. [DOI] [Google Scholar]
  70. Turner A. J.; Fung I.; Naik V.; Horowitz L. W.; Cohen R. C. Modulation of Hydroxyl Variability by ENSO in the Absence of External Forcing. Proc. Natl. Acad. Sci. U.S.A. 2018, 115 (36), 8931–8936. 10.1073/pnas.1807532115. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

es4c01289_si_001.pdf (5.7MB, pdf)

Articles from Environmental Science & Technology are provided here courtesy of American Chemical Society

RESOURCES