Jet Stream-surface tracer relationships: Mechanism and sensitivity to source region

,


Introduction and Motivation
Concentrations of near-surface air pollutants and greenhouse gases exhibit large day-to-day variations, driven by a combination of variations in emissions, chemistry, and transport.Understanding the cause of this variability is paramount for interpreting measurements and trends in pollutants (e.g., Cooper et al., 2014;Dawson et al., 2014;Kerr et al., 2019) and greenhouse gases (e.g., Keppel-Aleks et al., 2011;Miller et al., 2013Miller et al., , 2015;;Randazzo et al., 2020).
Several studies have highlighted the importance of transport in explaining the daily variability of near-surface composition.For example, daily variations of ozone (O 3 ) have been linked to transport-related phenomena such as horizontal and vertical advection and frontal systems (Jacob et al., 1993;Kerr et al., 2019;Porter & Heald, 2019;Kerr et al., 2020), while Keppel-Aleks et al. (2011) and Torres et al. (2019) have shown that the variability of carbon dioxide (CO 2 ) attributed to the prevailing synoptic-and mesoscale weather is of similar magnitude to the variability from local diurnal fluxes.Moreover, variations in the meridional, or north-south, position of the upper-level jet stream and its e↵ect on transient atmospheric eddies and frontal zones have been linked to variability in near-surface particulate matter (Ordóñez et al., 2019), CO 2 (Randazzo et al., 2020;Pal et al., 2020), methane (Guha et al., 2018), and O 3 (Barnes & Fiore, 2013;Shen et al., 2015;Kerr et al., 2020).
A recent study by Kerr et al. (2020) provided further support for a link between variability in the upper-level jet and surface-level O 3 but also showed substantial spa-tial variations in the relationship.They showed that the daily variability in surface-level O 3 during boreal summer (JJA) is significantly correlated with the meridional position of the jet across the Northern Hemisphere mid-latitudes, but the sign of the relationship di↵ered between land and ocean (with O 3 increasing over land but decreasing over the oceans when the jet is in a poleward position).Furthermore, the jet-O 3 relationship is weak or non-existent at high and low latitudes.
The findings from the aforementioned studies raise several important questions: What mechanisms connect flow aloft to near-surface composition and variability?Why does the jet-O 3 relationship vary with latitude and between land and ocean?How do species' lifetimes and source regions a↵ect the relationship?The last question is important when considering the jet's role in the variability of greenhouse gases and surface-level particular matter whose lifetimes and source regions di↵er.Increases in anthropogenic greenhouse gas emissions will likely shift the mean jet latitude poleward and modulate jet speed later in the twenty-first century (Barnes & Polvani, 2013).These projected changes warrant an improved understanding of how flow aloft impacts near-surface composition, which could improve our projections of how future pollutant distributions could change.
We address these questions by performing chemical transport model (CTM) simulations of a suite of idealized tracers with di↵ering source regions.The simulations enable us to examine how the Northern Hemisphere tracer-jet relationships vary with source region and under what condition(s) there are land-ocean or seasonal variations.Idealized tracers can aid in understanding and interpreting the impact of the jet stream on near-surface composition while avoiding the complex interplay of non-linear gas-and particlephase chemistry and temporally-and spatially-varying precursor emissions (e.g.Orbe et al., 2016).
In Section 2, we describe the CTM simulations, reanalysis, and methodology used in this study.We document the relationship of the tracers with the jet in Section 3.1 and the impact of the jet on near-surface meridional wind in Section 3.2.We find simple balances that relate the connection of the jet stream with near-surface meridional wind to the meridional tracer gradient give a satisfying physical explanation to di↵erences in the sign of the tracer-jet relationships (Sections 3.2-4).

Data and Methodology
We use the GEOS-Chem CTM (version 12.0.2) to perform our tracer simulations (Bey et al., 2001; The International GEOS-Chem User Community, 2018, October 10).GEOS-Chem is driven by assimilated meteorology from the Modern Era-Retrospective Analysis for Research and Analysis, Version 2 (MERRA-2).Three-dimensional MERRA-2 fields are input to the CTM every three hours, while surface quantities and mixing depths are provided every hour.Specifically, our configuration of GEOS-Chem follows a passive simulation described in Liu et al. (2001).We perform this simulation at a resolution of 2 latitude x 2.5 longitude with 72 vertical levels (⇠ 15 hPa spacing below 800 hPa) for 2007 2010, and we discard the first year (2007) for spin up.
Previous studies have demonstrated the accuracy of transport in GEOS-Chem and the assimilated meteorological product, MERRA-2, driving the CTM.Bosilovich et al. (2015) showed that magnitude of MERRA-2 zonal and meridional wind fields as well as the location of wind maxima are well-constrained by observations and other reanalyses.GEOS-Chem yields realistic mixing ratios and seasonal and latitudinal variations of other tracers such as lead and beryllium with no significant global bias (Liu et al., 2001).However, Yu et al. (2018) recently pointed out that the use of o✏ine CTMs, such as GEOS-Chem, together with an archived assimilated meteorological product can lead to vertical transport errors due, in part, to loss of transient advection (resolved convection).While potential biases and errors are important to keep in mind, the extensive body of liter- ature on the reliability of GEOS-Chem supports its suitability as the framework to address our research questions.
Within GEOS-Chem, we implement a suite of nine passive tracers that di↵ers only in their source regions, which are prescribed as constant flux boundary conditions (i.e., emissions) in zonally-symmetric 10 latitudinal bands.Tracers are herein denoted 1 2 , where 1 is the latitude corresponding to the southern boundary of the source region and 2 is the northern boundary.All tracers decay uniformly at a loss rate of ⌧ = 50 days 1 .
Tracers with the same loss have been used in prior studies (e.g., Shindell et al., 2008;Orbe et al., 2017Orbe et al., , 2018;;Yang et al., 2019).Although not the primary focus of our analysis, we also explore how the lifetime of tracers impacts their relationship with the jet by simulating 40 50 with loss rates of ⌧ = 5, 25, 100 and 150 days 1 .Unless indicated, all analyses use daily mean near-surface (1000 800 hPa) tracer mixing ratios.
In addition to driving the GEOS-Chem simulations, we use MERRA-2 to characterize the meteorology responsible for tracer variability (McCarty et al., 2016;Gelaro et al., 2017).MERRA-2 is output on a global 0.5 x 0.625 grid with 72 vertical levels.Specifically, we obtain 3-hourly 1000 800 hPa meridional wind (V ) and 500 hPa zonal wind (U ) from MERRA-2 and average these data to daily mean values, consistent with our treatment of tracers from GEOS-Chem.The horizontal resolution di↵ers between GEOS-Chem and MERRA-2, and we degrade the resolution of MERRA-2 to match that of GEOS-Chem using xESMF, a universal regridder for geospatial data (Zhuang et al., 2020).
We locate the latitudinal position of the jet stream ( jet ) daily at each longitude by finding the latitude (restricted to 20 70 N) of maximum 500 hPa U .A simple convolutionbased smoothing is applied in longitudinal space to address potential longitudinal discontinuities in the jet's position (i.e., "jumps" in jet ) using a box-shaped function with a width of ⇠ 10 longitude.Identifying jet using 500 hPa winds follows previous work by Barnes and Fiore (2013) and Kerr et al. (2020).
The temporal correlation between jet and near-surface tracer mixing ratios or V is quantified with the Pearson product-moment correlation coe cient, indicated by r(X, Y ), where X and Y are the time series of interest.We assess the significance of the correlation coe cient using the non-parametric moving block bootstrapping method, which preserves much of the temporal correlation in the time series and makes no a priori assumptions about the time series' distributions.In essence, time series X and Y are randomly reordered by sampling continuous blocks of data with length = 10 days, and r(X, Y ) is thereafter recalculated.We conduct 10000 realizations of this reordering, and significance is determined with a two-tailed percentile confidence interval method at the 0.05 significance level (Wilks, 1997;Mudelsee, 2003;Wilks, 2011).
We also generate composites of tracer mixing ratios and V on days when the jet stream is poleward (PW) and equatorward (EW).The PW (EW) composite is defined locally (i.e., at each longitude) as the average value of the field of interest for days where jet exceeds (is less than) the 70th (30th) percentile.We define a "positive" relationship to mean that the PW (EW) movement of the jet is associated with increased (decreased) mixing ratios or V .The opposite is true for a "negative" tracer-jet relationship.

Relationship between the jet stream and tracers
Before we examine the tracers' relationship with jet we briefly discuss the mean tracer distributions and their daily variability.Zonally-averaged tracer mixing ratios peak within their source regions and diminish to roughly half of their maximum value ±5 outside their source regions (Figure 1a).Tracers with source regions at latitudes ( ) north of 60 N have higher mixing ratios within their source regions compared with tracers emitted at lower latitudes (Figure 1a), supporting an isolated Arctic lower troposphere and the "polar dome" as a barrier to transport (Law & Stohl, 2007).
Despite zonally-symmetric emissions, there are zonal variations in tracer mixing ratios (Figure 1b-d).The latitudinal range with high tracer mixing ratios (> 0.8 ppm) is larger over the ocean basins for tracers with high and mid-latitude sources (e.g., 70 80 , 40 50 ; Figure 1b-c).These ocean regions coincide with the Atlantic and Pacific storm tracks.High mixing ratios of tracers with source regions in the tropics (e.g., 10 20 ) are more di↵use over land and more restricted over the tropical ocean (Figure 1d).Spatial variations in the tracers' daily variability (as measured by the standard deviation) are similar to spatial variations in their mean distribution, with highest variability near the tracer source region and decreasing to the north and south (not shown).
Furthermore, the ratio of each tracer's standard deviation to its mean is ⇠ 50% near the source region and diminishes to ⇠ 20% well outside the source region (not shown).
To assess the impact of the meridional movement of the jet on daily tracer variability, we examine composites of tracer mixing ratios when the jet is PW and EW (see Section 2).As is shown in Figure 2, there is a significant tracer-jet relationship for all tracers during JJA and boreal winter (DJF) within the mid-latitudinal range over which the jet traverses.However, the sign of the relationship hinges on the meridional gradients of the tracers (@ /@ ).Tracers with source regions at low latitudes ( < 40 N) have a negative gradient (@ /@ < 0) within the latitudinal range of the jet and increase in the mid-latitudes when the jet is PW (Figure 2a itudes ( > 60 N) are characterized by @ /@ > 0 in the mid-latitudes and decrease in the mid-latitudes when the jet is PW (Figure 2a-b).
Beyond the mid-latitudes and these three tracers, impact of source region on the tracer-jet relationships for all the GEOS-Chem tracers can be easily seen in the zonal mean (Figure 3a-b).The tracer-jet relationships all exhibit an oscillatory pattern, but tracers with source regions south of the range of the jet are positively correlated with the jet in the mid-latitudes and are flanked by negative correlations (although generally not significant) outside the mid-latitudes.Tracers with source regions north of the jet have a negative correlation with the jet in the mid-latitudes and a positive, but non-statistically significant, correlation outside the mid-latitudes (Figure 3a-b).
The variations in tracer mixing ratios related to the meridional oscillations of the jet are a sizable fraction of the overall daily tracer variability discussed earlier in this section.For example, the ratios of the jet-associated variations in 10 20 , 40 50 , and 70 80 to the overall variability (standard deviation) zonally-averaged over the mid-latitudes (40 < < 60 N) are 58%, 35%, and 47%, respectively.
In a gross sense, the relationship between the jet stream and our tracers does not change in DJF compared to JJA, but further inspection indicates that there are nuanced di↵erences in the relationships (Figure 2).For example, the change in mid-latitude mixing ratios of 40 50 due to the meridional movement of the jet is varied in sign and strength during JJA, while the DJF change is largely negative (Figure 3b-c).
The jet is an important source of variability for near-surface trace species spanning a range of lifetimes.The relationship of the jet with 40 50 for loss rates spanning 5 to 150 days 1 is similar in sign and significance to 40 50 with the 50 days 1 loss rate discussed elsewhere in this study, although the precise magnitude of the variability associated with the jet changes with tracer lifetime (Figure S3).We note that, although these findings hold for tracers with zonally-symmetric emissions, tracers with more realistic emissions (e.g., land-ocean contrasts, urban-rural di↵erences) may have a more complicated relationship with the jet.With that said, the results presented here indicate a strong relationship between trace gas variability and the jet absent these other confounding factors.

Mechanisms
The analysis presented in Section 3.1 has shown that a large fraction of daily tracer variability is related to meridional movement of the jet but does not show the mechanism(s) involved or why the signs of the tracer-jet relationships varies.Kerr et al. (2020) suggested that the jet stream a↵ects surface-level O 3 by altering the near-surface meridional flow (V ).We test this hypothesis using our suite of tracers.We first examine the V -jet relationship and then how this impacts the tracers.
Figure 3c indicates that southerly flow increases in the mid-latitudes (around the latitudinal range of the jet stream) when the jet is PW during JJA and DJF; however, it does not show the magnitude.As is shown in Figure 4a-b, V increases over 5 m/s in parts of the mid-latitudes when the jet is PW.This stands in sharp contrast to time-averaged V , which is generally weak ( 2 < V < 2 m/s) over the vast majority of the mid-latitudes.
It is exceedingly rare for time-averaged V to have the same magnitude changes in V linked to the jet (contours in Figure 3a-b).Outside the mid-latitudes, the relationship between V and jet is largely non-significant and weak (Figures 3c, 4a-b).
The V -jet relationship is not zonally-symmetric (Figure 4a-b).For example, the JJA V -jet relationship is negative over the mid-latitude oceans on the windward shores of the continents but is positive over the mid-latitude continents and the leeward shores (Figure 4a).The spatial extent of regions with a positive V -jet relationship increases in DJF compared to JJA, and the jet has a significant positive relationship with near-surface V over a majority of the Pacific and Atlantic Ocean basins in DJF (Figure 4a-b).
In the zonal mean, the latitudes, or nodes, where r( , jet ) = 0 are well-aligned with the latitudes where the jet stream and V are not correlated (Figure 3).This result is especially clear for tracers with northern source regions, while tracers with source regions in the mid-latitudes (e.g., 30 40 , 40 50 ) are slightly o↵set from the latitudes where r(V, jet ) = 0.The only node where r(V, jet ) = 0 does not coincide with r( , jet ) = 0 occurs during DJF north of the jet (Figure 3b).In this case, the latitude where r(V, jet ) = 0 lies north of r( , jet ) = 0 by ⇠ 5 , and other processes such as changes in zonal winds or convection could be important for the tracer-jet relationships in this region and season.These results support Kerr et al. (2020) and provide strong evidence linking the tracerjet relationships to (1) the source region of the tracers and (2) the V -jet relationship (Fig- ure 3).
Thus, the impact of a given change in V is expected to depend on the local tracer gradients.If @ /@ is weak, then smaller tracer changes are expected compared with locations with stronger @ /@ .It also follows that the same change in V operating over @ /@ < 0 versus @ /@ > 0 would result in changes of tracer mixing ratios with different signs.Given this, we postulate that the expected sign of the tracer-jet relationships (sgn[r( , jet )]) shown in Figures 2-3 can be approximated by: In practice, this balance implies that the anomalous southerly flow in the mid-latitudes that accompanies a PW-shifted jet (r(V, jet ) > 0) will advect higher tracer mixing ratios from lower latitudes if @ /@ < 0, yielding a positive expected tracer-jet relationship (i.e., sgn[r( , jet )] > 0).
The simple balance in Equation 1 robustly captures the large-scale di↵erences in the sign of the relationship between the jet and all tracers.We illustrate this for 40 50 in Figure 4c-d.The application of Equation 1 can explain the widespread negative 40 50jet relationship in mid-latitudes during DJF (Figure 4d) but also the di↵erences in sign on much smaller spatial scales during JJA (Figure 4c).Moreover, we note that Equa-  Our results help to shed light on whether the variability of an observed near-surface trace species is due to transport.If an observational site does not lie within the seasonal range of the jet, jet-driven variability is likely not a major factor.However, if the observational site is within the seasonally-dependent range of the jet, it is likely that there will be transport-driven variability with the sign of the jet-trace species relationship dependent on the sign of the meridional background gradient and the local V -jet relationship.
We show that the mechanism that connects the upper-level jet to variability nearsurface composition is changes in near-surface meridional flow that result from the meridional movement of the jet stream.This mechanism explains (1) the variation in sign of the jet-tracer relationship with tracer meridional gradients, (2) the land-ocean di↵erences in the jet-tracer relationship for tracers with mid-latitude sources, and (3) seasonal differences in the jet-tracer relationship.Furthermore, this mechanism explains both the latitudinal and land-ocean di↵erences in the JJA jet-O 3 relationship reported in Kerr et al. (2020) and also helps explain seasonality in the jet-O 3 relationship.Although not shown in Kerr et al. (2020), the sign of the jet-O 3 relationship over North America and Eurasia changes from positive during JJA to negative in DJF, which is broadly consistent with 40 50 .
The jet-tracer relationships found in our simulations hold for a range of tracer lifetimes (5 to 150 days; Figure S3) and for mass-weighted tropospheric column mixing ratios.Thus, our results may be useful for interpreting variations in a host of species, including the total column measurements commonplace among satellite products.Contemporaneous studies have found that variations in meteorology can explain a substantial portion of total column observations of greenhouse gases, comparable to the impact of regional variations in surface fluxes (e.g., Keppel-Aleks et al., 2011).Di↵erentiating whether patterns in satellite observations are due to transport versus variations in surface fluxes may help explain di↵erences in trace gas distributions due to large-scale transport.Future studies should test this possibility.
Our study has documented a major driver of near-surface composition variability (i.e., transport associated with the jet stream) and linked this driver with the location of emissions.This finding is relevant for understanding possible future changes of tracer variability, as models predict that the jet stream will migrate north (e.g., Barnes & Polvani, 2013), which will modify the poleward transport of air pollution and greenhouse gases via its regulation of the near-surface meridional flow.Recently there has been a redistribution of anthropogenic emissions from the mid-latitudes (developed nations) to low latitudes (developing nations) (Zhang et al., 2016), which may change meridional tracer gradients and the daily variations connected to the jet.Further research is needed to quantify (1) how seasonal variations and non-uniform chemical loss of tracers a↵ect their relationship with the jet and (2) the impact of changes in the the position of the jet and the source region of emissions on the variability of near-surface trace species.

Figure 1 .
Figure 1.(a) Zonally-averaged tracer mixing ratios in JJA.(b) JJA-averaged mixing ratios of (b) 70 80, (c) 40 50, and (d) 10 20.Scatter points and vertical bars in (b)-(d) represent the mean position and variability of the jet stream in JJA, respectively.Note that the thicker lines in (a) correspond to the tracers featured in (b)-(d).

Figure 2 .
Figure 2. The di↵erence in composites of JJA (a) 70 80, (c) 40 50, and (e) 10 20 for days with a PW versus EW jet stream.Hatching denotes tracer-jet correlations that are not statistically significant.Scatter points and vertical bars represent the mean position and variability of the jet stream in JJA, respectively.(b), (d), and (f) are the same as (a), (c), and (f) but for DJF.

Figure 3 .
Figure 3.An illustration of how jet impacts near-surface V and tracers.(a) The JJA zonally-averaged correlation between jet and individual tracers (colors) and the mean position and and variability of the jet stream (scatter point and horizontal bars).(b) same as (a) but for DJF.(c) Zonally-averaged r(V, jet).Dashed vertical lines in (a)-(b) denote the latitudes where r(V, jet) = 0 for each season.Dashed horizontal lines separate positive from negative correlations.
tion 1 captures the land-ocean contrasts present in the JJA 40 50 -jet relationship (Fig-ure 4c).The application of Equation 1 does not capture the sign of the 40 50 -jet relationship in the vicinity of the Atlantic and Pacific storm tracks (Figure4c-d), and this is the case for other tracers as well (not shown).Since our tracer mixing ratios are roughly zonallysymmetric (Figure1b-d), the e↵ect of changes in U are negligible to first order.However, the jet stream exerts an influence on near-surface U(Woollings et al., 2010), especially near the exit region of the these storm tracks.To account for this, future studies could consider the impact of both the V -jet and U -jet relationships.The zonal variations in the tracer-jet relationships previously discussed could stem from zonal variations in the response of V to the movement of the jet or zonal variations in the tracer gradients.To explore this, we have isolated the terms in Equation 1 by separately fixing each to its zonal mean value and thereafter recalculating sgn[r( , jet )] to gauge which exerts a stronger influence on the tracer-jet relationships (not shown).Recalculating Equation 1 with @ /@ fixed to its zonal mean value and r(V, jet ) varying as in Figure4a-b yields expected tracer-jet relationships with zonal variations that resemble the relationships shown in Figure4c-d.This sensitivity test together with the analysis performed in Figure4c-d confirm spatiotemporal variations in the V -jet relationship are the most important factor in explaining the tracer-jet coupling, followed by the latitudinal tracer gradient.The importance of the jet stream and meridional flow on daily tracer variability is not restricted to only near-surface mixing ratios but holds for tropospheric column abundances.To support this, we repeat the analyses shown in Figures3-4but with V and mass-weighted tracer mixing ratios from 1000 200 hPa (FiguresS1-S2) to show that the V -jet relationship not only explains variations in near-surface mixing ratios but also in tropospheric column tracer mixing ratios.

Figure 4 .
Figure 4. (a-b) Di↵erences in composites of V for days with a PW versus EW jet stream (colors).Time-averaged V is illustrated for 5 m/s (solid black contour) and 5 m/s (dashed black contour).Hatching denotes statistically non-significant V -jet correlations.(c-d) The correlation coe cient calculated between 40 50 and jet (colors).As denoted in the legend beneath (c), stippling and hatching show the expected sign of the correlation, E[r( 40 50, jet)], determined using Equation 1. Scatter points and vertical bars in all subplots represent the mean position of and variability of the jet stream, respectively.