Contribution of Asian emissions to upper tropospheric CO over the remote Pacific

Abstract. By analysing the global distribution of the highest 2% of daily CO mixing ratios at 400hPa derived from the MOPITT satellite instrument for 20 years (2000-2019), we detect very regularly regions with very high CO values (i.e. mixing ratios belonging to the globally highest 2%) over the remote northern hemispheric (NH) Pacific. Such events of elevated CO over the upper tropospheric NH-Pacific occur throughout the year, with a surprisingly high regularity and frequency (70% of all days during winter, 80% respectively during spring). During winter, most of these pollution events are detected over the north5 eastern and central NH-Pacific, during spring over the central NH-Pacific and during summer over the western NH-Pacific. We detect most pollution events during spring. To link each individual pollution event detected by the 2% filtering method with a specific CO source region, we performed trajectory calculations using MPTRAC, a lagrangian transport model. To analyse transport pathways and uplift mechanisms we combine MOPITT data, the trajectory calculations and ERA-Interim reanalysis data. It becomes apparent, that air masses 10 from China being lifted along a frontal system into the free troposphere are the major CO source throughout the year. The contribution of other source regions and uplift mechanisms shows a strong seasonal cycle: NE-Asia in relation with upward transport of air masses in the warm conveyor belt of a midlatitude cyclone is a significant CO-source region during winter, spring and summer while India is an important source region mainly during spring and summer and SE-Asia mainly during spring. 15


Introduction
The long-range transport of trace gases and aerosols from East Asia across the Pacific has been subject of investigation for many years as pollution plumes can be lifted to the free troposphere where they are quickly transported to the North Pacific and North American west coast by midlatitude storm tracks (Yienger et al., 2000;Liang et al., 2004;Holzer et al., 2005;Liang et al., 2005;Wuebbles et al., 2007;Turquety et al., 2008). A lot of effort has been spent in the last years to quantify the contri-20 bution of trace gases and aerosols of Asian origin to North American pollution levels (e.g. Zhang et al., 2008;Yu et al., 2008;Hu et al., 2019;Yu et al., 2019). However, these studies mainly focus on the chemical processing of the pollution plume rather than dynamical transport aspects. Several studies use satellite derived CO data to better understand vertical and horizontal transport mechanisms in relation to the prevailing synoptic situation. Ding et al. (2015) use data from numerical models and the MOPITT satellite instrument to investigate high levels of CO in the upper troposphere over the eastern Pacific while Liu 25 branch of the plume reached the western US while the second branch headed southward towards the tropical western Pacific and no transport towards the Arctic was observed. Liang et al. (2005) and Reidmiller et al. (2010) also find that enhanced transpacific transport is characterised by the combined effects of a strong Pacific High and a strong low over Alaska.
Even though many studies investigate the effect of long-range transport, they are mostly based on case studies like discussed above, use a composite approach or focus on meteorological conditions leading to uplift of polluted air masses into the free 65 troposphere.
Our study presents a detailed analysis of the spatial and temporal distribution of elevated CO level as a pollution tracer in the mid and upper troposphere over the Pacific using 20 years of MOPITT data. We create a climatology of severe pollution episodes and use trajectory calculations to link each particular pollution event detected in MOPITT satellite data with a distinct source region. A second objective is the investigation of different uplift regions and uplift types, in particular WCB-related 70 upward transport. We analyse each trajectory linking a pollution event detected by MOPITT with a CO source region individually. We create a seasonal statistic about different transport pathways and uplift types depending on the location of elevated upper tropospheric CO, its source region, the uplift region and uplift type of polluted air masses. To detect pollution outflow events from the Asian continent, we use thermal infrared level 3 data from the version 8 product of CO measurements derived from the MOPITT instrument (Deeter et al., 2019). Level 3 products are available as daily mean values on a 1x1 • global grid. The Terra satellite carrying the MOPITT instrument is flying in a sun synchronous polar orbit at an altitude of 705km. MOPITT splits the earth in pixels of a size of 22km 2 . By using a cross-tracking scanning method it sees the earth in a swath of about 640 km consisting of 29 pixels. A near-complete global coverage of the measurements would be 80 reached after three days. Pixels with a high cloud content are filtered out. As MOPITT uses gas correlation spectroscopy of the thermal infrared radiation emitted from the earths surface, it can retrieve vertical profiles for almost two independent layers of CO. Data products are available on 10 level with a vertical resolution of 100hPa (surface, 900hPa -100hPa). A retrieval algorithm is applied to the MOPITT data which is based on optimal estimation using a priori information to obtain additional constrains (Deeter et al., 2015). The averaging kernels (see supplement) applied to the in-situ profiles give confidence that the 85 vertical resolution of the MOPITT data in our study region is sufficient to use this dataset to address our scientific objective.
2.2 Detection of long range transport events using MOPITT CO data As we are primarily interested in upper tropospheric pollution episodes, we focus our analysis on the 400hPa data from MOPITT which is chosen as it is reliably in the troposphere throughout the year. The 300hPa level is frequently located above the tropopause (north of about ∼45 • ) during winter (Wilcox et al., 2012) where CO mixing ratios decrease quickly. 90 Note further, that the 400 hPa MOPITT CO has been shown to be insensitive to potential long-term bias drifts (Deeter et al., of Terra) at two selected days and the filtered grid points following the 2% criterion (black crosses). We refer to those data points as CO maxima. Regions with at least three neighbouring maximum-grid points are defined as a CO-maximum cluster and are included in our analysis. It turned out, that throughout the year, these CO maxima are found surprisingly regularly over the remote Pacific, far away from CO source regions and therefore, long-range transport of CO must be responsible for these observations ( Fig. 1(b)).
It is hypothetically possible that on days with strong and widespread biomass burning at any region outside the studying region (i.e. the NH-Pacific), the pollution signal over the NH-Pacific is weak and we underestimate the number of severe pollution events. Though, due to the incomplete global coverage of the satellite data, MOPITT only sees a fraction of the potential biomass burning area (Fig. 1, ( Fig. 3). The seasonal distribution of LRT events is very similar as the one found by Luan and Jaeglé (2013). Since the total number of events selected by our filter is rather large and certainly sufficient to compose a statistics, our final results regarding the trajectory analysis (section 3) would not be impacted severely if we missed single LRT events.

115
CO maxima events represent periods of elevated CO mixing ratios compared to background level. The difference between mean CO mixing ratios considering only CO maxima events and mean CO mixing ratios considering all data is rather high over almost the entire NH-Pacific during all seasons (Fig. 2). This indicates, that by our selection of data points we really capture periods with extraordinary high pollution level in the upper troposphere compared to background level. An exception is clearly visible over the northern Pacific during winter (Fig. 2, DJF) where the difference in CO mixing ratios is rather small 120 and of no statistical significance. Thus we can assume, that the maximum CO-levels (according to our criterion) are within the variability of the climatological mean in this region during this time. They thus rather represent this mean than extreme pollution events. This conclusion can also be drawn for the central NH-Pacific during spring (Fig. 2, MAM). Regions with a difference in CO mixing ratios of low statistical significance are also found during summer east of Japan (Fig. 2, JJA) where many maxima events occur. Though, this region is much closer to potential CO sources than the northern/central NH-Pacific 125 (where background CO level are lower) and it is less surprising that pollution outflow events strongly impact the overall climatological mean at this site.
CO maxima events occur throughout the year over the NH-Pacific with autumn showing an exception with much less events than during the rest of the year (Fig. 3). A seasonal shift of the regions where most CO maxima occur at 400hPa is clearly visible (also at 500hPa and 300hPa). In winter time (Fig. 3, DJF), CO maxima occur more often over the north-eastern NH-

130
Pacific while they are found over the central NH-Pacific during spring (Fig. 3, MAM). The total number of CO maxima events is much higher during spring than in other seasons which is in agreement with other studies showing a maximum of pollution export from Asia during spring (e.g. Luan and Jaeglé, 2013;Holzer et al., 2005;Zhang et al., 2008;Yu et al., 2008;Liang et al., 2004). During summer time (JJA) the occurrence of midlatitude cyclones is generally less frequent than during winter and spring while convection and the Asian summer monsoon gain in importance as transport pathways. are then found much closer to the continent over the western NH-Pacific, Siberia and Kamchatka.
We detect CO maxima over the NH-Pacific on ∼70% of all days during winter and summer, ∼90% during spring and ∼40% of all days during autumn ( Fig. 4(C)). The high temporal frequency of LRT events during spring stays unchanged during the analysed time period while the number of CO maxima days decreases from 2000 until 2019 in winter, summer and autumn. In particular, during winter time the decrease of CO-maxima days is statistically significant. We find CO-maxima on ∼80% of all points are over the Pacific, distributed among two cluster, Fig. 1 (d)).
During the 20 years of daily data, which we analysed, the number of CO maxima grid points slightly increases during spring (slope of regression line: 0.14) and decreases during winter (slope: -0.12, Fig. 4 (A)) while the number of cluster ( Fig. 4 (B)) and the frequency of LRT events ( Fig. 4   deviation of the mean number of CO maxima grid points found in one season is very high (especially for spring). Secondly, we cannot directly conclude, that the spatial coverage of severe pollution episodes increased, though it is likely. Regarding our filtering method, an increase of the number of CO maxima grid points over the Pacific during spring could hypothetically be also the result of e.g.: (i) weaker biomass burning in other regions of the world, (ii) CO emission reductions in other regions of the world, (iii) a faster dilution of CO plumes in other regions of the world, (iv) less clouds over the Pacific or more clouds over 155 other polluted areas or (v) instrumentational reasons. In addition, it is well known that precipitation patterns over the Pacific and the position of the jet stream can be altered during El Niño episodes (Breeden et al., 2021), and thus transport pathways of pollution plumes. Therefore, we considered a correlation between warm phases of ENSO and our selected LRT events. Years with an extraordinary large spatial extension of the upper tropospheric CO maxima cluster during MAM (e.g. 2004MAM (e.g. , 2010MAM (e.g. , 2016MAM (e.g. , 2019 are indeed related with El Niño periods ( Fig. 4 (A)). This result is in agreement with the work by Breeden et al.

160
(2021) who find an increase in storm track activity during La Niña years compared to neutral or El Niño years.

Transport characteristics
One of the questions we address in this study is the potential importance of vertical transport of pollution by extratrocpical cyclones and particularly warm conveyor belts as part of long-range transport events. Closely following Madonna et al. (2014), we define a criterion for WCB related upward transport: trajectories must be located within a two-dimensional surface cyclone 185 field for at least one 6-hourly time step during the ascend phase. i.e. trajectories must cross the cyclone within the time between two time steps prior or after the lift above the 800hPa level (Fig. 8).
This condition does not only ensure that trajectories rise in the vicinity of a cyclone but it also excludes accidental consideration of convectively uplifted air masses (Madonna et al., 2014). We require uplift from below 800hPa to 400hPa (the altitude we mainly analyse, see section 2.2) but we give no time limit for the uplift. Our analysis reveals however, that by far the majority 190 of all trajectories is lifted to 400hPa within 48 hours which is the time frame for WCB-type uplift used in the study by Madonna et al. (2014).
To define the position and size of a cyclone we follow the method of Wernli and Schwierz (2006) Madonna et al. (2014) our algorithm also allows the merging of two cyclones which are very close to each other. The algorithm calculating the horizontal extension of a cyclone tends to underestimate rather than overestimate the size of a cyclone.
In addition to WCB type uplift, we determine trajectories being lifted along a frontal zone. The classification of trajectories belonging to this category follows the procedure described above but trajectories must cross a frontal zone during their ascent.

200
The definition of a front follows in general one of the approaches discussed by Schemm et al. (2017): a frontal zone is identified by enhanced θ e gradients (at least 4K/100km). We slightly modified this criterion by requiring a gradient of at least 3K in one grid box (0.75 • ) in the ERA-INTERIM data set. As discussed by Thomas and Schultz (2018)  detail, it can be assumed that the basic regional lofting will be present in lagrangian trajectory simulations using input from global circulation models. It has to be expected that the mean rate of vertical transport is underestimated.

Case study
As described in section 3.1, we use lagrangian trajectory calculations and ERA-INTERIM reanalysis data to analyse the source regions of upper tropospheric CO detected over the Pacific and related uplift mechanisms. To better illustrate the trajectory 220 analysis, we will discuss in more detail two particular days (03.01.2001, 21.01.2001) with CO maxima being observed over the remote Pacific by MOPITT (Fig. 5). On both days a CO-maximum cluster was found near the Aleutian peninsula (Fig. 5 b,d) and over the central NH-Pacific (Fig. 5 a,c). On 03.01, a third cluster was found over the gulf of Alaska (Fig. 5 e).
Trajectory simulations link all three CO maxima cluster north of 40 • N (Fig. 5 ,b,d,e) mainly with NE-Asia (dark red trajectories) and NE-China (green trajectories) as source regions while the cluster over the central NH-Pacific (Fig. 5 a,c)  trajectory simulations indicate significantly different source regions: On 21.01 (Fig. 5 a, Fig. 6 a), 65% of all trajectories come from China, 15% from NE-Asia and SE-Asia and a small fraction of trajectories come from India (4%). While the CO maximum observed on 03.01 (Fig. 6 d) is mainly fed with CO from SE-Asia (46% of all trajectories) and India (50%) while China 230 seems to play a minor role as source region (3%) on that particular day.
Most of the trajectories reaching the CO maximum over the Aleutian peninsula on 21 January, are lifted into the free troposphere in the vicinity of an eastward moving cyclone (66% of all trajectories, Fig.5 b, Fig.6 c). Fig.7 shows exemplarily the position of the cyclone at 18:00 UTC, 14.01.2001 (a) and 00:00 UTC on 15. 01.2001 (b). Once being lifted into the free troposphere, air masses are transported north-eastward, circulate in the Aleutian low, and finally move further north towards the 235 Arctic or eastward towards the American west coast (not shown). This transport pathway is typical of LRT by fast uplift of air masses in the warm conveyor belt of a midlatitude cyclone over the Pacific with subsequent eastward motion in the free troposphere (Madonna et al., 2014). In contrast, air masses reaching the second cluster on 21 January over the central NH-Pacific ( Fig.5 a) are mainly lifted into the free troposphere over the continent (Fig. 6 b,c) along a front.
Air masses reaching the CO-maximum cluster over the Aleutian peninsula on 03 January do not show such a uniform transport 240 pattern as seen on 21 January. Even though a large fraction of trajectories experience WCB-type uplift (43% of all trajectories, Fig.6 d), the uplift time and uplift position has a stronger temporal and spatial variability compared to 21 January. Air masses need on average 3.3 days to reach the CO maxima cluster after rise into the free troposphere (4.8 days on 21.01.). Parts of the air masses are lifted faster to higher altitudes (∼ 400hPa 6-12 hours after uplift, Fig. 7 e, trajectories at the edge of the cyclone at ∼ 165 • E), compared to 21 January, but transport patterns in the free troposphere are rather diffuse (Fig.7 d,e). On both case 245 study days, a strong Aleutian low has built up. Though its position differs on both days. It is located over the gulf of Alaska on 03 January (Fig.7 f) and over the Aleutian peninsula on 21 January (Fig.7 c).
The CO maxima cluster over the gulf of Alaska (Fig. 5 e) is half composed of air masses which were lifted along a front near CO emission regions (NE-China, NE-Asia) and eastward transport took place in the free troposphere while the other half crossed almost the entire Pacific in the lower most (marine) troposphere. Those air masses are finally lifted into the free troposphere 250 close to the observed CO-maximum cluster in the vicinity of a cyclone. As a consequence it can be assumed that both plume composites experience significantly different chemical processing and mixing during transport.

Statistics of long-term observational data set (2000-2018)
We extended the above presented statistics to all CO maxima cluster detected in our simulation period (03/2000-12/2018) and to all trajectories being calculated for each single CO-maximum cluster. As described in section 3.1, backward trajectories are 255 assigned to a certain source region, if they descend below 850 hPa, are less than 1.5 km above ground and CO emissions exceed a certain threshold. Throughout the year, trajectories reveal China as the dominant source region (Fig. 8 d) of upper tropospheric CO. The contribution of other regions shows a stronger seasonality: during winter and summer almost the same percentage of trajectories come from NE-Asia (33.7% DJF, 27.9% JJA) as from China (36.8% DJF, 29% JJA, Fig. 8 d). During spring however, India becomes an important upper tropospheric CO source region (32% source contribution). During summer the total 260 percentage of trajectories coming from India is smaller than during spring. Though, we find that the individual contribution of India to a single pollution event is stronger during summer than during spring.
The yellow area marked in Fig. 8 a) covering Indonesia, has a source region contribution of less than 1% and is therefore (almost) not visible in the bar charts in Fig. 8.
Both case studies presented above indicate, that pollution plumes reaching a CO-maximum over the NE-Pacific during winter coming from NE-Asia and are predominantly lifted into the free troposphere in the warm conveyor belt of a mid-latitude 270 cyclone over the Pacific. Statistics including all 19 years of trajectory calculations confirm this result: More than 90% of all trajectories from NE-Asia ascend over the ocean during winter (almost 80% during spring) (Fig. 8 b, DJF) and 26% of these trajectories rise by WCB-type uplift (Fig. 8 c, DJF). We found, that 90% of the trajectories, which fit our criteria of vertical transport, are classified as WCB-type.
The fraction of WCB-related upward transport is much smaller for all other source regions throughout the year. This finding 275 is in agreement with the detailed study by Madonna et al. (2014) who show that by far the largest fraction of trajectories experiencing WCB-type uplift over the Pacific come from NE-Asia especially during winter. This can be explained with the position of a semi permanent low pressure system over Japan during winter time (Liang et al., 2005;Madonna et al., 2014).
For trajectories from all other source regions, uplift along a front line is the dominant uplift process. The uplift region shows a strong seasonality for all source regions. Only trajectories coming from China are predominantly lifted over land throughout 280 the year (more than 60% with a rather small standard deviation, Fig. 8 b).  The time between rise into the free troposphere and the arrival of air masses at the CO cluster is longer for trajectories coming from China (3.7 days DJF, 6.6 days JJA), compared to NE-Asia (2.8 days DJF, 5.7 days JJA) corresponding to different uplift regions (land/sea). The residence time in the free troposphere is longer in summer than in winter for all source regions even though CO maxima are located much closer to the continent during summer compared to winter. It is not surprising that 285 trajectories coming from India have the longest residence time in the free troposphere before reaching the CO maxima site (6.5 days DJF, 9 days JJA).

Conclusions
By analysing the global distribution of the highest 2% of daily CO mixing ratios at 400hPa derived from MOPITT for 20 years, we (i) detect very regularly regions with very high CO mixing ratios (i.e. belonging to the globally highest 2% values) over 290 the remote NH-Pacific. Based on an extended analysis of lagrangian trajectories in combination with ERA-INTERIM data, we (ii) conclude that long range transport events from east Asia are responsible for the observed CO maxima. We (iii) can link the majority of CO maxima not only to a specific source region but also with a distinct transport and uplift mechanism.
In agreement with the findings by Luan and Jaeglé (2013); Liang et al. (2004Liang et al. ( , 2005Liang et al. ( , 2007 we see a strong seasonality of pollution events over the Pacific with most CO maxima being observed during spring, followed by winter and summer. Our 295 results further extend the conclusion of these authors since we analysed the effect of LRT and vertical transport on the composition of the upper troposphere region. Notably, we see a spatial shift of the occurrence of the CO maxima related to different transport pathways with most events being detected over the north-eastern Pacific during winter, over the central NH-Pacific during spring and over the western NH-Pacific during summer.
The contribution of different CO emission regions to the observed upper tropospheric CO plumes change among the seasons.

300
Though China is the major CO source region throughout the year. In relation with uplift in the warm conveyor belt of a midlatitude cyclone, NE-Asia is the second most important source region for upper tropospheric CO observed over the north-eastern Pacific in particular during winter.
In general, we find a correlation between the occurrence of high levels of CO in the upper troposphere over the remote Pacific (e.g. over the Aleutian peninsula/north-eastern NH-Pacific during winter) and the synoptic situation. We identify common up-305 lift mechanisms for trajectories having a common source region and a common start region (thus the region where CO maxima are found). Though, we find strong differences when comparing in detail individual pollution events with each other (regarding: location of CO maxima, location, strength and development of the Aleutian low, transport time from a CO-source region over the continent to the observation location of elevated CO over the Pacific, location of uplift from the boundary layer to the upper troposphere, position and strength of cyclones leading to uplift, residence time in the boundary layer/free troposphere during 310 transport). Even though we describe transport patterns for one of our case studies (21.01.2001, CO maxima cluster over the Aleutian Peninsula) as rather uniform, the time of uplift of air masses in the vicinity of a midlatitude cyclone into the free troposphere differs by more than 24 hours for single trajectories, corresponding to hundreds of kilometres in distance. This is due to the fact, that the relevant synoptic systems may exist for several hours or days thereby travelling over large distances which results in a broad temporal and spatial range between source emission and vertical uplift of potential polluted air masses.

315
The observations of CO maxima at 400hPa confirm the crucial role of east Asian emissions for the pollution of the lower UTLS. Notably this also holds for non-monsoon seasons. Our analysis presents evidence for a surprisingly regular and highly frequent occurrence of these long-range transport events. Taking CO as a general marker of pollution and given the regularity of the transport processes, our study highlights the global role of the region also for other chemical constituents in the upper troposphere.