Oceanic moisture sources contributing to wintertime Euro-Atlantic blocking

Although conventionally attributed to dry dynamics, increasing evidence points to a key role of moist dynamics in the formation and maintenance of blocking events. The source of moisture crucial for these processes, however, remains elusive. In this study, we identify the moisture sources responsible for latent heating associated with the wintertime Euro-Atlantic blocking events detected over 31 years (1979-2010). To this end, we track atmospheric particles backward in time from the blocking centres for a period of 10 days, using an offline Lagrangian dispersion model applied to atmospheric reanalysis 5 data. The analysis reveals that 28 55% of particles gain heat and moisture from the ocean over the course of 10 days, with higher percentages for the lower altitudes. Via large-scale ascent, these moist particles transport low potential vorticity (PV) air of low-altitude, low-latitude origins into the upper troposphere where the amplitude of blocking is the most prominent, in agreement with previous studies. The PV of these moist particles remains significantly lower compared to their dry counterparts throughout the course of 10 days, preferentially constituting blocking cores. 10 Further analysis reveals that approximately two thirds of the moist particles source their moisture locally from the Atlantic, while the remaining one third sources it from the Pacific. There is also a small fraction of moist particles that takes up moisture from both the Pacific and Atlantic basins, which undergoes a large-scale uplift over the Atlantic using moisture picked up over both basins. The Gulf Stream and Kuroshio and their extensions, as well as the eastern Pacific northeast of Hawaii, not only provide heat and moisture to moist particles but also act as “springboards" for their large-scale, cross-isentropic ascent, 15 where its extent strongly depends on the humidity content at the time of the ascent. While the particles of Atlantic origin swiftly ascend just before their arrival at blocking, those of Pacific origin begin their ascent a few days earlier, after which they carry low PV air in the upper troposphere while undergoing radiative cooling just as dry particles. A previous study identified a blocking maintenance mechanism, whereby low PV air is selectively absorbed into blocking systems to prolong blocking lifetime. As they used an isentropic trajectory analysis, this mechanism was regarded as a dry process. We found that these 20 moist particles that are fuelled over the Pacific can also act to maintain blocks in the same manner, revealing that what appears to be a blocking maintenance mechanism governed by dry dynamics alone can, in fact, be of moist origin.

blocking to the imposed sea surface temperature (SST) and humidity over the North Atlantic basin in a high-resolution model, such that higher SST and humidity content promote larger-scale block formation. In line with this result, Scaife et al. (2011) found that a reduced mean Atlantic SST bias considerably improves blocking climatology over the Atlantic in a coupled climate model, although this improvement may be model dependent (Davini and D'Andrea, 2016). A recent study by Kwon et al. (2020) further corroborated the role of SST by showing that the multidecadal North Atlantic SST variability modulates the 60 frequency of wintertime European blocking, whose covariability with the North Atlantic SST was first identified by Häkkinen et al. (2011). Kwon et al. (2020) suggested that this modulation of blocking is achieved through the mechanism discussed in Novak et al. (2015), in which poleward transient eddy heat flux in the lower troposphere is regulated by the underlying sharp SST gradients, resulting in the shift of the eddy-driven jet and a change in the dominant type of Rossby wave breaking.
Other studies have attributed the oceanic influence on blocking specifically to the western boundary currents. Western bound- Using FLEXPART, we release 20 particles randomly placed in every 2.5 degree by 2.5 degree horizontal grid cell where and when Euro-Atlantic blocks have been identified. Assuming that blocking has an equivalent barotropic structure, the particles are released from altitudes between 7,000 and 8,000 m above the sea level (asl; corresponding to approximately 350 hPa 120 on average), given that the amplitude of blocking anomalies maximizes in the upper troposphere (Dole and Gordon, 1983;Schwierz et al., 2004). As such, we have released a total of over five million particles from the interior of the DJF Euro-Atlantic blocking events over 31 years. After their release, each of these particles is tracked backward in time for a duration of 10 days. Multiple atmospheric variables, including temperature, specific humidity, and air density, are interpolated to the particle's position at each time step, while additional meteorological quantities useful for the subsequent analyses, such as PV 125 and PBL height, are computed in FLEXPART along each trajectory. An additional criterion has been applied such that PV along the particles must remain below 2 PVU (potential vorticity unit; 1 PVU ≡ 1.0 × 10 −6 m 2 s −1 K kg −1 ), in order to exclude those particles that stay in the stratosphere.
Moreover, we have also released particles from higher altitudes between 9,500 and 10,500 m asl, approximately corresponding to 250 hPa, and from lower altitudes between 4,500 and 5,500 m asl as well we between 3,000 and 4,000 m asl, 130 approximately corresponding to 500 hPa and 700 hPa, respectively. To test how the blocking condition differs from the time when there is no block, we have also released 20 particles from every grid cell within the Atlantic domain defined in Section 2.2 when there is no blocking found within ±10 days from 7,500 m asl, which we refer to as non-blocking particles. We have applied a chi-squared independence test in order to compute the statistical significance of the difference between these non-blocking particles and blocking particles. 135 Furthermore, following Yamamoto et al. (2015) and Yamamoto and Palter (2016), whenever the particles reside within the atmospheric PBL over the ocean, we interpolate surface heat flux (SHF) and latent heat flux (LHF) to the particle's 6-hourly location and time. These turbulent heat fluxes are useful in evaluating where the particles that eventually reach the Euro-Atlantic blocks are likely to be supplied with heat and moisture from the ocean. Applying the same technique, Yamamoto et al. (2015) showed that up to 88% of the along-trajectory potential temperature variability of the particles released from western Europe 140 can be explained alone by the accumulated turbulent heat fluxes along the trajectories, indicating that this method adequately captures along-trajectory diabatic changes.

Distinct characteristics of moist and dry particles
In order to elucidate the role of the ocean in providing moisture to the atmospheric particles en route to the Euro-Atlantic blocks, we first partition the individual particles into moist and dry particles. Here, moist particles are defined as those that 145 have been subject to evaporation from the ocean, marked with positive LHF, for at least one time step over the course of 10 days. Otherwise, the particles are regarded as dry particles. As our goal lies in identifying the oceanic sources of moisture for the blocking particles, we opt for this definition based on moisture gain from the ocean over that used in Pfahl et al. (2015) and Steinfeld and Pfahl (2019), who defined diabatic particles as those that are heated by more than 2 K within three days and seven days prior to their arrival in the blocking regions.

150
With our definition, we have identified 41.3% of the total particles as moist particles and 58.7% as dry particles. This fraction of moist particles in our study falls within the range of diabatic particles of 30 -45% that gain more than 2 K in the last 3 days before their arrival in the blocking system as defined by Pfahl et al. (2015). The fraction is underestimated relative to their fraction of 60 -70% of particles that gain more than 2 K in the 7 days before their arrival in the blocking system (Pfahl et al., 2015). When we apply the same definition as Pfahl et al. (2015), we find that 51.8% and 69.5% of particles are identified as 155 diabatic particles which undergo diabatic processes in the last 3 days and 7 days, respectively. We speculate that the difference between Pfahl et al. (2015)'s results and ours might arise from the differences in the blocking definitions, Lagrangian tracking methods, and/or reanalyses used in two studies. Meanwhile, the fraction reported by the previous studies in the maintenance stage seems fairly consistent with the current study (see Figure 6 of Steinfeld and Pfahl, 2019 for the non-blocking particles is 37.1%, which is slightly lower than the blocking case. We have found that this difference is 160 significant at the 99% confidence level, indicating that blocking particles preferentially contain more moist particles compared to non-blocking conditions. When we release particles from lower altitudes, the fraction of moist particles increases to 49.9% and 55.0% for the 5,000 m asl and 3,500 m asl release, respectively, while the fraction drops to 27.8% for the particles released from 10,000 m asl.

165
The spatial distributions of moist and dry particles 2, 5, and 9 days prior to their arrival at the blocking locations are shown in Figure 2, where the particle release locations (i.e. where blocking occurs, corresponding to day 0) are indicated by red contours.
As shown in Figure 2, the particles in both categories are mostly advected eastward by the prevailing westerlies. Compared to their moist counterpart, dry particles tend to travel slightly faster, reflecting their tendency to be situated at higher altitudes (see Section 3.3), where the background westerlies tend to be stronger.

170
Upon further inspection of Figure 2, one notices that moist particles 9 days prior (upper-right panel) tend to be situated within two distant regions separated by the Rockies, seemingly corresponding to the two major storm tracks in the wintertime Northern Hemisphere. Meanwhile, the maximum particle density 2 days prior is found over the midlatitude North Atlantic just southwest of the blocking region, which is an important ascent region for some of these moist particles, as will be shown later.
In contrast, the corresponding locations of the dry particles appear to be less concentrated. As such, we speculate that these 175 two categories of particles tend to undergo different dynamical processes.
3.2 Spatial distribution of mean properties along moist particles within and above the marine PBL We further partition moist particles into the time when they are located within the atmospheric PBL over the ocean (Figure 3) and the rest of the time (Figure 4), in order to unambiguously identify the role of the ocean in fuelling moist particles. As such, Figure 3 illustrates where the particles are located and how the mean properties along them are distributed when moist particles 180 can exchange heat and moisture with the underlying ocean, while Figure 4 shows when those particles are not in direct contact with the ocean. Note that the latter also includes the time when the particles are located within the PBL over land. heat and moisture sources for the moist particles that are going to be advected into the Euro-Atlantic blocks. One notable exception to this overall tendency is found off the west coast of North America, where the particles undergo topographical lifting ( Figure 3d) followed by drying ( Figure 3c) and warming ( Figure 3b) due to latent heating. As will be shown later in Section 4.3, this region turns out to be a common location for the organized large-scale ascent of the particles that gain moisture from the Pacific.

190
In stark contrast, the dominant tendency for moist particles when they are situated above the PBL (Figure 4) is such that an increase in potential temperature is concurrent with a decrease in specific humidity, which is accompanied by their ascent, while also the opposite is true (Figure 4d). This result means that latent heating and evaporative cooling are the dominant processes that control the property changes of the particles travelling above the PBL. The preferred locations for the particle ascent are found just upstream of the blocking over the North Atlantic 2 days prior to their arrival at the blocking region, as 195 well as off the west coast of the North American continent, found 5 and 9 days prior to the arrival at the blocking.
Meanwhile, dry particles barely undergo any property change, except for a general cooling trend associated with radiative processes (not shown).

Time evolution of variables along the trajectories of moist and dry particles
In order to further understand different properties associated with dry and moist particles, we compare the mean temporal  , only those grid cells that contain over 0.005% of the total particle's 6 hourly density are plotted. SHF and LHF are computed as the total SHF and LHF along moist particles found within each 2 o × 2 o grid cell on the specified days, divided by the total number of moist particles, indicating the average LHF contributing to blocking per particle.
Parallels are plotted for every 10 • whereas meridians are plotted for every 30 • on each plot. 0.7 K day −1 , consistent with the radiative cooling reported by Steinfeld and Pfahl (2019). Specific humidity of dry particles remains fairly low over the course of 10 days en route to the blocking regions, as these particles do not pick up any moisture from the ocean by definition. Despite its large standard deviation, dry particles' PV is maintained around approximately 1.5 PVU, slightly lower than a typical PV value of 2 PVU that marks the dynamical tropopause.
Moist particles (red lines in Figure 5) display a fairly distinct picture from their dry counterparts: While also originating from the mid-latitudes and travelling northward, moist particles tend to originate from the lower troposphere and then undergo a substantial uplift over the course of 10 days. This uplift and the northward motions are particularly enhanced about 3 days heat fluxes (magenta) supplied to moist particles from the ocean when the particles reside within the marine PBL. (Second to last rows) mean potential temperature (second row), mean specific humidity (third row), mean potential vorticity (fourth row), mean pressure (fifth row), and mean latitude (last row) of dry (cyan lines) and moist particles (red lines) along their respective trajectories. The envelopes around each line indicate ±0.5 standard deviation. The particle age on the x-axis denotes the days before the particles arrive at the blocking system (i.e., day 0 means the arrival time at the blocks). The closed (open) circles indicate where a given property distribution between moist and dry particles is significantly different at the 99% (95%) confidence level. The maximum standard errors are 0.05 W m −2 for turbulent fluxes, 0.009 K for potential temperature, 0.002 g kg −1 for specific humidity, 0.001 pvu for PV, 0.2 hPa for pressure, and 0.01 • for latitude.
prior to their arrival at the blocking system, coinciding with notable drying and warming due to intense latent heating during this period. The mean PV value of moist particles tends to retain fairly low values throughout without undergoing any substantial changes, reflecting their origins at the lower altitudes and latitudes. SHF and LHF are on average steadily supplied to moist particles until about 3 days prior to their arrival at the blocks, before undergoing strong uplift.
Despite the aforementioned significant differences in their origins and their evolution between dry and moist particles, these exception is PV, which on average undergoes little change during the course of 10 days, retaining its initial values along the trajectories of both dry and moist particles even at day 0. This result is consistent with the previous study by Methven (2015), who showed that the average PV of an upper-tropospheric outflow of a strongly ascending air stream is approximately equal to that of the lower-tropospheric inflow on average, as PV increases below the heating maximum but decreases above it. This 220 PV increase and decrease, however, are not visible in Figure 5 after averaged over millions of particles, as the timing of ascent differs for individual particles. Figure 5 indicates that the mean PV values are significantly higher for dry particles than for moist particles at the 99% confidence level even at day 0. Given that blocks are high pressure systems associated with low PV values, feeding of lower PV air is expected to be crucial for their maintenance (Yamazaki and Itoh, 2013b). Our result thus suggests that these moist particles play an important role in transporting significantly lower PV into the block system, further 225 corroborating the findings of the previous studies (Pfahl et al., 2015;Steinfeld and Pfahl, 2019).
This significant difference between dry and moist particles in their mean PV values seen at day 0 in Figure 5 further suggests that there could be a difference in their spatial distribution at the time of their release from the blocks. Figure 6 illustrates the spatial distributions of the dry (left panels) and moist particles (middle panels) separately, as well as their difference (right panels). Each row denotes different release altitudes: 10,000-m asl (∼250 hPa), 7,500-m asl (∼350 hPa), 5,000-m asl (∼500 230 hPa), and 3,500-m asl (∼700 hPa). The figure reveals that regardless of the release altitudes, the density of moist particles at the centre of the blocks tends to be higher than that of dry particles. In other words, moist particles carry low PV air upward across isentropic surfaces and thereby preferentially occupy the upper-level blocking core region.
In the meantime, a closer look at the difference plots (right panels) highlights that moist particles tend to be situated poleward of dry particles among those particles released from the lower altitudes (5,000 m asl and 3,500 m asl). Previous studies showed 235 that geopotential height anomalies associated with blocks tend to be located equatorward of the corresponding PV anomalies (e.g., Fig. 10 in Steinfeld and Pfahl (2019)), suggesting that if we repeat the analysis with the blocks being defined with PV anomalies, the signature of moist particles making up of the blocking cores may be more clearly seen. Confirmation of this speculation, however, should be a topic of a future study.

Moisture sources for moist particles 240
In the preceding section we have found that heat and moisture for the moist particles are not only provided by the Atlantic, but the North Pacific also likely acts to supply necessary heat and moisture to some particles en route to the Euro-Atlantic blocking ( Figure 3). In this section, we further separate the moist particles depending on their moisture sources, in order to uncover the characteristics associated with their different pathways and their roles in blocking maintenance.  Figure 6. The number density of dry (left column) and moist particles (middle column) and their difference after being normalized by the total particle number of each category (right column) for each 2 • × 2 • grid cell at the time of their release from the block regions, separately for the release altitudes of 10,000 m asl (top), 7,500 m asl (second row), 5,000 m asl (third row), and 3,500 m asl (last row). The numbers indicated on the bottom left corner for the left and middle columns denote the percentages of dry and moist particles released from each altitude, respectively. Only the grid cells where statistical difference is significant at the 95% confidence level are shaded on the third column.
Parallels are plotted for every 10 • whereas meridians are plotted for every 30 • on each plot.
from the ocean over the Atlantic south of 65 • N and the Mediterranean Sea are termed "Atlantic particles", while those particles subject to LHF from the ocean over the Pacific south of 65 • N as well as the Sea of Japan are termed "Pacific particles". As such, 250 we have found that approximately two thirds of the moist particles (i.e. 25.8% out of 41.3% of the total moist particles) source their moisture from the Atlantic basin, while the vast majority of the remaining one third source it from the Pacific (i.e. 13.6% out of 41.3% of the total moist particles), based on the particle tracking backward in time for a period of 10 days ( Figure 7a).
Intriguingly, a minute fraction of the particles (1.4% of the total moist particles) receive heat and moisture from both the Pacific and Atlantic basins (denoted "two-basin particles"). In comparison, 23.0% of the non-blocking particles is constituted 255 by Atlantic particles, 12.3% by Pacific particles, and 1.2% by two-basin particles out of the 37.1% of the total moist particles.
Although the differences seem small, especially for the two-basin particles, all of these percentages are significantly different from the blocking particles at the 99% confidence level. Note that we do not consider the moist particles fuelled over the Indian Ocean in the rest of the analyses, which only account for less than 1% of the total moist particles. These particles do not additionally pick up moisture from the other ocean basins (not shown).

260
When we extend the duration of the particle tracking from 10 days to 20 days (Figure 7b), however, the fraction of moist particles increases by about 15%, especially escalating the fraction of the Pacific particles and two-basin particles, while the 20-day tracking accounts for almost the same fraction for the Atlantic origin as the 10-day tracking. The twofold increase of the Pacific pathway particles indicates that the 10-day period is not long enough for some particles to be traced back to their moisture source in the Pacific. The over sixfold increase of the two-basin pathways is explained primarily by the detection 265 of additional particles of Atlantic origin that previously stayed in the PBL over the Pacific between 10 and 20 days prior to blocking events, and to a lesser extent by some particles of Pacific origin that are tracked back further to the Atlantic, across Eurasia. The fraction of particles sourcing their moisture from the Indian Ocean remains lower than 1% even with the 20-day tracking, indicating that the Indian Ocean does not act to provide much moisture for the wintertime Euro-Atlantic blocking events.

270
The Atlantic particles tend to originate locally 10 days prior to the arrival at the block locations (the particle distribution at -10 days indicated by the red contours in Figure 7a), with their maximum concentration found over the lee side of the Rockies over the North American continent. These particles then pick up a substantial amount of moisture from the Gulf Stream and its extension. A similar picture holds when they are tracked for 20 days (Figure 7b), although more particles tend to stem from the east of the block locations over Eurasia.

275
The Pacific particles identified through the 20-day tracking tend to originate from farther upstream, primarily receiving turbulent heat fluxes from the Kuroshio and its extension, with a clearer maximum concentrated along the Kuroshio Extension.
Still, the LHF maximum along the Kuroshio Extension is only about a half of that along the Gulf Stream, despite no notable difference in the climatological turbulent heat fluxes between them. It remains to be investigated whether this suppressed LHF would be enhanced with particle tracking beyond 20 days. Further analysis indicates that these two factors both contribute, suggesting that the particles are preferentially situated over this location and that they are subject to a larger amount of LHF than its climatology (not shown). As discussed later in Section 4.3, 290 this region coincides with starting locations of the large-scale ascent for those particles supplied with moisture from the Pacific, thus acting as a "springboard".

Time evolution of variables along the trajectories of particles of different moisture sources
In order to assess whether the physical processes involved depend on the particles' moisture sources, we repeat the same analysis on the temporal evolution of variables along the particle trajectories, but this time after separating the moist particles 295 into different moisture sources (Figure 8). Additionally, a three-dimensional view of the mean positions of these different pathways along with their mean PV value are shown in Figure 9 in order to illustrate their typical trajectories relative to their geographical locations. In contrast, the Pacific particles (orange lines in Figure 8) stem from the lower latitudes and altitudes. On the following days (day 9 -day 4) these particles keep ascending and experiencing latent heating, especially off the west coast of North America (Figure 9), which tends to occur earlier than the Atlantic particles by approximately 4 -5 days. After nearly all the moisture is lost by day -4, most of the properties of these particles become almost indiscernible from those of dry particles (denoted by 305 cyan lines), travelling along similar pressure and isentropic surfaces as well as at similar latitudes. The figure indicates that these particles also undergo radiative cooling at the same rate as dry particles from then on. Thus, the Pacific particles are transported in the same manner as the dry particles once lifted up into the mid-troposphere. In the meantime, it is notable that PV along these Pacific particles increases as they ascend approximately from day -5, which is the opposite to the expectation that PV decreases above the heating maximum. Although further investigation is necessary, we speculate that the reason behind this PV gain may be that they are under the influence of PV mixing with the higher upper-tropospheric PV near the tropopause (e.g., Hoskins, 1991).
The two-basin particles evolve in a similar manner to the Pacific particles at the beginning (magenta lines in Figure 8), as the particles originate from relatively low latitudes and altitudes with relatively high moisture content and then undergo ascent over the Pacific, especially off the west coast of North America. The ascent is, however, only by about 150 hPa till day -7, 315 leading to slight latent heating. As will be shown in Section 4.4, this lesser extent of uplift off the west coast of North America stems from the lower humidity content of those particles compared to the Pacific counterpart at the time of the ascent. These particles then start descending on the lee side of the Rockies (Figure 9) and receive heat and moisture from the Atlantic, before ascending very rapidly into the blocking system in the last 2 days (Figure 8).
PV values of moist particles, regardless of their moisture sources, are significantly lower than those of dry particles. This is 320 the case even for the Pacific particles that undergo a slight increase in PV after day -5 as noted earlier. While they appear to two-dimensional locations (green). The former is based on 10.1% of the Pacific pathways that reached above 400 hPa and whose potential to be in the vicinity of cyclones to be identified as WCB particles in our definition. Our WCB particles may thus include those particles that are not strictly classified as the conventional WCBs. A detailed analysis of individual events, nonetheless, revealed that these events are almost always accompanied by cyclonic systems at the time of their ascent (not shown). Consistent with Madonna et al. (2014), the average increase in potential temperature associated with the WCB particles is found to be about 20 345 K, while the average decrease in specific humidity is about 7 g kg −1 . Table 1 indicates that approximately 30% of the moist particles (i.e. about 10% of the total number of particles) are identified as WCB particles at some point along their 10-day trajectories, regardless of their moisture sources. This number (∼10%) is fairly similar to 9.7% reported by Steinfeld and Pfahl (2019), based on 7-day period statistics. Meanwhile, this fraction is much larger than the finding by Madonna et al. (2014), who reported only 0.36% of the 2-day trajectories released for the period 350 from 1979 to 2010 from the entire lower troposphere to be identified as WCB trajectories. While this difference in the fraction of WCB trajectories from the climatology may stem from different definitions used in detecting WCBs, it may also suggest the possibility that WCB trajectories are preferentially channeled towards the blocking systems. The fraction of moist particles identified as WCBs exceeds 60% (85%) of the total moist particles when the WCB definition is loosened to an ascent over 500 hPa (400 hPa) within 48 hours for the 10-day tracking trajectories (Table 1).  Figure 11a illustrates that the WCB particles undergo ascent preferentially around the western boundary currents and their extensions, as well as over the eastern Pacific to the northeast of the Hawaiian Islands, just as was the case for the locations of the maximum LHF that the particles receive (Figure 7). These locations, including the region to the northeast of Hawaiian Islands agree with the reported climatological locations of the wintertime WCB occurrence   Figure   4d). This close collocation between the particles' acquisition of turbulent heat fluxes and their ascent is in line with the results

360
shown by Boutle et al. (2010) in their idealized extratropical cyclone simulation. They showed that the augmented continuous evaporation is found in the vicinity of the WCB as a consequence of the continual moisture export from these regions by the horizontal divergence forced by boundary-layer drag. This moisture, in turn, converges at the base of the WCB forced by surface drag and large-scale ageostrophic flow, thereby providing a substantial amount of moisture continuously to the WCB.
Thus, the collocation of the LHF from the ocean and particle distributions is likely to indicate an active involvement of the 365 moisture convergence associated with extratropical cyclones. While the result shown in Figure 11a is for the 20-day tracking since the signature over the Pacific is more visible compared to the 10-day tracking, the same conclusion also holds for the 10-day tracked particles (not shown).

30.6%
(a) WCB Occurrence Atlantic Particles To identify the moisture sources necessary for these large-scale ascents, we have applied a moisture source diagnostic following Pfahl et al. (2014). In this method, we assign a weight to moisture uptake at each time step based on the ratio of 370 the moisture increase to the specific humidity value that each particle has at the starting point of the WCB ascent, such that each weight indicates a contribution of each moisture uptake for the WCB ascent. The weight at the previous moisture uptake is reduced when the specific humidity along the trajectory decreases and hence indicates a moisture loss due to precipitation. Figure 11b shows the result of this moisture diagnostic applied to the moist particles of different moisture sources separately.
Here we show the result for the 20-day tracking since the moisture source region particularly over the Pacific is more visible 375 compared to the 10-day tracking, while the 10-day tracking also shows similar results. The figure clearly indicates that the Gulf Stream and Kuroshio, together with their extensions, act to primarily provide moisture necessary for the WCB ascent. For the two-basin particles, we have separately evaluated the ascent over the Atlantic and Pacific. The diagnostic reveals that the two-basin particles retain the moisture taken from the Pacific basin until they are lifted over the Atlantic basin. This moisture uptake over the Pacific explains 23.5% and 41.6% of the necessary moisture for the WCB ascent over the Atlantic basin for the 380 10-day and 20-day tracking trajectories, respectively.

Typical synoptic conditions for individual moist particle pathways
Thus far we have shown distinct properties along different pathways of moist particles and how they evolve differently from one another. The question, however, remains as to what kind of synoptic conditions possibly give rise to the different pathways that these moist particles are likely to take. To answer this question, we have analyzed the particle trajectories of different moisture 385 sources with the corresponding upper-tropospheric synoptic eddies. Great case-to-case spatiotemporal diversity is found in the associated synoptic fields, hampering their meaningful composite analyses (not shown). Instead, Figure 12 illustrates a more or less representative case that exhibits the common features found for a given moist particle category. Many cases are found to exhibit roughly similar properties to these representative cases.
One common feature for the vast majority of the pathways that the Atlantic particles take is that the moist particles experience 390 an organized descent over the North American continent in travelling southeastward in the vicinity of a synoptic-scale lowpressure system, before they gain moisture from the North Atlantic. These features well correspond to those associated with a dry intrusion, a deep descent of dry air from the tropopause level into the lower troposphere often associated with extratropical cyclones (Browning, 1997), as clearly presented in Figure 12 for January 22 1992 (top right-most panel). A climatological study conducted by Raveh-Rubin (2017) illustrates that the North American continent is one of the most prominent locations 395 for dry intrusion air to start descending, before interacting with the warmer North Atlantic underneath approximately 48 hours later. Intrusion of dry, cold atmospheric particles is likely to enhance SHF and LHF from the North Atlantic, especially over the warm Gulf Stream, effectively moistening the particles before they are lifted up into an upper-level blocking anticyclone.
Both the behaviour of trajectories and the associated synoptic conditions are fairly dissimilar for the Pacific particles. Subsequent to a large-scale ascent into the upper troposphere over the northeastern Pacific, the particles do not undergo a deep 400 descent over the North American continent unlike the Atlantic particles. Instead, the particles appear to follow the upper-level westerly waveguide in the vicinity of blocking (the second row of Figure 12). Although further investigation is required, this aspect appears to be consistent with the finding by Yamazaki and Itoh (2013b), who suggested that upper-level westerlies upstream of a blocking high can act as a waveguide to effectively carry synoptic anticyclones towards the blocking high.
All cases of the two-basin particles are fairly analogous to the Atlantic particles but extended farther upstream into the Pacific 405 10 days prior to the arrival at the blocking events. Compared to the Pacific particles, they do not ascend as high off the west coast of North America, and as a result, they do not experience a strong descent over the North American continent compared to the Atlantic particles either. We found that this lack of a strong ascent over the Pacific is likely attributable to the difference in moisture content of the particles. At the time of the ascent, the moisture content of the Pacific particles is significantly higher than that of the two-basin particles at the 95% confidence level, with the mean difference amounting to 1. ( Figure 13a). This relationship between the ascent level and moisture content becomes more evident when a different ascent criterion is applied (Figure 13a, dashed and dotted lines). The figure reveals a clear link between the moisture content and the ascent level of a particle, in such a way that the higher the particle's moisture content at the base of the ascent, the higher the particle tends to ascend subsequently. In contrast, no such clear relationship is discernible with the sea-level pressure (SLP; Figure 13b) or with the SLP tendency ( Figure 13c) found at the base of the particle ascent, despite that a moderate to strong 415 relationship has been reported between cyclone intensification and WCB strength by a previous study (Binder et al., 2016).
Although more careful analyses, including an investigation of the lowest SLP found in the vicinity of the ascent as included in the WCB detection , are indispensable to draw a more solid conclusion, these results point to the key role of the particles' moisture content at the base of the ascent in determining their ascending altitude. Same as the panel (a) but for the sea level pressure found at the time of the ascent at the particle's horizontal location. (c) Same as the panel (a) but for the sea level pressure tendency.

420
Motivated by previous studies which pointed towards the significance of latent heating for atmospheric blocking events, we have investigated the oceanic moisture sources for wintertime Euro-Atlantic blocking events. To this end, we have utilized an atmospheric Lagrangian dispersion model, which allows us to investigate the processes taking place within the turbulent planetary boundary layer (PBL). We found that 27.8 -55% of the total particles released from wintertime Euro-Atlantic blocking highs undergo diabatic processes, with the moist particle percentages decreasing with altitude. Consistent with the previous 425 studies, these moist particles transport significantly lower PV compared to dry particles from the PBL to the upper-tropospheric blocking highs via large-scale ascent. In stark contrast, dry particles travel pseudo-isobarically in the upper troposphere, while being gradually cooled radiatively ( Figure 5). PV values remain significantly different between dry and moist particles throughout their evolution even when they reach the blocking highs ( Figure 5). As a consequence, there is a tendency for moist particles to be preferentially concentrated in the blocking cores, indicating an important role of moist processes in the blocking systems 430 ( Figure 6).
Depending upon their moisture sources, moist particles can be broadly divided into two categories: those moistening over the Atlantic and those moistening over the Pacific. The former, the particles that source moisture locally over the Atlantic, correspond to those that have been extensively studied in the previous studies (Pfahl et al., 2015;Steinfeld and Pfahl, 2019), which undergo a rapid ascent a few days prior to their arrival at the blocking systems. We have found that these particles tend to experience a large-scale descent over the North American continent, before efficiently gaining moisture from the North Atlantic (Figure 12 top row). Those particles that source their moisture remotely from the Pacific and then undergo ascent also transport low PV into the Euro-Atlantic blocking, but almost in the same manner as the dry particles that travel along a westerly waveguide in the upper troposphere. Still, PV values of these moist particles remain significantly lower than the dry counterparts throughout, even all the way to their arrival at the blocking (Figure 8). In this sense, these moist particles 440 can be regarded as a part of the class considered by Yamazaki and Itoh (2013a), such that anticyclonic low-PV particles are preferentially absorbed into blocking systems, acting to prolong blocking lifetime. In their study, Yamazaki and Itoh (2013a) identified the mechanism by analyzing isentropic trajectories, and thus the involvement of moist processes remained uncertain.
The current study, however, suggests that the same mechanism can be at play regardless of the source of low PV, thereby bridging the gap between the dry and the moist dynamical frameworks.

445
Our results further highlight that the paths that these moist particles preferentially follow while gaining moisture are concentrated over the western boundary currents as well as over the eastern Pacific to the northeast of the Hawaiian Islands ( Figure 3 and Figure 7). These regions also coincide with the most frequent starting points of large-scale ascent within warm conveyor belts associated with extratropical cyclones (Figure 11). This collocation is likely linked with continuous evaporation from the warm sea surface and the subsequent transport toward the base of the warm conveyor belt associated with extratropical cyclones 450 (Boutle et al., 2010). The western boundary currents are known for supplying substantial amounts of heat and moisture to the overlying atmosphere and thereby acting to anchor storm tracks by maintaining mean baroclinicity .
Thus, given the primary role of extratropical cyclones, it is not surprising that the enhanced latent heat fluxes and the starting positions of warm conveyor belts are collocated over the western boundary currents. Meanwhile, the region to the northeast of the Hawaiian Islands is not a region that is climatologically known for large upward latent heat fluxes nor a maximum in 455 extratropical cyclone activity; rather, it corresponds to a region with an amplified vertically-integrated moisture flux divergence driven primarily by advection associated with intraseasonal variability of the Aleutian low (Newman et al., 2012;Kwon and Joyce, 2013). Whether this anomalous moisture advection associated with the intraseasonal variability of the Aleutian low is linked with moisture convergence at the base of the WCB as discussed by Boutle et al. (2010) requires further investigation in future analyses.

460
Outstanding questions include whether the temporal variability of the western boundary currents on seasonal and longer time scales can affect blocking activities through modulations of moist processes. In this study we have focused on the climatological aspect of direct diabatic processes associated with blocking maintenance and shown that the western boundary currents provide substantial amounts of heat and moisture to the atmospheric particles en route to the Euro-Atlantic blocking systems.
Observational evidence indicates that both the Kuroshio (Qiu and Chen, 2005;Kwon et al., 2010) and . Whether these oceanic variabilities can modulate the direct diabatic processes associated with the blocking events described in the current study, however, is yet to be known. Hence in a subsequent study we will investigate the effect of the oceanic temporal variability. Given a recent study that showed the projected strengthening of the Kuroshio and weakening of the Gulf Stream under future climate scenarios side of the anomaly maximum. The reversal criterion is met if the height gradient is negative at any longitude within a longitudinal range (∆λ). 4. The anomaly is identified as a blocking event if the above three criteria are met during a given period (D).
We have set the arbitrary thresholds to be identical to those used by Dunn-Sigouin et al. (2013): the amplitude threshold A of 1.5 standard deviations of Z over 30 -90 • N computed for a period of 3-months centred at a given calendar month; the spatial scale S of 2.5 × 10 6 km 2 ; the overlap criterion O of 50%; the meridional ∆φ and zonal ∆λ scales of 15 • in latitude and 10 • in longitude, respectively; and the duration period D of 5 days. These choices of criteria are rather restrictive and, as a result, 505 tend to highlight the mature/maintenance stage of blocking. The usage of these criteria can suppress the erroneous detection of quasi-stationary ridges and immature systems as blocks, while well capturing omega-shaped blocking, thereby overcoming the shortcomings of the previously used indices (Dunn-Sigouin et al., 2013).

Appendix B: Atmospheric dispersion model FLEXPART
FLEXPART is a comprehensive offline atmospheric Lagrangian dispersion model, widely used for diverse atmospheric trans-510 port applications. FLEXPART computes the particle trajectories under the assumption of zero acceleration as where t denotes time, ∆t the time increment, and X the particle's position vector. v = v +v t +v m , is the three-dimensional wind vector consisting of grid-scale winds (v), turbulent wind fluctuations (v t ), and mesoscale wind fluctuations (v m ). Turbulent wind fluctuations are evaluated by solving Langevin equations (Thomson, 1987) in assuming Gaussian turbulence.

515
Mesoscale wind fluctuations, whose spectral interval falls between the resolved winds and the turbulent wind fluctuations, are approximated by solving an independent Langevin equation for the mesoscale wind velocity fluctuations in a similar manner to Maryon et al. (1998).
In addition, in FLEXPART, one can optionally evaluate the effect of moist convective transport that takes place within convective clouds by turning on the moist convection scheme by Emanuel and Živković-Rothman (1999), at the expense of the 520 substantial increase in the computational time (Stohl et al., 2005). This transport is grid-scale in the vertical but subgrid-scale in the horizontal. A recent study by Oertel et al. (2020) has highlighted the importance of convection embedded in WCBs by using a high-resolution model with a 0.02 • horizontal resolution. Switching on the moist convection scheme in FLEXPART, however, does not qualitatively or quantitatively alter our results, suggesting that the importance of the subgrid-scale horizontal convection is only secondary to the resolved grid-scale convection in the current study with a much coarser resolution (not 525 shown). For this reason, we turn off the moist convection scheme in our analyses.