The concept of atmospheric river (AR) as a filamentary structure in the atmospheric water vapour transport has been around since the 1990s1,2,3, and can even be dated back to the conceptual models of warm conveyor belt in the 1970s4,5,6 and moist tongue in the 1930s7,8. ARs are typically associated with a low-level jet stream ahead of the cold front of an extratropical cyclone, and frequently lead to heavy precipitation when they are forced upward by mountains or frontal lift9,10. Along the west coast of North America (NA), high-impact ARs often arrive in the cold season when weather is dominated by the strong onshore flow of Pacific airstreams2,9,11,12,13,14. Some of these systems are also known as pineapple express storms for their apparent origin in the tropical area near Hawaii15,16,17. Their landfall on mountainous areas can produce extremely heavy precipitation and may trigger hazardous floods, mudslides, and avalanches.

Intense ARs are rare over the northeast Pacific in the warm season, when a ridge of high pressure tends to reside off the west coast of NA. In this article, we investigate an anomalous summer AR that moved across the North Pacific (NP) and made a prolonged landfall over Southeast Alaska during 24–26 June 2021. A snapshot of this AR is shown in Fig. 1, which displays the distribution of satellite-derived total precipitable water18, also known as vertically integrated water vapour (IWV, see “Methods”), valid at 0600 UTC 24 June 2021. Because the origin of this AR can be traced to the Southeast Asian summer monsoon (SEASM), we call it an oriental express to emphasize this unique connection. Our focuses are on its evolution and possible impacts on an extreme heatwave developed over western NA.

Fig. 1: Morphed integrated microwave total precipitable water (MIMIC-TPW2)18 valid at 0600 UTC 24 June 2021.
figure 1

The image shows an atmospheric river with origin in Southeast Asia moving across the North Pacific and making landfall over Southeast Alaska.

As an important component of the Asian monsoon system, the SEASM is characterized by strong convection and moisture transport across the South China Sea–Philippine Sea regions19,20. The associated precipitation pattern has an east-west banded structure with alternate dry and wet conditions over the subtropical western Pacific and East Asia21. Previous studies found that a northward propagation of the intraseasonal oscillation in the SEASM has a far-reaching influence on the global atmospheric circulation22,23. In a recent study24, we showed that the SEASM precipitation anomaly in late June 2021 was able to trigger an extratropical Rossby wave train across the NP, leading to a synoptic condition favourable for an extreme heatwave in western NA. We also noted that a realistic representation of the AR-induced moisture transport appears crucial for the extended-range forecast of this heatwave. As a follow-up study, here we present an in-depth analysis of this oriental express and its impacts on the NA heatwave in late June 2021. We utilize a three-dimensional global reanalysis dataset25 to diagnose the structure and evolution of this AR event. It is shown that this system formed a super-long channel for moist and warm air to flow from the Southeast Asia to NA. The track for this express was controlled mainly by an unseasonably deep Aleutian Low in conjunction with a stable NP High. Its landfall produced heavy precipitation and local flooding in Southeast Alaska26.

As the AR-heatwave relationship is concerned, we analyze the AR-induced moisture and heat transport across the Pacific Coast Ranges. It is shown that the convergence of sensible heat flux in northern British Columbia (BC) and southern Yukon formed a direct link between the landfalling AR and the northward expansion of the heatwave. In addition, given that water vapour is the dominant short-lived greenhouse gas in the atmosphere27,28,29,30,31, we propose another possible mechanism linking the injected moisture with the heatwave development in western NA in late June 2021.

Results and discussion

Oriental express as a trans-Pacific AR from Southeast Asia

ARs are often defined based on the vertically integrated vapour transport (IVT) derived from specific humidity and wind profiles (see “Methods”)1,2,3,32,33,34. Based on the 5th generation global reanalysis of the European Centre for Medium-Range Weather Forecasts (ERA5)25, the IVT distributions over the NP basin valid at 0000 UTC 21–28 June 2021 are shown in Fig. 2. The corresponding distributions of sea level pressure (SLP), 500 hPa geopotential height (Z500), and IWV are shown in Supplementary Figs. S1S3. Here one can apply either the IVT250 or the IVT500 scheme (see “Methods”) for AR identification. In Fig. 2, we see the northeastward advance of a strong AR across the NP basin. In general, this AR was driven mainly by a twin-engine system consisting of an anomalously deep Aleutian Low and a stable subtropical high-pressure system in the NP. It was linked to the SEASM system through most of its lifespan. This connection is the inspiration for calling it the oriental express, analogous to the pineapple express that brings warm and moist air from tropical sources near Hawaii to the west coast of NA. It can be considered as an important mechanism of teleconnection between the Southeast Asian intraseasonal oscillation and the NA weather anomalies.

Fig. 2: The development of an oriental express across the North Pacific.
figure 2

ah The snapshots of IVT distributions valid at 0000 UTC during 21–28 June 2021. Data are from the ERA5 reanalysis. The vector represents the integrated water vapour flux \(({Q}_{u},{Q}_{{v}})\) defined in “Methods”. \({{{{{\rm{IVT}}}}}}={({Q}_{u}^{2}+{Q}_{v}^{2})}^{1/2}\). The major high and low centres of SLP in the Pacific Ocean are marked by letters “H” and “L” with their local maximum/minimum SLP values (in hPa).

Two ARs can be identified in the NP at 0000 UTC 21 June 2021 (Fig. 2a). The one in the east was driven by a cyclone in the Gulf of Alaska (L1003) with a strong anticyclone to its south (H1029). The major AR was the oriental express driven by a major cyclone (L979) into the northeast Pacific from the west. The minimum SLP at the cyclonic centre (979 hPa) was 31 hPa below the climatological mean value (1010 hPa, Fig. 3a) at that location. On the northern flank of this AR there were also two secondary cyclones (L1001 and L995) centred at (28°N, 125°E) and (39°N, 148°E), respectively. They were interacting with the AR and pumping moisture from the South China Sea into the AR system. The NP High spawned a secondary centre west of the 180th meridian (H1021). Anticyclonic flow around this western centre exported warm moist air from the tropical western Pacific. These features were maintained through the following two days as the major cyclone moved eastward across the 180th meridian and gradually weakened (Fig. 2b, c).

Fig. 3: The anomalies of 5-day (21–25 June 2021) mean SLP, 500 hPa geopotential height, and IWV (colour contours) with respect to their 30-year (21–25 June, 1991–2020) climatology (line contours).
figure 3

The intervals of line contours are 4 hPa in a, 60 m in b, and 5 kg m−2 in c, respectively. All data are from the ERA5 reanalysis. The yellow band in a highlights the Pacific Coast Ranges that separate Alaska from western Canada.

The major cyclone re-strengthened rapidly through 23 June with the minimum SLP falling to 973 hPa around 0000 UTC on the 24th as the AR made landfall on Southeast Alaska (Fig. 2d). With a strong and stable anticyclone to its southeast, this AR remained quasi-stationary in the northeast Pacific for more than 48 h, and began to dissipate and disconnect with the Asian monsoon system around 0000 UTC 27 June (Fig. 2g). It was capable of transporting substantial heat and moisture inland across the Pacific Coast Ranges (including the Boundary Ranges and the Saint Elias Mountains, highlighted by the yellow band in Fig. 3a).

Anomalous atmospheric patterns

To reveal the atmospheric processes responsible for driving the oriental express across the NP, we calculated the 5-day (21–25 June 2021) mean anomalies of SLP, IWV, and Z500 with respect to their 30-year (1991–2020) climatology. As shown in Fig. 3a, the NP High off the NA west coast dominates the SLP distribution in the climatological map (line contours) for this early summer period. Its western extension into the subtropical northwest Pacific forms an important component of the SEASM19. Around this time, as the polar jet stream migrates northward, the Aleutian Low has often reached its minimum intensity or virtually disappeared35. This was not the case in 2021. In Fig. 3a, the lowest 5-day mean SLP anomaly is −21.5 hPa in the extratropical NP. This anomalous cyclone in conjunction with the stable NP High created a strong horizontal pressure gradient that forced the northeastward advance of the oriental express from the subtropical northwest Pacific.

The anomaly pattern of Z500 in Fig. 3b reveals a well-defined Rossby wave train propagating from the subtropical western NP into the extratropical Northern Hemisphere. It provides a physical explanation to the intensification of the Aleutian Low that drove the oriental express across the NP basin. Previous studies have shown that this wave train can be linked to phase 7 of the boreal summer intraseasonal oscillation in Southeast Asia, when the tropical convection is suppressed near the South China Sea and the Philippines23. The reduction of latent heat released from the convection in this subtropical region can trigger an extratropical Rossby wave train with a cyclonic circulation in the central NP and an anticyclone in the eastern NP24.

The two bands of large positive IWV anomalies in Fig. 3c indicate that the subtropical export from the SEASM was capable of contributing a large amount of moisture to the oriental express as it moved across the NP basin. Some authors have argued that ARs may be considered as the footprints left behind on the cyclone pathway, and local sources are primarily responsible for the high moisture content within the AR core36,37,38. The current understanding is that strong ARs are capable of pumping moisture out of tropical regions and incorporating midlatitude moisture sources along their paths39,40.

Orographic precipitation and cross-mountain moisture transport

Landfalling ARs are capable of producing heavy precipitation in mountainous areas, especially in the cold season with water vapour supplied from warm tropical sources9,16,34,41,42,43,44,45,46. ARs in the warm season often produce less orographic precipitation because the warmer atmosphere can contain more water vapour. The oriental express depicted in Figs. 1 and 2 was intercepted by the Pacific Coast Ranges when it made landfall over Southeast Alaska during 24–27 June 2021. Figure 4a shows the 4-day total precipitation amounts from this AR event. The maximum amount of 122.4 mm was in Southeast Alaska. Note that June is the driest month in Southeast Alaska with an average monthly precipitation around 100 mm47. There were a few flooding events in this region that could be attributed to this AR, not only because of heavy rainfall but also from warm advection associated with the system leading to a large build-up of ice and snow melt26,48. Also note that there was a weak AR making landfall in this region earlier (20–21 June, see Fig. 2a), which resulted in a wet condition of soil moisture prior to the oriental express event. This antecedent condition might influence infiltration versus runoff of precipitation that fell and affect latent heat release that could shape local or regional temperatures.

Fig. 4: The precipitation distribution and the atmospheric river profile.
figure 4

a The 4-day precipitation amount during 24–27 June 2021 (see “Methods” for data description); the 96 h period ends at 0600 UTC in Canada and at 1200 UTC 27 June in the United States. b An along-AR vertical cross section (along the white line in the embedded box) valid at 0000 UTC 25 June 2021 based on the ERA5 reanalysis; colour-filled contours represent the specific humidity, red dashed lines are potential temperature contours (unit: K, intervals: 5 K), blue solid lines are elevations above sea level (unit: m, intervals: 1000 m), and winds are indicated by the conventional wind barbs.

Another feature often associated with a landfalling AR is the spillover of moisture and hydrometeors from the windward side to the leeside of the mountains45. As shown in Fig. 2 (and Supplementary Fig. S3), the oriental express in late June 2021 penetrated deep into western Canada after its landfall. Figure 4b presents an along-AR vertical cross-section valid at 0000 UTC 25 June. It shows that, although the AR core on the windward side was located near 1000 m above sea level, the air column between 2000 and 5000 m contained noticeable amounts of water vapour available for transport across the Pacific Coast Ranges. Some condensates generated by the orographic uplift on the windward side within this layer might also be carried to the leeside where the descending airstreams would tend to dissipate cloud and lessen any potential rainfall through re-evaporation. The leeward cloud dissipation can be seen in the satellite imagery given in Supplementary Fig. S4.

Possible mechanisms of AR-heatwave interactions

An unprecedented heatwave developed in western NA from late June through mid-July 2021, causing millions overheated with more than 1400 deaths, failures of power grids, and widespread wildfires24,49,50,51,52. The responsible weather system was a strong high-pressure ridge known as heat dome, under which the warm air was trapped and the associated subsidence caused further warming by compression49,53. As shown in Fig. 3c, the heat dome on the west coast of NA was associated with a Rossby wave train across the NP. This wave pattern was also responsible for driving the oriental express across the NP. Figure 5 shows the anomalies of 5-day averages of 2 m air temperature (T2m) and low-level mean temperature (\(\bar{T}\)) through late June. It appears that the extreme heatwave started in Washington and spread rapidly southward into Oregon and northward across western Canada. This event was characterized by some of the highest temperatures ever recorded in this region, including the new Canadian record of 49.6 °C measured at Lytton, BC on June 29, 202124,49,50. This village is marked by a small circle in Fig. 5.

Fig. 5: The 5-day mean anomalies of the low-level averaged temperature (\(\bar{T}\) defined in “Methods”, line contours with intervals of 2 °C) and the 2 m temperature (T2m, colour-filled).
figure 5

The four overlapped 5-day periods are a 22–26, b 23–27, c 24–28, and d 25–29 June 2021. Data are from the ERA5 reanalysis. The anomalies are calculated based on the 30-year (1991–2020) climatology. The small circle indicates the location of Lytton, BC, where the highest temperature ever measured in Canada was recorded at 49.6 °C on 29 June 2021.

The oriental express could exert its influence on this heatwave event by transporting warm and moist air from the Pacific Ocean across Southeast Alaska into western Canada. Figure 6 shows the vertically integrated sensible and latent heat fluxes (ISHF and ILHF) and their convergences averaged over the 4-day period (24–27 June) when the AR made landfall over Southeast Alaska. The AR-induced heat transport was dominated by the ISHF over the northeast Pacific Ocean, where the maximum magnitudes of ISHF and ILHF were 3491 × 106 and 1947 × 106 J m−1s−1 respectively. Although the magnitudes of ISHF were much smaller over the continent, the onshore transport led to strong ISHF convergences on the leeside of the Pacific Coast Ranges (Fig. 6c). The converged sensible heat was capable of warming the lower troposphere, and thus could be directly responsible for the heatwave expansion to northern BC and Alberta in late June 2021 (Fig. 5d). For reference, we also show the daily means of ISHF and its convergence during 24–27 June 2021 in Supplementary Fig. S5. On the other hand, the release of latent heat can occur only when the convergence of ILHF leads to condensation. In Fig. 6d, noticeable ILHF convergence can be found in Southeast Alaska, where the orographic uplift could lead to condensation and release of latent heat. On the leeside of the mountains, the diabatic heating related to the ILHF convergence was negligible as compared to heating caused by the ISHF convergence.

Fig. 6: Sensible and latent heat transport in the lower troposphere averaged over 24–27 June 2021.
figure 6

a The integrated sensible heat flux (ISHF). b The integrated latent heat flux (ILHF). c The ISHF convergence. d The ILHF convergence.

There could be another mechanism for this AR to influence the heatwave through radiative heating. The water vapour transported by the AR into western Canada was trapped under a ridge of high pressure (see Supplementary Figs. S3, S6). The added moisture could act as greenhouse gas to trap solar radiation and make the lower atmosphere even warmer27,28,29,30,31. One may argue that more moisture tends to increase clouds and enhance precipitation, which reduces, instead of increases, surface temperature. For this event, however, the moisture was injected into a warm and dry air mass under a ridge of high pressure in western Canada. The unsaturated conditions resulted in generally clear skies after 25 June 2021 (see Supplementary Fig. S4). Figure 7 shows the anomalies of 5-day mean IWV and temperatures in western North America during 26–30 June 2021, together with their local correlation coefficients based on 43-year (1979–2021) data from the ERA5 reanalysis. The positive IWV anomalies (Fig. 7a) in western Canada can be attributed jointly to the AR-induced moisture transport from the NP Ocean and the local evaporation or evapotranspiration under the heat dome. The intensity and extent of the heatwave are indicated by the significant positive temperature anomalies in western NA (Fig. 7b). It is shown in Fig. 7c, d that across western Canada the air temperatures are positively correlated with the local IWVs during the period of 26–30 June. In some regions, including northern BC on the leeside of the Pacific Coast Ranges, the correlation coefficients are higher than 0.80. Although correlation does not necessarily imply causation, it is reasonable to assume that the moisture injected by the AR into western Canada could have a positive feedback loop to amplify the heatwave in late June 2021. On the other hand, one could also argue that the IWV build-up in western Canada was coincidental, because the advection in of moisture from the AR happened due to the building upper ridge that provided the trajectory for the moisture. In addition, it is possible to have more water vapour in an air parcel at a higher temperature, because the saturation vapour pressure is uniquely determined by temperature (with little pressure dependence). Our lagged correlation analysis cannot provide convincing evidence to resolve this causality issue. We will come back to this point later in the discussion.

Fig. 7: The anomalies of 5-day (26–30 June 2021) mean IWV and air temperatures, together with the correlation coefficients between them.
figure 7

All data are from the ERA5 reanalysis. The anomalies are calculated based on the 30-year (26–30 June, 1991–2020) climatology. In a, the IWV anomalies over western Canada can be attributed to the water vapour transport from the oriental express. In b, the lower-tropospheric averaged temperature (\(\bar{T}\)) anomalies are represented by line contours with intervals of 2 °C, and the 2 m temperature (T2m) anomalies are colour-filled. In c, d, the correlation coefficients are calculated from 43-year (1979–2021) data; the coefficient values between −0.30 and 0.30 are statistically insignificant with p-values beyond 0.05 based on the two-tailed Student t-test.

An indirect support for our argument of possible AR contributions to the heatwave development through both heat and moisture transports comes from our verification of the operational numerical weather predictions of Environment and Climate Change Canada (ECCC). Figure 8 shows the IVT distributions valid at 0000 UTC 25 June 2021 as analyzed and predicted by the Global Deterministic Prediction System (GDPS) of ECCC54. The analyzed AR over the northeast Pacific (Fig. 8a) was well predicted by the numerical weather prediction (NWP) model with lead times (LT) up to 96 h (Fig. 8b, c). The predictions with LT = 144 and 192 h (Fig. 8d, e) could still be considered as good guidance for the arrival time and landfall location of the AR. However, the deep inland-penetration of the AR into western Canada was not correctly predicted by these model runs, which were initialized at 0000 UTC 19 and 17 June, respectively. The model run initialized 10 days earlier (Fig. 8f) gave a poor prediction of landfall location of the AR at 0000 UTC 25 June.

Fig. 8: Distributions of the IVT valid at 0000 UTC 25 June 2021 based on the Canadian GDPS operational outputs.
figure 8

a The GDPS analysis. bf The GDPS predictions with different lead times (LT).

Figure 9 shows the analyzed and predicted T2m valid at 0000 UTC 27 June. The temperature pattern in western Canada was well predicted with LT = 144 h, i.e., from the NWP model initialized at 0000 UTC 21 June (Fig. 9a vs Fig. 9b). It is shown in Fig. 8 that the inland-penetration of the AR at 0000 UTC 25 June was also correctly predicted by the model run initialized as early as 21 June. The model initialized at 0000 UTC 19 or 17 June provided poor guidance for both the AR penetration (Fig. 8d, e) and the temperature distribution (Fig. 9c, d) in western Canada.

Fig. 9: Distributions of the 2 m temperature valid at 0000 UTC 27 June 2021 based on the Canadian GDPS operational outputs.
figure 9

a The GDPS analysis. bd The GDPS predictions with different lead times (LT).

The above argument about the AR-heatwave relationship in the forecast could be questionable, because it is not surprising that the forecasts from earlier initializations have poor skill due to the increase of model error. The argument can be better justified by using the same initialization but different perturbed simulations (ensemble members) or different models24. Figure 10 shows four sub-ensemble forecasts from the ECCC Global Ensemble Prediction System (GEPS) initialized at 0000 UTC 17 June 2021. Based on their predicted IWV patterns valid at 0000 UTC 25 June, the forecasts from the 21 GEPS ensemble members were separated into four groups. Group A consists of 6 members that predicted the AR landing over the south coast of Alaska (Fig. 10a, b). Each of the other three groups consists of 5 members with AR landing over Southeast Alaska (Fig. 10c, d), the north-central coast of BC (Fig. 10e, f), and the south coast of BC (Fig. 10g, h), respectively. Based on the analysis in Fig. 2e (and Supplementary Fig. S3e), the best forecast of the IWV pattern at 0000 UTC 25 June (LT = 192 h) was given by the sub-ensemble mean of Group B (Fig. 10c), which predicted the AR landfall over Southeast Alaska. As compared with the analysis in Fig. 9a, Group B also achieved the best forecast for the T2m pattern valid at 0000 UTC 27 June (Fig. 10d). The other three groups under-predicted the heatwave intensity in western Canada. A similar but less skilful ensemble prediction initialized at 0000 UTC 15 June 2021 is shown in Supplementary Fig. S7.

Fig. 10: Sub-ensemble mean predictions of IWV and T2m with the ECCC GEPS run initialized at 0000 UTC 17 June 2021.
figure 10

The 21 ensemble members are grouped into four categories based on the similarities of their predicted IWV patterns valid at 0000 UTC 25 June. a, b Group A with 6 ensemble members. c, d Group B with 5 members. e, f Group C with 5 members. g, h Group D with 5 members. The group-mean IWV distributions valid at 0000 UTC 25 June (LT = 192 h) are plotted on the left panel. The group-mean T2m valid at 0000 UTC 27 June (LT = 240 h) are plotted on the right panel.

Concluding remarks

We have described an anomalous warm-season AR moving across the NP Ocean in late June 2021 and dubbed it the oriental express. This system brought tropical moisture to the west coast of NA from the SEASM. It was driven by an anomalous atmospheric pattern forced by the boreal summer intraseasonal oscillation in Southeast Asia23,24. The landfall of this AR over Southeast Alaska transported large amounts of moisture and heat energy into western Canada. The injected water vapour and sensible heat accumulated under a ridge of high pressure and could influence the heatwave development over western NA during late June and early July 2021. A rapid attribution analysis has pointed out that the occurrence of this extreme heatwave is virtually impossible without human-caused climate change50; also see a related study on the impact of climate change on heatwaves in Metro Vancouver55. Our results provide evidence that AR-induced moisture and heat transports should also be considered when assessing heatwave development in a changing climate. Because ARs are narrow corridors of strong horizontal water vapour and heat transports in the atmosphere, their landfalls not only have the potential to produce heavy orographic precipitation, but also are possible to affect the development of an inland heat dome in summer.

Our analysis suggests that the AR-induced sensible heat flux convergence in northern BC and southern Yukon could be an important reason for the heatwave expansion to the area from the south. A more difficult issue is to assess whether the AR-induced moisture transport into western Canada has a large effect on the heatwave through greenhouse effect. Our result cannot quantify this impact to an order of magnitude. Some questions remain to be addressed, such as (1) how much of the anomalous water vapour was transported in, and how much was contributed locally through evaporation and evapotranspiration? (2) what is the timescale over which adding water vapour to the atmosphere leads to the increase of temperatures locally at the surface? It should be pointed out that the statistical connections between AR and heatwave in Fig. 7 alone cannot resolve the causality related to the greenhouse effect argument. Probably this issue can be addressed in a quantitative manner by running a model with and without the water vapour feedback into western Canada and analyzing what differences arise in the heatwaves produced by the model. Another possible direction is conducting a similar reforecast but replacing the radiative transfer scheme with grey radiation56,57. This could be a more straightforward way to rule out the short-lived greenhouse effect by water vapour but still keep the influence of convective adjustment. These approaches, however, are beyond the scope of this study.

There have been a number of studies regarding the responses of ARs to climate change12,58,59,60,61. It is understood that, under the global warming conditions, AR events would intensify through increased moisture flux in the future in most regions, with some of them being projected to increase the magnitude of heavy precipitation events62. However, little attention has been directed to understanding the AR-heatwave interactions in a changing climate. Our case study may serve to suggest yet another important topic to add to the “blind spot” list of future climate projections and weather forecasting, as discussed by Alex Hall in his AGU2019 Turco Lectureship63.

We close this article with a brief comparison of the 2021 western NA heatwave with a similar but weaker event during 28–30 July 2009, which was linked to an elevated mortality with 122 to 156 excess deaths in Metro Vancouver55. As shown in the Supplementary Fig. S8, the 2009 heatwave affected the west coast of NA from Oregon to Yukon. Most of this area was also covered by positive IWV anomaly during 28–30 July. The excess moisture in Supplementary Fig. S8b could be linked to the antecedent water vapour and sensible heat transports from an oriental express across the NP (Supplementary Figs. S9S11). The apparent similarity between these two events warrants further study to determine the connection between historical ARs and extreme heatwaves in NA.


Atmospheric river identification methods

The IVT and integrated water vapour (IWV) are two commonly used variables in AR analysis. They can be defined as1,3,34

$${{{{{\rm{IVT}}}}}}=\sqrt{{Q}_{u}^{2}+{Q}_{\upsilon }^{2}},{{{{{\rm{with}}}}}}\,({Q}_{u},\,{Q}_{\upsilon })={g}^{-1}{\int }_{{p}_{{{{{{\rm{t}}}}}}}}^{{p}_{{{{{{\rm{b}}}}}}}}(u,\,{v})q\,{{{{{\rm{d}}}}}}p,$$
$${{{{{\rm{IWV}}}}}}={g}^{-1}{\int }_{{p}_{{{{{{\rm{t}}}}}}}}^{{p}_{{{{{{\rm{b}}}}}}}}q\,{{{{{\rm{d}}}}}}p.$$

In the above equations, g = 9.81 m s−2 is the acceleration due to gravity, q is the specific humidity, p is the air pressure, u and υ are the zonal and meridional velocities. The vertical integration is from the top (\(p={p}_{{{{{{\rm{t}}}}}}}\)) to the bottom (\(p={p}_{{{{{{\rm{b}}}}}}}\)) of the atmosphere. \({Q}_{u}\) and \({Q}_{{v}}\) represent the zonal and meridional components of the integrated water vapour flux, respectively.

There have been many AR identification and tracking algorithms in the literature based mainly on the distribution of IVT, IWV, or a combination of the two64. In this study we choose a simple method called the IVT250 scheme11, which defines an AR as a contiguous region ≥ 2000 km in length with IVT ≥ 250 kg m−1s−1. For warm-season ARs, one may also increase the criterion to 500 kg m−1s−1 to get an IVT500 scheme.

Vertically integrated sensible and latent heat fluxes

Vertically integrated sensible and latent heat fluxes (\({{{{{{\bf{Q}}}}}}}^{{{{{{\rm{S}}}}}}}\) and \({{{{{{\bf{Q}}}}}}}^{{{{{{\rm{L}}}}}}}\)) are defined as65,66

$${{{{{{\bf{Q}}}}}}}^{{{{{{\rm{S}}}}}}}={g}^{-1}{c}_{{{{{{\rm{p}}}}}}}{\int }_{{p}_{{{{{{\rm{m}}}}}}}}^{{p}_{{{{{{\rm{b}}}}}}}}(u,\upsilon )T\,{{{{{\rm{d}}}}}}p,\,{{{{{{\bf{Q}}}}}}}^{{{{{{\rm{L}}}}}}}={g}^{-1}{L}_{{{{{{\rm{c}}}}}}}{\int }_{{p}_{{{{{{\rm{m}}}}}}}}^{{p}_{{{{{{\rm{b}}}}}}}}(u,\upsilon )q\,{{{{{\rm{d}}}}}}p,$$

where \({c}_{{{{{{\rm{p}}}}}}}=1004\,{{{{{\rm{J}}}}}}{{{{{{\rm{K}}}}}}}^{-1}{{{{{{\rm{kg}}}}}}}^{-1}\) is the specific heat of dry air at constant pressure, T is temperature,\({L}_{{{{{{\rm{c}}}}}}}=2.5\times {10}^{6}\,{{{{{\rm{J}}}}}}{{{{{{\rm{kg}}}}}}}^{-1}\) is the latent heat of condensation, and pm is a specified pressure level (pm = 500 hPa in this study). The convergences of \({{{{{{\bf{Q}}}}}}}^{{{{{{\rm{S}}}}}}}\) and \({{{{{{\bf{Q}}}}}}}^{{{{{{\rm{L}}}}}}}\) are given as \(-\nabla \cdot {{{{{{\bf{Q}}}}}}}^{{{{{{\rm{S}}}}}}}\) and \(-\nabla \cdot {{{{{{\bf{Q}}}}}}}^{{{{{{\rm{L}}}}}}}\), respectively.

The mean temperature in the lower troposphere

Meteorologists often express the horizontal pressure gradient force in the mid-troposphere as the gradient of geopotential along the 500 hPa pressure surface. This level is, on average, around 5.5 km above sea level. It is also common to define mean temperature (\(\bar{T}\)) in the lower troposphere as a function of the geopotential thickness between this level and the 1000 hPa level,


where R = 287 Jkg−1K−1 is the gas constant for dry air, Z500 and Z1000 are the geopotential heights at the levels of 500 and 1000 hPa, respectively.

Data sources

The main data sources for this study are the ERA5 global reanalysis25 and the operational NWP outputs from the Canadian Global Deterministic/Ensemble Prediction System (GDPS/GEPS)54. The ERA5 provides hourly output on a 0.25° × 0.25° latitude-longitude grid; we only use the 0000 and 1200 UTC data in June from 1979 to 2021. The GDPS has an approximate horizontal spacing of 15 km. It provides weather forecasts up to 10 days, and is run twice daily initialized at 0000 and 1200 UTC, respectively. The GEPS consists of 21 ensemble members with an approximate horizontal spacing of 39 km.

The morphed integrated microwave total precipitable water (MIMIC-TPW2) in Fig. 1 is the satellite-derived IWV. It is the product of morphological compositing of microwave-frequency sensor data from several operational polar-orbiting satellites18. The daily precipitation data used to produce Fig. 4a are from the Climate Prediction Centre (CPC) Global Unified Gauge-Based Analysis of Daily Precipitation67 on a 0.5° × 0.5° latitude-longitude grid.