Zonal and meridional wind on pressure levels was accessed from the Japanese Reanalysis-55 dataset (Kobayashi et al., 2015), at 2.5×2.5∘ horizontal resolution at 6 h time intervals, from February–June 1958–2017, and the European Centre for Medium-Range Forecasting (ECMWF) ERA-Interim dataset (Dee et al., 2011) at 1×1∘ horizontal resolution at 6 h time intervals from February–June 1979–2017. The lower horizontal resolution of JRA-55 was used due to our interest in large-scale patterns of variability, and the resolution difference does not influence the resultant analysis. The 60-year JRA-55 reanalysis is used to compare the spring transition for as many ENSO events as possible (Fig. 8), while ERA-Interim is used to be consistent with the transport and tropopause fold diagnostics that were derived from ERA-Interim reanalysis. Both datasets use four-dimensional variational data assimilation schemes, and for JRA-55 there is good consistency between the pre- and post-satellite era values (Kobayashi et al., 2015). The fold and transport diagnostics, described below, were determined using ERA-Interim on the original 60 hybrid model levels, deemed suitably high vertical resolution for tracking these small-scale features (Škerlak et al., 2014, 2015). The state of ENSO was evaluated using the monthly Oceanic Niño Index (ONI) which is based on a threshold of ± 0.5 ∘C (positive indicating El Niño, negative indicating La Niña) averaged over the Niño 3.4 region (5∘ N–5∘ S, 120–170∘ W; NOAA CPC).

STT-PBL, calculated and presented in Škerlak et al. (2014), was accessed at 1×1∘ horizontal resolution and 6 h time intervals from February–June 1980–2016. STT-PBL is defined to occur when trajectories that originate in the stratosphere cross the tropopause and reach a pressure value greater than that of the PBL top. First, kinematic trajectories were determined every 24 h globally from 50 to 650 hPa using the tool introduced in Wernli and Davies (1997). Stratospheric trajectories, determined using an advanced three-dimensional labeling algorithm (Škerlak et al., 2014), that crossed the tropopause were then identified, and their maximum pressure value tracked as they evolved forward in time. If a trajectory's pressure value exceeded the pressure of the PBL top, then it was flagged and converted to an amount of mass: Δm≈1gΔx2Δp≈6.52×1011 kg, where Δx=80 km and Δp=30 hPa, in the extratropics (Škerlak et al., 2014). PBL height was determined using the 6 h forecast by the ECMWF model. Using PBL height instead of a fixed pressure level to define deep STT improves identification of deep STT over mountainous regions, particularly important for this study.

For a measure of the depth of stratospheric intrusions that lead to STT-PBL over western North America, the tropopause fold identification scheme developed by Škerlak et al. (2015) was used at 6 h temporal resolution and 1×1∘ horizontal resolution. Building upon the fold identification scheme introduced by Sprenger and Wernli (2003), tropopause folds were identified by first distinguishing between air of stratospheric and tropospheric origin on a potential vorticity basis using a three-dimensional labeling scheme. Folds of stratospheric origin are then identified when there were multiple crossings of the tropopause (defined using surfaces of the ± 2 PVU (potential vorticity unit, 1 PVU = 10-6 K m2 kg-1 s-1) surface and 380 K potential temperature, tropopause marked at whichever surface is lower), observed within a single vertical profile. At locations where folds were identified, the minimum and maximum pressure the fold reached was recorded (Fig. S1). To examine how the jet transition, and later ENSO phase, affects the depth of tropopause folds, we tracked the maximum pressure a fold, if identified over the western US (box in Fig. S1), reached for each 6 h time step. The size of folds deeper than 400 hPa was also tracked by counting the number of grid points over the region that were characterized by a maximum pressure greater than 400 hPa. The number of grid points meeting this criterion was then divided by the number of grid points over the western US domain. The number of time steps characterized by a fold with a maximum pressure greater than 400 hPa was also recorded, for a measure of the frequency of tropopause folds under various large-scale conditions. Any mention of tropopause fold frequency therefore refers to only this subset of folds, as folds shallower than 400 hPa are unlikely to affect STT-PBL. Similar results are found when folds larger or deeper than 500 hPa are selected, although differences in the size of intrusions are more difficult to discern given the small-scale nature of intrusions this deep (e.g., Knowland et al., 2017).

We note that the 2-PVU surface, which is used to define the maximum pressure folds reached and marks the “dynamic tropopause” boundary (Hoskins et al., 1985), does not always mark the terminus of the fold, and that ozone originating in the stratosphere more closely follows the 1-PVU surface which penetrates further downward, as shown in Albers et al. (2018; their Fig. 2). Shapiro (1980) also discussed how ozone associated with a tropopause fold in March 1978 reached farther into the troposphere than the dynamic tropopause, indicative of cross-isentropic mixing. Knowland et al. (2017) examined cross sections of two stratospheric intrusions that led to enhanced surface ozone concentrations in Colorado in spring 2012 and showed that the dynamic tropopause only reached to the mid-troposphere, with only small filaments of the intrusion reaching deeper than 500 hPa, while high stratospheric-ozone values extended to the surface. For these reasons, we believe that tracking when the dynamic tropopause is deeper than 400 hPa captures the structures often associated with transport deep into the troposphere. This is confirmed by the strong relationship between STT-PBL and the fold characteristics discussed in Sect. 3.3. Changes in these measures of fold depth, size, and frequency will be evaluated during the phases of the spring transition, as well as during El Niño, La Niña, and ENSO-neutral conditions.

We use the leading empirical orthogonal function (EOF1) and principal component (PC1) time series of the daily mean 200 hPa zonal wind over the North Pacific basin (100–280∘ E, 10–70∘ N), smoothed with a 5 d running mean, to track the seasonal evolution of the jet. Zonal wind anomalies used for EOF analysis were calculated with respect to the February–June 1958–2017 average using JRA-55 reanalysis (black contours, Fig. 1a), with the resultant anomalies intentionally including changes associated with the seasonal cycle (in contrast to the more canonical approach, in which the seasonal cycle is removed). To be consistent with the STT-PBL and tropopause fold datasets, which are based upon ERA-Interim reanalysis (Sprenger et al., 2017), we recomputed the 200 hPa zonal wind EOFs in the same manner using ERA-Interim 200 hPa zonal wind (the correlation between PC1 using JRA55 and ERA-Interim for their common period, 1979–2017, is 0.99).

For a measure of the eddy characteristics and horizontal Rossby wave energy propagation during various phases of the spring transition, the horizontal E vector (Eq. 1; Hoskins et al., 1983) was calculated using daily zonal and meridional wind anomalies that have (a) the 60-year daily climatology and (b) the 11 d running mean removed. Regions where E points eastward (westward) are characterized by meridionally (zonally) elongated eddies (Fig. S2). Negatively tilted anomalies, indicative of cyclonic wave breaking, correspond to a northward-pointed E and energy propagation, while positively tilted anomalies indicate anticyclonic wave breaking and correspond to southward-pointed E and energy propagation. 1E=[v′2-u′2‾,-u′v′‾]

For a measure of confidence in the differences in fold characteristics and STT-PBL during different jet phases and ENSO conditions, mean values were bootstrapped using a sample size Neff= N/tautocorr, where tautocorr is the number of time steps at which the autocorrelation of the variables decreases to below 0.5 and N represents the number of samples in the smallest group being compared. For example, when comparing STT-PBL over North America during May El Niño, La Niña, and ENSO-neutral conditions, with corresponding samples sizes (N) equal to 20, 9, and 7 years, respectively, STT-PBL for the three groups is resampled using N=7 years × 124 time steps/year = 868 time steps and tautocorr = 6 time steps (36 h), which equates to Neff=144 time steps. To fairly compare the confidence intervals for each ENSO group, every group was resampled using the reduced sample size Neff to calculate a new mean STT-PBL value. This process was repeated 10 000 times to determine the 95th and 5th percentiles of the mean value for each group. A similar approach was applied to each variable considered.

The leading EOF pattern of 200 hPa zonal wind tracks the seasonal evolution of the North Pacific jet from February through June each year (Fig. 1). A positive PC1 value represents the stronger wintertime state (Fig. 1a), which gradually weakens on average from about March through June, as shown by the transition of PC1 from positive to negative each year (Fig. 1b). There is greater spread among the PCs of individual years during February, March, and April than there is during May and June, indicating the transition from winter to spring is more variable than the transition from spring to summer. The composite zonal wind on days when PC1 >1σ, hereafter referred to as the winter phase, most often occurring in February–March, is characterized by a strong jet extending well past the date line (Fig. 2a). During the winter phase, high-frequency eddy kinetic energy (EKE) is greatest in the jet exit region in the central Pacific, representing the wintertime Pacific storm track and tendency for eddies to amplify via deformation in the jet exit region (Rivière and Joly, 2006; Breeden and Martin, 2018). The prevalence of equatorward-pointed E, signifying positively tilted waves, over North America is consistent with the frequency of positively tilted troughs and ridges identified during boreal winter by Schemm and Sprenger (2020).

As PC1 decreases to values between ±5σ, which we define as the transition phase, the jet core weakens substantially, while the jet exit region shifts northward (Fig. 2b). The storm track is more energetic throughout the Pacific–North American region compared to the winter phase and shifts northward with the jet exit region. In the east Pacific, a distinct secondary jet maximum develops near Hawaii in the subtropics, creating a double-jet structure in the Pacific Basin which differs substantially from the strong, merged wintertime jet. The formation of this secondary zonal wind maximum was also observed to develop in April by Newman and Sardeshmukh (1998). The magnitude of E increases during the transition phase, with meridionally elongated, positively tilted waves dominating the structure in the midlatitude Pacific. Such characteristics are related to frequent anticyclonic Rossby wave breaking associated with the formation of the two jet maxima (Peters and Waugh, 2003; Pan et al., 2009; Breeden and Martin, 2018) observed in the transition phase. A distinct region of nearly zonal E is observed over the eastern Pacific or western US, indicating waves in this region are more amplified meridionally compared to when the jet occupies the winter phase. Zonal wind, EKE, and eddy amplitude proceed to weaken by late spring or early summer, when PC1 <-1σ, hereafter referred to as the summer phase (Fig. 2c), with the subtropical jet in the eastern Pacific essentially disappearing altogether. Over the western US, eddies are still meridionally amplified but less so than during the transition phase, characteristic of the weakened storm track.

To examine the variability in the spring transition, we tracked the date on which PC1 dropped below +5σ and remained below that value for the remainder of the season. We target this transition in particular given the high variability of PC1 early in the season and the marked invigoration of the storm track associated with PC1 decreasing from strongly positive to neutral. The mean transition date over the 60-year record is 4 April, with a standard deviation of ± 12 d. To test if there are dynamic differences in the transition if it occurs earlier or later than normal, we grouped each spring into early, neutral, and late transition years, requiring early (late) transition years to have a transition date at least 5 d earlier (later) than the 60-year average (Table 1). The average timing of the transition for the three groups differs by about 2 weeks, with the early group transitioning on average in mid-March, the late group in mid-April. Comparing the composite February–June evolution of PC1 for the early and late groups (the neutral transition years fall in-between), PC1 in the early group begins to decrease near the beginning of March, although these differences are not significant until later in the month (Fig. 3). During April, the late group PC1 value is roughly 5σ higher than the early group, with an average zonal wind difference of 10 m s-1 within the jet core, while by May the two groups are indistinguishable from one another. This indicates that an early winter-to-spring transition is not associated with an early spring-to-summer transition, with PC1 decreasing below -5σ in mid-May for all transition groups. To test whether early transitions are more abrupt (and therefore more dynamically disruptive) than later transitions, we compared the composite evolution of PC1 with respect to each year's transition date and did not find any significant differences in the vigor of the transition (Fig. S3).

This section will show how the spring transition modulates STT-PBL over western North America. We find that earlier transitions enhance the amount of the time the jet occupies its transitional phase, corresponding to a more invigorated storm track, more folds, and therefore more STT-PBL than later transitions. Early in the season, deeper folds enhance STT-PBL, while later in the season more expansive folds enhance STT-PBL.

STT-PBL is modulated by the phase of the jet and corresponding invigoration of the storm track (Fig. 4). Transport increases by roughly threefold when the jet is in its transition phase, compared to the composite STT-PBL during both the winter and summer phases. STT-PBL was averaged over western North America (box in Fig. 4b) for each day in the record and subsequently binned by PC1, confirming STT-PBL is strongest when the jet is closer to its transitional phase than at either extremity (Figs. 5a, S4). Both the highest STT-PBL days in the record and the highest median STT-PBL values occur during the transition phase, while the distributions during the winter and summer phases are indistinguishable from one another (Fig. 5a). Consistent with the STT-PBL changes, tropopause folds reach farthest into the troposphere during the transition phase, on average to 450 hPa, in contrast to median values near 400 hPa during the winter phase and 300 hPa during the summer phase (Fig. 5b). During the winter phase, shallow (<300 hPa) and deep (>400 hPa) folds are equally likely, reflecting high wintertime jet variability (Athanasiadis et al., 2010), while deep folds are more frequent than shallow folds during the transition phase. Shallow folds are overwhelmingly more likely during the summer phase, which might be related to the weaker jet and associated ageostrophic circulation during summer. Thus, while the STT-PBL distributions during the winter and summer phases are indistinguishable from one another, the fold depth distributions differ substantially.

In addition to changes in tropopause fold depth during the spring transition, the daytime PBL height in the interior west increases dramatically, meaning shallower folds can reach the top of the boundary layer. Consistent with the STT data, 6 h forecasts of PBL height valid at 18:00 UTC were averaged over western North America and grouped by jet phase (Fig. 5c), confirming the PBL deepens as the jet transitions. Thus, while folds deeper than 500 hPa still occur somewhat frequently during the winter phase, the PBL is far lower, meaning a smaller subset of folds is deep enough to penetrate the boundary layer compared to the transition phase. Conversely, when the jet occupies the summer phase, despite a very deep boundary layer, there is limited transport due to a relative lack of intrusions deeper than 350 hPa. The transition phase is associated with higher STT-PBL through the coincidence of both more frequent deep tropopause folds and a deepening PBL. We note that STT-PBL can be displaced spatially from the position of the tropopause-level folds measured here and can be aided by lower-tropospheric vertical motions such as those occurring around frontal zones and convection (Škerlak et al., 2019, their Fig. 1).

Reconsidering the eddy characteristics associated with the three jet phases (Fig. 2), it appears that tropopause folds deep enough to produce STT-PBL occur most often when waves are highly amplified and the storm track is most energetic. Highly amplified Rossby waves are associated with strong curvature, particularly on the western edge of troughs, producing the subsidence that forms deep tropopause folds and STT-PBL (e.g., Sprenger et al., 2007). Amplified waves propagating over western North America, which occur most often during the transition phase, bring more folds over the high terrain of the Rocky Mountains as the PBL deepens, leading to the STT-PBL maximum observed in boreal spring.

Given the longer period of time the jet is within its transition phase (when PC1 ± 0.5σ), hereafter referred to as the residence time, (Fig. 3), we hypothesize that early transition years are characterized by more STT-PBL than late transition years. To that end, we compared monthly mean STT-PBL for the early and late transition groups, revealing there is indeed more STT-PBL during early transition years in March, April and May, coinciding with more frequent folds deeper than 400 hPa (Fig. 6a–b). In February and March, folds are deeper and larger in early transition years as well, while in later months folds are larger but not deeper (Fig. 6c–d). The residence time of the jet is much greater during early years by definition, and monthly mean EKE over the North Pacific (180–250∘ E, 40–60∘ N; box in Fig. S6) is greater in April during early years (Fig. 6e–f). Compositing each variable with respect to each year's transition date reveals an upward shift in STT-PBL in the 2 weeks following the transition in both groups, coincident with a marked increase in tropopause folds, residence time, and EKE (Fig. 6g, h, k, l). Fold depth and area, conversely, are not systematically affected by the transition (Fig. 6i, j).

The relationship between STT-PBL and each related but distinct fold characteristic – maximum depth, fold area, and frequency – evolves over the course of the spring transition. This is evident from time series of the correlation between monthly mean STT-PBL, February–June, with each fold characteristic, residence time of the jet, and median PBL height (Fig. 7; Fig. S4). During February and March, STT-PBL has the strongest correlation with fold depth and frequency, consistent with intuition. In April, however, the relationship between STT-PBL and fold depth diminishes, while fold frequency maintains a strong relationship with transport through June. In contrast to fold depth, the relationship between fold size and STT-PBL is strongest in May, when it has the second-strongest relationship after frequency. A longer residence time of the jet within the transition phase enhances STT-PBL in March and February, a relationship which disappears later in spring, in part because PC1 continues decreasing towards its summertime state. Daytime PBL height and STT-PBL are modestly correlated in February and March, with no relationship in later months when the daytime PBL has deepened to several kilometers and appears to no longer be the limiting factor for deep transport (Fig. 5c; Seidel et al., 2012; Langford et al., 2017). Since fold frequency maintains a strong relationship with STT-PBL throughout the transition, we correlated fold frequency to EKE, confirming that a more energetic storm track is associated with more folds and supporting the relationship between storm track variability, folds, and STT-PBL. The correlation drops off by May, however, for reasons which are not immediately clear but might reflect the more convective nature of transport during this time of year, which can be important for transport to the surface (Langford et al., 2017; Škerlak et al., 2019). Overall, fold frequency maintains the strongest relationship with STT-PBL throughout the transition, while fold depth and area also affect STT-PBL early and late in the transition, respectively.

What drives the substantial variability in the timing of the spring transition (Fig. 3)? While prior research has alluded to a connection between ENSO and STT-PBL, the precise nature of the ENSO–fold–STT relationship during boreal spring is not fully understood. Here we demonstrate that ENSO conditions do influence the jet, tropopause fold characteristics, and STT-PBL and that this influence evolves throughout the spring transition.

ENSO markedly affects the jet from February–April, with La Niña conditions corresponding to a much lower PC1 value than neutral or El Niño conditions, while in May and June the differences are weaker (Fig. 8). There is some asymmetry in the PC1 response, with La Niña weakening the jet more substantially than El Niño strengthens it. Given the positive relationship between STT-PBL and residence time of the jet, we hypothesize that La Niña conditions are associated with enhanced STT-PBL, which is broadly confirmed in Figs. 9–10 and is consistent with the conclusions of Lin et al. (2015). In February and March, El Niño conditions are associated with a zonally extended jet that connects to the jet over North America, while the jet is zonally confined to the central Pacific during La Niña (Fig. 9a–f). STT-PBL is overall weak in February but is strongest during La Niña years, consistent with the most neutral PC1 value. STT-PBL increases in all three ENSO groups during March, as PC1 values decrease. Note that the jet has already transitioned during many of the La Niña and some of the ENSO-neutral years (Table 1). In April, the jet transition is either underway or has already occurred, and correspondingly STT-PBL peaks for the El Niño and ENSO-neutral groups and remains elevated for La Niña (Figs. 9g–i, 10a). During May and June, STT-PBL remains elevated during La Niña years, although the difference compared to ENSO-neutral is somewhat uncertain with the number of samples available (Figs. 9j–o, 10a). The (presumably eddy-driven) jet core in the western Pacific is stronger during La Niña years, reflecting an increase in storm track activity compared to neutral and El Niño conditions coincident with a more negative PC1 value (Figs. 10f,  S6).

Which of the various tropopause fold characteristics explored in the prior section do ENSO conditions affect? Just as the influence of ENSO on PC1 evolves over the course of the transition, so too does the influence of ENSO on tropopause folds. During February and March, La Niña conditions are characterized by significantly deeper and more frequent folds, driving an increase in STT-PBL (Fig. 10a–c). Folds are also larger, particularly in May when the relationship between fold area and STT-PBL is the strongest (Fig. 10d). While STT-PBL during April is similar in all three groups, folds are still more frequent and potentially deeper during La Niña (Fig. 10b–c). STT-PBL is elevated during La Niña in May, when folds are more common and larger in areal extent. Mean fold depth, in contrast to fold size, is insensitive to ENSO phase in May and June. Finally, the residence time of the jet within its transition phase is significantly enhanced in February and March (as suggested in Fig. 8), while it is reduced in May when PC1 is more negative (Figs. 10e, 9l). EKE during La Niña is most enhanced in April, similar to the early transition years (Figs. 10f, 6f), due to the notable increase in EKE following the transition (Fig. 10l, 6l). The E-vector differences indicate more amplified and positively tilted waves during La Niña, reflecting the enhanced anticyclonic Rossby wave breaking occurring during this phase (Fig. S6). Note that during May, EKE is enhanced during La Niña in a smaller region over the eastern Pacific (Fig. S6d), accompanied by a more zonal E vector, consistent with elevated STT-PBL observed in Fig. 10a but not reflected in Fig. 10f.

In summary, in late winter or early spring, the teleconnection to the extratropics during La Niña projects onto the seasonal transition of the jet represented by PC1, often expediting the transition. This large-scale modulation, in turn, enhances the depth of tropopause folds and fold frequency over western North America, enhancing STT-PBL. In May, La Niña conditions continue to increase the frequency and size of folds and therefore STT-PBL, also through invigoration of the storm track as in February and March (Fig. S6). An opposite response is observed during El Niño conditions.