The Role of Local and Remote Processes for Wintertime Surface Energy Budget Extremes over Arctic Sea Ice

a. Data

The study is based on ERA5, the latest reanalysis dataset produced by the European Centre for Medium-Range Weather Forecasts (ECMWF; Hersbach et al. 2020). We use hourly analysis fields on model levels for extended winters (November–March) between 1979 and 2020, with the first winter starting in November 1979. Fields have been interpolated in the horizontal to a 0.5° × 0.5° longitude–latitude grid.

b. SEB life cycle events

We use multiday life cycle events (LCEs) of extreme positive SEB anomalies over Arctic sea ice as introduced by Murto et al. (2023). In the following, we briefly outline the approach for the identification of LCEs; for details we refer to the aforementioned study. Each LCE consists of a sequence of coherent regions of extreme positive SEB anomalies, hereafter referred to as SEB anomaly patches. These SEB anomalies are defined relative to an SEB climatology that takes the seasonal cycle and the long-term trend into account, which ensures a uniform distribution of events across the study period. SEB anomaly patches are then defined as spatially coherent areas where the standardized SEB anomalies exceed a seasonally varying 95th-percentile threshold. That way, only the most extreme SEB anomalies are selected. These patches are then tracked in space and time to form LCEs. This is done iteratively starting from the patches with the largest mean SEB anomalies and area. Specifically, a candidate patch is connected to a track if it satisfies the following two criteria with respect to a parent patch that has already been connected to the track. First, the two patches need to be spatially close to each other; that is, the centers of mass must be within two equivalent radii of the parent patch, whereby the equivalent radius of a patch corresponds to the radius of a circle with the same area as the patch. Second, the centroids of air parcels arriving at 200 hPa above ground level (AGL) at the two patches must be located within 1500 km from each other two days prior to the occurrence of the patches. Following this procedure, we identify 142 LCEs, 72 of which could be attributed to an Atlantic airmass origin and 60 to a Pacific origin based on the location of the trajectory centroids two days prior to arrival (cf. Figs. 2a and 5e in Murto et al. 2023). For simplicity, LCEs with airmass origin over Siberia (2 events) or North America (8 events) are not considered here. The LCEs have a typical lifetime of 3–7 days and affect an area of several 10⁶ km². If not mentioned otherwise, the analysis presented in this study includes all SEB anomaly patches attributed to all or selected LCEs. Note that even though we do not explicitly consider the time evolution of LCEs (i.e., the life cycle of events from onset to decay), we still use the term LCE for consistency with Murto et al. (2023).

c. Air-parcel trajectories

For the analysis of the origin of the air in the tropospheric column collocated with LCEs and for quantifying the contributions of different processes to the generation of θ* associated with these events, we compute kinematic backward trajectories from the tropospheric column collocated with the SEB anomaly patches using the Lagrangian Analysis Tool (LAGRANTO; Wernli and Davies 1997; Sprenger and Wernli 2015). For each SEB anomaly patch, trajectories are initialized on an equidistant grid with 50 km grid spacing in the horizontal and at 50, 100, 200, 300, 400, 500, and 600 hPa AGL in the vertical.¹ Initialization times of the trajectories are at 0300, 0900, 1500, and 2100 UTC on the day when the SEB anomaly patch occurs. Trajectories are calculated backward in time for 10 days and trajectory positions (longitude, latitude, pressure) as well as additional variables interpolated to these positions (i.e., potential temperature, potential temperature climatology) are stored in 3-hourly intervals. Note that we use hourly wind fields for the computation of the trajectories.

Complementary to the set of LCE trajectories, an additional set of reference trajectories is computed, which is meant to approximate the climatology of air parcel trajectories in the Arctic (pseudoclimatology). To account for the spatially nonuniform distribution of SEB anomaly patches across the Arctic, reference trajectory starting points are chosen to be identical to those associated with the SEB anomaly patches. For the reference trajectory calculation, we use data from the same calendar day but a randomly chosen year other than the actual year when the patch occurred. To increase the representativeness, this procedure is repeated twice such that the pseudoclimatology is obtained from averaging over twice as many trajectories as contained in the LCE trajectory dataset.

d. Potential temperature anomaly budget

Finally, the diabatic tendencies are further decomposed into contributions from latent heating in the free troposphere, heating in the planetary boundary layer, residual diabatic heating, and diabatic cooling. For that purpose, incremental changes Δθ_i = θ(t_i+1, x_i+1, p_i+1) − θ(t_i, x_i, p_i) are uniquely attributed to one of these categories according to the following criteria:

1. Latent heating: Δθ_i > 0 K and Δq_i < 0 g kg⁻¹ h⁻¹ and $p_i-p_i^{\text{pbl}}\le 0\text{hPa}$ and Δp_i < 0 hPa h⁻¹

2. Boundary layer heating: Δθ_i > 0 K and $p_i-p_i^{\text{pbl}}>0\text{hPa}$

3. Diabatic cooling: Δθ_i ≤ 0 K

where q is specific humidity and p^pbl the pressure at the top of the boundary layer. The latter is interpolated from the diagnostic height of the planetary boundary layer available in the ERA5 archive. The rationale behind the first criterion is that condensation in the free troposphere is associated with ascending air and a decrease of specific humidity. The second criterion captures situations where an air parcel is located in the planetary boundary layer and experiences heating, for example by sensible heat fluxes such as in a marine cold air outbreak. The remaining positive θ changes that are not attributed to the latent heating or boundary layer heating categories are denoted as residual diabatic heating. A caveat of this approach is that each incremental change of θ is uniquely attributed to one category. In reality, however, several physical processes are at work at the same time; for example, longwave radiative cooling will occur concomitant with latent heating partially compensating each other. Note that our use of these categories primarily lies in studying diabatic heating contributions in specific flow situations rather than due to different physical processes. Finally, we define t_genesis as the last time step along the trajectory when θ* has the opposite sign as ${\theta }^{*}(t=0\mathrm{h})$ (or is exactly 0 K). In the evaluation of the θ* budget we will mostly consider the time interval from t = t_genesis to t = 0 h.

e. Interpretation of the θ* budget

The processes generating positive θ* along air parcel trajectories directed toward the Arctic are schematically illustrated in Fig. 2. The transport terms represent the gain of θ* resulting from the transport of air from a region with climatologically high to a region with climatologically low potential temperature. Thereby, we distinguish the crossing of climatological isentropes in the horizontal (dark blue arrow in Fig. 2) and the vertical (light blue arrow in Fig. 2) to separate the effect of transport from lower latitudes into the Arctic from subsidence or ascent. Note that in this framework, flow along slanted climatological isentropes is associated with equal but opposite contributions from horizontal and vertical transport.

View Full Size

Schematic illustrating processes along air parcel trajectories causing positive potential temperature anomalies (θ*) in the Arctic. The transparent gray surface indicates a climatological isentrope and arrows illustrate various processes generating positive θ*, i.e., horizontal and vertical transports across climatological isentropes (dark and light blue, respectively), as well as diabatic processes including latent heat release (red) and turbulent fluxes into the boundary layer (orange).

Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0883.1

• Download Figure
• Download figure as PowerPoint slide

For quantitatively understanding the relevance of horizontal transport, it provides insight to consider Fig. 3a, which quantifies the θ* acquired due to horizontal (isobaric) transport at 700 hPa (see Fig. 1 in the online supplemental material for additional levels) from a given geographical location into the polar cap poleward of 80°N assuming no diabatic processes (e.g., radiative cooling and/or mixing occurring along the transport). For instance, θ* of an air parcel moving from the British Isles to the high Arctic increases by around 20 K. The figure clearly reveals the climatological dome of polar cold air. Air parcels crossing this dome in the poleward direction experience a substantial gain in θ*. Furthermore, the pattern also reveals the well-known stationary wave patterns with the polar dome extending farther south over the eastern continents and farther north over the eastern ocean basins. The stationary wave pattern further leads to a zonal component of the gradient over the North Atlantic and to a lesser extent over the North Pacific such that air parcels moving zonally in these regions experience a decrease of θ* due to horizontal transport (cf. with black contours). In addition, the climatological potential temperature contrast between the subpolar oceans and the high Arctic is considerably stronger in the lower than the upper troposphere (cf. supplemental Fig. 1).

View Full Size

Generation of θ* due to horizontal and vertical transport. (a) θ* acquired by an air parcel transported isobarically at 700 hPa from a given geographical location into the polar cap poleward of 80°N with the 6- and 18-K and the 12-K contours highlighted by thin and thick black lines, respectively. The figure is obtained by subtracting the polar cap (≥80°N) average of the potential temperature climatology θ_c from θ_c at a given geographical location. Note that variations of θ_c within the 80°N polar cap are small as compared to θ* acquired due to transport from lower latitudes into the polar cap. (b) θ* acquired by an air parcel subsiding from 700 to 800 hPa at a fixed geographical location. Both panels show averages over extended winters 1979–2020 and the polar cap 80°N is indicated by a black circle.

Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0883.1

• Download Figure
• Download figure as PowerPoint slide

Similarly, Fig. 3b shows the gain in θ* due to subsidence from 700 to 800 hPa at a given geographical location. An air parcel descending from 700 to 800 hPa gains θ* of about 5–7 K due to vertical transport in most of the regions. Variations in the efficiency of subsidence in creating θ* are determined by variations of the climatological stratification, i.e., $\partial {\theta }_c/\partial p$. Consequently, subsidence is particularly efficient in creating large θ* along the slopes of Greenland’s ice sheet and the coldest regions over the eastern continents where the climatological stratification is high, whereas it is considerably less efficient over the Nordic seas and near Kamchatka.

Finally, θ* can increase due to diabatic heating such as the release of latent heat, for example as warm and humid midlatitude air impinges on the dome of polar cold air, ascends, and cools, followed by cloud formation (red arrow in Fig. 2), or sensible heat fluxes from the ocean to the atmosphere when Arctic air sweeps over a comparatively warm ocean surface (orange arrow in Fig. 2).

a. Life cycle event 44

The first case, LCE 44 [numbering according to Murto et al. (2023)] in December 1986, was accompanied by one of the most severe near-surface warm extremes in the Arctic in recent decades (Messori et al. 2018). The affected area extended in a narrow plume from the ice-covered portions of the Greenland Sea across the North Pole toward Siberia [see also Murto et al. (2023) for a discussion of its evolution and characteristics]. The LCE formed in conjunction with a moist-air intrusion from the Atlantic sector that reached the area affected by the LCE (purple hatched in Fig. 4) at 200 hPa AGL and above (see supplemental Fig. 2 and animations in the online supplemental material).

View Full Size

Case study of LCE 44 showing the positions of trajectories initialized at 50 hPa AGL from all patches associated with the LCE. Maps are shown for lags of (a) −8, (b) −6, (c) −4, and (d) −2 days prior to 1200 UTC 12 Dec 1986, corresponding to the peak of the LCE. Trajectory positions are colored according to pressure. Additionally shown are sea level pressure (gray contours at intervals of 10 hPa; blue and gray shading for values below 990 hPa and above 1030 hPa, respectively) and the 5000 m (blue thin), 5300 m (blue thick), and 5600 m (blue thin) isohypses of 500 hPa geopotential height. The purple hatched area indicates the region affected by the LCE and the solid gray line shows the 60°N latitude circle. Note that only positions from every second trajectory are drawn for better visibility.

Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0883.1

• Download Figure
• Download figure as PowerPoint slide

In contrast to the air farther aloft, the near-surface air originated in the Arctic. Figure 4 shows the location of air parcels during the days prior to arriving at 50 hPa AGL at the LCE. Eight days prior to the LCE’s peak, air parcels were mainly located poleward of 60°N at an elevation above 850 hPa (Fig. 4a). Subsequently, they descended and moved westward over open ocean in the Barents Sea (Figs. 4b,c), turned poleward and re-entered the Arctic upon which they moved across the pole in a narrow plume (Fig. 4d). In addition, a few air parcels originated over the Canadian Archipelago and went around the southern tip of Greenland, thus leaving and returning to the Arctic. Furthermore, others descended from above the Greenland ice sheet (Figs. 4b,c). Two days prior to the peak of the LCE, near-surface air parcels have merged into two coherent blobs, one near Fram Strait and the other north of Siberia, already deep within the high Arctic (Fig. 4d). Thus, the near-surface air associated with the exceptional warmth had a very different history than the air farther aloft, which has the characteristic evolution of midlatitude moist-air intrusions (cf. Woods and Caballero 2016). As a result, the mechanisms that caused θ* likely differ between the near-surface air and the air farther aloft.

Figures 5a and 5c show the fractions of trajectories with t ≥ t_genesis for trajectories arriving at 50 and 500 hPa AGL and the evolution of θ* (black line). Recall that t_genesis refers to the last time when θ* has the opposite sign than the final anomaly. For most of the trajectories θ* started to form less than 4 days prior to the LCE with a longer tail (i.e., earlier onset of θ*) for lower tropospheric air parcels (gray shading in Figs. 5a,c). For about 50% of the air parcels, the anomalies formed only within 48 h prior to the LCE. Furthermore, the final θ* (at t = 0 h) associated with near-surface air parcels was substantially larger than for air parcels aloft (more than 10 K vs roughly 7 K).

View Full Size

θ* budgets for LCEs (left) 44 and (right) 137. Shown is the temporal evolution of θ* (black) and of accumulated contributions to θ* from horizontal transport (dark blue), vertical transport (light blue), latent heating (red), boundary layer heating (orange), residual diabatic heating (yellow), and diabatic cooling (purple) for trajectories initialized at (a),(b) 500 and (c),(d) 50 hPa AGL (axis on the right) from all SEB anomaly patches associated with the given LCE. For each trajectory, terms are set to zero for t = t_genesis and accumulated for the period 0 ≥ t ≥ t_genesis. The resulting time series are then averaged over all trajectories. Gray bars show the fraction of trajectories with t ≥ t_genesis (axis on the left). t = 0 h corresponds to the time of arrival in the LCE and t = t_genesis to the time when θ* was last ≤0 K.

Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0883.1

• Download Figure
• Download figure as PowerPoint slide

The temporal evolution of the contributions of the θ* budget terms accumulated over the period from t_genesis to t = 0 h shows that θ* of near-surface air parcels formed mainly through horizontal and vertical transport, as well as diabatic heating in the planetary boundary layer (Fig. 5c), which occurred when the air parcels transited open ocean in the Barents Sea (see the animation in the supplemental material). Note that in this region, horizontal transport is particularly efficient in creating θ* due to the large near-surface climatological temperature contrast between the Barents Sea and the high Arctic (supplemental Figs. 1a,b). Substantial amounts of these positive contributions were balanced by diabatic cooling. Similar to near-surface air parcels, air parcels aloft gained θ* primarily via horizontal transport, followed by latent heating (Fig. 5a). Since latent heating is largely associated with ascent and adiabatic cooling, the latent heating contribution was entirely consumed by a negative contribution from vertical transport.

b. Life cycle event 137

The second case, LCE 137, was associated with meridional transport from the Pacific sector aloft and a persistent anticyclone within the Arctic (Fig. 6). Note that the anticyclone was confined to the lower troposphere with no upper-level anticyclone present (cf. SLP contours and 500 hPa isohypses in Fig. 6). With a very persistent anticyclone located within the Arctic, this event differs from the typical LCE as described in Murto et al. (2023), yet it illustrates how diverse the processes can be that lead to extreme SEB anomalies over Arctic sea ice.

View Full Size

As in Fig. 4, but for LCE 137 with lags relative to the peak of that LCE at 1200 UTC 3 Jan 1994.

Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0883.1

• Download Figure
• Download figure as PowerPoint slide

Air parcels aloft originated in the Pacific sector and were episodically injected into the high Arctic associated with transient extratropical cyclones and tropospheric cutoffs (see supplemental Fig. 3 and supplemental animations). In contrast, the bulk of near-surface air parcels was located within the Arctic 8 days prior to the peak of the LCE (Fig. 6a). Three bundles of air parcels can be discerned: one collocated with the anticyclone at roughly 700 hPa, one over Siberia also at midtropospheric elevations, and one over and north of Alaska located at low altitudes. The air parcels associated with the latter two bundles were gradually swept up by the anticyclone over the Beaufort Sea, in which they remained trapped and subsided (Figs. 6b–d).

Considering the θ* budget, Figs. 5b and 5d show that θ* formed over a longer time period compared to LCE 44, as evident from the longer tail of the fraction of trajectories with t ≥ t_genesis. Furthermore, final θ* was almost twice as large aloft than near the surface. While aloft the θ* budget terms show a qualitatively similar evolution as in the case of LCE 44 (Figs. 5a,b), positive contributions for near-surface air parcels were dominated by vertical transport, while horizontal transport was less important, and diabatic warming was negligible (Fig. 5d). Thus, in this case, the near-surface air acquired its θ* primarily via subsidence in the anticyclone. A substantial portion of the θ* increase due to vertical transport was, however, compensated by diabatic cooling.

c. Summary of case studies

Comparing the evolution of the θ* budget terms for air parcels near the surface and aloft for the two case studies reveals several interesting characteristics. First, either horizontal or vertical transport dominates the positive contributions at all levels. In particular, this also applies to the near-surface air with Arctic origin, which is required to reach an area with substantially higher θ_c than in the high Arctic before θ* starts to form. Second, θ* arises as a residual between opposing processes, whose individual contributions typically exceed the magnitude of θ*. Third, heating in the planetary boundary layer can contribute to the warming of the near-surface air, but its contribution to the final θ* remains smaller than that from transport. That said, heating in the planetary boundary layer, especially due to surface turbulent heat fluxes over open ocean, might be more important for transforming the originally Arctic near-surface air while it moves into a region with warmer θ_c before θ* starts to form, as well as for eroding an initially negative θ*. This possibility will be explored to a greater extent in the climatological part of the study.

a. Air parcel origin and evolution of θ*

Before turning attention to the climatological θ* budgets, we first consider when and where positive values of θ* form. The so obtained temporal and spatial scales of θ* formation will be helpful in interpreting the θ* budgets. Figure 7 shows the temporal evolution of θ* and histograms of t_genesis for trajectories initialized at 500 and 50 hPa AGL.

View Full Size

Histograms of t_genesis for all Atlantic and Pacific LCEs (gray bars) and temporal evolution of mean (solid lines) and interquartile range (shading) of θ* for Atlantic (blue) and Pacific (red) LCEs for trajectories initialized at (a) 500 and (b) 50 hPa AGL. Dashed black lines show the 75th percentile of θ* associated with climatological trajectories (see section 2c). Bars are shown every 6 h and the leftmost bar includes trajectories with t_genesis ≤ −240 h.

Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0883.1

• Download Figure
• Download figure as PowerPoint slide

Near-surface air parcels tend to be anomalously cold 10 days prior to arrival at the LCE. The initial cold anomaly is slightly more pronounced for Atlantic than Pacific events and the anomalies form over a shorter time period in the former group. The anomalies emerge between t = −72 and −48 h in the mean (Fig. 7b) and they exceed ordinary variations of θ* as described by the 75th percentile of θ* along reference trajectories (dashed line) at about t = −48 h. Hence, air parcels become unusually warm only 2 days prior to arrival. The fraction of air parcels for which t_genesis ≤ −240 h (i.e., for which the generation of θ* cannot be fully captured by 10-day backward trajectories) is acceptably small (3%–4% of all trajectories). Considering the distribution of t_genesis further reveals a long tail indicating that for some trajectories the formation of the anomalies occurs over considerably longer time periods than in the mean, such as in the case of LCE 137 discussed in section 3b. Unlike lower tropospheric trajectories, mid- and upper tropospheric trajectories associated with Atlantic LCEs feature only weak initial cold anomalies and none for Pacific LCEs (Fig. 7a). Furthermore, anomalies tend to form slightly earlier and the distribution of t_genesis is wider with an earlier peak.

Complementary to the temporal scales over which θ* forms, we consider the geographical origin of LCE air parcels and their location at t = t_genesis. For that purpose, Fig. 8 shows trajectory densities at t = −240 h and at t = t_genesis for trajectories associated with Atlantic and Pacific LCEs and initialized at 50, 200, and 500 hPa AGL. At t = −240 h, near-surface air parcels are mainly located in the Arctic, including the high Arctic, Eurasia, and the Canadian Archipelago (Figs. 8a,d) consistent with the two case studies. Only a small portion of air parcels originate over the North Atlantic and North Pacific Oceans.

View Full Size

Spatial distribution of trajectories initialized at (a),(d) 50, (b),(e) 200, and (c),(f) 500 hPa AGL associated with (a)–(c) Atlantic and (d)–(f) Pacific LCEs. Trajectory positions are shown for t = −240 h [red hatched; contours at 1% and 2% (10⁶ km²)⁻¹], t = t_genesis (shading), and t = 0 h [black; contour at 1% (10⁶ km²)⁻¹]. Further shown in (b) and (e) are the 6-, 12-, and 18-K contours of the gain of θ* due to horizontal transport at 700 hPa from a given geographical location into the 80°N polar cap (thin and thick black lines; as in Fig. 3a).

Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0883.1

• Download Figure
• Download figure as PowerPoint slide

For Atlantic LCEs and near-surface air parcels, the anomaly starts to form mainly over open ocean in the Nordic seas and the Barents Sea, while for Pacific LCEs anomaly genesis occurs in the northern North Pacific, east of Kamchatka and near the Bering Strait. All these regions are climatologically warmer than the predominant origin of the air parcels (i.e., their location at t = −240 h) as well as the regions with Arctic sea ice (see black contours in Figs. 8b,e and 3a). It is worth noting that trajectory densities at t = t_genesis are sharply delimited by the sea ice edge and landmasses. This hints at the relevance of diabatic heating via turbulent surface heat fluxes for the erosion of the originally negative θ* and counteracting the effect of horizontal transport from the cold Arctic origin to the climatologically warmer ocean regions at t = t_genesis.

At t = −240 h, air parcels initialized at 200 and 500 hPa AGL are located only 5°–15° latitude equatorward of the latitude at which they arrive at the LCE (supplemental Fig. 4b). However, they undergo a stronger equatorward excursion (supplemental Fig. 4b) and the great circle distance between origin and arrival is considerably larger as compared to near-surface air parcels (supplemental Fig. 4a) since the former originate predominantly upstream of the Atlantic and Pacific gateways into the Arctic. More specifically, air parcels originate primarily over Canada, Labrador and Greenland for Atlantic LCEs (Figs. 8b,c) and eastern Eurasia for Pacific LCEs (Figs. 8e,f). During the first 5–7 days, they tend to move equatorward, followed by a steep poleward trajectory during the final days (supplemental Fig. 4b). The locations of air parcels at t = t_genesis arriving at 200 and 500 hPa AGL lie downstream of the southern inflection point of the trajectory in the respective sector of the Arctic (Figs. 8b,c) and the North Atlantic and the North Pacific ocean basins (Figs. 8e,f), respectively.

In line with the case studies, the different origins of near-surface air parcels and air parcels further aloft suggest that LCEs do not result from deep vertically coherent injections of midlatitude air into the Arctic. Instead, they are systematically caused by the vertical superposition of originally Arctic air in the lowermost troposphere and midlatitude air aloft.

This finding is further corroborated by the evolution of the great circle and latitudinal distances between the centroids of trajectories initialized near the surface and further aloft (supplemental Figs. 4c,d). The great circle distance decreases from originally more than 2500 km between near-surface air parcels and air parcels at 500 hPa AGL to below 500 km only between t = −36 and 0 h (supplemental Fig. 4a). Thus, the coherent transport of the airstreams across the tropospheric column is only a short phase right before the occurrence of an LCE. Likewise, the evolution of the latitudinal distance shows a similar rapid decrease during the final 2 days (supplemental Fig. 4b). Furthermore, the latitudinal distance between near-surface air parcels and air parcels aloft has pronounced peaks at about t = −120 and −72 h for Atlantic and Pacific LCEs, respectively, reflecting that air parcels aloft first move from their origin equatorward before turning poleward (supplemental Figs. 4b,d; also compare Figs. 8c,f).

b. θ* budgets

Given the different origins of air parcels and the location where θ* starts to form, we now aim to shed light on the processes that cause θ*. For that purpose, we first consider the θ* budget for the reference trajectories (see section 2c). There, a balance exists between positive and negative contributions to θ* (supplemental Fig. 5a). This balance is essentially set by the mean circulation in the Arctic: midlatitude air is transported poleward, whereby it ascends into the mid- and upper troposphere (negative contribution from vertical transport), indicating that the flow is at least partially along slanted climatological isentropes. Adiabatic cooling during ascent facilitates condensation, which in turn compensates part of the negative contribution from vertical transport. From the mid- and upper troposphere, the air then subsides within the Arctic into the lower troposphere while it cools radiatively such that the first-order balance is between diabatic cooling and vertical transport. Any θ* forming along LCE trajectories must, thus, result from departures from this balance between positive and negative contributions to θ* along reference trajectories (supplemental Figs. 5b,c).

Figures 9a and 9b show the magnitude of θ* at the time of the LCEs (t = 0 h; diamonds) and the θ* budget accumulated over the time period t = t_genesis to t = 0 h (bars) for Atlantic and Pacific LCEs, comprising on average about 3 to 4 days (see section 4a). This reveals that θ* peaks in the lower troposphere with more than 10 K. In the case of Atlantic LCEs, θ* declines with altitude to 6–7 K at 600 hPa AGL, whereas the decrease is more modest for Pacific LCEs (see also Fig. 7a in Murto et al. 2023).

View Full Size

(a),(b) Accumulated contributions of horizontal transport (dark blue), vertical transport (light blue), latent heating (red), boundary layer heating (orange), residual diabatic heating (yellow), and diabatic cooling (purple) to the final θ* (diamonds) for all (a) Atlantic and (b) Pacific LCEs. For each trajectory, the terms are accumulated from t = t_genesis to t = 0 h. (c),(d) As in (a) and (b), but for the difference between trajectories with the 20% largest (upper quintile) and smallest (lower quintile) θ* at t = 0 h. Quintiles are defined by level.

Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0883.1

• Download Figure
• Download figure as PowerPoint slide

For LCEs, horizontal transport dominates the positive contributions at all levels. This is consistent with air parcel locations at t = t_genesis (Fig. 8), which implies that air parcels must cross the climatological dome of cold air as they track poleward (see also the 12-K contour shown in Figs. 8b,e). This crossing of the climatological dome of polar cold air occurs during the final ∼48 h prior to an LCE when lower and upper tropospheric air parcels approach each other and move fairly coherently poleward (cf. supplemental Fig. 4a). Similar to the reference trajectories, vertical transport contributes positively in the lower and negatively in the upper troposphere reflecting descent and ascent, respectively, and with stronger positive contributions in the case of Pacific LCEs compared to Atlantic LCEs. Further positive contributions stem from latent heating in the mid- and upper troposphere, as well as from heating in the planetary boundary layer, the latter more for Atlantic than Pacific LCEs. Finally, diabatic cooling balances a substantial part—but not all—of the positive contributions across all levels, with larger magnitudes in the lower troposphere. Comparing with the reference trajectories, this suggests a decisive role for horizontal transport for the formation of θ* associated with LCEs.

Insight into the spatial variability of the θ* budgets for LCEs is obtained from Lagrangian forward projections of the individual θ* budget terms (Fig. 10 and supplemental Fig. 6). Lagrangian forward projections are computed by first assigning a given characteristic quantity of a trajectory, such as the θ* budget contributions, to the trajectory location at t = 0 h (Liniger and Davies 2003), and then averaging over all values assigned to the same location. Thus, the maps in Fig. 10 show, for a given location and altitude, the average magnitude of θ* and how much a particular process contributes to its formation. This shows that lower tropospheric θ* is largest near the sea ice edge, reaching up to 16 K, and declines with fetch over sea ice (Figs. 10a,i). The reasons for the existence of this gradient are twofold. First, the horizontal transport contribution declines as the air traverses the high Arctic and eventually moves across climatological isentropes from cold to warm (Figs. 10b,f,j,n). Second, diabatic contributions become more negative with fetch (Figs. 10d,l) due to prolonged diabatic cooling and the absence of any notable diabatic warming (supplemental Figs. 6d,h,l,p). This gradient of θ* amplitude is much less pronounced in the upper troposphere (Figs. 10e,m), primarily due to weaker diabatic cooling contributions (Figs. 10h,p) and a reduced or even reversed gradient in the contributions of horizontal transport across the Arctic (Figs. 10f,n). Finally, positive contributions to lower tropospheric θ* due to vertical transport tend to peak along the coastlines and orography, particularly north of Alaska and Greenland (Figs. 10c,k), while in the upper troposphere, vertical transport contributions are almost uniformly negative (Figs. 10g,o).

View Full Size

Lagrangian forward projections of θ* budget terms for (a)–(h) Atlantic and (i)–(p) Pacific LCEs and trajectories initialized at (a)–(d) and (i)–(l) 50 and (e)–(h) and (m)–(p) 500 hPa AGL. Shown are (first column) θ* at t = 0 h, and the contributions of (second column) horizontal transport, (third column) vertical transport, and (fourth column) diabatic processes to θ* accumulated over t ≥ t_genesis. The black contour shows the spatial distribution of trajectories at t = 0 h [contour at 1% (10⁶ km²)⁻¹].

Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0883.1

• Download Figure
• Download figure as PowerPoint slide

c. Peculiarities of SEB trajectories

As for the case studies, the θ* budgets again reveal a clear compensation between the contributions of different processes such that the final θ* emerges as a residual, which is notably smaller than the contributions from the individual processes. Compensation results from the anticorrelation between the contributions from certain physical processes: Latent heat release is closely linked to ascent, such that in the upper troposphere positive contributions from latent heat release are often counteracted by negative contributions due to vertical transport. Furthermore, diabatic cooling often goes along with subsidence and as a result vertical transport compensates some of the diabatic cooling in the lower troposphere. Nevertheless, this decomposition of θ* by processes does not provide quantitative information about which processes have unusual contributions as compared to the average Arctic air parcel, which does not obtain a strongly positive θ*.

Some insight into the unusualness can be gained, however, by stratifying trajectories according to the magnitude of the final anomaly and comparing the contributions. We do so by considering the differences of the contributions of LCE trajectories with the 20% largest (upper quintile) and smallest (lower quintile) θ* at t = 0 h among all LCE trajectories at a given level (Figs. 9c,d). Note the pronounced difference of θ* at t = 0 h of 10–17 K between the two sets (diamonds). Even though some cancellation between terms remains, the cancellation is clearly less than the magnitude of the anomalies, with the exception of the most elevated air parcels.

Across levels, horizontal transport is the dominant term contributing additional θ* (Figs. 9c,d). For upper tropospheric trajectories, the figures further reveal more negative contributions from vertical transport for trajectories with large θ*. Hence, upper tropospheric trajectories with a large θ* (upper quintile) ascend more since t_genesis than trajectories with a weaker θ* (lower quintile). Related to the enhanced ascent, the former also experience more condensation, as reflected by additional positive contributions from latent heating. In the lower troposphere, there is much less cancellation and the additional θ* is mainly caused by stronger horizontal transport contributions, and in the case of Pacific LCEs also vertical transport. The latter is likely linked to anomalously high frequency of blocking over Alaska preceding Pacific LCEs and associated subsidence (supplemental Fig. 7d). Interestingly, also diabatic cooling is generally more intense for warmer trajectories, which on one hand could be due to stronger longwave emission resulting from higher temperature and water vapor content, and on the other hand also a longer timespan over which the anomaly forms.

In summary, this analysis emphasizes the relevance of horizontal transport across the climatological horizontal temperature gradient associated with the polar dome of cold air for the generation of the θ* between t = t_genesis and t = 0 h. Other processes contribute less to the formation of θ* and they can enhance or oppose the formation of θ* depending on the vertical level and geographical region.

d. Role of airmass transformations en route

Even though the θ* changes prior to t_genesis do not contribute to the final θ*, their analysis is nevertheless insightful for understanding the chain of events that leads to large positive near-surface θ* during LCEs for several reasons. First of all, near-surface air parcels tend to have a slightly negative θ* 10 days prior to an LCE (Fig. 7). More importantly, though, the Arctic origin of near-surface air parcels (location at t = −240 h; red hatching in Figs. 8a,d) implies that air parcels must first move from a region that features comparable climatological temperatures as the sea ice–covered portion of the Arctic to the climatologically warmer subpolar oceans, where the air parcels preferentially reach a zero θ* for the last time before the LCE (location at t = t_genesis; shading in Figs. 8a,d). Only then can horizontal transport contribute toward a positive θ* at t = 0 h. In addition, radiative cooling, which in the Arctic typically amounts to 1–2 K day⁻¹ (e.g., Papritz 2020), would inevitably contribute to the formation of negative θ*. Consequently, warming processes must compensate for the initial negative θ* and the negative contributions due to transport and radiative cooling in order for the air parcels to reach a zero θ* at t = t_genesis. In the following, we shall shed light on these warming processes prior to t = t_genesis.

For the purpose, Fig. 11 shows the θ* budget accumulated from t = −240 h to t = t_genesis for the entire troposphere, that is, the θ* budget for the time before the final positive anomaly starts to form. Note that the initial θ* is negative but relatively small in comparison to the other budget terms, such that positive and negative contributions nearly balance each other. In line with equatorward transport into climatologically warmer regions, horizontal transport contributions are negative for lower tropospheric air parcels associated with Atlantic LCEs (Fig. 11a). For Pacific LCEs, negative contributions from horizontal transport occur across the entire tropospheric column (Fig. 11b), which is in line with the more poleward origin of air parcels aloft as compared to Atlantic LCEs (see supplemental Fig. 4b). For LCEs in both regions, vertical transport and diabatic heating contribute positively and, hence, compensate negative contributions due to diabatic cooling and horizontal transport (Figs. 11a,b).

View Full Size

θ* budget as in Figs. 9a and 9b, but for terms accumulated from t = 240 h to t = t_genesis. In this case the trajectories obtain a nonzero initial anomaly (gray).

Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0883.1

• Download Figure
• Download figure as PowerPoint slide

While at upper levels additional diabatic heating is mainly due to latent heating, heating in the planetary boundary layer becomes increasingly important toward the lower troposphere, more so for Atlantic than for Pacific LCEs. Given that at t = t_genesis the near-surface air is preferentially located over open ocean (Figs. 8a,d), the heating likely occurs in a marine cold air outbreak type of flow, where cold and dry Arctic air is advected over a comparatively warm ocean, resulting in upward surface sensible and latent heat fluxes. The relevance of cold air outbreaks is further corroborated by the geographical location of the maximum 24-hourly boundary layer heating (Fig. 12).

View Full Size

Spatial distribution of trajectories at the time of maximum 24-hourly diabatic heating in the planetary boundary layer during their 10-day evolution for trajectories initialized between 50 and 200 hPa AGL associated with (a) Atlantic and (b) Pacific LCEs. The 50% sea ice concentration contour is indicated by a solid black line.

Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0883.1

• Download Figure
• Download figure as PowerPoint slide

For Atlantic LCEs, several maxima are apparent: the Labrador Sea, the Greenland Sea south of Fram Strait, and most notably the Barents Sea (Fig. 12a), while for Pacific LCEs the maximum heating occurs in the Sea of Okhotsk and the Bering Sea (Fig. 12b). These regions are well known for cold air outbreaks (Fletcher et al. 2016; Papritz and Spengler 2017). Furthermore, they are fairly close to the location of air parcels at t = t_genesis (Figs. 8a,d), namely the last time when they transition from negative to positive θ*.

Provided that near-surface air parcels associated with LCEs are of Arctic origin and anomalously cold initially, we conclude that in the absence of such diabatic heating, diabatic cooling and equatorward transport would inevitably lead to more negative θ* along the trajectories, preventing them from attaining a zero θ* at t = t_genesis, let alone a strong positive θ* during the LCEs. Such airmass transformations can, thus, be seen as a precursor and prerequisite to the formation of a positive near-surface θ* by horizontal transport associated with LCEs. Despite this, the budget shown in Fig. 11 is not a unique feature of near-surface trajectories associated with LCEs but instead appears as a frequent characteristic of trajectories over subpolar oceans (not shown).

a. Synthesis

Synthesizing the results from the trajectory analyses, a three-dimensional conceptual picture of the airstreams constituting the anomalously warm tropospheric column during LCEs and their characteristic thermodynamic evolution emerges. A robust feature of Atlantic and Pacific LCEs is that air arriving near the surface (i.e., below 200 hPa AGL) has Arctic origin, whereas the air farther aloft originates at midlatitudes and is transported over larger distances. To simplify the discussion in the following, the former will be referred to as near-surface air and the latter as air aloft. Typical representations of these airstreams are schematically depicted in Fig. 13. Regarding the questions posed at the end of the introduction, our conclusions are as follows:

1. What is the geographical origin of the air at different levels in the tropospheric column? Ten days prior to an LCE, air aloft is located at midlatitudes, predominantly in the lower troposphere, a feature that applies equally to Atlantic and Pacific LCEs. Typically, the air associated with Atlantic LCEs is transported from the North American continent across the North Atlantic and ascends as it moves poleward east of Greenland (upper airstream in Fig. 13 left). Similarly, the air associated with Pacific LCEs originates over Eurasia and ascends over the central North Pacific (upper airstream in Fig. 13 right). In contrast to the air aloft, the near-surface air originates in the Arctic and undergoes modest equatorward excursions en route, though remaining at subpolar latitudes. Thereby, the air traverses open ocean areas in the Barents Sea and the Nordic seas (Atlantic LCEs; lower airstream in Fig. 13 left), as well as east of Kamchatka and south of Bering Strait (Pacific LCEs; lower airstream in Fig. 13 right).

View Full Size

Schematic illustrating typical pathways of near-surface air and air aloft (see text) associated with LCEs in the (left) Atlantic and (right) Pacific sectors. Airstreams are colored according to θ* from blue (negative θ*) to red (positive θ*) and arrows indicate where horizontal transport (dark blue), vertical transport (light blue), latent heating (red), and boundary layer heating (orange) contribute to the generation of θ*. Furthermore, black dots indicate the typical location of trajectories at t = t_genesis (i.e., the time when θ* was last 0 K), corresponding approximately to 2–4 days before arrival at the LCE. The brown dashed contour shows the sea ice edge.

Citation: Journal of Climate 36, 21; 10.1175/JCLI-D-22-0883.1

• Download Figure
• Download figure as PowerPoint slide

The latitudinal distance between near-surface air and air aloft is maximized between 3 and 5 days prior to the occurrence of an LCE when the air aloft is at its southernmost latitude and the near-surface air is still deep within the Arctic (cf. black dots in Fig. 13). The airstreams join over the subpolar and polar oceans and vertically coherent transport is limited to the final 1–2 days as the air is transiting through the principal Arctic gateways (i.e., Fram Strait and Bering Strait).

1. 2) Where and when do potential temperature anomalies (θ*) form along the pathway of the air associated with the SEB events? Across levels, positive θ* forms over a time period of 2–4 days prior to an LCE with earlier onset of θ* formation for air aloft. At the time when the air last obtained a zero θ*, near-surface air is located over the subpolar and polar oceans, whereas air aloft is located closer to or at midlatitudes over the North Atlantic and North Pacific Oceans (black dots in Fig. 13).
2. 3) How much do horizontal and vertical transports, as well as diabatic heating contribute to the formation of θ*? The θ* commonly forms as a residual between opposing processes, whose individual contributions exceed the final θ*. Horizontal transport and diabatic cooling—mostly due to radiation—have the largest magnitude of all terms in the θ* budget. Horizontal transport contributes positively across levels, whereas vertical transport has positive contributions for near-surface air due to descent and negative contributions for air aloft due to ascent. Latent heating and boundary layer heating have notable positive contributions above and below 200 hPa AGL, respectively.

Considering the differences in the contributions between air parcels with the 20% largest and smallest θ* emphasizes rapid horizontal transport from climatologically warmer to colder regions as the key process causing larger θ* at all levels. Latent heat release provides additional positive contributions to θ* of air aloft, which are opposed by more negative contributions of comparable magnitude due additional ascent. Subsidence can be an important cause for large θ* of near-surface air, especially for Pacific LCEs (see LCE 137). The more important role of vertical transport for Pacific as compared to Atlantic LCEs is consistent with the more frequent occurrence of blocking in the former case.

1. 4) What is the role of local airmass transformations for the generation of θ*? Given that near-surface air has a modest cold anomaly 10 days before an LCE occurs and that it originates in a region that is climatologically as cold or colder than the high Arctic where the LCE occurs, the air has to adjust to warmer climatological temperatures as found over the subpolar and polar oceans before it can attain a positive θ* via horizontal transport. Such an adjustment typically occurs in a marine cold air outbreak, where the air is warmed by upward turbulent heat fluxes from the ocean, or via vertical transport across the climatological potential temperature gradient. These airmass transformations can be seen as a necessary but not unique precursor for LCEs.

b. Concluding remarks

Based on our results, the view that the most extreme SEB anomalies over Arctic sea ice are associated with vertically coherent injections of warm and humid midlatitude air appears oversimplified. Instead, our findings show that warm and humid air from midlatitudes ascends and joins forces with originally Arctic near-surface air. Thereby, the air at different levels moves coherently only during the final 1–2 days. This is consistent with the idea that poleward flowing air tends to ascend along slanted moist isentropes over the polar dome (e.g., Orbe et al. 2015; Bozem et al. 2019), constituting the air aloft, while the air in the polar dome is trapped and recirculates at low altitudes between the high Arctic and the subpolar oceans, providing the near-surface air after diabatic transformation. As a consequence, the θ* spanning the tropospheric column collocated with the most extreme SEB anomalies over Arctic sea ice results from a combination of poleward transport from lower latitudes and local diabatic airmass transformations over the subpolar and polar oceans as well as subsidence within the Arctic.

The importance of local processes for the generation of near-surface temperature extremes over Arctic sea ice has previously been noted by Binder et al. (2017) and Papritz (2020), whereas the systematic Arctic origin of near-surface air during extreme SEB anomalies has not previously been recognized. This finding may be of particular relevance for understanding the aerosol composition during moist-air intrusions into the Arctic, which in turn plays an essential role in cloud formation processes and associated radiative effects (e.g., Dada et al. 2022). Despite the relevance of local warming processes, the driving dynamical processes in many cases may have an origin at lower latitudes. This includes on one hand synoptic weather systems as identified in many previous studies (e.g., Sorteberg and Walsh 2008; Messori et al. 2018; Nygård et al. 2019; Murto et al. 2022) and confirmed here (supplemental Fig. 7), and on the other hand also established teleconnections with the tropics and midlatitudes (e.g., Baggett et al. 2016; Hofsteenge et al. 2022).

Here, we focused on the origin of large positive θ* in the entire tropospheric column, whose presence is one of the most striking characteristics of the air collocated with extreme SEB anomalies. Temperature directly influences surface sensible heat fluxes and downwelling longwave radiation. However, water vapor and liquid clouds also play an essential role for longwave radiative anomalies (Curry et al. 1995; Shupe and Intrieri 2004; Persson et al. 2017; Vargas Zeppetello et al. 2019; Clark et al. 2021), which strongly contribute to extreme SEB anomalies (Murto et al. 2023). In the presence of moisture sources such as over open ocean, we argue that water vapor is in the first place slaved to temperature via the Clausius–Clapeyron relationship such that anomalously warm air will inevitably have higher moisture content. This is confirmed for the LCEs by vertical profiles (Murto et al. 2023; see also supplemental Fig. 8) and justifies our focus on θ*. Furthermore, this mechanism appears to be especially important for originally Arctic near-surface air transiting over open ocean, since in associated marine cold air outbreaks ocean evaporation immediately follows warming by sensible heat fluxes (Papritz and Spengler 2017). Finally, the fact that a significant amount of the downward longwave radiation at the surface is emitted in the lowermost 1 km of the atmosphere (Ohmura 2001; Persson et al. 2017; Vargas Zeppetello et al. 2019) underpins the relevance of the near-surface air, and thus, the role of local processes for extreme SEB anomalies over Arctic sea ice (see also supplemental Fig. 8).

In summary, these findings suggest that intrusions of midlatitude air into the Arctic alone are not the single cause of collocated SEB anomalies over Arctic sea ice. Rather, the large-scale flow patterns are such that midlatitude air aloft coincides with anomalously warm, humid, and cloudy near-surface air of Arctic origin, which has been transformed by local processes. This results in a vertically deep intrusion of warm and moist air with the near-surface air likely exhibiting the strongest forcing on the SEB.