Analysis of an Arctic cold air outbreak during autumn and related air mass transformations forced by surface changes and advection in higher altitudes

The meridional transport of air masses between the central Arctic and the lower latitudes profoundly impacts the atmospheric composition and energy budget. One prominent example are warm and moist air intrusions (WAIs). They can reach deep into the Arctic and strongly affect the humidity content, clouds, and aerosol concentrations (Dada et al., 2022), the vertical thermodynamic structure (You et al., 2021a; Kirbus et al., 2023; Svensson et al., 2023), as well as the surface energy budget of the entire Arctic (Johansson et al., 2017; Tjernström et al., 2019; You et al., 2021a; Murto et al., 2022; Kirbus et al., 2023; Svensson et al., 2023; Wendisch et al., 2023b). The combination of these effects related to WAIs can significantly accelerate the sea-ice melt (Tjernström et al., 2015; Mortin et al., 2016; Woods and Caballero, 2016; Persson et al., 2017; Yang and Magnusdottir, 2017). In opposite direction, the southward transport of cold and typically dry Arctic air from the sea ice toward the much warmer oceans is termed marine cold air outbreaks (MCAOs). During MCAOs, the near-surface air temperatures can rise by up to 30 K within a few hours (Pithan et al., 2018). The accompanying air mass transformations include a rapid heating and moistening over the marginal sea-ice zone (MIZ) and open ocean, the formation of cloud streets due to roll convection, and their transition into open cells further downstream (Lloyd et al., 2018; Tornow et al., 2021). These air mass transformations are primarily driven by the swift change in the underlying surface properties during the meridional air mass transport (Pithan et al., 2018; Wendisch et al., 2023a).

Typically, MCAO studies focus on winter and/or early spring (Fletcher et al., 2016; Dahlke et al., 2022). Then, the strong near-surface horizontal temperature gradients between sea ice and the open ocean generate intense MCAOs (Fletcher et al., 2016; Papritz and Spengler, 2017; Dahlke et al., 2022). Accordingly, 60%–80% of the wintertime ocean heat loss in the Nordic Seas can be attributed to strong MCAO events (Papritz and Spengler, 2017). However, our current understanding of the air mass transformations occurring during other seasons and/or weaker MCAOs is limited, in particular with regard to cloud evolution (Sanchez et al., 2022; Wu and Ovchinnikov, 2022; Murray-Watson et al., 2023). The latter is relevant as changes in cloud cover and cloud thickness have a large impact on the radiative surface energy budget (Li et al., 2011; Sanchez et al., 2022). The Arctic amplification observed in recent decades (Wendisch et al., 2023a) may interact with the changes in MCAO patterns and strength. The observed bottom-heavy warming has led to substantial decadal decreases of strong MCAO events in Fram Strait in mid-winter (Dahlke et al., 2022) as well as in the western Russian Arctic, including the Barents Sea region, in winter and late autumn (Narizhnaya et al., 2020). Future projections for the Nordic Seas based on climate models also forecast significant decreases in the MCAO index in winter (Landgren et al., 2019). As the Arctic sea-ice extent is decreasing, the spatial patterns of MCAOs start shifting (Landgren et al., 2019). Finally, moistened and heated MCAO air parcels might be more frequently circulated back into the Arctic, for example, in subsequent WAIs (Wendisch et al., 2017; Wendisch et al., 2023a).

The Fram Strait is an especially interesting “hot spot” for studying MCAOs. Here, the warm North Atlantic Current pushes the sea-ice edge northward. Air cooled over Arctic sea ice moves onto open ocean with high sea-surface temperatures (Dahlke et al., 2022). To obtain an improved understanding of processes involved in MCAOs, the full history from their formation over the sea ice, their movement over the MIZ, and finally, the open ocean should be captured (Papritz and Spengler, 2017; Pithan et al., 2018; Wendisch et al., 2023a). This can be achieved using a Lagrangian approach, which provides valuable insights into the crucial air mass transformation processes (Pithan et al., 2018; Wendisch et al., 2021). Relying on the fifth generation of atmospheric reanalyses of the global climate (ERA5; Hersbach et al., 2020) operated by the European Centre for Medium-Range Weather Forecasts (ECMWF), Papritz and Spengler provided a thorough investigation into the life cycle of wintertime MCAO air masses for distinct geographic sectors. In Fram Strait, they report a persistent median diabatic cooling rate of about −1 K day⁻¹ over sea ice, which transforms into rapid diabatic warming rates of around +6 K day⁻¹ during the first 24 h over the open ocean (Papritz and Spengler, 2017). Lagrangian trajectory analysis of MCAOs is typically based on reanalysis data, often ERA5 (e.g., Papritz and Spengler, 2017; Papritz et al., 2019; Dahlke et al., 2022). However, it is crucial to also include observations. This has been done using satellite measurements (Wu and Ovchinnikov, 2022; Mateling et al., 2023; Murray-Watson et al., 2023), atmospheric soundings (Dahlke et al., 2022; Geerts et al., 2022; Michaelis et al., 2022), or airborne observations (Sanchez et al., 2022). Such measurements are also required to constrain simulations. This applies to models over a wide range of spatiotemporal scales, including single column models, Large Eddy Simulations (LES), limited-area models, and reanalysis (Pithan et al., 2016; Geerts et al., 2022; Sanchez et al., 2022; Svensson et al., 2023). Particularly challenging processes to model include thermodynamic phase transitions and precipitation formation in atmospheric boundary layer (ABL) clouds (Pithan et al., 2018), turbulent processes and impacts of radiative energy fluxes (Papritz and Spengler, 2017; You et al., 2021a, 2021b; Chechin et al., 2023) as well as the emerging complex vertical interactions between stacked cloud layers (Li et al., 2015; Yao et al., 2020; Truong et al., 2022). Latter are additionally relevant as satellite-based studies tend to filter out cases with more than one cloud layer (Murray-Watson et al., 2023).

In this work, we present a detailed analysis of an MCAO case study based on quasi-Lagrangian measurements. We are using data from the shipborne Multidisciplinary drifting Observatory for the Study of Arctic Climate (MOSAiC; Nicolaus et al., 2022; Rabe et al., 2022; Shupe et al., 2022) and the MOSAiC Airborne observations in the Central Arctic campaign (MOSAiC-ACA; Herber et al., 2021; Mech et al., 2022a). The air mass was first observed by the research vessel (RV) Polarstern in the central Arctic. After a 2-day southward drift of the air mass, the Polar 5 aircraft sampled the same air mass over the open ocean of the Fram Strait. We select this case for analysis because it features several interesting aspects. First, the 2 sets of observations allow a thorough characterization of the atmospheric states at the start and end points of the drift. Moreover, as the boundary layer winds were oriented almost exclusively meridionally, the selected case constitutes the most direct link between the central Arctic and the marine environment of Fram Strait that was observed during the 2 concurrent campaigns. Furthermore, the case occurred in autumn, a season that is understudied due to typically lower MCAO intensities. And finally, this specific case is characterized by complex vertical interactions between multiple cloud layers, often not accessible by satellite studies. Our study aims to address 3 specific questions. (i) How did the thermodynamic, cloud, and radiative properties of the air mass evolve over time as it moved from the sea ice covered central Arctic to the open ocean at Fram Strait? (ii) How did differentially advected midlevel clouds affect the ABL clouds below? and (iii) Was the observed air mass transformation mainly driven by the change of surface properties from sea-ice to open ocean—as our conception suggests?

The study is structured as follows. Section 2 provides a general overview of the case and an introduction of the relevant observations gathered during 2 Arctic expeditions MOSAiC and MOSAiC-ACA. A description of selected output from ERA5 reanalysis follows. In Section 3, we analyze the evolution of air masses based on ERA5 data and compare the reanalysis output to observations. Finally, the air mass transformations are quantified using temperature and humidity tendencies extracted from ERA5.

The initial atmospheric conditions were characterized by observations from aboard RV Polarstern. The drift of RV Polarstern was conducted within the MOSAiC expedition (Shupe et al., 2022). Figure 1 shows the location of RV Polarstern on September 11, 2020, over closed sea ice close to the North Pole. Table 1 gives an overview of the measured properties. These include vertical profiles of thermodynamic air mass characteristics obtained from radiosondes (Maturilli et al., 2021), vertical profiles of cloud liquid water and frozen hydrometeor quantities derived via the Cloudnet algorithm (Engelmann et al., 2023), and near-surface upward and downward broadband irradiances (solar and terrestrial; Riihimaki, 2021). In order to evaluate the spatiotemporal variability of the observed atmospheric state, a 60-min time window was centered on the observational point on September 11, 2020, at 14 UTC. Statistical means and standard deviations of investigated variables were calculated.

The Polar 5 research aircraft of the German Alfred-Wegener-Institute (Wesche et al., 2016) conducted airborne observations on September 13, 2020, as part of the MOSAiC-ACA (Herber et al., 2021; Mech et al., 2022a). Polar 5 was equipped with a wide palette of instruments for both in situ sampling as well as remote sensing measurements (see Table 1; Mech et al., 2022b). Polar 5 took off from Svalbard at 9:20 UTC and headed toward Fram Strait to investigate cloud structures above the ocean and the MIZ (see Figure 1). During the flight back, which began at 13:40 UTC, the aircraft probed the ABL. It first conducted horizontal flight legs at low altitudes of 0.3–0.45 km (see Supplemental Figure S1). Cloud total and liquid water content were measured by a Nevzorov probe (Moser et al., 2022). Solar and terrestrial broadband irradiance were also recorded continuously (Becker et al., 2021b). The aircraft then quickly ascended from 0.45 km to 3.4 km altitude, passing through a cloud deck. At 14:17 UTC and the final altitude of 3.4 km, a dropsonde was released (Becker et al., 2021a). A 10-min time window around the release of the dropsonde was chosen to average Microwave Radar/radiometer for Arctic Clouds radar data (Mech et al., 2022b). Similar to other studies, a power law was used to correlate the measured radar reflectivities Ze to the vertical distribution of frozen hydrometeors, meaning the combined cloud ice and snow water content q_ice+snow (Shupe et al., 2005):

Here, b = 0.63 is constant, and a varies according to the climatology. Comparisons with the Nevzorov data during ascent showed good fit for a = 0.04, which lies close to the median value reported for the Arctic in September (Shupe et al., 2005).

Based on the broadband radiation measurements, the solar and terrestrial cloud radiative forcings (CRFs) were calculated (Hartmann, 2015). We followed the same approach that has been applied for several recent Arctic airborne campaigns (e.g., Stapf et al., 2021a; Stapf et al., 2021b; Becker et al., 2022). Using the measured solar or terrestrial net irradiance F_{all_sky}, the CRF F_CRF can be expressed as:

The libRadtran library for radiative transfer was used to simulate F_{clear_sky} (Emde et al., 2016). At the MOSAiC site, libRadtran was initialized with the vertical thermodynamic profiles obtained from radiosondes and the liquid water path retrieved from the Humidity And Temperature PROfiler, a component of the Cloudnet suite. At the area investigated by Polar 5, CRF calculations were initialized using thermodynamic profiles from the dropsonde (altitude below 3.4 km; Becker et al., 2021a) as well as nearby radiosoundings in Ny-Ålesund (altitude above 3.4 km; Maturilli, 2020). The liquid water path was retrieved continuously from measured and simulated downward irradiances, consistent with the approach taken by previous studies (Stapf et al., 2021a; Stapf et al., 2021b; Becker et al., 2022).

The evolution of the air mass of the MCAO at and in between the observational points was primarily analyzed using ERA5. Output fields are available with a frequency of 1 h, a horizontal resolution of 0.25° latitude/longitude, and on 137 vertical levels from the surface up to the top of the atmosphere. ERA5 was chosen for its general good performance in the Arctic region (Graham et al., 2019b), which notably includes temperature and wind fields in the Fram Strait (Graham et al., 2019a). Many recent studies that apply trajectory analysis to the Arctic domain are based on ERA5 (e.g., Papritz and Spengler, 2017; Ali and Pithan, 2020; Papritz, 2020; You et al., 2021a, 2021b). The reanalysis data were also used in this article for calculating trajectories based on the 3-dimensional wind fields. Furthermore, the air temperature, specific humidity, CRF, and other cloud-related parameters were also extracted. To estimate spatiotemporal variability, ERA5 data were averaged for a grid box surrounding the respective measurement locations. The box spanned ±1 grid points north/south, corresponding to about 28 km. Zonally, grid points were selected east/west of the observation sites, so that an equal distance was covered. In all vertical profiles of atmospheric variables at RV Polarstern and Polar 5, the solid lines represent the profiles of the box means, while the standard deviations within the box are indicated as colored shading. Similarly, the values of ERA5 surface variable relating to CRF and turbulent fluxes are provided as the mean ± the standard deviation within this box.

Lagrangian trajectories were computed using Lagranto (Sprenger and Wernli, 2015). To detect links between the central Arctic and the region covered during MOSAiC-ACA, 2-day backward and 3-day forward trajectories were started above the location of RV Polarstern with a temporal resolution of 1 h. Horizontally, the trajectories were initialized in a circle of 20 km radius around the ship with a 2-km latitude/longitude spacing. This yielded 49 starting points per altitude. Vertically, trajectories were initialized every 10 hPa between 10 and 100 hPa above the ground, which corresponds to an approximate height range of 0.1–0.9 km. This range was chosen to focus on the dynamics of the lower troposphere, including the ABL. A link between the central Arctic (RV Polarstern) and Fram Strait (Polar 5) was identified for air parcels initialized above the location of RV Polarstern on September 11, 2020, at 14 UTC. Figure 1 depicts the vertically integrated density of the respective air mass trajectories as blue contours. During the previous 2 days, the low-level air masses observed at the MOSAiC site had resided over sea ice. In the course of the next 48 h, the flow toward the MOSAiC-ACA site was mostly oriented meridionally. At all times, air parcels remained at altitudes below 1 km (not shown). They crossed the MIZ approximately 42 h after the departure at RV Polarstern. Then, within few hours, they were sampled again by the Polar 5 aircraft. From the large ensemble of trajectories, a median trajectory representing the average drift was calculated (thick colored line in Figure 1). Height-time cross sections of relevant meteorological parameters, such as air temperature and specific humidity, were extracted along the median trajectory.

The synoptic conditions during the drift of air masses were analyzed using ERA5 data (Figure 2). On September 11, the sea-level pressure distribution indicated weak and stationary Greenland and central Arctic high-pressure systems. In the course of the next 2 days, a dynamic low-pressure system initially positioned around Novaya Zemlya (74°N, 56°E) shifted toward the North Pole. The resulting pressure dipole induced northerly flow and pushed air from the central Arctic southwards toward the open waters of Fram Strait. This drift was characterized by an increase in the 2-m air temperature from −6.6°C at RV Polarstern to +0.8°C at the location sampled by Polar 5.

Figure 3a and b depicts the vertical profiles of relative humidity (RH), cloud liquid water content (q_liq), and frozen hydrometeor contents (q_ice+snow) at the start of the trajectory. In the central Arctic at RV Polarstern, a low-level mixed-phase cloud was identified by both ERA5 as well as observations from Cloudnet and radiosondes. It should be noted that there were no soundings directly at 14 UTC, so the observations from 11 UTC and 17 UTC are shown. The upper part of the observed cloud mostly consisted of liquid water, and the lower part contained very little ice. While ERA5 generally captured the cloud structure, it greatly overestimated the total cloud liquid water content relative to the observations, which are affected by uncertainties as well. The ERA5-derived liquid water path of (81 ± 4) g m⁻² was almost 5 times higher than the value of (16 ± 3) g m⁻² assessed by Cloudnet. However, it should be mentioned that remote sensing retrievals of cloud properties are especially challenging under mixed-phase cloud conditions (Bühl et al., 2016). In the Cloudnet data, the uncertainty of the liquid water content retrieval from the microwave radiometer alone is given as 20 g m⁻², plus 25% of the retrieved value. In the case here, this yields a total systematic uncertainty of about 24 g m⁻², in addition to random errors and spatiotemporal variability of the atmospheric state. Another source of uncertainties of cloud liquid water and frozen hydrometeor observations can be uncertain cloud-base and cloud-top height assumptions. Still, the wide variety of observations collected during MOSAiC gives confidence in the Cloudnet results. The RH profiles as obtained from radiosondes confirm the location of clouds in the lowest 500 m. Independent estimates of surface CRF reinforce the notion that the liquid clouds in the reanalysis are too thick. The terrestrial CRF based on ERA5 provided a value of (70 ± 3) W m⁻², while observations yielded (44 ± 4) W m⁻² (see Figure S2). As terrestrial CRF quickly saturates for liquid water paths larger than 30 g m⁻² (Shupe and Intrieri, 2004), misrepresentations in cloud properties can lead to significant inaccuracies also in representing the surface energy budget. In contrast to the terrestrial CRF, the solar CRF at RV Polarstern was negligible. This result is caused by the high surface albedo combined with the large solar zenith angle of 85°. Finally, it should be mentioned that the Cloudnet product and radiosondes represent single measurements, while the reanalysis output has a spatial resolution of 30 km.

To analyze the air mass evolution of the investigated MCAO event in a Lagrangian perspective, height-time cross sections of ERA5 cloud q_liq and q_ice were extracted along the median trajectory in hourly steps (Figure 3c). Please note again that this trajectory was computed for the near-surface flow (approximately lowest 1 km), while the cloud structure is shown for up to 3.5-km altitude. During the first 24 h of the drift, the ABL cloud system was in a quasi-steady state. Then, starting at 27 h, a mixed-phase cloud appeared at altitudes of about 1.3–2.7 km. At the same time, the cloud located at the top of the ABL started ascending and glaciating. After crossing the MIZ and over open ocean, ice clouds were found throughout the whole lowest 3.5 km. The boundary layer height had increased steadily, from initial 0.35 to 0.6 km.

Figure 3d and e shows the vertical profiles of cloud properties as observed by Polar 5 and based on ERA5. At 0.3–0.45-km altitude, measurements from the Nevzorov probe and dropsondes show thin remaining mixed-phase cloud layers. Throughout the lowest 3 km, profiles derived from radar data show the presence of frozen hydrometeors. ERA5 captured this structure well, but with a slightly different vertical distribution. Estimates of surface CRF yield comparable terrestrial forcings. At Polar 5, (76 ± 2) W m⁻² are found, and in ERA5 (70 ± 2) W m⁻². Considering the higher variability, solar CRF was captured reasonably well by ERA5, but with a slight underestimation. Polar 5 measurements yielded (−156 ± 10) W m⁻², versus ERA5 (−129 ± 9) W m⁻². Note that over the daytime ocean, the net CRF was negative, and thus, clouds led to a radiative cooling at the surface.

Figure 4a depicts the vertical profile of specific humidity obtained from radiosonde launches at RV Polarstern compared to ERA5 data. 360° panoramic pictures taken from RV Polarstern (Nicolaus et al., 2021) and the measured irradiances (Riihimaki, 2021, both not shown) reveal that 11 UTC still featured cloud-free conditions, while low-level clouds started being advected over the ship at around 12:30 UTC. These clouds persisted until the end of the day. Thus, the following paragraphs focus on the 17 UTC radiosonde measurements.

The central Arctic state was characterized by a moist layer below the temperature inversion, with specific humidity generally lower than 3 g kg⁻¹. Layers further up were drier, with average values mostly below 1 g kg⁻¹. The air masses experienced a slow but steady reduction in water vapor at the height of the ABL cloud during its drift toward the Fram Strait (Figure 4b). This trend abruptly changed at 24 h past the start of the drift at RV Polarstern. Here, large amounts of water vapor were suddenly appearing. As will be shown later, this was caused by the differential advection at higher altitudes. In Fram Strait (Figure 4c), both dropsondes and ERA5 indicated a moistened lower troposphere, with values exceeding 2 g kg⁻¹ in all altitudes and the highest humidity of 3.3 g kg⁻¹ found at the surface. When comparing the profiles obtained from the airborne with shipborne observations 48 h apart, the strongest net change in moisture of about +2.5 g kg⁻¹ had occurred at around 1.0-km altitude.

Figure 5a shows the air temperature profiles as measured at RV Polarstern. The observations revealed a shallow, mixed surface layer in the lowest 0.15 km. Above, a temperature inversion of about 1 km thickness was identified. Since here the temperature profile did not follow an adiabatic lapse rate, no coupling to the surface could be inferred. As the soundings were assimilated into ERA5, the reanalysis captured the thermodynamic structure very well. Nevertheless, in the lowest 0.35 km above ground, the shape of the observed surface inversion was not represented properly. Here, ERA5 showed a strongly coupled ABL and an elevated temperature inversion. This might in part be linked to the overestimation of cloud liquid water mentioned earlier, causing an overestimation of the cloud top inversion and exaggerated mixing in the lowest layers.

Figure 5b shows the evolution of air temperature during the drift. In the first 24 h, a slight cooling throughout the lowest 3.5 km of the troposphere was observed. However, at 24–36 h and altitudes of 1.5–3.5 km, air masses were suddenly up to 8 K colder. In the MIZ and over open ocean, a general warming of air masses sets in, most notably near the surface. In Fram Strait, vertical profiles based on the Polar 5 dropsonde and ERA5 (Figure 5c) agreed on the presence of a coupled marine ABL in the lowest 0.5 km, that is, air temperatures following an adiabatic lapse rate. While ERA5 simulated a cloud top inversion at the top of the ABL, in the dropsonde, this was seen about 1 km higher, at 1.5 km. Finally, ERA5 and observations indicated a cloud top inversion occurring at 2.7 km. A comparison of the Polar 5 dropsonde data with the profile measured 2 days prior (gray line in Figure 5c) shows that the the strongest net heating of around +8 K had occurred near the surface. On the contrary, air masses above 0.5 km were up to 5 K cooler than at RV Polarstern.

By extending the trajectory to 24 h after leaving the sea ice, a diabatic heating rate of +7 K day⁻¹ is found. Considering that the MCAO event occurred in autumn, this seems relatively high compared to the median heating rate of +6 K day⁻¹ reported by Papritz and Spengler (2017) for winter MCAOs in Fram Strait. However, here we focus solely on the dynamics of the lowest 90 hPa. In the mentioned study, the starting points of trajectories were placed over open ocean, vertically every 25 hPa between 1,000 and 500 hPa. Starting points were retained as long as the MCAO index of the underlying grid cell was positive, and the potential air temperature at the respective altitude was larger than the potential sea surface temperature. Particularly for strong MCAO events, this can apply to altitudes much higher than 90 hPa above the ground. As Papritz and Spengler (2017) report a column average, the surface heating first had to be mixed into higher layers. Considering wintertime temperature contrasts between sea ice and open ocean often in excess of 30 K (Pithan et al., 2018), this might also explain why their heating rates appear relatively small.

To conclude this section, both observations and ERA5 indicated a transformation from a rather dry, surface-based inversion into a moistened marine state. ERA5 captured the cloud structures at the start and end locations reasonably well, although with some deviations in the vertical locations and thickness of clouds. Of special interest for this case is the appearance of moist and cloudy higher layers 24 h into the drift, that is, still 15 h upstream of the MIZ.

To identify and quantify the processes shaping the Arctic air mass transformations, ERA5 humidity and temperature tendencies were used. The tendency of specific humidity (∂q/∂t) due to all the parameterized processes (such as evaporation from the ground and evaporation of falling precipitation, condensation into clouds; see ECMWF, 2016) could be directly retrieved from the ERA5 reanalysis data. Furthermore, the specific humidity tendency due to advection of water vapor was calculated in a similar way as in previous studies (Randall and Cripe, 1999; You et al., 2021b). Figure 6 shows the humidity tendencies due to different mechanisms. Initially, at the altitude of the low-level clouds and especially the maximum liquid cloud water, the tendency due to parameterized processes (Figure 6a) showed a steady loss of water vapor. After the appearance of the midlevel clouds approximately 27 h into the drift, this pattern was extended. A partial water vapor loss was again found in the altitude of the maximum cloud water. But also an increase of water vapor was indicated below clouds, possibly by the evaporation of falling precipitation. At Fram Strait, the open sea surface acted as a source of water vapor due to evaporation. Here, the double cloud deck showed the pattern of water vapor loss inside clouds compared to slight water vapor gain below clouds via evaporation of precipitation.

Especially in the lowest 0.5 km, these local processes were in first order balanced by advection (Figure 6b). However, above the ABL, the high relevance of advection for this case becomes apparent. At 24–33 h, a strong confluence of 2 different air masses occurred due to wind shear. While the near-surface air masses were of central Arctic origin, the air masses further up were of marine origin, and due to backing, winds had been advected from the Kara Seas and Siberian coastline (see Figure S3). This differential advection of air masses also caused a westward drift in the ABL, which was already visible on the case overview (Figure 1, for 24 h ≤ t ≤ 33 h). In addition to injecting large amounts of water vapor, the mid-level clouds were directly introduced (Figure S3). As these mid-level clouds continued their flow across the low-level trajectory, they vanished in the vicinity of the MIZ. The importance of differential advection is supported by the fact that the strongest net moisture change of +2.5 g kg⁻¹ after the 48-h drift between the Arctic and Fram Strait was not found close to the surface, as would initially be expected, but at altitudes of around 1 km (Figure 4c, gray vs. blue line).

The vertical profiles recorded over Fram Strait showed a significant over- and underestimation of the specific humidity for altitudes of 0.5–1.3 km and 1.3–2.7 km. This means that the just discussed net humidity tendencies in these altitudes might be biased toward values too high and too low. Of special concern is the representation of snowfall formation from the advected mixed-phase cloud system, which was shown to be incorrect in exactly the same altitude ranges (Figure 3d and e). The overall deviations of ERA5 with regard to observations might thus be caused by misrepresentations of differential advection, but also of the ERA5 snowfall scheme.

Following the approach by You et al. (2021a, 2021b), the net temperature tendency can be split into 4 components:

Therefore, the net temperature tendency (∂T/∂t)_net consists of contributions from the solar (∂T/∂t)_sol and terrestrial radiation (∂T/∂t)_terr, latent heat release caused by cloud formation (∂T/∂t)_lhr, as well as turbulent sensible and latent heat fluxes (∂T/∂t)_slhf. The net temperature tendency shown in Figure 7a was characterized by persistent cooling in the ABL cloud system, which transformed into steady warming toward the open ocean. Especially above the ocean surface, this was in first order balanced by advection (Figure 7b). However, advection generally did not balance the local processes, which can be seen as the indication of a gradual transformation of air masses (You et al., 2021b).

The vertical profile of air temperature recorded by the dropsonde from Polar 5 showed a significant over- and underestimation of the air temperature for altitudes of 0.5–1.3 km and 1.3–2.7 km. As for the humidity tendency, this means that the net temperature tendencies in these altitudes might be biased toward values too high and too low. Furthermore, an exaggerated amount of snow falling through a lower dry layer can directly lead to an exaggerated evaporative cooling effect and vice versa.

While it is important to be aware of possible biases in the ERA5 tendencies for specific altitude ranges, the net temperature tendency was separated into the 4 different components (Equation 3). This allows an assessment of the processes governing the air mass transformation, and how their relative contributions changed throughout the drift. In the first 24 h of the air mass drift, terrestrial radiation (Figure 7c) mostly drove a marked cloud top cooling, which was reaching values as low as −2 K h⁻¹. As ERA5 likely overestimated the liquid water content in the central Arctic, the real cooling rate can be expected to be of slightly smaller magnitude. During the differential advection of midlevel clouds, this terrestrial radiative cooling shifted to higher altitudes of 1.5–2.5 km, where the top of these liquid-bearing clouds was situated. Over Fram Strait, the cloud-top cooling rate of around −1 K h⁻¹ modeled by ERA5 was in a good agreement with the observations from aboard Polar 5 (see Figure S4). The altitude of maximum cloud top cooling was overestimated in ERA5 by around 0.4 km and small-scale fluctuations were not captured, which however could not be expected from a data set on the scale of roughly 30 km. In contrast to terrestrial radiation, solar radiation (Figure 7d) contributed only negligibly to the transformation.

Especially in the first 24 h, latent heat release due to cloud formation (Figure 7e) partially balanced the terrestrial cooling near the cloud top. Furthermore, it contributed also some minor heat release in the mid- and higher level ice clouds.

The sign of the temperature tendency due to turbulence (Figure 7f), that is, heating versus cooling, was a function of the vertical location of air parcels relative to the temperature inversion. Turbulence was tightly coupled to the cloud-top cooling, which drove entrainment. During the advection of the midlevel clouds, turbulence in the ABL cloud deck weakened and partially shifted to the midlevel cloud top above. Further into the drift, the changing surface type became increasingly more important. Still approximately 50 km north of the dropsonde launch site, a low flight leg conducted by Polar 5 80 m above the open ocean yielded an observed turbulent sensible plus latent heat flux of (70 ± 5) W m⁻². While this is close to ERA5’s corresponding value of 75 W m⁻² in that area (data not shown), the real heat fluxes at the surface, that is, 80 m below Polar 5, can be assumed to be larger. As the sea-surface temperatures increased by around 2 K toward the location of the dropsonde launch, ERA5 calculated an increased sensible plus latent heat flux of 150 W m⁻² as the driver of turbulent heating in the ABL. These values are considerably lower than the maximum of over 500 W m⁻² reported for the strongest wintertime MCAO (Papritz and Spengler, 2017). However, the fluxes observed during September 2020 were sufficient to contribute a heating effect of about +1 K h⁻¹ over the open waters of Fram Strait.

In this study, the transformation of a cloudy Arctic air mass on its 2-day pathway during an MCAO from the central ice-covered Arctic Ocean toward the ice-free Fram Strait during autumn 2020 was investigated. The atmospheric state was constrained at both ends of the trajectory by shipborne and airborne observations. The measurements were supplemented by ERA5 reanalysis data and trajectory simulations. Some deficiencies of the ERA5 data in representing the air mass properties at both ends became apparent. For example, the reanalysis overestimated the liquid water path and, as a result, the magnitude of the cloud-top inversion at the initial state of the pathway. In part, these deviations comparing the large-scale ERA5 to point observations might be attributed to meso-scale variability. Nonetheless, ERA5 generally captured the vertical thermodynamic and cloud structures. Using ERA5, the air mass transformation was analyzed in a Lagrangian perspective along the 48-h trajectory following the ABL. We arrive at the following conclusions, answering our initial research questions.

• Temporal evolution: Along the southward drift and still upstream of the marginal sea ice zone, differential advection of the air mass originating from the central Arctic at low levels occurred with marine-origin air parcels higher above. This introduced moisture and midlevel clouds at high altitudes, ultimately leading to the dissipation of boundary-level clouds. Over the open water of Fram Strait, the ABL was moistened through surface fluxes and the differential advection, facilitating further evolution of a multilayer cloud system.

• Vertical interactions: The differentially advected higher clouds affected the ABL by displacing cloud-top cooling, entrainment, and turbulent processes upward. Furthermore, precipitation in the form of snow was falling through the ABL cloud deck. The combination of these factors led to the dissolution of the ABL clouds.

• Driver of air mass transformations: Contrary to our original expectation, a rapid transformation of the cloudy ABL was initiated still over sea ice. It was caused by the differentially advected midlevel clouds and moisture, mediated through the mechanisms listed in (ii). Accordingly, the comparisons of the soundings placed the greatest net moisture gain after 48 h in altitudes of around 1 km, not close to the ocean surface as expected. Nonetheless, surface forcing played a major role after the air mass drifted over the open water. It led to turbulent sensible plus latent heat fluxes of about 150 Wm⁻², causing a net increase in near-surface air temperatures by about 8 K over 48 h, a partial recovery of the ABL clouds, and a continued warming and moistening also beyond the airborne observation point.

The discussed case study observed in autumn does not represent an idealized MCAO, where typically the intense surface forcing dominates the air mass transformation. However, as outlined in the introduction, the Arctic amplification observed during recent decades is leading to significant changes in the spatiotemporal pattern and intensities of MCAOs. We highlight that such untypical cases, with their complex vertical interactions often not accessible by satellite measurements, should not be overlooked. Moving forward, the wealth of observations available at both ends of the trajectory motivates a follow-up study, where we further investigate the impact of advected cloud masses on the transformation of the ABL. We aim to address the research questions related to uncertainties in ERA5: What is the relative importance of the displaced cloud-top cooling and the falling precipitation; and to quantify their dependence on the changes in the hydrometeor sizes in the advected clouds. In order to study the sensitivity that was not directly observed, we are therefore currently employing a Lagrangian LES as a virtual laboratory.

The majority of the observational data used in this study can be accessed through the PANGEA data repository. This includes all radiosonde data from RV Polarstern (https://doi.org/10.1594/PANGAEA.928833) and from AWIPEV at Ny-Ålesund (https://doi.org/10.1594/PANGAEA.926804). Additionally, the following data sets collected by Polar 5 are publicly accessible: dropsonde thermodynamic profiles (https://doi.org/10.1594/PANGAEA.933581), MiRAC radar reflectivities (https://doi.org/10.1594/PANGAEA.944507), broadband irradiances (https://doi.org/10.1594/PANGAEA.936232), as well as in situ cloud measurements (https://doi.org/10.1594/PANGAEA.940557). The Cloudnet data used in this article were generated by the European Research Infrastructure for the observation of Aerosol, Clouds and Trace Gases (ACTRIS) and are available from the ACTRIS Data Centre (https://hdl.handle.net/21.12132/2.9e078b4e14494412). The surface radiation data are available from the DOE ARM archive (https://doi.org/10.5439/1608608). All data related to ERA5 can be found on the ERA5 data repository (https://www.ecmwf.int/en/forecasts/dataset/ecmwf-reanalysis-v5, see also Hersbach et al. [2020]).

The supplemental files for this article can be found as follows:

Figures S1–S4. PDF

The authors would like to express their gratitude for the valuable data used in this manuscript, parts of which were generated by the international Multidisciplinary drifting Observatory for the Study of the Arctic Climate (MOSAiC) with the tag MOSAiC20192020. They would like to extend their gratitude to the Alfred-Wegener-Institute (AWI), the dedicated crews of RV Polarstern and Polar 5, as well as the entire MOSAiC and MOSAiC-ACA science teams. Their efforts made the field campaigns possible, and they also played a crucial role in processing the collected measurements. We furthermore acknowledge the European Research Infrastructure for the observation of Aerosol, Clouds and Trace Gases (ACTRIS) for generously providing access to the CLU (2023) Cloudnet data set. This data set is available for download from the Finnish Meteorological Institute (https://cloudnet.fmi.fi/). Surface radiation data were obtained from the Atmospheric Radiation Measurement (ARM) User Facility, a U.S. Department of Energy (DOE) Office of Science User Facility Managed by the Biological and Environmental Research Program.

The authors acknowledge funding by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation)—project 268020496 TRR 172, within the Transregional Collaborative Research Center “ArctiC Amplification: Climate Relevant Atmospheric and SurfaCe Processes, and Feedback Mechanisms (AC)³.” The publication of this article was funded by the Open Access Publishing Fund of Leipzig University supported by the German Research Foundation within the program Open Access Publication Funding.

The authors have declared that no competing interests exist.

Conceived and designed this study: BK, JC, AE, MW.

Analyzed and interpreted the different data sets: BK, JC, AE, SB.

Drafted and revised the article: BK, JC, RN, MS, AE, MW.

Approved the submitted version for publication: All authors.

How to cite this article: Kirbus, B, Chylik, J, Ehrlich, A, Becker, S, Schäfer, M, Neggers, R, Wendisch, M. 2023. Analysis of an Arctic cold air outbreak during autumn and related air mass transformations forced by surface changes and advection in higher altitudes. Elementa: Science of the Anthropocene 11(1). DOI: https://doi.org/10.1525/elementa.2023.00079

Domain Editor-in-Chief: Detlev Helmig, Boulder AIR LLC, Boulder, CO, USA

Associate Editor: Joël Savarino, Laboratoire de Glaciologie et Géophysique de l’Environnement, CNRS/Grenoble University, Saint-Martin d’Hères, France

Knowledge Domain: Atmospheric Science

Part of an Elementa Special Feature: The Multidisciplinary Drifting Observatory for the Study of Arctic Climate (MOSAiC)