This study focuses on a CAO event over WNAO observed on 1 March 2020 during the Aerosol Cloud meTeorology Interactions oVer the western ATlantic Experiment (ACTIVATE) field campaign (Sorooshian et al., 2019). The CAO event was preceded by two cold front passages on 27 February (primary) and 29 February (secondary), with northerly winds advecting dry, cold air over the warm GS water. This triggered intense surface heat fluxes (> 300 W m⁻²) over the GS and mesoscale solenoidal circulation. Post-frontal clouds developed after the frontal passage, emerged at 08:00 Eastern Standard Time (EST) on 1 March 2020, and persisted for several hours. Chen et al. (2022) validated their WRF simulations against GOES-16 satellite retrievals and ACTIVATE dropsonde measurements (09:45–10:45 EST), demonstrating consistency in cloud morphology (e.g., cloud street alignment) and boundary layer thermodynamics (e.g., inversion height and moisture profiles), despite a slight cold bias in the simulated near-surface temperatures that were attributed to the ERA5 reanalysis inputs.

This study builds on the same CAO case analyzed by Chen et al. (2022), leveraging its robust validation with both satellite and in situ observational datasets. The well-documented cloud evolution over the GS, characterized by cloud structures influenced by SST structure, for this case provides an ideal framework for investigating SST-driven impacts on PFC morphology.

The Weather Research and Forecasting (WRF v4.2) model was configured with two nested domains. The outer domain covers a 1650 km × 1650 km region of the WNAO, with simulations conducted at 3 km horizontal resolution. The inner domain corresponds to a 450 km × 450 km at 1 km horizontal resolution, and specifically intended for studying the cloud street dynamics. The simulations are conducted at 150 vertical levels (up to 100 hPa), with finer resolution in the lower troposphere (∼ 46 m layer thickness up to 6 km) versus above. The model simulations were performed for the period of 06:00 UTC 1 March–00:00 UTC 2 March. Key physics schemes include the Morrison two-moment cloud microphysics scheme (Morrison et al., 2005), the Yonsei University PBL scheme (Hong and Lim, 2006), and the Rapid Radiative Transfer Model for GCMs (RRTMG) longwave and shortwave radiation schemes (Iacono et al., 2008). Subgrid convection parameterization is disabled in the simulations to better resolve mesoscale processes at gray-zone resolutions (Field et al., 2017). Initial and boundary conditions are derived from ERA5 reanalysis. This configuration aligns with the control simulation in Chen et al. (2022), which demonstrated strong consistency with satellite and dropsonde observations, thereby validating the model's ability to capture PFC evolution. The other details of the simulation setups have been described in Chen et al. (2022).

Figure 1a–c shows the spatial distribution of SST and surface fluxes from the outer domain. The SST enhancement associated with the GS leads to a large SST gradient at the north edge of the GS. There is a strong surface flux band along the GS, and strong gradients are present in both surface sensible heat flux (Fs) and latent heat flux (Fl). Hereafter, the simulation with this SST setup is referred to as “Ctrl”.

Two sensitivity experiments are conducted to isolate the individual aspects of SST changes and their impacts on PFCs:

Mean temperature experiment: SST is uniformly increased by 4 K across the domain, including the GS core region, to mimic extreme warming scenarios (the scenario with intensive fossil fuel burning and rapid economic growth), which is expected to enhance boundary layer moisture and reduce stability (hereafter referred to as “Plus4”).

We also use the FLEXPART-WRF model (version 3.3.2) to track the backward trajectory of airmass reaching the model domain (Brioude et al., 2013). FLEXPART-WRF is a Lagrangian particle dispersion model coupled with WRF enabling high-resolution simulations of atmospheric transport and dispersion. By using WRF's output – including wind fields, temperature, humidity, and turbulence parameters – FLEXPART-WRF tracks the trajectories of numerous air parcels (treated as particles) to model particle transport and to ultimately give insight into source-receptor relationships. For backward trajectory analysis, the model reverses the temporal integration of wind fields using the WRF output in the time frequency of 15 min, allowing particles to move backward in time from a receptor location (e.g., ACTIVATE aircraft sampling location) to identify potential emission sources. We use the turbulence option (TURB_OPTION=1 and CBL_SCHEME=1) that internally calculates boundary layer turbulent mixing using the Hanna turbulence scheme (Hanna, 1982) and allow the skewed turbulence (Brioude et al., 2013).

The cloud structure behind the frontal system exhibits diverse horizontal spatial patterns. Based on the classification criteria consistent with Chen et al. (2022), six zones are classified in each of the three experiments (Fig. 2) according to thresholds of smoothed liquid water path. The cloud water path is smoothed by a uniform filter of 30 grids, which removes the high-frequency variations in the complex cloud structure. Then we separate out the clear coastal and zones 3–6 with the values of smoothed water path of < 40, 40–200, 200–500, and > 500 g m⁻², respectively. Zone 1 corresponds to the eastern coastal landmass of the United States. Zone 2 represents a cloud-free region over the ocean. Zones 3–5 correspond to regions with PFCs, whereas zone 6 identifies the frontal system, which is dominated by extensive cloud cover.

PFCs west of the frontal zone (Zones 3–5) display distinct cloud morphologies. Between the clear-sky region (Zone 2) and the distinct cloud streets (Zone 4), Zone 3 serves as a transition zone where cloud streets appear from the upwind clear skies. Zone 4 features well-defined, continuous cloud street structures aligned with prevailing winds by a small angle (Chen et al., 2022). Near the frontal boundary (Zone 5), cloud streets transition into fragmented patches, manifested as open-cellular convection.

We test the hypothesis that variations in SST patterns influence coverage of each zone (Fig. 2b, c) by comparing the different experiments. Relative to the Ctrl experiment, the Plus4 simulation shows a 20.8 % reduction in spatial area of Zone 2, a 16.3 % reduction in Zone 3, a 12.7 % reduction in Zone 4, a 2.6 % expansion in Zone 5, and a 14.3 % expansion in Zone 6 (Fig. 3a). Also, the total area of PFCs (Zones 3–5) remains largely unchanged overall (only a 4.5 % reduction), despite a spatial shift of clouds toward the coastline. These changes suggest a significant enhancement in frontal cloud development (Zone 6) and a contraction of clear-sky areas (Zone 2). This cloud redistribution is likely driven by warmer SSTs and elevated boundary layer moisture, which promote convective cloud formation and frontal zone intensification.

In the Gradplus experiment, the areal coverage of Zone 5 (cloud street zone) decreases by 19.3 %, while Zone 2 (clear zone) expands by 20.2 %, reflecting a decline in organized PFC structures and an increase in coastal clear-sky coverage. Zones 3, 4, and 6 show small changes (1.5 % reduction, 6.4 % reduction, and 2.1 % expansion, respectively), compared to the Ctrl simulation. These results indicate that intensified SST gradients suppress cloud organization within PFCs, favoring localized clear-sky expansion at the expense of coherent cloud structures.

Figure 3b–e illustrates changes in cloud object size under the altered SST patterns. Cloud objects are defined here as regions where the ratio of LWP to smoothed LWP exceeds 0.8 (Chen et al., 2022). In the Plus4 experiment, PFCs exhibit a reduced number of cloud objects but a larger cloud size, while frontal clouds show minimal changes of cloud number despite a reduced cloud size. The aspect ratio in Fig. 3d, which quantifies cloud street elongation, increases by 122.2 % in Zone 4 (distinct cloud streets) indicating enhanced longitudinal organization of cloud streets in Zone 4. Mean LWP decreases by 26.0 % in Zone 4 and 15.5 % in Zone 5 but rises sharply by 24.4 % in Zone 6, reflecting more vigorous frontal clouds. These results suggest that warmer SSTs promote fewer but larger PFCs with elongated cloud streets in Zone 4, while frontal clouds become vertically deeper (evidenced by LWP increases) despite reduced horizontal extents.

In the Gradplus experiment, both PFCs and frontal clouds exhibit larger individual cloudy areas but a smaller number of clouds compared to the Ctrl simulation. The aspect ratio shows an overall reduction, except in Zone 4 where it demonstrates a slight increase. Mean LWP values remain consistently lower than those in the Ctrl simulation. The meteorological phenomena in Zones 5 and 6 may be attributed to their position within positive SST anomaly regions, where the primary SST gradients associated with the GS are displaced farther northwestward.

This section explores the mass and energy transport in frontal systems and PFCs. Previous studies have relied on isentropic analysis of moist convection to identify energy transport through diabatic processes, such as convection (Chen et al., 2023; Pauluis and Mrowiec, 2013). This method involves averaging the mass flux and other examined variables at the isopleths of equivalent potential temperature (θe) to isolate upward and downward motions that contribute to energy transport through their net effects. In other words, adiabatic vertical motions with oscillations (such as gravity waves) are filtered out by this averaging to derive irreversible energy/mass fluxes. Here, we apply the same method to Zones 4–6 in Fig. 4a–c to illustrate energy transport in these regions. We also show the mass flux at the isopleths of qv (Fig. 4d–f) and θ (Fig. 4g–i) to describe the properties of the air parcels.

Zones 5–6 exhibit a primary upward band at high θe and a downward band at low θe, consistent with previous studies on moist convection (Chen et al., 2023; Pauluis and Mrowiec, 2013). As parcels with high θe ascend and those with low θe descend, the net energy transport is upward. However, in Zone 4, two distinct upward bands are observed at different θe values: one at ∼ 287.75 K (similar to the upward band in Zone 5) and another at 278.25 K, which is notably lower. We speculate these two bands relate to different processes, which will be discussed below. Additionally, in Zones 4 and 5, a downward mass flux band is present above the clouds, terminating at lower altitudes near the cloud top. This may be attributed to large-scale downward motion within PFCs.

Since θe comprises two components, potential temperature (θ) and water vapor mixing ratio (qv), we also display the mass flux at isopleths of these variables in the second and third rows of Fig. 4, respectively. When using qv isopleths, the mass flux magnitude is significantly larger than with isentropic coordinates or θ isopleths, suggesting a stronger transport of qv by mass flux than θ, because upward and downward transports compensate on isopleths, and the mass fluxes defined on isopleths represent the fluxes that effectively transport the corresponding quantities. Cloud cores (black contours) appear directly above the upward bands in the qv analysis, implying that these clouds draw their moisture from the boundary-layer upward motion. In Zone 4, two separate cloud cores align with the two upward bands. When mass flux is averaged over θ isopleths, it is large within and above the clouds, indicating that thermal energy transport occurs primarily at the altitudes of clouds.

The two separate upward mass flux bands in Fig. 4a, d, and g suggest that Zone 4 involves distinct convection processes. We locate the two θe values associated with large mass fluxes on the spatial distribution of surface latent heat fluxes in Fig. 5. The higher θe band is located south of the GS region, at a distance from the GS itself, while the lower θe band is positioned at the northern GS edge. Because the higher θe value (287.75 K) in the southern upward band is comparable to the θe values in the upward bands of Zones 5 and 6, we speculate that the southern upward energy transport band may represent an extension of the frontal system, indicative of large-scale frontal influences. The low θe upward mass flux band at the northern GS edge is likely linked to local SST gradients, where mesoscale circulations may develop and influence the broader region, potentially triggering new convective upward energy transport. We further explore the secondary circulation associated with the SST gradients in Sect. 3.3.

In Fig. 6, we explore the impacts of SST variations on energy transport in Zones 4–6, with mass flux averaged below 2 km. In the Plus4 experiments, energy transport bands shift to higher isopleth values of θe, qv and θ across all zones. This suggests that warmer SST scenarios drive convective energy transport to higher energy levels, accelerating the process. The energy transport magnitudes in the Gradplus experiments are similar to those in the Ctrl experiments, but mass flux peaks are reduced in the lower θe band in Zone 4, indicating a weakened transport efficiency under the stronger SST gradients. The underlying mechanisms are discussed in Sect. 3.3. Zones 5 and 6 exhibit no significant differences between Ctrl and Gradplus experiments, likely because these zones lie outside the GS-influenced region.

In summary, the analysis of mass fluxes on isentropic coordinates indicates two distinct influences on the development of the cloud street zone: one influenced by the frontal system and the other by the SST gradient. The upward mass-flux band at the northern edge of the Gulf Stream illustrates the role of the SST gradient in affecting the post-frontal cloud system.

In this section, we focus on the inner domain where a contrast in SST anomalies is present. Figure 7 shows the vertical profiles of boundary layer properties in the negative-anomaly region (north of the GS) and the positive-anomaly region (south of the GS) at 09:00 EST. The Plus4 experiment results in higher boundary layer qv, θ, hydrometeor mixing ratio, and wind speed in both regions (Fig. 7a, b, c, and e). Additionally, the fraction of liquid-phase hydrometeors is reduced, indicating that clouds are elevated with more ice-phase hydrometeors in both regions (Fig. 7d). Wind direction shifts clockwise with the increased SST (Fig. 7f), likely due to a strengthened Coriolis effect by the enhanced wind speed.

Figure 8 displays the time series of differences between the Plus4 experiment and the Control simulation. The elevated boundary layer qv, θ, cloud liquid and ice water mixing ratio, and wind speed persist from sunrise to sunset. In the early morning before sunrise, liquid-phase cloud content is slightly higher in the Plus4 experiment, but the differences diminish after sunrise. The all-ice-phase hydrometeor content in Plus4 is significantly larger than in the Control experiment, suggesting that the higher SST promotes deeper cloud development.

Figure 7 demonstrates that SST anomaly impacts in the Gradplus experiment are notably weaker than in Plus4 at 09:00 EST. The temporal variations in Gradplus are less straightforward, as shown in Figs. 9–11, with the magnitude of the differences being much smaller than in the Plus4 experiment and the sign being less consistent. The GS region (“middle” in Figs. 9–11) corresponds to latitudes with maximum SST per longitude, with areas north and south of it labeled accordingly.

In Fig. 9, moving from north to south, the intensified SST gradient shifts their effects from negative to positive in the values of qv and θ within the boundary layer, indicating suppression of temperature and moisture in the north (due to a 25 % SST reduction) and enhancement in the south (due to an SST increase). At the top of the boundary layer, strong negative effects on qv are observed in the southern and middle regions, with altitude decreasing over time, consistent with declining boundary layer heights (Chen et al., 2022). Coupled with positive effects from Gradplus on wind speeds at these altitudes (Fig. 10a–c), we speculate that drier air advection from the north causes the negative qv anomaly at higher altitudes. Additionally, strong divergence at the top of the boundary layer in Fig. 10g–i provides evidence of drier air transport by intensified northerly winds.

However, despite stronger advection observed above the boundary layer, vertical transport of colder air is largely confined to the boundary layer in the northern region (Fig. 9d), leaving positive anomalies at higher altitudes. Meanwhile, stronger convection lifts warm air from the surface to above the boundary layer between 04:00 and 08:00 EST in the southern region (Fig. 10e–f). Stronger convergence observed during 04:00–08:00 EST in the southern region aligns with the enhanced vertical velocity (Fig. 10g–i). This suggests that transport of thermal energy is from south to north, opposite to the direction of moisture transport.

Two processes govern interactions between air above the negative SST anomaly region in the north and the positive SST anomaly region in the south: (1) North-to-south advection of smaller θ and qv, and (2) south-to-north diffusion of both θ and qv driven by their gradients (i.e., larger values in the south). The net effects observed in Fig. 9 result from these opposing processes. In Sect. 3.2, we show that the one of the upward mass flux bands with cloud formation is at the northern edge of GS, suggesting cloud processes interfere with the interactions of θ and qv between regions north and south of the GS. We hypothesize that higher qv lifted from surface to above the boundary layer in the southern region by convection is consumed via stronger phase changes with the intensified SST gradient, weakening the south-to-north diffusion of qv. Consequently, the north-to-south advection of drier air dominates (Fig. 9a–c). We hypothesize that stronger phase changes associated with the intensified SST gradient from water vapor to hydrometeors release more latent heat, warming the air and intensifying south-to-north heat diffusion (Fig. 9d–f).

We examine the hypothesis of cloud interference described above in Fig. 11, which shows that a stronger low-level convergence (below ∼ 2 km) and larger vertical velocities before 08:00 EST in Gradplus (Fig. 10f and i) closely associate with larger ice-phase hydrometeor mixing ratios during the same period. An increase of liquid-phase hydrometeors in Gradplus dominates within the boundary layer of the middle and southern regions (Fig. 11k and l). These results suggest that, in the Gradplus experiment (vs. Ctrl), enhanced latent heating from the increased ice-phase hydrometeors is key to modulating θ and qv interactions across GS regions above the boundary layer. This implies that microphysical processes – specifically, transitions from warm clouds to mixed-phase/ice-phase clouds – critically influence θ and qv values above the boundary layer.

Section 3.1–3.3 demonstrate that the Plus4 and Gradplus experiments introduce variations in local mesoscale circulation around the GS region. Beyond influencing thermal and moisture fields, we hypothesize that SST variations also modify airmass sources, which determine the type of aerosols (continental vs. marine) transported over the WNAO. To test this hypothesis, we use Lagrangian tracers to track backward trajectories of airmasses ending up within clouds. This approach aims to determine whether changes occur in the origin of airmasses that ultimately affect clouds.

We release tracers at four locations (Fig. 12a) and seven altitudes. The locations align with wind directions corresponding to cloud lines. The selected altitudes reflect their vertical position relative to cloud layers: (1) ocean surface, (2) 500 m above sea level and within the boundary layer, (3) cloud base, (4) midpoint between cloud base and the height of maximum qc, (5) height of maximum qc, (6) midpoint between the height of maximum qc and cloud top, and (7) cloud top. We calculate tracer concentrations for all 28 configurations of tracer release in the Ctrl, Plus4, and Gradplus experiments. Results show that, for six out of seven altitudes, tracer concentrations are similar at 30 min and 2 h prior to release time between sensitivity and control simulations (not shown). The exception occurs for the altitude between the height of maximum qc and cloud top (Squares in Fig. 12b).

In the Ctrl experiment, the primary air source originates from west of the tracer release location, as tracer positions at 2 h prior to release (cold color in Fig. 13) are situated farther west compared to their positions at 30 min before release (warm color in Fig. 13). When tracers are released within Box 2, two distinct airmass sources are observed: one from the cloud top and the other from the ocean surface. From Box 2 to Box 4, the cloud top airmass becomes increasingly dominant, suggesting that secondary circulation, associated with cloud streets that enhance tracer transport from the ocean surface, weakens progressively from Box 2 to Box 4.

In the Plus4 experiment, the airmass fraction originating from the ocean surface increases, likely due to the higher SSTs that drive a stronger turbulent transport within the boundary layer. In contrast, the contribution of airmass from the cloud top across all four boxes is enhanced in the Gradplus experiment, likely because stronger northwesterly winds above the boundary layer accelerate transport (Fig. 10a–c).

Aerosols linked to clouds in the examined regions are influenced by a mix of continental and marine sources. Our results regarding the air parcel origins implicitly suggest that SST variations associated with the GS trigger distinct shifts in aerosol composition. Specifically, marine aerosols influence Boxes 2–4 in Ctrl experiments and all four boxes in Plus4 experiments via air parcels originated from sea surface, while continental aerosols influence the other boxes through air parcels originated from continents. This subsequently alters aerosol-cloud interactions over the WNAO region.