The ERA5 reanalysis dataset (C3S, 2017; Hersbach et al., 2019, 2020) from the ECMWF was used to describe the three-dimensional large-scale flow situation during isoTrades in 2018, to calculate back trajectories for the extended period January–February 2009 to 2018 and to derive several Lagrangian transport diagnostics as further described below. In this study, the hourly ERA5 reanalysis data are interpolated to a regular horizontal grid with 0.5∘ spacing.

First evaluations of the difference between ECMWF's predecessor reanalysis product ERA-Interim and ERA5 show that spatial transport deviations between the two datasets can be up to an order of magnitude larger than those caused by parameterised diffusion and subgrid-scale wind fluctuations (Hoffmann et al., 2019). Hoffmann et al. (2019) found differences of up to 30 % in specific humidity along back trajectories calculated with the two datasets after 1 d. They showed that in addition to the improved spatial resolution in ERA5, changes in the forecast model, available observations and the data assimilation system all play a role in the observed deviations in transport and thermodynamic conditions along back trajectories between the two datasets. Here the added value of hourly data in ERA5 (ERA-Interim is only available 6-hourly) and the increased spatial resolution are the primary reasons for using the new reanalysis dataset.

The BCO (https://barbados.mpimet.mpg.de/, last access: 10 January 2021) was set up on a promontory near the most windward point of the island of Barbados, at Deebles Point (13.1626∘ N, 59.4287∘ W; see Fig. 1a), in 2009 (Stevens et al., 2016) and is operated as a cooperative research project of the Max Planck Institute for Meteorology, the Caribbean Institute for Meteorology and Hydrology, and the Museum of Barbados. Situated on a cliff at 17 m a.s.l., the BCO is directly exposed to the North Atlantic trade winds coming in from the east or north-east. Previous work has not been able to identify a significant island effect on the measurements. During December to May, Barbados is exposed to a typical trade wind flow with prevailing low-level easterlies and large-scale subsidence at upper levels. Shallow cumulus clouds can be observed throughout the year (Nuijens et al., 2014), organised in a variety of mesoscale cloud patterns (Stevens et al., 2020). Medeiros and Nuijens (2016) showed that observed and simulated clouds near Barbados are representative of much of the tropical oceans. Barbados therefore represents an ideal study site of shallow cumulus clouds in the trades.

Meteorological data from a Vaisala WXT-520 mounted on a 3 m mast at the BCO were used for calculating mean conditions during the different flow regimes. Cold pools affecting the BCO were identified following the method introduced in Vogel (2017) based on 1 min surface temperature data from the BCO. Radio-sounding data from the Grantley Adams Barbados Airport (15 km from the BCO) were used in the two case studies presented in Sect. 3.4 to characterise the contrasting lower-tropospheric thermodynamic conditions in the two extratropical flow regimes.

In preparation of the large international field campaign “Elucidating the role of clouds–circulation coupling in climate” (EUREC4A) in January–February 2020 (Bony et al., 2017; Stevens et al., 2021), a 24 d field experiment focusing on stable water isotope measurements (isoTrades) was carried out at the BCO. The water vapour isotope measurements and event-based precipitation samples from isoTrades served as a basis for the planning of the multi-platform isotope measurements on four ships, two aircraft and at the BCO performed during EUREC4A-iso by several European and US American teams.

The stable water isotope composition of a water sample is usually quantified by the δ notation (Craig, 1961a): δ [‰] = (Rsample/RVSMOW-1)⋅1000, where R is the isotopic ratio of either H218O or 2H1H16O (with R representing the ratio of the concentration of the heavy molecule to the concentration of H216O). The δ notation expresses the relative deviation of the isotopic (molecular) ratios from the internationally accepted primary water isotope standard, that is, the Vienna Standard Mean Ocean Water (VSMOW2; IAEA, 2017; with 2RVSMOW2 = 3.1152×10-4 and 18RVSMOW2 = 2.0052×10-3). The second-order isotope parameter deuterium excess (d=δ2H-8⋅δ18O; Dansgaard, 1964) serves as a tracer for non-equilibrium fractionation (Craig and Gordon, 1965; Pfahl and Wernli, 2008), in particular for events of strong large-scale ocean evaporation (Aemisegger and Sjolte, 2018).

The high-resolution temporal variability in water isotopes can be measured directly in the gas phase owing to recent advances in laser spectrometric devices that have reached sufficiently high precision (Baer et al., 2002; Kerstel, 2004; Crosson, 2008) and have been widely used in the field (Wei et al., 2019). As a part of isoTrades, a customised fast-response version of an L2130 Picarro cavity ring-down laser spectrometer characterised in detail in Aemisegger et al. (2012) and in the supplement of Thurnherr et al. (2020b) was installed in a temperature-regulated container (Tcontainer=24±2 ∘C) at the BCO. The set-up is described in Fig. 1b. In short, the inlet is shielded from rainfall and sea spray by a funnel through which a heated inlet line (80 ∘C, Winkler, Germany) guides the ambient air into the container. The inlet line material is PTFE; it has an outer diameter of 12 mm and a length of 9 m, with 5.5 m outside and 3.5 m inside the container (red line in Fig. 1b). The heated inlet line is flushed with the inlet pump (P1; KNF HN022AN.18). The sub-sample of the ambient gas is guided to the cavity ring-down system through a 30 cm isolated PTFE sample line with an outer diameter of 1/4 in. (green lines) by the instrument external pump (P2; KNF N920AP.29.18). Altogether the residence time in the system is 6 s, with 3 s in the inlet system and 3 s in the instrument. For the normalisation scale based on Vienna Standard Mean Ocean Water and Standard Light Antarctic Precipitation Water (referred to as the VSMOW–SLAP scale in the following) and for drift correction, the isotope composition of known liquid standards was measured daily for 20 min by the cavity ring-down system using a vaporiser and a standard delivery module from Picarro. The isotope composition of the standards used was (4.77±0.2) ‰, (-11.4±0.2) ‰ and (-34.5±0.1) ‰ for δ18O and (42.7±0.4) ‰, (-82.1±0.6) ‰ and (-267.8±0.1) ‰ for δ2H, respectively. Subsamples of the liquid standards were taken on the first and last day of the campaign to correct for a potential drift in the standard's isotope composition, which was determined in the same way and by the same laboratory as the precipitation samples (see below). The vapour isotope data are post-processed and normalised to the VSMOW–SLAP scale (Gonfiantini, 1978; IAEA, 2017) following the procedure described in Aemisegger et al. (2012). The measurement uncertainty based on a conservative estimate using error propagation is 0.8 ‰, 1.7 ‰ and 1.9 ‰ for δ18O, δ2H and d, respectively. From the daily calibration runs performed during the campaign, a more realistic estimate of the uncertainty can be obtained based on the root mean square deviation of the calibrated data to the reference standard values, which yields 0.3 ‰, 0.7 ‰ and 0.8 ‰ for δ18O, δ2H and d, respectively. For the combination with the hourly ERA5 back-trajectory analysis, the 1 Hz water vapour isotope data were averaged to hourly data (available online; see Aemisegger and Graf, 2020).

The precipitation samples were collected as soon as possible after a rainfall event using a precipitation sampling system that is especially designed to avoid post-sampling re-evaporation (PALMEX RS1). The same sampling system has also been used by the IAEA in its Global Network for Isotopes in Precipitation. The uncertainty due to small-scale variability in rainfall was assessed in Europe in a dedicated study on 10 precipitation events with an array of similar samplers and found to be <2 ‰ in δ2H and <0.3 ‰ in δ18O (Fischer et al., 2019). The isotope composition of the collected water samples was analysed by cavity ring-down laser spectrometry (Picarro L2130-i, Picarro Inc., Santa Clara, CA, USA) in “high-precision mode” in the laboratory of the Chair of Hydrology at the University of Freiburg, Germany. This laboratory regularly participates successfully in water isotope inter-comparisons from the IAEA (Wassenaar et al., 2018). Samples were filtered via syringe filters (0.45 µm) prior to analysis if they were muddy. Of each sample, 1 mL was filled into autosampler vials. According to the manufacturer's handbook, six injections per vial were analysed with the isotope analyser, and raw data of the first three injections were discarded to keep memory effects from one sample to the next at a minimum. Mean and standard deviation of the last three injections were calculated. In case there was still a memory effect, and the standard deviation was larger than 0.08 ‰ for δ18O or larger than 0.30 ‰ for δ2H, the fourth injection was also discarded, and only the last two injections were averaged. Calibration of the raw data was then conducted using three in-house standards with distinct isotopic compositions: -14.86 ‰, -9.47 ‰and 0.30 ‰ for δ18O and -107.96 ‰, -66.07 ‰and 1.53 ‰ for δ2H, referenced to the international VSMOW–SLAP scale (Craig, 1961a, b). The standards were each analysed in triplicates and averaged. The light and the heavy standards – embracing the samples – were used for a two-point calibration; the third standard was used for validation. Long-term post-calibration accuracy of the validation standard was ±0.05 ‰ for δ18O and ±0.35 ‰ for δ2H. A conservative estimate of the overall measurement uncertainty, comprising sampling and analytical uncertainty, yields 2.5 ‰ in δ2H and 0.5 ‰ in δ18O.

Even though Barbados lies close to the Equator, at 13.16∘ N, our systematic climatological-trajectory analysis reveals that during the winter months January–February, the sub-cloud layer air at the BCO very rarely originates purely from the tropics. Only for 12 % of the hourly time steps did the back trajectories never visit the subtropics or extratropics, i.e. never reached poleward of 23.5∘ N, in the last 10 d (Fig. 2). A total of 55 % percent of the air parcels arriving in the sub-cloud layer in Barbados have spent at least 1 d in the extratropics (north of 35∘ N), highlighting the climatological importance of transport pathways from the extratropics towards Barbados. Of the remaining 45 %, two-thirds of the air parcels have spent at least 1 d in the subtropics. Given the importance of the Intertropical Convergence Zone (ITCZ) for the tropical circulation as well as the midlatitude jet for the subsidence from the extratropics, it is likely that their respective latitudinal positions both play an important role in shaping the interannual variability in the residence times of air parcels in the tropics vs. the extratropics.

The isoTrades campaign period was characterised by an unusually large influence of air parcels originating from relatively far north compared to the climatology. This anomalous behaviour is linked to a poleward shift in the midlatitude jet in January–February 2018 over the western North Atlantic and a positive North Atlantic Oscillation index (NAO; see Supplement S2). Several studies have shown that anticyclonic Rossby wave breaking (ARWB) over the North Atlantic occurs preferentially in the positive phase of the NAO (Benedict et al., 2004; Rivière and Orlanski, 2007). Furthermore, January–February 2018 was characterised by a relatively weak deep-convective activity in the northern part of South America as well as a southward shift in the ITCZ over the South Atlantic compared to climatology (see Supplement S2). Both the northward shift in the midlatitude jet and the anomalies of the ITCZ may have played a role in the anomalously strong contribution of air parcels from the extratropics during isoTrades.

When classifying hourly time steps into the three flow regimes (see Sect. 2.1.2), the extratropical trade wind and dry intrusion regimes are climatologically about equally frequent (28 % and 27 %; Table 1), albeit with relatively large interannual variability (e.g. 62 % and 32 % in 2018; Table 1). The timing and strength of the subsidence clearly differ in the two extratropical flow regimes (Fig. 4; see also Table 2). Extratropical dry intrusions experienced pronounced subsidence in the subtropics between 3 and 5 d before arrival, whereas in the extratropical trade wind regime the strongest subsidence occurs between 5 and 7 d before arrival (Table 2). Subsidence within the tropics, i.e. for the tropical flow regime, can also be relatively fast in rare cases; however the median subsidence within 2 d is up to 60 hPa larger for extratropical dry intrusions 4 d prior to arrival (Fig. 4b). In the period 8–9 d before arrival, the subsidence rate of extratropical trade wind air parcels is similar to the one of the tropical air parcels. This might be due to the relatively strict definition of the extratropical flow regimes with some subtropical air parcels that experience a similar descent pathway as extratropical trade wind air parcels being classified into the tropical flow regime. Since the tropical flow regime is not the focus of this paper, we leave the more detailed analysis of the processes involved in this regime to future dedicated research.

The reasons for the difference in subsidence behaviour of the two extratropical flow regimes are related to the flow configuration over the North Atlantic (Fig. 5). The air parcels arriving in the sub-cloud layer in Barbados are advected predominantly with the typically observed easterly trade winds (see e.g. Fig. 5c, grey wind barbs for the situation in 2018). However, in some cases, a pronounced meridional transport of air occurs into the tropical lower troposphere, which interrupts the trades in the western North Atlantic due to an extratropical perturbation, as is shown in detail in a case study in Sect. 3.4. Given the pronounced meridional temperature gradient in the lower troposphere in winter, such a meridional transport event is expected to occur in a slantwise descending manner towards the tropics along the sloping isentropes (or even slightly steeper due to radiative cooling and interaction with low-level cloud tops). Four days prior to arrival, extratropical dry intrusion air parcels are located further north, with a median cardinal direction of 66∘ compared to the extratropical trade wind and the tropical flow regimes, in which the air parcels are located further east of Barbados (with a median cardinal direction of 74 and 76∘, respectively). Compared to the two other flow regimes, extratropical dry intrusions thus generally follow a meridionally slantwise descending pathway towards Barbados with anomalously large subsidence rates extending over a time period going slightly beyond the 4 d time window used for their definition (Fig. 4). In contrast, the extratropical trade wind regime shows anomalously weak subsidence with a substantial share of air parcels that are slightly ascending 1–2 d prior to arrival (negative subsidence in Fig. 4b).

The two transport pathways from the extratropics studied here leave a distinct fingerprint in the isoTrades 2018 campaign mean North Atlantic flow conditions (Fig. 5). In particular, two regions of enhanced subsidence can be observed in the upper-level (320 K) and mid-level (500 hPa) vertical wind ω (Fig. 5a and b, orange contour lines): one centred over the Canary Basin of the eastern North Atlantic and the second in the western North Atlantic, centred at 25∘ N and 55∘ W. These regions of enhanced upper- to mid-level subsidence reflect the two transport pathways of extratropical air towards Barbados: the time-averaged peak in the vertical pressure velocity near 20∘ W is associated with the extratropical trade wind regime, and the peak near 55∘ W is associated with the extratropical dry intrusion regime, respectively. A detailed overview of the day-to-day variability in transport between 25 January and 17 February 2018 is provided in the Supplement S3. Two types of large-scale features shown in Fig. 5 play an important role in steering the air parcels from the extratropics: the midlatitude jet stream and the North Atlantic surface anticyclones.

The high surface pressure in the subtropical North Atlantic found in the campaign mean (Fig. 5c) is a well-known climatological feature arising from the presence of North Atlantic subtropical surface anticyclones. The maximum surface pressure in the eastern North Atlantic is the signature of the quasi-stationary Azores' high-pressure system. The westward extension of the zonally elongated high-pressure feature in the campaign mean emerges due to more transient anticyclones, which also occur in the western part of the basin and propagate eastward from the North American east coast (see Supplement S3; Davis et al., 1997). The position of the surface anticyclone is strongly influenced by the position of the midlatitude jet stream and the zonal location of ARWB events (Thorncroft et al., 1993). Synoptic-scale ARWB events occur in situations with a meridionally extended undulation of the jet stream and appear in the shape of meridionally elongated and narrow tongues of potential vorticity (PV) on isentropic surfaces (McIntyre and Palmer, 1984; Appenzeller and Davies, 1992). The frequency of ARWB occurrence is higher over the eastern North Atlantic (20 %–30 %) than over the central North Atlantic (5 %–10 %; Wernli and Sprenger, 2007; Fröhlich and Knippertz, 2008; Martius and Rivière, 2016). During ARWB, the formation of elongated PV filaments (streamers) or isolated regions with stratospheric air (i.e. stratospheric cut-offs) can dynamically induce quasi-geostrophic vertical motion in a region with ambient baroclinicity. Quasi-geostrophic descent is typically induced at the westward side of the PV streamer or cut-off low. In the campaign mean for isoTrades in 2018, two subtropical maxima in PV at 320 K can be observed, one with PV values of 1–1.5 pvu (1 pvu = 10-6 m2 s(-1) K kg(-1)) near 20∘ N, 50∘ W and one with PV values of 1.5–2.5 pvu around 30∘ N, 15∘ W, respectively, in both cases east of the maximum in ω (Fig. 5a).

On the extratropical trade wind regime day (29 January), the moist trade wind layer is particularly deep, with nearly saturated air and easterly winds reaching up to 600 hPa (Fig. 9a). The shallow-cumulus-cloud pattern on that day features Gravel-like structures (Stevens et al., 2020), with many cold pools developing in the vicinity of Barbados (Fig. 10a and b). The radio sounding from 2 February (Fig. 9b) illustrates again the much drier near-surface conditions, the stronger northerly wind component up to 400 hPa and enhanced lower-tropospheric stability (between 950 and 600 hPa) typical for extratropical dry intrusions. In this regime, cloud-free conditions are prevailing in a large area around Barbados (Fig. 10c and d). The dynamical environment in which the air parcels arriving on these 2 exemplary days travel towards Barbados is analysed in more detail in this section.

On both 29 January and 2 February, the air parcels arriving in the upper troposphere above Barbados (trajectories with red and orange dots in Fig. 11) are associated with an approximately zonal westerly flow. However, for air parcels arriving at low levels, the transport pathways on the 2 d strongly differ. On 29 January, representative of the extratropical trade wind regime, sub-cloud and cloud layer air parcels (blue and green) descend in front of the western European and North African west coast in the context of an eastern North Atlantic surface anticyclone (Figs. 11a and 12a

For the sake of readability only air parcels with arrival times every 3 h starting at 00:00 UTC are shown.

The extratropical dry intrusion air parcels arriving in Barbados between 1 and 4 February subside in the context of a central North Atlantic ARWB (Fig. 131 and the animation in Supplement S5). These air parcels descend slantwise from the north within an airstream that consists of two branches (Figs. 11b and 13):

Branch 1 (B1) includes a few air parcels that come from the north and the east at low levels and ascend rapidly (Fig. 13a–d), likely due to deep convection, in the vicinity of the PV streamer that then develops into a PV cut-off on 28 January near 30∘ N, 47∘ W in the vicinity of the line C1 in Fig. 11b. Subsequently, these air parcels descend towards Barbados at the western edge of the PV cut-off (see animation in Supplement S5).

The dynamics of the large-scale descent of the air parcels arriving in Barbados in the two case studies is influenced by several extratropical weather systems developing in the context of the ARWB events. For example, during the extratropical dry intrusion, a large area of enhanced subsidence can be observed near the eastern edge of the transient western North Atlantic anticyclone (Fig. 11b) and west of an extratropical cyclone located below the PV cut-off on 29 January (Fig. 11b).

The different zonal location of the two ARWB events studied here has a distinct impact on the evolution of the thermodynamic properties of the air parcels associated with the extratropical trade wind and dry intrusion cases. Figure 15 shows the Lagrangian evolution of the two contrasting cases in the phase space of temperature (T) vs. potential temperature (θ). Visualising the evolution of trajectories in this θ–T phase space serves to distinguish adiabatic vs. diabatic processes that occur along the flow (e.g. Bieli et al., 2015; Papritz et al., 2019). The differences in the thermodynamic behaviour of the sub-cloud layer air parcels from the case studies on extratropical trade wind and dry intrusion flow regimes can thereby be summarised. The descent voyage of the extratropical dry intrusion shown by the red line in Fig. 15 can be split into three stages. During the first stage, 5–10 d before arrival, the air parcels are moving relatively fast within the midlatitude jet, and the thermodynamic properties of the air parcels remain approximately constant. During the second stage, 3–5 d before arrival, rapid descent (200 hPa (2 d)-1; Fig. 15, red line) occurs, and the airstream warms adiabatically in the upper troposphere (motion towards the right in Fig. 15 at relatively constant θ of about 304 K) in a largely cloud-free region (Figs. 14a, c and 15, red line). During the third stage of the descent, starting 3 d before arrival, the sinking air is further warmed adiabatically due to the descent but also cooled diabatically, in particular during nights (motion towards the bottom right in Fig. 15). In this third stage, the air parcels are located at the top of tropical low-level clouds (Fig. 14b and d), where they are most probably diabatically cooled by either microphysical (cloud evaporation) or radiative (inversion or cloud top radiative cooling) processes. In this stage, the air parcels start to take up substantial amounts of water vapour (Figs. 14d and 15, dots coloured with specific humidity). In contrast, the extratropical trade wind airstream (blue line in Fig. 15) begins to be moistened 7 d before arrival and is continuously diabatically heated by surface fluxes within the sub-cloud layer as it travels across the North Atlantic during the 5 d before arrival at the BCO.

In summary, for both case studies, ARWB plays a key role in setting the scene for a rapid descent of air parcels that reach Barbados either directly from the north in the case of central North Atlantic ARWB (dry intrusion) or from the east in the case of eastern North Atlantic ARWB (trade wind) after a 6 d voyage within the subtropical and tropical boundary layer. This contrast in the location of ARWB and in the pathway of the air parcels into the trades strongly influences the thermodynamic evolution of the air parcels arriving at the BCO near the surface.

During isoTrades, the variability in moisture transport pathways and local conditions discussed above lead to two very interesting summarising relations: (1) between the d and the moisture source distance (Fig. 16a) and (2) between d and the cloud patterns (Fig. 16b). These relations and their significance are shortly discussed in the following.

A strong anticorrelation between the d and the distance to the moisture uptake region is found during isoTrades (Fig. 16a). The smaller distance to the source during extratropical dry intrusions (500–1000 km) is indicative of enhanced and rapid moisture uptake (because of the large humidity deficit in the rapidly subsiding air) with little time for interaction with rainfall and clouds underway. Much longer transport distances are associated with the trade wind flow regime, in which longer-range transport across an increasing SST gradient is favourable for the formation of clouds and rainfall underway, leading to below-cloud interaction between sub-cloud layer vapour and falling rain drops. The d therefore can be seen as a measure for the time (and place) of the air parcels' entry into the boundary layer, with low d indicating “old” boundary layer air and high d “new” boundary layer air of extratropical origin. A clear observational linkage between the extratropical transport pathways and the isotope signals from the BCO is thereby obtained, which is not available from traditional humidity measures. Note that the relations between RH or q and the distance to the moisture source are much weaker than for d, and no physical rationale exists for their dependence to the distance to the moisture uptake region.

From the analysis of the impact of the different transport regimes on the trade wind water cycle detailed above, the question arises whether different transport pathways favour the occurrence of specific cloud patterns. One way to discriminate among different cloud patterns around Barbados using the surface wind speed and the lower-tropospheric stability has been presented in Bony et al. (2020b). The dominant cloud patterns during the isoTrades campaign were Gravel and Fish, with some occurrence of the Sugar pattern (Fig. 16b, Supplement S3; see also Stevens et al., 2020). The high-d anomalies during the extratropical dry intrusions are associated with the Fish cloud pattern, while the low-d anomalies in the trade wind flow regime are associated with the Gravel cloud pattern (Fig. 16b). These characteristic d fingerprints in the Fish and Gravel cloud patterns suggest important differences in the water vapour cycling for these two patterns. Based on the isoTrades campaign data, the Fish cloud pattern with anomalously low surface wind speeds and high LTS seems to be associated with extratropical dry intrusions behind trailing cold front passages. The calm conditions in this flow regime are likely due to the interruption of the normally near-zonal trade wind flow by the southward moving air. The high LTS might be due to the cold advection in the sub-cloud layer and the enhanced subsidence aloft. The potential link between enhanced stability and the dry intrusion has to be investigated in more detail in a future study by analysing the heat budget along back trajectories. The anomalously low d associated with the Gravel cloud pattern is due to the cloud processing and the in general more remote moisture uptake as discussed above. The influence of cold pool passages on the short-term variability in isotope signals in the trade wind regime, specifically during gust front passages, seems to be important. However, this short-term variability has a smaller amplitude than the high-d anomalies produced at the synoptic timescale during the extratropical dry intrusion flow regime. The anomalous d signal during the dry intrusion regime thus provides observational evidence of the altered large-scale flow conditions compared to the trade wind regime and to some extent validates our findings based on the back-trajectory analysis.

Given the influence of low-cloud amount and potentially cloud organisation on cloud radiative feedbacks (Bony et al., 2020b), the sensitivity of the cloud patterns to the different large-scale flow regimes in the trades has to be studied in more detail. In particular, the frequency of occurrence and persistence of the four cloud types identified in the surroundings of Barbados are likely tied to the prevailing flow regimes. This provides a promising starting point for a more detailed analysis of the link between different flow regimes and cloud patterns. In particular, the importance of the extratropical origin of air parcels during prolonged episodes of the Fish pattern will be further investigated using the EUREC4A isotope datasets in a follow-up study.