How Rossby wave breaking modulates the water cycle in the North Atlantic trade wind region

The interaction between low-level tropical clouds and the large-scale circulation is a key feedback element in our 10 climate system, but our understanding of it is still fragmentary. In this paper, the role of upper-level extratropical dynamics for the development of contrasting shallow cumulus cloud patterns in the western North Atlantic trade wind region is investigated. Stable water isotopes are used as tracers for the origin of air parcels arriving in the sub-cloud layer above Barbados, measured continuously in water vapour at the Barbados Cloud Observatory during a 24-day measurement campaign (isoTrades, 25 January to 17 February 2018). This data is combined with a detailed air parcel back-trajectory analysis using hourly ERA5 15 reanalyses of the European Centre for Medium Range Weather Forecasts. A climatological investigation of the 10-day air parcel history for January and February in the recent decade shows that 55% of the air parcels arriving in the sub-cloud layer have spent at least one day in the extratropics (north of 35°N) before arriving in the eastern Caribbean at about 13°N. In 2018, this share of air parcels with extratropical origin was anomalously large with 88%. In two detailed case studies during the campaign, two flow regimes with distinct isotope signatures transporting extratropical air into the Caribbean are investigated. 20 In both regimes, the air parcels descend from the lower part of the midlatitude jet stream towards the equator, at the eastern edge of subtropical anticyclones, in the context of Rossby wave breaking events. The zonal location of the wave breaking, and the surface anticyclone, determines the dominant transport regime. The first regime represents the “typical” trade wind situation with easterly winds bringing moist air from the eastern North Atlantic into the Caribbean, in a deep layer from the surface up to ~600 hPa. The moisture source of the sub-cloud layer water vapour is located on average 2000 km upstream of 25 Barbados. In this regime, Rossby wave breaking and the descent of air from the extratropics occurs in the eastern North Atlantic, at about 33°W. The second regime is associated with air parcels descending slantwise by on average 300 hPa (6 d)-1 directly from the northeast, i.e., at about 50°W. These originally dry airstreams experience a more rapid moistening than typical trade wind air parcels when interacting with the subtropical oceanic boundary layer, with moisture sources being located on average 1350 km upstream to the northeast of Barbados. The descent of dry air in the second regime can be steered towards 30 the Caribbean by the interplay of a persistent upper-level cutoff low over the central North Atlantic (about 45°W) and the associated surface cyclone underneath. The zonal location of Rossby wave breaking, and consequently, the pathway of https://doi.org/10.5194/wcd-2020-51 Preprint. Discussion started: 12 October 2020 c © Author(s) 2020. CC BY 4.0 License.

extratropical air towards the Caribbean, is shown to be relevant for the sub-cloud layer humidity and shallow cumulus cloud cover properties of the North Atlantic winter trades. Overall, this study highlights the importance of extratropical dynamical processes for the tropical water cycle and reveals that these processes lead to a substantial modulation of stable water isotope 35 signals in the near-surface humidity.

Introduction
Understanding and correctly predicting the patterns of shallow cloudiness over the trade-wind dominated tropical oceans is one of the current big challenges of climate research (Bony and Stevens, 2012). The interaction of these low-level clouds with their large-scale environment and their impact on Earth's radiation budget are key feedback elements in our climate system 40 (Bony et al., 2015). A large part of the uncertainty in current climate models is thought to be related to the models' representation of the strength of vertical mixing of water vapour in the lowest kilometres of the atmosphere, in particular between the boundary layer and the free troposphere (Sherwood et al., 2014). The balance between convective drying by mixing in of free tropospheric dry air and turbulent moistening of the boundary layer by ocean evaporation under different large-scale forcing situations is an important element in the process chain of shallow cumulus cloud formation. Furthermore, 45 evaporation of falling rain drops below low-level clouds into unsaturated downdrafts can trigger density currents spreading out at the surface and leading to the formation of convective cold pools. The gust fronts of propagating cold pools force environmental air to rise and can thereby trigger new convection, clouds and precipitation (Purdom, 1976;Weaver and Nelson, 1982;Zuidema et al., 2012;Torri et al., 2015). The three main components of the boundary layer moisture budget (Risi et al., 2019) namely (1) ocean evaporation, (2) convective drying, and (3) moistening by hydrometeor evaporation carry a distinct 50 signature in their stable water isotope composition (Aemisegger et al., 2015, Benetti et al., 2015, Sodemann et al., 2017, Aemisegger and Sjolte, 2018.
Given the strong contrasts in the isotope signature of the different sources mentioned above, isotope meteorology could be used as an observation-based measure to evaluate the influence of the large-scale circulation on shallow cumulus cloud patterns. The recent studies by Scholl and Murphy (2014a) and Torri et al. (2017) took first steps towards using stable water 55 isotopes to link convection in the tropics to the large-scale circulation. Due to the scarcity of available data with high precision and high time resolution in the tropics, Torri et al. (2017) relied on monthly precipitation isotope information from the Global Network of Isotopes in Precipitation stations of the International Atomic Energy Agency (IAEA) and Scholl and Murphy (2014a) used weekly collected precipitation and a few cloud water samples (Scholl et al., 2009, Scholl et al., 2014b. However, at these aggregated timescales a large part of the interesting dynamics of convective activity that is responsible for cloud 60 organisation is lost. Studies focussing on short timescale variability are thus in great need. In this paper, the hourly stable water isotope signature of the tropical trade wind sub-cloud layer measured at the Barbados Cloud Observatory (BCO, black cross in Fig. 1, Stevens et al., 2016) during a 24-day campaign in January-February 2018 (isoTrades) is used in combination with a detailed air parcel back-trajectory analysis based on ERA5 reanalysis data from the https://doi.org/10.5194/wcd-2020-51 Preprint. Discussion started: 12 October 2020 c Author(s) 2020. CC BY 4.0 License. follows: in Section 2, a description of the used Lagrangian diagnostics based on ERA5 reanalyses is provided as well as a summary of the measurement setup and postprocessing of the isotope data; in Section 3, the results addressing the four 100 objectives formulated above are presented and discussed; and a concluding summary is given in Section 4.

Large-scale flow characteristics and Lagrangian diagnostics
The ERA5 reanalysis dataset (Copernicus Climate Change Service, 2017;Hersbach et al., 2019;Hersbach et al., 2020) from the ECMWF was used to describe the three-dimensional large-scale flow situation during isoTrades in 2018, to calculate 105 backward 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 is 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 110 parameterised diffusion and subgrid-scale wind fluctuations after one day (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 one day.
They showed that in addition to the improved spatial resolution in ERA5, changes in the forecast model, available observations and in 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 6-115 hourly) and the increased spatial resolution is the primary reason for using the new reanalysis dataset.

Trajectory calculation
Ten-day backward trajectories were calculated with the Lagrangian Analysis Tool (LAGRANTO, Wernli and Davies;1997, Sprenger and based on the three-dimensional wind fields from ERA5. The trajectories were started hourly above the geographical position of the BCO (13.16°N, 59.43°W, Fig. 1) as well as 4 other points displaced zonally and meridionally 120 by 0.5° from the BCO to account for the uncertainty of trajectory calculations with the resolved wind. The starting points were vertically stacked every 7.5 hPa between 1000 hPa and 940 hPa for January-February in the period 2009and up to 200 hPa in 2018 to be able to also look at the transport at higher levels. For the case studies during isoTrades, we separately considered two layers with trajectory arrival points: the sub-cloud layer (p ³ 940 hPa) and the cloud layer (940 > p ³ 700 hPa).
Even though the vertical extent of the sub-cloud and cloud layers is variable in time and depends on the strength of the shallow 125 convective activity, we choose to use fixed vertical layers. The reason for this simplifying choice is that the temporal variability is not large, and the precision of our subsequent calculations would not benefit from accounting for this variability. We defined the cloud base height at 940 hPa (~630 m) and the cloud top height at 700 hPa (~3 km). The chosen cloud base level is in close https://doi.org/10.5194/wcd-2020-51 Preprint. Discussion started: 12 October 2020 c Author(s) 2020. CC BY 4.0 License. agreement with the estimate of Nujiens et al. (2014), who found a cloud base level from the BCO humidity and temperature measurements at 700±150 m. The top of the cloud layer is chosen slightly above the mean level of the trade wind inversion 130 (2 km). With this setup, the sub-cloud layer air consists of 40 air parcels per hourly time step, and the cloud layer of 150 air parcels. For the majority of the analyses conducted in this paper, in particular for the combination with the near-surface stable water isotope measurements, the trajectories arriving in the sub-cloud layer are used. Note that the spread of the trajectories calculated from different starting points in a given vertical layer allows us to take into account the uncertainty of the trajectory calculation and to capture the effect of mixing of air parcels from different origin. 135

Trajectory-based diagnostics
The following four types of trajectory-based diagnostics are used in this paper and introduced here: 1) Air parcel residence times in different spherical caps: The residence time (tl) of air parcels in different Northern Hemisphere caps (north of l=23.5°N, 30°N, 35°N, 40°N, 50°N, 60°N), is calculated based on the position of the air parcels within 10 days before their arrival at the BCO. Thus, e.g. the 140 residence time north of 23.5°N is referred to as t23.5. The residence time in the extratropics is defined as t35. The exceedance probabilities of t, i.e. the occurrence frequency of air parcels with t ≥ x days in a given latitudinal band provides a climatological measure of the origin of air parcels arriving in Barbados. The exceedance probability of t for the range 0 to 10 days in the defined spherical caps is calculated for a climatology of the recent 10 years as well as separately for 2018, in order to place the year of the isoTrades campaign in a climatological context. The result of this 145 analysis is shown in Fig. 3 and discussed in Section 3.1.

2) Three flow regimes:
Air parcels arriving in the sub-cloud layer in Barbados are classified into three regimes: the extratropical dry intrusion regime, the extratropical trade wind regime and the tropical regime (see schematics in Fig. 2). For the moistening of the sub-cloud layer air parcels, their latitudinal origin and the characteristics of their subsidence pathway is of particular 150 importance. Therefore, the definition of the three flow regimes used in this study is based on the median residence time in the extratropics ̃! " and the median amplitude of the subsidence (∆ & ) of air parcels. The tropical flow regime is distinguished from the two extratropical flow regimes based on ̃! " < 1 day. If ̃! " ≥ 1 day, the time step is classified into one of the extratropical flow regimes, which are differentiated based on ∆ & #$ in the four days prior to arrival of the air parcels in Barbados, with ∆ #$ = %&'( − %&'()#$ , where p is the air parcel pressure and tBCO the arrival time in 155 Barbados. If ∆ & #$ ≥ 100 hPa (4 d) -1 the date is classified into the extratropical dry intrusion flow regime. If ∆ & #$ < 100 hPa (4 d) -1 , the date is classified into the extratropical trade wind flow regime with sub-cloud layer air originating from the extratropics but having experienced limited subsidence in the four days prior to arrival in Barbados. A more detailed justification for the chosen flow regime classification framework can be found in Appendix A.
https://doi.org/10.5194/wcd-2020-51 Preprint. Discussion started: 12 October 2020 c Author(s) 2020. CC BY 4.0 License. The occurrence frequencies of the different regimes during the isoTrades campaign and in the climatological period 2009-160 2018 are summarised in Table 1 and discussed in Section 3.1. To characterise the subsidence behaviour of the air parcels arriving in the sub-cloud layer in Barbados in the three defined flow regimes, the hourly median pressure ( *) and two-day subsidence rate Δ & *$ are calculated for 1 to 10 days before arrival in Barbados. Additionally, for the campaign period in January-February 2018, several flow regime average characteristics are calculated from the ERA5 trajectories and from the local meteorological and isotope variables measured at the BCO. These values are shown in Tables 2 and 3 and shortly 165 discussed in Section 3.2. Given the low occurrence frequency of the tropical flow regime in 2018 (6%), this flow regime is not further analysed in terms of its associated isotope signature and sub-cloud layer properties due to the small sample size.
3) Characteristics of maximum slantwise descent: To investigate the dynamical environment in which the subsidence from the extratropics occurs, the two-day period of 170 maximum subsidence is identified for each trajectory (max(Dp2d)). Time, pressure, latitude, longitude, specific and relative humidity are calculated for the start and end of the period of maximum subsidence for all the trajectories arriving in the sub-cloud layer and the average for each flow regime is calculated and shown in Table 2. 4) Moisture sources and moisture uptake characteristics: The moisture sources of the air parcels arriving in the sub-cloud layer are identified using the method of Sodemann et al. 175 (2008). In short, this method considers the mass budget of water vapour in an air parcel. Moisture uptakes are registered, whenever the specific humidity along an air parcel increases. The weight of each uptake depends on its contribution to the final humidity of the trajectory. If precipitation occurs (i.e. a decrease of specific humidity along the trajectory happens) after one or several uptakes, the weight of all previous uptakes is reduced proportionally to their respective contribution to the loss. The moisture sources identified for each trajectory are subsequently weighted by the air parcel's specific 180 humidity at the arrival in the sub-cloud layer. This method has been used extensively, in particular for the interpretation of water isotope signals in vapour and precipitation (e.g., Pfahl and Wernli, 2008;Aemisegger et al., 2014;Aemisegger, 2018;Thurnherr et al., 2020). In addition to the moisture source regions, the moisture uptake during the two-day period of maximum subsidence as well as during the period after maximum subsidence until arrival in Barbados is calculated and the average for each regime is given in Table 2. 185

Meteorological data from the Barbados Cloud Observatory
The BCO (https://barbados.mpimet.mpg.de/) 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. 1) 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 190 in from the east or northeast. 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-https://doi.org/10.5194/wcd-2020-51 Preprint. Discussion started: 12 October 2020 c Author(s) 2020. CC BY 4.0 License. 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., 2020a). Medeiros and Nuijens (2016) showed that observed and simulated clouds near Barbados are representative for much of the tropical oceans. Barbados therefore represents an ideal 195 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 was 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 Section 3.4 to characterise the contrasting 200 lower tropospheric thermodynamic conditions in the two extratropical flow regimes.

Stable water isotope measurements at the Barbados Cloud Observatory
In preparation of the large international field campaign "Elucidating the role of clouds-circulation coupling in climate" (EUREC 4 A) in January-February 2020 (Bony et al., 2017, Stevens et al., 2020b, a 24-day field experiment focusing on stable water isotope measurements (isoTrades) was carried out at the BCO. The water vapour isotope measurements and event-based 205 precipitation samples from isoTrades served as a basis for the planning of the multi-platform isotope measurements on four ships, two aircrafts, and at the BCO performed during EUREC 4 A-iso by several European and U.S. American teams.
The stable water isotope composition of a water sample is usually quantified by the δ notation (Craig, 1961a): δ [ ‰] = ( +,-./0 / 123(4 − 1) ⋅ 1000, where R is the isotopic ratio of either H2 18 O or 2 H 1 H 16 O (with R representing the ratio of the concentration of the heavy molecule to the concentration of H2 16 O). The δ notation expresses the relative deviation 210 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 2 RVSMOW2=3.1152×10 −4 and 18 RVSMOW2=2.0052×10 −3 ). The second order isotope parameter deuterium excess ( = δ * H-8 • δ 56 O, 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). 215 The high-resolution temporal variability of 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 a L2130 Picarro cavity ring-down laser spectrometer characterised in detail in Aemisegger et al. (2012) and in the supplement of Thurnherr et al. (2020) was installed in a temperature-regulated container (Tcontainer = 24 ± 2 °C) at the BCO. The setup is 220 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 ¼ inch (green lines) by the instrument external 225 https://doi.org/10.5194/wcd-2020-51 Preprint. Discussion started: 12 October 2020 c Author(s) 2020. CC BY 4.0 License. pump (P2, KNF N920AP.29.18). All together 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 to the international IAEA VSMOW-SLAP scale 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. 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 is post-processed and normalised to the VSMOW-VSLAP2 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 d 18 O, d 2 H, and d, respectively. From the daily calibration runs performed 235 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 d 18 O, d 2 H, and d, respectively.
For the combination with the hourly ERA5 backward trajectory analysis, the 1 Hz water vapour isotope data was 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 240 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 of 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 d 2 H and < 0.3 ‰ in d 18 O (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 245 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 250 the last three injections were calculated. In case there was still a memory effect and the standard deviation was larger than 0.08 ‰ for δ 18 O or larger than 0.30 ‰ for δ 2 H, 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 d 18 O, -107.96 ‰, -66.07 ‰, and 1.53 ‰ for d 2 H referenced to the international VSMOW-SLAP scale (Craig, 1961a,b). The standards were analysed in triplicates each and averaged. The light and the heavy standards 255 -embracing the samples -were used for a 2-point calibration, the third standard was used for validation. Long-term postcalibration accuracy of the validation standard was ± 0.05 ‰ for δ 18 O and ± 0.35 ‰ for δ 2 H. A conservative estimate of the overall measurement uncertainty comprising sampling and analytical uncertainty yields 2.5 ‰ in d 2 H and 0.5 ‰ in d 18 O. https://doi.org/10.5194/wcd-2020-51 Preprint. Discussion started: 12 October 2020 c Author(s) 2020. CC BY 4.0 License.

Results and Discussion
The results of this study are presented in five parts, beginning with the climatological perspective on the relevance of transport 260 pathways from the extratropics towards Barbados in Section 3.1, followed by a short overview of the atmospheric flow conditions during isoTrades in January-February 2018 in Section 3.2. The impact of the two flow regimes with extratropical influence on the stable water isotope signals, as well as on other properties of the water cycle are discussed in Section 3.3. The midlatitude dynamical flow configuration of the two extratropical transport regimes is described in a detailed comparative case study in Section 3.4. Finally, a short summary on the link between isotope signals and the extratropical transport regimes is 265 given in Section 3.5.

Importance of air with extratropical origin for the sub-cloud layer over Barbados
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, the backward trajectories never visited the subtropics or extratropics, i.e., never reached 270 poleward of 23.5°N, in the last 10 days (Fig. 3). Fifty five percent of the air parcels arriving in the sub-cloud layer in Barbados have spent at least one day 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 has spent at least one day in the subtropics. As shown in Fig. 3, the year 2018 is associated with an unusually large influence of air parcels originating from relatively far north compared to the climatology. When classifying hourly time steps into the three flow regimes (see Section 275 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). Given the importance of the intertropical convergence zone (ITCZ) for the tropical circulation as well as the midlatitude jet on 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. 280 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 days before arrival, whereas in the extratropical trade wind regime the strongest subsidence occurs between 5 and 7 days 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 days is up to 60 hPa larger for extratropical dry intrusions four days prior to arrival (Fig. 4b). In the period 285 8-9 days 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 tropical flow regimes to future dedicated research. 290 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 will be shown in detail in a case study in Section 295 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). 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 300 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 four-day 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 days prior to arrival (negative subsidence in Fig. 4b). 305

Overview of atmospheric flow conditions during isoTrades
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 upperlevel (320 K) and mid-level (500 hPa) vertical wind w (Fig. 5a,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 310 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 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 Supplementary material S1. Two types of large-scale features shown in 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, 320 which also occur in the western part of the basin and propagate eastward from the North American east coast (see Supplementary material S1, 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 anticyclonic Rossby wave breaking (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 325 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 cutoffs) 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 cutoff low. In the campaign mean 330 for isoTrades in 2018, two subtropical maxima in PV on 320 K can be observed, one with PV values of 1-1.5 pvu near 50°W, 20°N and one with PV values of 1.5-2.5 pvu around 15°W, 30°N, respectively, in both cases east of the maximum in w (Fig.   5a).

Contrasts in air parcel transport characteristics 335
In January-February 2018, the extratropical trade wind and extratropical dry intrusion regimes are both associated with ARWB over the North Atlantic, however, as discussed in Section 3.2, at different longitudes. The time of the ARWB events is estimated by searching for the period with the strongest two-day subsidence of the trajectories. For trajectories in the trade wind regime, the strongest descent of 403 hPa (2 d) -1 occurs 6-8 days prior to arrival from 47°N, 35°W towards North Africa (Table 2). They reach the boundary layer (i.e. the trade wind inversion) in front of the North African coast (35°N, 20°W), 340 where they start their pathway across the North Atlantic within 6 days in the subtropical and tropical boundary layer. For trajectories in the extratropical dry intrusion regime, ARWB in the central North Atlantic leads to a maximum subsidence of on average 440 hPa (2 d) -1 starting at 41°N, 51°W towards 29°N, 43°W (Table 2). Both regimes can thus involve rapid descent as defined by Raveh-Rubin (2017) in her dry intrusion climatology, for some air parcels exceeding 400 hPa (2 d) -1 . The occurrence of such a rapid descent is rarer for extratropical air parcels forming a trade wind flow towards Barbados (53%) 345 compared to air parcels forming extratropical dry intrusions as defined here (72%). This is likely due to the weaker baroclinicity in the eastern North Atlantic compared to the western part of the basin. Note that the strength of the strongest descent of 400 to 450 hPa (2 d) -1 in the two extratropical regimes associated with ARWB is about one order of magnitude larger than the climatological vertical pressure velocity (25 hPa d -1 ) resulting from adiabatic compression due to subsidence balancing radiative cooling (~1 K d -1 , see, e.g. Holton and Hakim 2013) 350 The moisture source regions associated with these two regimes are shown in Fig. 6 with two maxima of moisture uptake at the southwestern edge of the anticyclone between 30°W and 60°W for the trade wind regime and between 50°W and 60°W for the dry intrusion regime. These two maxima correspond to the two maxima in time-mean surface evaporation south of the two maxima in w (Fig. 5c). They reflect the impact of enhanced meridional subsidence of dry air towards the ocean surface, which increases the near-surface vertical humidity gradient. The presence of an upper-level PV streamer northeast of the moisture 355 uptake maximum in the extratropical dry intrusion regime composite (Fig. 6b) is a first indication for the role of extratropical dynamical forcing for the enhanced vertical winds, descent and moisture uptake to the northeast of the Caribbean.
The two extratropical transport regimes thus have a contrasting impact on the properties of the North Atlantic trade wind water cycle (Fig. 6, Table 2). In the trade wind regime, the sub-cloud layer air parcels are continuously warmed, due to the zonal SST gradient across the North Atlantic, leading to continuous moisture uptake along their low-level westward pathway. These 360 air parcels generally take up 73% of their final humidity (Table 2) from ocean evaporation and below-cloud evaporation of rainfall. The latter humidity is taken up at the south-eastern edge of North Atlantic anticyclones on average 2028 km upstream of Barbados and within 6 days prior to arrival (Table 2). Extratropical dry intrusion air parcels take up their moisture on average 1348 km upstream to the north of Barbados (Table 2, Fig. 6). After their fast adiabatic descent, the extratropical dry intrusion air parcels reach a relative humidity of 55% and increase their specific humidity from 1.4 g kg -1 to 4.9 g kg -1 (Table 2). Twenty-365 seven percent of their humidity uptake occurs in the rapid descent phase of the air parcels and can be due to horizontal mixing or due to convective injection of moisture into the air parcels ( Table 2). The rest (63%) of the humidity uptakes occurs after the period of maximum subsidence and is likely due to enhanced surface fluxes, as well as convective and turbulent mixing in the boundary layer during the slow final descent of the air parcels towards Barbados (Table 2). A more detailed study of the physical processes involved in the moistening of extratropical trade wind compared to dry intrusion air streams is out of scope 370 here but is planned using process-specific moisture tendency output from a simulation with the IFS model (see, e.g., Spreitzer et al., 2019).
Over the isoTrades campaign period, substantial temporal variability appears in the Lagrangian transport diagnostics shown in Fig. 7, particularly in the four-day subsidence rate and the longitudinal location of the weighted mean moisture source location. At the beginning of the extended period of the extratropical dry intrusion in early February the residence time of the 375 air parcels in the extratropics is particularly large with 7 days (Fig. 7). These air parcels experience a strong descent of 300 hPa (4 d) -1 and take up their humidity relatively close to Barbados in the northeast of the island (Fig. 7). Compared to the persistent nature of the extratropical trade wind regime (prolonged periods with blue bars in Fig. 7), extratropical dry intrusions are generally more transient (short spells of red bars and short peaks in Dp4d in Fig. 7). The prolonged period of the extratropical intrusion at the beginning of February is an exception. During the rest of the isoTrades campaign, enhanced subsidence into 380 the sub-cloud layer associated with extratropical dry intrusions mostly occurs during short episodes. These events are often induced by a short-lived central North Atlantic upper-level PV streamer or weak cutoffs reaching particularly far south (see Supplementary material S1). Furthermore, often these events are associated with wider distributions of the Lagrangian diagnostics (shaded areas in Fig. 7), particularly for Dp4d revealing the importance of mixing of air parcels with different transport histories. 385

Contrasts in local stable water isotope and meteorological conditions
The two extratropical flow regimes characterised above in terms of their transport characteristics are associated with remarkably contrasting stable water isotope signatures as will be discussed in this section along with the differences in local meteorological conditions in Barbados for the two regimes in 2018. The stable water isotope signals and meteorological conditions during isoTrades (Fig. 8) show three striking periods: 1) the positive anomalies in the d-values and a local minimum 390 in d between 28 and 30 January (trade wind regime case study), 2) a prolonged period of anomalously low specific humidity with about half of the air parcels arriving in the sub-cloud layer showing characteristics of extratropical dry intrusions and the other half following the pathway that is typical for extratropical trade wind air parcels. This more complex event is likely also more sensitive to uncertainties in the trajectory calculation and will therefore not be further analysed in this study.
The extratropical trade wind regime leaves a distinct signature in the short-term variability of water vapour isotope signals with anomalies of up to +4 ‰ in d 2 H, +1 ‰ in d 18 O, and -2 ‰ in d (Fig. 8a). This regime is associated with positive anomalies 400 in the total column water vapour, intense cold pool activity with more than 3 cold pool passages per day, leading to intense short rain showers (Fig. 8c, Table 3). The relative humidity on 28 to 30 January is high (75-85%) and strong easterly winds (8-10 m s -1 , Fig. 8b) prevail. Due to below-cloud interaction of rainfall droplets with ambient vapour, the cold pool passages leave a characteristic signature in the short-term variability of water vapour isotope signals. Campaign mean precipitation isotope compositions are d 2 Hp=8.1 ‰, d 18 Op=-0.4 ‰ and dp=11.4 ‰ (Table 4), which is in the same range as the winter 405 precipitation measured in eastern Puerto Rico (Scholl and Murphy, 2014). Rapid total re-evaporation of rain droplets (i.e. no net fractionation) e.g. at cold pool gust fronts may thus have contributed to the positive anomalies in d values and negative anomalies in d observed during the trade wind regime periods of the campaign. Furthermore, moisture input from sea spray evaporation (Thurnherr et al., 2020), a process which is certainly enhanced during extratropical trade wind flow situations due to higher wind speeds (Table 3) can also lead to an increase in d 18 O and d 2 H and a decrease of d in sub-cloud layer vapour. 410 During the particularly strong extratropical intrusion event at the beginning of February, a depletion in heavy isotopes by up to 6 ‰ in d 2 H, 1 ‰ in d 18 O and an increase by 6 ‰ in d as was measured (Fig. 8a). In this regime, the total column water vapour is reduced by 10 mm and the lower troposphere is stabilised (+1 K in lower tropospheric stability, Klein and Hartmann, 1993) compared to the trade wind regime (Table 3). During the extratropical dry intrusion period at the beginning of February, the lower tropospheric stability was particularly large (~16 K, Fig. 8c). The strong drying of the free troposphere during such 415 an event (Fig. 10b) is likely to enhance radiative cooling at the top of the boundary layer, which thereby strengthens the inversion . There is hardly any cold pool activity and large-scale areas of clear-sky conditions dominate (-9% in total cloud cover compared to the conditions during extratropical trade winds, Table 3) due to the continuous supply https://doi.org/10.5194/wcd-2020-51 Preprint. Discussion started: 12 October 2020 c Author(s) 2020. CC BY 4.0 License.
of dry upper-level extratropical air into the tropical lower free troposphere. The entrainment of the dry air into the boundary layer lowers the surface humidity at the BCO (-2% in RH and -0.6 g kg -1 in q in the composite mean compared to the campaign 420 mean in Table 3) thereby increasing the near surface vertical humidity gradient, and intensifying ocean evaporation (+0.5 mm d -1 ). A clear minimum in the BCO wind speed can be observed (2-4 m s -1 , Fig. 8b) as well as a very low near-surface relative humidity of 55% compared to the campaign mean of 71% (Fig. 8b). No precipitation was registered at the BCO for a period of 5 days during the extratropical dry intrusion case at the beginning of February (Fig. 8b). The combination of these processes led to an increase of near-surface d due to enhanced non-equilibrium fractionation effects during ocean evaporation and a 425 lowering of the d in water vapour.
Concluding this section, we would like to note that, even though many processes affect the variability of water vapour isotopes, the clear contrast in isotope signals on our chosen case study days provides a robust foundation for their use as proxies for the two transport pathways from the extratropics in 2018. 430

Involved dynamical processes: a comparative case study
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 with many cold pools developing in the vicinity of Barbados (Fig. 10a,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 435 lower tropospheric stability (between 950 hPa and 600 hPa) typical for extratropical dry intrusions. In this regime, cloud free conditions are prevailing in a large area around Barbados (Fig. 10c,d). The dynamical environment in which the air parcels arriving on these two exemplary days travel towards Barbados is analysed in more detail in this section.
Both on 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 440 levels, the transport pathways on the two days 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 (Fig. 11a, Fig. 12a 1 ). This anticyclone is associated with an ARWB that develops over western Europe and the Mediterranean. The air parcels that descend fastest are located in an area with enhanced subsidence at the southwestern tip of a PV streamer that moves over the Atlas (Fig. 12a and the  445 animation in the Supplement S2). After their descent during the wave breaking along the northwest African coast and over North Africa, the air parcels travel across the North Atlantic within the boundary layer towards Barbados.
The extratropical dry intrusion air parcels arriving in Barbados between 1 to 4 February subside in the context of a central North Atlantic ARWB (Fig. 13 1 , and the animation in Supplement S3). These air parcels descend slantwise from the north within an airstream that consists of two branches (Figs. 11b and 13): 450 -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 cutoff on 28 January near 47°W, 30°N in the vicinity of the line C1 in Fig. 11b. Subsequently, these air parcels descend towards Barbados at the western edge of the PV cutoff (see animation in Supplement S3).
-Branch 2 (B2) with a majority of air parcels that first travel within the midlatitude jet (Fig. 13a) and then subside 455 meridionally in an area with enhanced subsidence at the right jet exit (Fig. 13a,b and animation in Supplement S3), and a few days later west of the PV cutoff (in the vicinity of line C2 in Fig. 11b, Fig. 13c,d).
The air parcels from B1 thus experience strong ascent at the eastern edge of the PV streamer (blue dots in Fig. 13c), whereas the air parcels from B2 experience strong descent at the western flank of the PV streamer (red dots in Fig. 13c) and subsequently of the PV cutoff (red dots in Fig. 13d). Interestingly, the two branches join along the PV streamer and the air parcels from B1 460 and B2 gather below and around the PV cutoff between 28 and 29 January (Figs. 13d, 14a,c). Two cross sections along lines C1 and C2 in Fig. 14 illustrate the environment in which the air parcels from B1 and B2 gather and achieve their concerted final meridional descent into the tropics.
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 465 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 cutoff 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 (q). 470 Visualizing the evolution of trajectories in this q -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 subcloud layer air parcels from the extratropical trade wind and dry intrusion flow regime case studies 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 days before arrival, the air parcels are moving relatively fast within the midlatitude jet and the thermodynamic 475 properties of the air parcels remain approximately constant. During the second stage, 3-5 days 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 q of about 304 K) in a largely cloud-free region (Fig. 14a,c and Fig. 15 red line). During the third stage of the descent, starting three days 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, 480 the air parcels are located at the top of tropical low-level clouds (Fig. 14b,d), where they are most probably diabatically cooled https://doi.org/10.5194/wcd-2020-51 Preprint. Discussion started: 12 October 2020 c Author(s) 2020. CC BY 4.0 License. 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 (Fig. 14d, Fig. 15 dots coloured with specific humidity). In contrast, the extratropical trade wind airstream (blue line in Fig. 15) begins to be moistened seven days before arrival and is continuously diabatically heated by surface fluxes within the sub-cloud layer as it travels across the North Atlantic during the 485 five days 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 either reach Barbados 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 six day 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 490 thermodynamic evolution of the air parcels arriving at the BCO near the surface.

Linking the deuterium excess to the moisture transport pathways and cloud patterns around Barbados
During isoTrades, the above discussed variability in moisture transport pathways and local conditions 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. 495 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 500 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 high 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.
From the above detailed analysis of the impact of the different transport regimes on the trade wind water cycle, the question 505 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 sugar (Fig. 16b, Supplement S1, see also Stevens et al., 2020a). The high d anomalies during the extratropical dry intrusions are associated with the fish cloud pattern, while the low d anomalies are associated with the gravel cloud pattern. This provides 510 a promising starting point for a more detailed investigation on the importance of the extratropical origin of air parcels during prolonged episodes with fish clouds, which will be performed using the EUREC 4 A isotope datasets. https://doi.org/10.5194/wcd-2020-51 Preprint. Discussion started: 12 October 2020 c Author(s) 2020. CC BY 4.0 License.

Conclusions and Outlook
In this paper, we show that in winter, the dynamics of the large-scale descent in the subtropics is essential for the variability of the stable water isotopes, the sub-cloud layer humidity, and the low-level cumulus cloud cover in the western North Atlantic 515 trade wind region. The stable water isotope signals in vapour and precipitation from a 24-day measurement campaign in We show that in January-February 2018, the extratropical trade wind and dry intrusion regimes are both associated with anticyclonic Rossby wave breaking (ARWB) over the North Atlantic. The longitude of the ARWB determines which of the two pathways is active. ARWB in the eastern North Atlantic close to the West African coast leads to the formation of a low-535 level easterly trade wind flow towards Barbados. The air parcels in this first regime typically start their descent eight days before arrival at 47°N, 35°W, and reach the boundary layer in front of the North African Coast (35°N, 20°W) in the context of an eastern North Atlantic anticyclone six days before arrival. These air parcels then cross the North Atlantic at low levels and their specific humidity increases substantially from about 6 to 10 g kg -1 during this passage. A deep layer establishes with easterly winds from the surface up to ~600 hPa above Barbados such as on 29 January 2018. 540 In contrast, ARWB over the central North Atlantic favours the occurrence of descending extratropical dry intrusions directly into the sub-cloud layer of Barbados. After exiting the upper-level extratropical jet stream in the right exit region near 41°N, 51°W, these extratropical air parcels start their three-stage equatorward descent six days prior to their arrival in Barbados. The first rapid descent stage is mainly adiabatic during which the largest part of the descent of 440 hPa (2d) -1 towards 29°N, 43°W is accomplished. During the slower second stage, the air parcels are located on top of tropical low-level clouds, they also 545 experience diabatic cooling compensating some of the adiabatic warming, as shown for the case of 2 February 2018, and they reach the sub-cloud layer in Barbados within four days. During this period, their specific humidity increases very strongly from about 3 to 10 g kg -1 . For extratropical dry intrusion air parcels to directly reach Barbados from the North, key ingredients are a quasi-stationary, central North Atlantic upper-level PV cutoff formed during the ARWB, potentially coupled to a subtropical surface cyclone underneath. The combined influence of the surface cyclone and the upper-level cutoff is to steer 550 the subsiding air of the dry intrusion towards the Caribbean. The two regions of preferred ARWB and descent from the extratropics leave a distinct imprint also in the campaign time-mean field of mid-tropospheric vertical motion over the North Atlantic.
The two extratropical transport regimes have a contrasting impact on the low-level cloud patterns, sub-cloud layer humidity and stable water isotope properties. In the extratropical trade wind regime, the sub-cloud layer air parcels take up 73% of their 555 final humidity from ocean evaporation and below-cloud evaporation of rainfall near the south-eastern edge of North Atlantic anticyclones on average 2028 km upstream of Barbados and within six days before arrival. Due to below-cloud interaction of rain droplets with ambient vapour, cold pool passages leave a distinct signature in the short-term variability of water vapour isotope signals with anomalies of up to +4 ‰ in d 2 H, +1 ‰ in d 18 O, and -2 ‰ in d, compared to the campaign mean. The important role of cold pools in the extratropical trade wind regime is reflected in their gravel-like signature on visible satellite 560 images. In the extratropical dry intrusion regime, the air parcels take up their moisture on average 1348 km upstream of Barbados. A smaller share of this moisture uptake (27%) occurs during the rapid descent stage due to convective and turbulent mixing of moisture into the descending airstream and the main uptake (63%) occurs during the slow descent stage when the dry air approaches and interacts with the cloud-topped marine boundary layer. The total column water vapour is smaller by 6 mm compared to the trade wind conditions, leading to large-scale areas of clear sky conditions (7% reduction of total cloud 565 cover compared to campaign mean conditions) due to the continuous supply of dry upper-level air into the tropical lower free troposphere. The penetration of the dry air parcels into the boundary layer lowers the humidity (-2% in RH and -0.6 g kg -1 in q compared to the campaign mean), thereby increasing the near-surface humidity gradient, and intensifying ocean evaporation (+0.5 mm d -1 ). The combination of these processes leads to a decrease in heavy isotopes by 6 ‰ in d 2 H, 1 ‰ in d 18 O and an increase by 6 ‰ in d, as was measured for a particularly strong extratropical dry intrusion event described in a detailed case 570 study.
The results of this study on the dynamics of atmospheric moisture uptake in the extratropical trade wind regime raise the question, whether the extratropical origin of these air parcels is relevant for their subsequent evolution. The thermodynamic properties of these air parcels are mainly determined within the trades, the dominant part of the moisture uptake occurs in the subtropical and tropical North Atlantic within 6 days prior to arrival and only a very small part of the remaining moisture in 575 Barbados actually originates from the extratropical upper troposphere. If the properties of the trade wind air masses are determined only after the entry of the air parcels into the trade wind region, their exact descent pathway prior to reaching the trades might be irrelevant. To more thoroughly address this question, a more detailed analysis of the cloud patterns associated with the trade wind regime is needed. It has yet to be investigated if air parcels with a descent history associated with an eastern North Atlantic ARWB event lead to different cloud patterns compared to tropical air parcels that descend over Africa as 580 outflows from the Intertropical Convergence Zone (ITCZ) deep convective systems. Deep convective outflow air parcels from the ITCZ might be moister and their subsidence rate, which is mainly controlled by radiative cooling (Salathé and Hartmann, 2000), might be smaller compared to the rapidly subsiding eastern ARWB air parcels. Such tropical origin air parcels would therefore probably take up less humidity within the trades. These reflections lead us to the overarching question about the size of the spatio-temporal window within which the atmospheric circulation is relevant for the observed variability in the cloud 585 cover and lower tropospheric humidity in the North Atlantic trades. Although the typical moisture residence time in the atmosphere is around 4-5 days in this region (Sodemann, 2020), the lifetime of weather systems within which the airstreams take up their humidity tends to be longer, in particular in the case of eastern subtropical anticyclones. Future, more detailed investigations will be needed to shed light on the relevance of the tropical vs. extratropical air parcel origin fuelling the winter North Atlantic trade wind layer. 590 Episodes with Saharan dust transport across the North Atlantic reaching the Caribbean are rarer in winter than in summer (Gläser et al., 2015) but might nevertheless play a role for the cloudiness over Barbados in January and February (Gutleben et al., 2019). Such events can be associated with eastern North Atlantic ARWB (Knippertz and Fink, 2006). During EUREC 4 A, multiple days with Saharan dust arriving above Barbados have been observed. Aircraft-based lidar (Chazette et al., 2020, Stevens et al., 2020b and stable water isotope measurements have been performed in this period, which provides the data 595 basis to investigate this aspect in more detail. In addition to the above mentioned open question on the relevance of the extratropical origin of the air in the trade wind regime, other caveats of this study are associated with the limited spatio-temporal coverage of the performed measurements and the subjectivity of the thresholds of subsidence and extratropical residence time involved in the flow regime definition. The results presented in this paper are based on a relatively short campaign including only near-surface measurements. The isoTrades 600 dataset already shows large variability in the transport patterns and local conditions at the BCO depending on the location of the ARWB over the North Atlantic. An extension with data from different surface and airborne platforms during EUREC 4 A in January-February 2020 as well as longer time series of isotopes measured at the BCO, including also other seasons, will be very valuable. Furthermore, here only the flow regime and thermodynamic history of sub-cloud layer air parcels has been studied. However, the vertical thermodynamic structure of the lower troposphere is also strongly influenced by differential 605 advection at higher levels. A future study on the impact of the different transport histories of air parcels arriving at various vertical levels in the troposphere is planned, using the comprehensive balloon sounding array compiled during EUREC 4 A (Stephan et al., 2020).
Finally, it remains open to what extent low-level cloud formation modulates the isotope signature of the sub-cloud layer vapour forming the cloud from convective updrafts and how much of the vapour isotope signature of the large-scale transport regime 610 is still reflected in precipitation reaching the surface. Previous studies have identified some influence of large-scale advection and moisture source signals in orographic clouds over land (e.g. Spiegel et al., 2012, Scholl et al., 2014. Measurements of the short-term isotope variability in low-level clouds and shallow convective precipitation are however very scarce. https://doi.org/10.5194/wcd-2020-51 Preprint. Discussion started: 12 October 2020 c Author(s) 2020. CC BY 4.0 License.
In summary, this paper provides new insight into the role of the large-scale circulation for the environment in which trade wind cumulus clouds form and thereby contributes to the ongoing climate research effort to elucidate the role of the coupling 615 between clouds and the circulation. Furthermore, this study underlines the importance of extratropical dynamics for the humidity and the isotope signature of the tropical trade wind region in winter.
Data availability. The ERA5 reanalyses used in this study can be accessed from the ECMWF website (https://www.ecmwf.int/en/forecasts/datasets/reanalysis-datasets/era5). The imagery of the Earth Observing System Data and 620 Information System (EOSDIS) can be obtained from the Worldview Snapshots application (https://wvs.earthdata.nasa.gov).
The isoTrades dataset is published in the ETH data collection, available online here: 10.3929/ethz-b-000439434. with sub-cloud layer air originating from the extratropics but having experienced limited subsidence in the four days prior to arrival in Barbados. The threshold value for the subsidence rate was chosen close to the ERA5 climatological value of free tropospheric subsidence (at 500 hPa) in the region of Barbados. This corresponds approximately to the vertical pressure 650 velocity that is expected from subsidence balancing radiative cooling (~25 hPa (d) -1 , with ~1 K(d) -1 longwave radiative cooling, see Holton and Hakim, 2013). This choice is based on the hypothesis that air parcels penetrating into the sub-cloud layer due to large-scale advection with a subsidence rate that exceeds its climatological free tropospheric value are likely to have a considerable impact on the sub-cloud layer properties. The time window for defining the subsidence threshold was chosen based on the typical lifetime of water vapour in the subtropics and winter tropics of four days (Sodemann, 2020). The 655 subsidence rate within four days thus seems to be an adequate choice to investigate the impact of the different flow regimes on the western North Atlantic trade wind water cycle.
Note that in comparison to the definition of dry intrusions in Raveh-Rubin (2017), with a slantwise subsidence rate of > 400 hPa (2 d) -1 , the extratropical dry intrusions defined here subside much more slowly. For our purpose, the important point is that our air parcels classified as extratropical dry intrusions show a pronounced subsidence within the time window that is 660 relevant for the water vapour lifetime in this region. In addition, for comparing with the climatology of Raveh-Rubin (2017), the percentage of hourly time steps with at least one backward trajectory from the BCO with a descent of > 400 hPa (2 d) -1 is computed for the two extratropical flow regimes.                   Table 1: Occurrence frequency of the three flow regimes defined in this study based on a classification of all hourly time steps in January-February in the period 2009 to 2018. The tropical flow regime is defined with " < 1 day. This regime also includes air 1015 parcels that descend from the subtropics. The share of purely tropical air parcels with " . < 1 day is shown in parentheses. The extratropical trade wind regime is defined with with " ≥ 1 day and a limited subsidence in the last 4 days prior to arrival in Barbados (< 100 hPa (4d) -1 ); and extratropical dry intrusions with a residence time of at least one day in the extratropics and enhanced subsidence (³ 100 hPa (4d) -1 ).   D p4d = p(tBCO) -p(tBCO -4 days) is the slantwise subsidence rate, max(Dp2d ) stands for the maximum subsidence within a 2 day period along each of the 10-day back-trajectories (see Section 2.1 and Appendix A). l is the 1025 latitude, f the longitude, RH the relative humidity, q the specific humidity, fmsd weighted mean moisture source longitude and Lmsd the distance to the centre of mass of the moisture source region. Mean conditions (±their standard deviations) for the isoTrades period are indicated in the table based on all the trajectories arriving in the sub-cloud layer at the BCO for the two regimes. Table 3: Campaign mean local conditions at the BCO as well as conditions associated with the extratropical trade wind and dry intrusion flow regimes, respectively, based on measurements (first 9 variables) and ERA5 data. The p-value of a Wilcoxon rank-sum test for the significance of the difference between the two regime's distributions is <0.01 except for T. q is the specific humidity, T the air temperature, P the rain rate, ncp the number of cold pools within a 12 hour time period centred at the respective hourly time steps, U is the wind speed, D the wind direction and RH the relative humidity measured at BCO. The evaporation rate (E), the total 1035 column water (TCW), the total cloud cover (TCC) and the lower tropospheric stability (LTS=q 700hPa-q 1000hPa) are extracted from ERA5 and averaged over a 2°x2° box around BCO (13 -15°N and 58 -60°W) to provide regionally averaged values of these variables. Note that the results do not change substantially if the variables are interpolated to the position of the BCO.