Lagrangian matches between observations from aircraft, lidar and radar in an orographic warm conveyor belt

Warm conveyor belts (WCBs) are important airstreams in extratropical cyclones, often leading to the formation of intense precipitation and the amplification of upper-level ridges. This study presents a case study that involves aircraft, lidar and radar observations in a WCB ascending from western Europe towards the Baltic Sea during the field experiments HyMeX and T-NAWDEX-Falcon in October 2012. Trajectories were used to link different observations along the WCB, that is to establish so-called Lagrangian matches be5 tween observations. To this aim, wind fields of the ECMWF ensemble data assimilation system were used, which allowed for a probabilistic quantification of the WCB occurrence and the Lagrangian matches. Despite severe air traffic limitations for performing research flights over Europe, the DLR Falcon successfully sampled WCB air masses during different phases of the WCB ascent. The WCB trajectories revealed measurements in two distinct WCB branches: one branch ascended from the eastern North Atlantic over southwestern France, while the other had its inflow in the western Mediterranean. Both branches passed 10 across the Alps, and for both branches, Lagrangian matches coincidentally occurred between lidar water vapour measurements in the inflow of the WCB south of the Alps, radar measurements during the ascent at the Alps, and in situ aircraft measurements by Falcon in the WCB outflow north of the Alps. An airborne release experiment with an inert tracer could confirm the long pathway of the WCB from the inflow in the Mediterranean boundary layer to the outflow in the upper troposphere near the Baltic Sea several hours later. 15 The comparison of observations and ensemble analyses reveals a moist bias in the analyses in parts of the WCB inflow but a good agreement of cloud water species in the WCB during ascent. In between these two observations, a precipitation radar measured strongly precipitating WCB air located directly above the melting layer while ascending at the southern slopes of the Alps. The trajectories illustrate the complexity of a continental and orographically influenced WCB, which leads to (i) WCB moisture sources from both the Atlantic and Mediterranean, (ii) different pathways of WCB ascent affected by orography, and 20 1 https://doi.org/10.5194/acp-2020-1019 Preprint. Discussion started: 28 October 2020 c © Author(s) 2020. CC BY 4.0 License.

able that the detailed properties of a WCB strongly depend on the humidity in its inflow. For numerical weather prediction, the correct representation of humidity particularly in strongly cloudy situations as in WCBs is still one of the main challenges 90 (Rodwell et al., 2018). However, most of the climatological investigations (e.g. Eckhardt et al., 2004;Madonna et al., 2014b) and case studies of WCBs (e.g. Joos and Wernli, 2012;Martínez-Alvarado et al., 2016) were based on reanalyses or model simulations only. Comparatively few studies used observations to evaluate the humidity in WCBs. Schäfler et al. (2011) investigated airborne lidar measurements of humidity in the low-level inflow of a WCB over Spain and found an overestimation of humidity in ECWMF analyses. A similar result was found by Schäfler and Harnisch (2015) for the inflow of a marine WCB 95 over the eastern North Pacific. Their sensitivity experiments with the ECMWF forecast model revealed that with the corrected low-level moisture in the initial conditions, the WCB outflow would have occurred at a lower potential temperature level and produced a less developed upper-level ridge. The humidity in the WCB inflow, however, is not only determined by boundary layer ventilation (Boutle et al., 2011;Pfahl et al., 2014). It can also be affected by a recycling of moisture within the WCB, which occurs, e.g. when raindrops from an elevated layer of the WCB fall into a sub-saturated lower layer of the WCB inflow 100 where they evaporate (Crezee et al., 2017;Attinger et al., 2019;Spreitzer, 2020).
Most observational studies of WCBs so far, e.g. based on surface radar measurements (Browning, 1971) or more recently on aircraft and satellite data (Oertel et al., 2019;Binder et al., 2020), provide limited information about the variability of the WCB characteristics along the ascent, which typically covers spatial and temporal dimensions of > 1000 km and 1 day, respectively.
To better address the Lagrangian nature of these airstreams, a few pioneering field campaign studies attempted to follow the 105 pathway of a WCB with an aircraft and to measure the same WCB air parcels multiple times (e.g. Stohl et al., 2003;Methven et al., 2006). In principle, such Lagrangian matches enable investigating the material evolution of thermodynamic variables along a WCB. A major challenge of such experiments is the fact that the planning of Lagrangian matches with aircraft must rely on air parcel trajectories using forecast wind fields, which are inherently uncertain. To cope with this uncertainty, the planning of Lagrangian matches is best done with data from ensemble forecasts (Schäfler et al., 2014;Schäfler et al., 2018). An 110 interesting observational approach to identify Lagrangian matches is the use of a physical tracer that is measured at consecutive times to experimentally corroborate the pathway of air parcels. An experiment in 2004 described in Methven et al. (2006) aimed at realising Lagrangian matches between airborne measurements in the free troposphere to study intercontinental transport of pollutants. One case of the campaign involved a WCB, for which they used the natural occurrence of their physical tracer to mark air parcels. The approach in this study is, for the first time, to investigate transport along a WCB by the release and WCBs and they contribute up to 60% to extreme precipitation events (Pfahl et al., 2014). In a case study, Buzzi et al. (1998) investigated the flood in the Piemont region south of the Alps in November 1994, which was associated with a WCB. Using nu-125 merical experiments with flattened orography, they could show that the WCB-related precipitation was distinctly enhanced by orography. But so far, no study investigated how the pathway of WCB trajectories is affected by the interaction with orography.
In October 2012, a team from the German Aerospace Center (DLR) in Oberpfaffenhofen and ETH Zurich organised a small aircraft research campaign, devoted to obtaining in situ measurements of moisture and thermodynamic parameters in different phases of WCBs over Europe (Schäfler et al., 2014). This campaign was termed T-NAWDEX-Falcon and occurred in parallel 130 to the comprehensive HyMeX Special Observation Period 1 (Ducrocq et al., 2014). The WCB presented in this study occurred during the campaign's IOP2, and it ascended from the western Mediterranean across the Alps towards the Baltic Sea. The WCB was successfully sampled by two Falcon flights. The analysis of the airborne observations benefits from additional surface lidar observations made in the framework of HyMeX and from operational radar observations by MeteoSwiss.
In addition, a tracer experiment with the physical release of a passive tracer by a small aircraft in the inflow of the WCB 135 was conducted as part of the T-NAWDEX-Falcon IOP2. The main objective was to "label" WCB air in the inflow and then later "catch" the same air during its ascent by the Falcon aircraft. The use of a passive tracer enables, in principle, a validation whether the air observed at a later time by the Falcon actually had its origin in the labelled WCB inflow region. However, such an experiment can only be successful: (i) if a tracer gas is released that otherwise does not exist in the atmosphere or has a very low atmospheric background concentration like PFCs, such that any observation of the tracer can be uniquely associated 140 with the release experiment; (ii) if a suitable and highly sensitive airborne sampling and analysis technique is available, which allows measuring potentially low concentrations after long-range transport and dilution; (iii) if the forecast is exact enough to reproduce a fairly realistic picture of the actual flow situation that serves as the basis for the flight planning; and (iv) if the flight planning method is appropriate such that the tracer is indeed released in the WCB inflow and that the subsequent Falcon flights targeting the ascending WCB air have a chance of sampling the tracer. In this study we report to what degree we were 145 successful with this challenging experiment.
As shown in detail below, the campaign successfully collected valuable observations in the WCB and benefited from overall good ECMWF forecasts and from the sophisticated multi-stage flight planning procedure (Schäfler et al., 2014). This set of observations will enable us to address the following specific questions related to this orographic WCB case study: 1. How can ensemble analyses be used to provide probabilistic information about the location of WCBs and about La-150 grangian matches between observations? 2. How well does the humidity and cloud structure of the WCB in the ECMWF analyses agree with observations from aircraft, lidar, and radar?
3. How closely does the transport in WCBs as given by trajectory calculations with ECMWF analyses correspond to the dispersion of an actually emitted passive tracer? The manuscript will continue with introducing the measurements and the analysis products, as well as the procedure to identify Lagrangian matches in section 2. The synoptic situation is described in section 3 before first the results from airborne obser-160 vations are discussed in subsection 4.1. Lagrangian matches with further observations are described in subsection 4.2, and the tracer experiment in subsection 4.3. Conclusions, including (partial) answers to the questions posed above, are given in section 5.

Data and Method
This study uses observations from in situ aircraft measurements, from ground-based lidar and radar, respectively, and from 165 a tracer release experiment. The observational datasets are briefly introduced in the following subsections. Then a special diagnostic method, based on ECMWF ensemble analysis data, is introduced, that serves to identify WCBs and Lagrangian matches between measurements.

T-NAWDEX-Falcon humidity measurements
For the T-NAWDEX-Falcon campaign in October 2012 the German DLR Falcon aircraft was operated from the base at DLR 170 in Oberpfaffenhofen near Munich (Schäfler et al., 2014). The Falcon flights on 15 October 2012 were conducted from 7:34 to 10:52 UTC for flight IOP2b and from 13:04 to 16:03 UTC for flight IOP2c. In this study in situ measurements of specific humidity are used that were taken from the basic instrumentation of the aircraft and from the tuneable diode laser system WARAN Kaufmann et al., 2018). The Falcon basic humidity measurements for water vapour result from a composite of three instruments, where emphasis is placed on the lyman-alpha absorption instrument 1 . For flight IOP2b, 175 WARAN was installed with a forward facing inlet so that it collected water of all phases and hence resulted in observations of total water. For flight IOP2c, the inlet was reversed such that no total water measurements were possible during this flight Kaufmann et al., 2016). For IOP2b, the cloud water content was calculated assuming saturation with respect to liquid water for T > −30 • C and ice for T ≤ −30 • C, respectively. Particle enhancement at the inlet is corrected assuming a mean particle radius of 20 µm (Krämer and Afchine, 2004). The sampling efficiency of much larger particles like raindrops 180 or snow cannot be quantified directly and thus imposes an additional uncertainty on the measured cloud water content. The two-second in situ measurements of WARAN are smoothed with a 5 min running mean, which corresponds to about the time the aircraft needs to intersect a horizontal grid box of 0.5 • (about 50 km) in the model. The same is applied to the one-second water vapour observations. Since the measurements are taken as water vapour mixing ratio and total water mixing ratio per volume they are converted to specific humidity Q v and cloud water content Q c , respectively, for comparison with model data. profile as well as a 1-hourly time running mean to account for comparison with the coarser resolved model data. Considering an integration time of 1 hour and a vertical resolution of 300 m, the statistical uncertainty affecting daytime water vapour mixing ratio measurements is found to be smaller than 0.5 g kg −1 (or 10%) up to 2.5 km and smaller than 3 g kg −1 (or 40%) up to 4 km. Nighttime performance is characterised by much smaller uncertainties, with a random error affecting vapour mixing ratio measurements not exceeding 0.01 g kg −1 (or 0.5%) at 4 km and of 0.035 g kg −1 (or 7%) at 10 km. The procedure applied

Monte Lema radar 205
The Monte Lema radar is part of the Swiss network of C-band (5.5 GHz) Doppler weather radars operated by MeteoSwiss and it is located on the southern slopes of the Alps at 8.83 • E, 46.04 • N, and 1626 m altitude. Monte Lema was upgraded in 2011 to dual polarisation which, together with other technical enhancements, provided increased data quality (Germann et al., 2015;Gabella et al., 2017). The volume scan geometry comprises 20 elevations from -0.2 • to 40 • with a maximum range of about 18 km vertically and 246 km horizontally, while the interleaved scan strategy provides every 5 min a new full volume scan.

210
The methods for reading and processing of the raw polar data from Monte Lema radar can be read in Figueras i Ventura et al. (2020).

Tracer experiment
A tracer experiment was performed using the perfluorocarbon tracer system PERTRAS. This instrument was newly designed for studying Lagrangian long-range transport and related dispersion of chemical species, e.g. in the Asian Monsoon system 215 (Ren et al., 2015). Perfluoromethylcyclopentane (PMCP C6F12) has very low background concentrations in the atmosphere of 6-7 ppqv in Europe (Watson et al., 2007) and is thus very well suited for a tracer experiment. For the T-NAWDEX-Falcon campaign, the release unit of PERTRAS for the inert tracer gas was installed on board of the light aircraft Partenavia P68 operated by enviscope 2 . The sampling unit of PERTRAS, on the other hand, was installed on board the Falcon. The intention to separate the release unit from the sampling unit was to exclude possible contamination throughout the re-sampling of the 220 tracer and to meet the logistic demands given by the relatively fast transport of air masses between the WCB inflow and outflow stage. The Falcon installation allowed to collect probes of air at an interval of about 5 min yielding a spatial resolution of 50 km.
Such an experiment could be conducted only once during the T-NAWDEX-Falcon campaign and had to be carefully planned based on the operational ECMWF forecasts (Schäfler et al., 2014). During IOP 2 on 14 October 2012 suitable conditions were given, the Partenavia was therefore transferred to Southern France and released in the morning hours between 09:09 and 225 09:39 UTC 6.2 l of the tracer in the target region. The liquid PMCP tracer was dispersed using a spray nozzle connected to the tracer reservoir and mounted outside the window of the Partenavia. Thereby, PMCP droplets with diameters less than 20 mm were released which evaporated readily. A day later, on 15 October, two Falcon flights (24 and 30 h after the release) above Germany were devoted to resample the tracer in the ascending air masses. For the tracer sampling, 64 adsorption tubes were used where the PMCP molecules were trapped and concentrated. The samples were analysed after the flights in the laboratory 230 using a gas-chromatic method described in Ren et al. (2014). Unfortunately, a technical issue concerning the manual time adjustment of the device likely occurred during the first sampling flight IOP2b.

Model data
For the calculation of WCB trajectories, we use the Ensemble of Data Assimilations (EDA) dataset from the European Centre for Medium-Range Weather Forecasts (ECMWF). The EDA represents the best possible estimate of the state of the atmosphere 235 considering uncertainties associated with observations and the ECMWF data assimilation system. It became operational in 2010, and in 2012, it became available as a set of 11 analyses that primarily serves to improve the initial conditions of the ensemble prediction system (Isaksen et al., 2010), with one control analysis (CTL) and 10 perturbed members. The slightly differing analyses are obtained by perturbing atmospheric observations, sea surface temperature fields and model physics in the 4D-Var data assimilation cycle. The original resolution of the EDA of T399L91 in 2012 is interpolated to a 0.5 • horizontal 240 grid on the original 91 model levels. In addition to the 6-hourly standard EDA times, the intermediate products at 03, 09, 15, and 21 UTC are also used to increase the time resolution.

Trajectories and WCBs probabilities
Kinematic offline trajectories are calculated using the Lagrangian analysis tool LAGRANTO (Wernli and Davies, 1997;Sprenger and Wernli, 2015). Here, we use trajectories to detect WCBs as well as matches between measurements in each 245 member of the EDA, and they are therefore a central part of this study.
To know where WCB air masses ascend during the IOP, trajectories are started from the grid points between 1025 and 700 hPa with vertical steps of 25 hPa between 1000 and 100 hPa. They are initialised every 3 h over several days and calculated 48 h forward in time with trajectory positions saved every 1 h. To identify WCB trajectories, only the ones are selected that ascend at least 600 hPa in 48 h (Madonna et al., 2014b). WCB probabilities are calculated using the WCB trajectories calculated for all 11 EDA members based on the method that has been adopted from Schäfler et al. (2014) and Rautenhaus et al. (2015).
More specifically, the trajectories are disassembled into individual air parcel positions and then, for a given time, sorted into grid boxes with a horizontal dimension of 0.5 • in the horizontal and 25 hPa in the vertical. For each grid box, the WCB probability (in %) then corresponds to the relative number of EDA members that have at least one WCB trajectory in the box. The method results in a time sequence of Eulerian three-dimensional fields of WCB probabilities.

255
EDA-based trajectories are also used to identify Lagrangian matches between aircraft measurements at one time and any type of measurements at an earlier or later time. To this end, forward and backward trajectories are calculated in each member of the EDA, starting every minute from the respective flight route (coloured example trajectories in Fig. 1). The time range of the trajectories is backward to 12 UTC 13 October and forward to 12 UTC 16 October to cover all measurements and the entire WCB ascent. In this set of trajectories, those that fulfil the ascent criterion described above are labelled as WCB trajectories.

260
Similar to the method to obtain WCB probabilities, the positions of the EDA trajectories at a time of interest are assigned to grid boxes (grey boxes in Fig. 1). Trajectory positions are hence available as a time sequence of Eulerian three-dimensional probability fields (number of contributing EDA members; boxes with different grey shadings in Fig. 1). Matches between measurements along the aircraft track from which the trajectories were started and any other measurements occur when the trajectory probability (calculated with trajectories that started from the flight track) in the grid box that corresponds to the time 265 and location of the other measurement is larger than zero (the observing instrument in Fig. 1, which, in the example shown, diagnoses differing matching probabilities with altitude). Hence, this procedure to identify matches accounts for the exact time (of the measurement) but with a tolerance in space, represented by the 0.5°grid box around the exact trajectory position. If the trajectory probability is high, then the Lagrangian match between the measurements can be regarded as more 'certain'.
If a matching trajectory also fulfills the WCB criterion, then we refer to this as a WCB match, i.e., then we know that two 270 measurements at different times and locations were sampling the same air parcel and that this air parcel was actually part of a WCB. These WCB matches are of particular interest for our study.

Synoptic situation
The two-day IOP2 of the T-NAWDEX-Falcon campaign was planned to take measurements along a WCB ascending from the western Mediterranean over the Alps towards the Baltic Sea. The WCB was induced by a low pressure system that originated in 275 the central North Atlantic before 13 October 2012. WCB air masses start from the warm sector already during this earlier phase of the cyclone while it moves eastward, steered at the southern flank of a complex upper-level PV cutoff (Fig. 2a, b, the cyclone is marked by a red 'L' in a). Downstream of the cyclone a zonal wind direction prevails at low levels over France at 18 UTC 13 October (Fig. 2a). North of the Pyrenees and further east over the Mediterranean, the low-level flow is orographically influenced and arrives with increased wind speed and reduced humidity at the northwestern Mediterranean, as reflected by low 13 October (see westernmost red marker for the lidar in Fig. 2). Within the westerly flow, WCB trajectories originating over the Atlantic are embedded that start to ascend prior to passing the lidar (Fig. 2b). At the same time, another WCB branch north of the one from the Atlantic starts to rise over France. These two northern WCB branches with inflow from the Atlantic are 285 directly steered by the surface cyclone that enters France from the Bay of Biscay at 06 UTC 14 October (not shown). The cloud band related to this ascending part of the WCB is visible east and north of the cyclone centre few hours later in the satellite image ( Fig. 2d and black marker at 12 UTC 14 October in Fig. 2b).
The cold front of the cyclone is located over the northwestern part of the Iberian peninsula and western France at 12 UTC 14 October, when the wind direction over the northwestern Mediterranean turned to a moist southwesterly flow onshore (Fig.   290 2c). In this flow, ahead of the surface cold front, another WCB branch is discernible at low levels ( Fig. 2b) while this region is covered by heavy convective clouds in the satellite image (Fig. 2b,d, see also Duffourg et al., 2018 about the associated heavy convective precipitation event). The inflow of this Mediterranean WCB branch passes the HyMeX lidar just southeast (Fig.   2b). On a smaller scale, the strong westerly wind ahead of the cold front initiates a secondary lee cyclone east of the Maritime Alps in the Italian Piemont region (red 'L 2 ' for the mature lee cyclone in Fig. 3b).

295
The preceding upper-level PV cutoff forms a streamer-like structure during the development of the main cyclone 'L' and subsequently extends southward over the western Mediterranean (Fig. 3b). The low-level lee cyclone 'L 2 ' intensifies due to the interaction with the upper-level PV streamer. The Mediterranean WCB branch shown in Fig. 2b moves northeastward ahead of the PV streamer. Between 00 and 06 UTC 15 October, the WCB branch passes cyclone 'L 2 ' (Fig. 3b) and impinges upon the Alps where it is forced to ascend. There the WCB air stream crosses the Monte Lema radar in the early phase of the ascent 300 (red marker in Fig. 3b). Within the next 6 hours, the Mediterranean WCB air parcels reach the middle troposphere and proceed with a moderate ascent rate over Germany towards the Baltic Sea where it approaches the Atlantic WCB branch. Along with the WCB, the cold front of the main cyclone 'L' proceeds into Central Europe. Together they form the stratiform and scattered In summary, Falcon flight IOP2b intersected the WCB with a high probability in the mid-troposphere during the ascent and at its outflow level. However, two horizontal legs (before 08:00 and after 10:30 UTC) just missed the center of the WCB as they were slightly too high and too low, respectively. In all clouds within or near the WCB the observed cloud condensate agrees 360 well with the EDA values. According to the EDA, the lower part of the WCB contains supercooled liquid water at temperatures between −5 and −10 • C. As found by radar observations in Gehring et al. (2020), the formation of supercooled liquid water in the phase of strongest WCB ascent facilitates aggregation and riming, and provides ideal conditions for rapid precipitation growth.

365
The flight in the afternoon of 15 October again went towards the Baltic Sea (see blue line in Fig. 3a) and crossed the cold front with WCB probabilities up to 40% at around 550 hPa (Fig. 5a). As for the previous flight, the ascent across the WCB leads to peaks of first SWC (at the lower edge of the WCB) and then LWC and IWC (within the WCB) according to the EDA (Fig.   5b). Here, water vapour is underestimated in EDA compared with the measurements and shows a local minimum where snow is likely to sublimate below the WCB as already seen in the flight before (Fig. 5d,b).

370
The subsequent flight leg at 350 hPa is located just above an extended region with high WCB probabilities. The observed air ascended by 300 hPa within 48 h according to Fig. 5c. The measured specific humidity in EDA is again too low (Fig. 5d).
Given the moderate ascent of this air and its location just above the WCB, it is plausible that this cloud layer corresponds to a so-called in situ cirrus, which often tops the liquid-origin cirrus produced by the strongly ascending WCB .
Shortly before 15:00 UTC, the aircraft descended to 500 hPa and, according to the EDA, intersected a region with WCB 375 probabilities up to 100% (Fig. 5a). Along this flight leg between 15:00 and 15:25 UTC, SWC and IWC are elevated in EDA (Fig. 5b). The observed water vapour is still higher than in the EDA, but with a smaller deviation than before. The bump of the 0 • C isotherm below the aircraft in Fig. 5a indicates that the aircraft crossed the surface cold front and entered the warm sector at 15:20 UTC before crossing the cold front again in the reverse direction at 15:35 UTC.
Taking both flights (IOP2b and 2c) together, the aircraft sampled WCB air with high probability during the ascent and 380 outflow of the WCB on several legs of the flights. Whenever high WCB probability was intersected, specific humidity and cloud condensate in EDA are increased. The magnitude and structure of cloud condensate is well represented in the EDA compared with the observations. Water vapour is often underestimated, in particular below increased WCB probability where precipitation is likely to sublimate or evaporate. Short periods with overestimated water vapour in EDA occurred near regions with large gradients of WCB probabilities, i.e. where the aircraft encountered regions with lower probability of WCB occurrence in the 385 EDA and where, in reality, most likely large gradients in humidity occurred.

Lagrangian matches of the aircraft-probed WCB air masses with ground-based measurements
In this section, the pathway of the aircraft-observed WCB air, i.e., of WCB trajectories that intersected the flight route, is considered in more detail. In addition, where possible, ground-based water vapour lidar and precipitation radar measurements south of the Alps are considered, which, unplanned, sampled some of the WCB air masses during the inflow and ascent phase.  This confirms that we successfully sampled a tracer, which was released in the boundary layer over the Mediterranean Sea, 535 30 h later and 1000 km further north at an altitude of 500 hPa. This long-range transport occurred with the Mediterranean WCB air mass close to the evolution of the example trajectory T2 described in subsection 4.2.2, which realized triple Lagrangian matches with observations. Falcon field experiment that happened in parallel to the HyMeX SOP1 in autumn 2012. The key object of this study is an orographically influenced WCB ascending with two inflow branches across the Alps towards the Baltic Sea. The WCB shows a rather complex flow behaviour in the vicinity of the Alps. There is low-level inflow of air into the WCB from both the North Atlantic and the Mediterranean, where the former mainly flows around and the latter rises above the Alps, with some of the lifting being clearly due to mechanically forced orographic ascent at the southern Alpine slopes. This is so far the most detailed 545 WCB case study near complex topography, illustrating that dynamic and orographic lifting can interact in the formation and evolution of a WCB. Assessing the generality of this finding for other WCBs near mountains is left for further research. Two inflow branches with moisture sources in the North Atlantic and the Mediterranean, respectively, were also found in studies of heavy precipitation events on the Alpine south side (Winschall et al., 2012), which emphasises the potential of WCBs in contributing to severe weather in this region (Buzzi et al., 1998). North of the Alps, where the two branches of the WCB further 550 ascend towards the Baltic Sea, two research flights intersected both branches, one at mid-level during ascent and the other later in the outflow.
In order to clearly attribute certain periods of the in situ aircraft observations to the WCB, we used a sophisticated trajectory approach. Kinematic air parcel trajectories are subject to uncertainties in the wind field and to numerical errors in the trajectory computation. To address these uncertainties and to provide more confidence about the identification of the WCB, we determined 555 the WCB probabilistically by calculating trajectories in each of the 11 members of the ECMWF's ensemble data analysis (EDA) -an operational dataset that quantifies uncertainties in the representation of the state of the atmosphere.
Observational evidence for the long-range transport in the WCB was obtained from an airborne tracer release experiment.
The experiment involved the release of an inert tracer gas in the WCB inflow region in the western Mediterranean by a small aircraft, and the sampling of the tracer by the Falcon aircraft 30 hours later and 1000 km further north over Germany.

560
The setup of this experiment was unique as previous releases of the same passive tracer were conducted in smaller-scale convective flow systems and included a much smaller spatial and temporal separation of the release and sampling. Before we summarise the results we need to consider that such an experiment requires (i) a tracer gas with very low natural atmospheric concentrations; (ii) if a suitable and highly sensitive airborne sampling and analysis technique; (iii) accurate forecasts for reliable flight planning; and (iv) a sophisticated flight planning method. We can report that increased values of the collected 565 tracer gas coincided with sections of the Falcon flights where trajectory calculations suggested increased probabilities for the tracer air mass. However, the locations of the sampled tracer were characterized by rather low probabilities (i.e. they were not confirmed by many EDA members) and the tracer sampling just occurred at the edge of the WCB. Since we regard our technical setup for the experiment and the flight planning as sophisticated (Schäfler et al., 2014), we attribute the relatively low tracer probabilities at the edge of the WCB -instead of the targeted high tracer probability in the core of the WCB -to 570 a non-perfect forecast as mentioned in (iii) above, to obviously substantial uncertainties in the EDA, and finally to the limited area where the tracer gas was released into the WCB. The fact, however, that the observed transport of the tracer qualitatively agrees well with the WCB pathway as indicated by EDA trajectories provides highly valuable evidence that WCBs, identified on the basis of trajectory calculations in many studies in the last 25 years, are meaningful Lagrangian flow features in the atmosphere.

575
WCB probabilities were also used to link the Falcon in situ measurements in space and time with earlier observations in the WCB, which we refer to as Lagrangian matches. This approach revealed that part of the WCB inflow was already observed at low levels by a HyMex lidar near Montpellier, and later at the southern slope of the Alps as the WCB passed a MeteoSwiss radar. These Lagrangian matches between in situ airborne and ground-based observations occurred within the WCB air as confirmed by the tracer. The Lagrangian matches were not planned beforehand, but were beneficially included later during 580 the analysis of this WCB. EDAs are also used for evaluating humidity and cloud properties with the measurements and to characterise the evolution of the moisture in the model. The key findings from this evaluation can be summarised as follows: -Lidar humidity measurements sampled both branches of the WCB inflow from the Atlantic and Mediterranean, respectively, during a time period of almost one day. Irrespective of the changing wind direction due to the approaching cyclone and the local wind circulation (Di Girolamo et al., 2009), the observations reveal an almost continuously too 585 moist boundary layer in the ECMWF analyses. This highlights the difficulty to correctly represent moisture in this complicated flow situation of the WCB inflow, which was influenced by multi-scale processes. Our results also agree with earlier observations by Schäfler et al. (2011) and Schäfler and Harnisch (2015) who also found a too humid boundary layer in WCB inflows in analysis data.
-While the North Atlantic WCB branch already started to ascend, the Mediterranean branch moved further east taking up 590 moisture along the coast before impinging upon the Alps. There, the MeteoSwiss radar observed intense surface rainfall and mid-level snow associated with the WCB. The particular WCB air parcels that were previously observed by the lidar and subsequently sampled by the Falcon are part of this precipitating cloud close to the melting layer. A few hours earlier, the radar observed WCB air from both the Mediterranean and the Atlantic branches and the two branches were vertically separated. Comparing the moisture and cloud formation along the WCB trajectories, the dryer Mediterranean 595 branch almost instantaneously formed dense high-reaching clouds when it was forced to ascend at the Alps, while the clouds of the Atlantic branch developed more gradually and were less deep. The radar-observed WCB cloud in our case is rather stratiform and does not exhibit characteristics of embedded convection, as observed in other (non-orographic) WCB cases.
-Later during the further ascent of the WCB north of the Alps, the Falcon measured in situ water vapour and condensate 600 in the mid and upper troposphere, including Lagrangian matches with the earlier observations. Several successful WCB intersections showed that water vapour in the cloudy WCB is often lower in the EDA compared with the measurements.
In regions below the WCB, where according to the EDA snow falls out of the WCB and presumably sublimates, water vapour seems to be distinctly underestimated in the model. Cloud condensate in the WCB was found to be of similar magnitude in the EDA compared to observed cloud condensate in the mid-troposphere (mixed-phase cloud) and slightly 605 underestimated further above (ice cloud). To our knowledge, these are the first reported in situ measurements of cloud condensate in a WCB, repeated two years later as part of the ML-Cirrus aircraft campaign (Voigt et al., 2017).
Apart from this appealing results, it is important to emphasise the limitations of this campaign-driven study. The limitations of the study are related to the fact that even with a sophisticated flight planning, perfect flight routes in terms of maximising WCB encounters and Lagrangian matches have not been possible for several reasons. As an effect, the number of Lagrangian 610 matches between flights and the duration of in situ observations within the WCB is limited. We think that this is the prize to pay for such a challenging observational study that essentially relies on flight planning, which in turn was based on nonperfect forecasts. However, we think that this multi-faceted study reveals a range of interesting and, for some of them, unique features of WCBs. This study presented the first case of a WCB with measurements following the Lagrangian ascent and describing the moisture and cloud evolution along the flow. The limited number of observations, however, only allowed to 615 analyse a limited set of processes contributing to the total moisture budget of this WCB. We expect that repeating this kind of Lagrangian measurements in a WCB with a more complete instrumental package will be rewarding. In addition, the theme of orographic WCBs definitively deserves further attention as an orographcally induced change in updraft velocities may modify the dynamics and the water partitioning within the WCB.