The lacustrine-water vapor isotope inventory experiment L-WAIVE

Abstract. In order to gain understanding on the vertical structure of atmospheric water vapour above mountain lakes and to assess the respective influence of evaporation and advection processes, the L-WAIVE (Lacustrine-Water vApor Isotope inVentory Experiment) field campaign was conducted in the Annecy valley in the French Alps in June 2019. This campaign was based on a synergy between a suite of ground-based, boat-borne, and airborne measuring platforms implemented to characterise the thermodynamic and isotopic state above the lake environment using both in-situ and remote sensing instruments. Two ultra-light aircrafts (ULA), one equipped with a Rayleigh-Mie lidar, solar fluxmeters and an optical counter, and one equipped with a Cavity Ring-down Spectrometer (CRDS) and an in-cloud liquid water collector, were deployed to characterize the vertical distribution of the main stable water vapour isotopes (H216O, H218O and H2H16O), and their potential interactions with clouds and aerosols. ULA flight patterns were repeated several times per day to capture the diurnal evolution as well as variability associated with different weather events. ULA flights were anchored to continuous water vapour and wind profiling of the lower troposphere performed by two dedicated ground-based lidars. Additional flights have been conducted to map the spatial variability of the water vapour isotope composition regarding the lake and surrounding topography. Throughout the campaign, ship-borne lake temperature profiles as well as liquid water samples at the air-water interface and at 2 m depth were made, supplemented on one occasion by atmospheric water vapour isotope measurements from the ship. The campaign period included a variety of weather events leading to contrasting humidity and cloud conditions, slope wind regimes and aerosol contents in the valley. The water vapor mixing ratio values in the valley atmospheric boundary layer were found to range from 3–4 g kg−1 to more than 10 g kg−1 and to be strongly influenced by the subsidence of higher altitude air masses as well as slope winds. A significant variability of the isotopic composition was observed within the first 3 km above ground level. The influence of the lake evaporation was mainly detected in the first 500 m of the atmosphere. Well-mixed conditions prevailed in the lower free troposphere, mainly above the mean altitude of the mountain tops surrounding the lake.



Introduction
The vertical structure of the water vapor field in the lower troposphere is only sparsely documented in mountainous regions and particularly above Alpine mountain lakes (AMLs). This is in part due to the complexity and fast-evolving nature of the low-level atmospheric circulation in Alpine-type valleys which is intimately linked to the orography surrounding the lakes 40 interacting with the synoptic scale circulation. Thermally driven wind systems may be induced by hilly terrain, such as slope, mountain, and plateau winds (Kottmeier et al., 2008). Such winds result in mountain venting phenomena that control the variability of the water vapour field in mountain catchment on very small-time scales. Furthermore, small-scale inhomogeneity in soil properties across a mountain valley, as well as lake breezes resulting from land-lake contrasts, may also induce the development of thermal circulations, particularly on clear-sky days, modifying the wind, humidity, and temperature fields on 45 small spatial scales. The interaction of slope-driven and secondary circulations can furthermore influence the thermodynamical environment in the valley by the formation of convergence lines. Such convergence lines may favour the formation of shallow clouds, and in some cases even deep convection (Barthlott et al., 2006). Interaction with the synoptic scale circulation can lead to the formation of strong, gusty down-valley winds such as foehn events (Drobinski et al., 2007) and gap flows (Flamant et al., 2002;Mayr et al., 2007) that can also contribute to rapid modifications of the water vapour field in mountain catchment 50 areas.
In light of the complexity around AMLs described above, two research questions arise: (1) What is the role of lakes on the local and regional atmospheric water cycle in AMLs? (ii) What is the relative contribution of evaporation from the lake to the atmospheric water content over and downwind of a lake? An early study based on water stable isotope measurements conducted in the US Great Lakes region suggested that in the summer up to 15% of the atmospheric water content in the 55 atmosphere downwind of the lakes is derived from lake evaporation (Gat et al., 1994). Stable water isotopes have long been used as a tool to study processes in lacustrine and hydrologic systems (see review by Gat, 2010), as well as evapotranspiration (e.g. Berkelhammer et al., 2016). Isotopic measurements of lake water have shown the relative roles of evaporation from the lake surface and transpiration from surrounding vegetation (Jasechko et al., 2013). The link between hydrology and evaporation has mainly been investigated using vapour and liquid water isotopes measurements gathered just above the Earth's 60 surface and samples from lake water and precipitation (e.g. Cui et al., 2016). Measurements from tall towers only provide incomplete information on the link between evaporation and atmospheric processes in the free troposphere, such as mixing and distillation (e.g. Steen-Larsen et al., 2013). He and Smith (1999) combined airborne measurements and surface sampling of the water isotope composition to study the evaporation process over the forests of New England, however before the advent of high-resolution laser-based spectrometers. More recent available airborne measurements of the isotope composition in the Page 3 sur 43 boundary layer either focused on areas above sea (Sodemann et al., 2017), or did not include measurements of the surface isotope composition (Salmon et al., 2019).
Hence, unexploited potential remains to use stable water isotopes for increasing our understanding of the influence of evaporation, boundary-layer processes, and the free troposphere for local and regional climate conditions in AMLs. For example, the depth of the atmospheric layer over which the influence of evaporation from the lake surface is detectable, and 70 how different factors control the depth of this layer are still largely unknown. Detailed and comprehensive analysis of smallscale factors, such as winds in a valley, and how they are related to the mesoscale and large-scale dynamics, such as synoptic scale subsidence in complex terrain, are therefore needed.
In order to get so insights into such aspects, the L-WAIVE (Lacustrine-Water vApor Isotope inVentory Experiment) field campaign was conducted in the Annecy valley (45°47' N, 6°12' E, in Haute-Savoie in the French Alps) around the Annecy 75 lake during the month of June 2019. Being the second largest natural, glacial lake in France, the Annecy lake is expected to play a substantial role for the regional hydrometeorology.
The overarching scientific objective of L-WAIVE is to study evaporation processes and their heterogeneity over the Annecy lake using an original a multi-platform instrumental approach based on continuous high-resolution vertical profiling of tropospheric water vapour, temperature and wind as well as aerosols in the valley, together with ship-borne and airborne 80 measurements of stable water isotopes (H2 16 O, H 2 HO and H2 18 O) in the lake, in the lower atmosphere as well as in-cloud and in precipitation. An additional objective is to construct a reference aerosol and water vapour stable isotope profiles database for ground, airborne, and satellite lidar simulators currently developed by the consortium of researchers involved in L-WAIVE, itself included in the WAVIL (Water Vapor and Isotope Lidar) project.
This paper provides an overview of L-WAIVE campaign in terms of experimental strategy (Section 2), instrumental platforms 85 operations, environmental variables monitoring (Section 3) and synoptic conditions between 12 and 23 June 2019 in the Annecy lake area (Section 4). Atmospheric water vapour and liquid water isotopes, as well as lake water isotope observations made across the valley are described in Section 5. In Section 6, we summarize and conclude.

L-WAIVE experimental strategy
To achieve the scientific and methodological objectives of the L-WAIVE project, the field campaign was implemented in the 90 southern part of the Annecy lake (the so called "Petit Lac"), in the vicinity of the city of Lathuile (Fig. 1). The Annecy lake is bordered by the city of Annecy to the north, the Massif des Bauges to the west (2217 m above mean sea levela.m.s.l.), the Massif des Bornes to the East (2438 m a.m.s.l.) and the depression de Faverges to the south (where Lathuile is located).
Lathuile is located east of the foothill of the Roc des Boeufs (1774 m a.m.s.l.) to the west of the "Petit Lac" and the Tournette summit (2350 m a.m.s.l.) to the east. The Annecy lake, at a mean altitude of 446.7 m a.m.s.l., covers an area of roughly as the Roc des Boeufs to the left, the Tournette summit to the right and the "Petit Lac" in between.

Measurement platforms
Three airborne, one ship-borne and one ground-based instrumented platforms were deployed in the vicinity of Lathuile in order to monitor humidity, temperature, wind, clouds and aerosols in the lower troposphere over the Annecy lake and the surrounding valley environment, as well as to conduct measurements at the interface between the atmosphere and the lake, and in the lake.

105
A brief overview of the platforms is given below, whereas the details on the instrumental payloads are given in Section 3: i) Airborne platforms: a. One ultra-light aircraft (ULA) was mainly dedicated to remote sensing measurements. It allowed exploring the two or three-dimensional structure of the lower troposphere thanks to a polarized Rayleigh-Mie lidar. It also carried a meteorological probe (pressure, temperature, relative humidity), an aerosol particle sizer and two visible flux meters for the upward and downward solar radiations to complement the lidar measurements and determine the surface albedo in the visible spectrum. This will be referred to as "aerosol ULA" (ULA-A) in the following.
b. A second ULA carried both a Cavity Ring-down Spectrometer (CRDS) water vapour isotope analyser, a meteorological probe for pressure, air temperature, Global Positioning System (GPS) location and relative 115 humidity, and a cloud water collector. The platform offers the opportunity to measure the vertical profiles of temperature, relative humidity, H 2 HO, H2 18 O and H2 16 O and to collect cloud water samples. We will refer to this platform as "isotope and cloud ULA" (ULA-IC) in the following.
c. An Unmanned Aerial Vehicle (UAV) acquired vertical profiles of temperature, relative humidity, and pressure in the surface layer (first 150 m above the ground level) using a meteorological probe. 120 ii) Ground-based platform: Simultaneous high-resolution vertical profiles of water vapour, temperature, aerosols and winds were acquired continuously from two co-located ground-based lidars.
iii) Ship-borne platform: An instrumented boat allowed sampling the lake water by vials at the surface film and underneath (~2 m deep) for the assessment of H 2 HO and H2 18 O. A probe was also used to assess the vertical profiles of water temperature below the surface down to a depth of 55 m. A CRDS water vapour isotope analyser 125 performed measurements during one day at the end of the experiment just above the lake surface in parallel with the lake water sampling.
These different platforms are presented in Fig. 1 together with a view of the experiment site.

Deployment
Depending on the weather conditions, airborne platforms were deployed several times a day to document the temporal 130 evolution of the atmospheric boundary layer over the lake. The days of operation of all platforms are summarized in Table A1 of Appendix A. The ground-based water vapour, temperature and aerosol lidar operated continuously between 12 and 21 June in the morning (gathering over 220 h of data), while the ground-based wind lidar (WL) operated continuously between 14 and 23 June in the morning (acquiring also over 220 h of data  17, 18, 19 and 20 June. These are considered as "golden days" and will be analysed in priority. A few less optimal days ("silver days") have also been defined based on the fact that 5 out 6 instruments were simultaneously in operations, namely: 13 and 21 June. 21 June stands out as being a day when both precipitation water samples, and ship borne CRDS were acquired and will also be an analysis priority.
On days when both ULA flew coordinated patterns (13, 16-20 June), flights typically began with a profiling sequence between 150 the surface and ~4 km a.m.s.l. which was carried out in the vicinity of the 2 ground-based vertically pointing lidars (see Fig. 2).
Soundings with levelled legs (see blue dotted line in Fig. 2a) were performed at a relatively slow ascent rate (~60 m min -1 ) to ensure that the instruments were as close as possible to equilibrium with the environment. Upon reaching 4 km a.m.s.l., the flight route of the 2 ULA differed, ULA-A performing a high altitude survey above the Annecy lake (see red dotted line in Fig. 2a), while ULA-IC was aiming for shallow cumulus clouds to sample cloud water droplets as illustrated in Fig. 2b showing 155 the instrumental synergy that took place during L-WAIVE. Liquid water sampling was performed via multiple passes through the clouds to accumulate enough material to conduct isotope analysis. At the end of the flight, both ULA performed race-track descents around the ground-based lidars on their way back to the airfield.
During ascent and descent, the airborne lidar ALiAS onboard ULA-A was pointing sideways to directly derive the aerosol extinction coefficient (Chazette et al., 2007;. For the exploration of the valley at a cruising altitude 160 between 3.5 and 4.5 km a.m.s.l., ALiAS was pointing to the nadir. The combination of both flight sequences thus allowed to survey the 3-dimensional structure of the lower troposphere over the lake and its surroundings. The individual flight characteristics (time, maximum altitude, type of exploration) are presented in Appendix B for the two ULAs (Tables B1 and   B2 for ULA-A and ULA-IC, respectively).

Airborne payloads
We used two Tanarg 912 XS ULA from the company Air Création . For each ULA (ULA-A and ULA-IC, Fig. 3), the maximum total payload is of approximately 250 kg including the pilot. Flight durations were between ~1 and 2 hours, depending on flight conditions, with a cruise speed around 85-100 km h -1 . The ULA location was provided by a 175 GPS and an Attitude and Heading Reference System, which are part of the MTi-G components sold by XSens. Meteorological probe. Part of ULA-A payload was a shielded meteorological probe VAISALA PTU-300 for measuring temperature, pressure and relative humidity. This probe measures the atmospheric pressure, with a 1-minute sampling time, within an uncertainty of 0.25 hPa, the air temperature within an uncertainty of 0.2 K and relative humidity (RH) within a 195 relative uncertainty of 2.5%.
Particle sizer. The granulometer used on board ULA-A was a FIDAS mobile manufactured by PALAS (https://www.palas.de/en/). The particle sizer operates on battery power with a volume flow of 1.4 l min -1 in environmental conditions of temperature, atmospheric pressure, and relative humidity (no drying). The particle size distribution is determined from 180 nm to 20 µm by means of an optical aerosol spectrometer using Lorenz-Mie scattered light analysis. The LED source 200 homogeneously illuminates an optically differentiated measurement volume with white light. Each particle moving through this volume generates a scattered light impulse detected at an angle of 85° to 95° degrees. The amplitude of the impulse is a measure of the particle diameter and the particle number corresponds the number of impulses. To allow in-flight measurement while limiting the loss of particles, a sampling head has been designed and printed in 3D in order to guarantee an isokinetic air flow at the entrance of the FIDAS. the first flight on 20 June. The anomaly was due to a saturated inlet system from condensate forming on the aircraft during a cloud sampling flight on 18 June. Flight periods affected by the saturated inlet were excluded from further analysis. Further details of the data processing and calibration procedure are described in the data report that accompanies the data set, and in a forthcoming publication.
In the afternoon of 19 June 2019, the two ULA performed a coordinated ascent, providing an intercomparison of response 230 times and measurement offsets regarding pressure, temperature, and relative humidity on the two aircraft ( Fig. 4). During the entire ascent sequence, there is a small, visually discernible temperature offset between the two instrumental packages ( Fig. 4b). This offset originates from the slower response time of ULA-A instrumental package (red line), lagging behind ULA-IC the faster and more exposed iMet probe (black line). The relative humidity measurement on ULA-IC is clearly faster, and resolves spikes in more detail (Fig. 4c, black line). A comparison of the specific humidity calculated from iMet and the 235 laser spectrometer revealed similar response times (not shown). The intercomparison between the two ULAs thus also provides a first-order in-flight validation of the water vapour measurements of the CRDS analyser.
A time resolution better than 0.1 Hz was obtained for specific humidity from the CRDS and iMet probe, providing a spatial resolution of 200-300 m in horizontal direction, and 10-50 m in the vertical, assuming a typical horizontal speed of 80-100 m s -1 and ascent rate of the aircraft of about 1-5 m s -1 . Due to more complex memory effects, the isotope composition has lower 240 actual time resolution (Sodemann et al., 2017;Steen-Larsen et al., 2014). We use here 10 s average data for all parameters on ULA-IC from upward profiles, filtered for rapid elevation changes, defined as exceeding 50 m ascend or 20 m descend within Cloud water collector. A pre-cleaned Caltech Active Strand Cloud Water Collector was mounted on ULA-IC, modified to efficiently collect cloud water at the speed of the ULA (Fig. 3b). The ULA-IC relative cruising speed is 85 to 100 km h -1 , in the operating range of the Caltech Active Strand Cloud Water Collectors (CASCC, Demoz et al., 1996). A CASCC was modified to efficiently collect cloud liquid droplets from the ULA. In order to sample droplets under the same conditions as 250 those obtained on the ground, the CASCC's fan was removed, and its inlet and outlet were prolonged with convergent and divergent High Density PolyEthylene (HDPE) cones to ensure an isokinetic air sampling. The flow through the probe must be steady and as turbulence-free as possible. The design of these modified inlet and outlet was calculated to get a constant mass of water droplets per time unit through the probe onboard the ULA. All sampling materials used were plastic or HDPE. The resulting probe was installed on the side of the ULA where the flow was assumed laminar and allowed air sampling with a 255 flow rate of 35 to 47 m 3 min -1 , ahead of the motor exhaust (Fig. 3b). The CASCC strings and inlet were pre-cleaned with deionized water prior each flight and covered with a clean plastic bag when not in-cloud (especially during take-off and landing). For a flight of 10 minutes inside a shallow cumulus cloud, the probe typically collected 41 to 48 g of cloud liquid droplets corresponding to a liquid water content of 0.10 to 0.16 g m -3 , typical for such cloud type (Herrmann, 2003).

260
The ground-based scientific facility hosted by the technical department of the city of Lathuile was mainly composed of the Raman lidar WALI and of a scanning Doppler lidar. WALI was embedded in the ground-based MAS (mobile atmospheric station; Raut and Chazette, 2009). The UAV was also operated from this site, close to the lidar. Figure 5 shows the two lidars and the UAV.
The lidar operates with an emitted wavelength of 354.7 nm and is designed to fulfil eye-safety standards (EN 60825-1). Its emitter is a pulsed Nd:YAG laser (Q-smart 450 by Quantel™) with a fibre laser injector to stabilize the emitted wavelength.
The acquisition system is based on a PXI (PCI eXtensions for Instrumentation) technology manufactured by the National 275 Instruments™ company, and contains 12 bits digitizers at 200 MS/s corresponding to a native vertical resolution of 0.75 m.
During the entire experiment, the acquisition was performed for mean profiles of 1000 laser shots leading to a native temporal sampling close to 1 minute. The UV pulse energy was ~70 mJ and the pulse repetition frequency was 20 Hz. The wide fieldof-view of ~2.3 mrad allows a full-overlap of the transmission and reception paths beyond ~ 200-300 m. Note that the pulse energy decreased linearly with time due to a laser failure, from 70 mJ to about ~50 mJ at the end of the campaign.

280
The receiver is composed of 2 distinct detection paths, both using small collector diameters of 15 cm, with a low full-overlap distance (~150-200 m). The first path is dedicated to the detection of the elastic molecular, aerosols and cloud backscatter from https://doi.org/10.5194/acp-2020-1194 Preprint. Page 13 sur 43 the atmosphere. Two different channels are implemented on that path to detect i) the total (co-polarized and cross polarized with respect to the laser emission) and ii) the cross-polarized backscatter coefficients of the atmosphere. The second path, a fibered achromatic reflector, is dedicated to the measurements of the atmospheric Raman scattering, namely the vibrational 285 signal for nitrogen (N2-channel) and water vapor (H2O-channel) and the rotational signal to derive the temperature (T-Channel).
The water vapor mixing ratio (WVMR) is retrieved with an absolute error less than 0.4 g kg -1 in the first 2 km above the ground level (a.g.l.) (Chazette et al., 2014;Totems et al., 2019). The calibration of the T-channel is derived from the methodology presented by Behrendt (2006) and leads to an absolute error on the temperature lower than 1 °C within the first 2 km a.g.l. The 290 final vertical resolution is set to 15 m below 1 km a.g.l. and 30 m above, and the temporal resolution is 0.5 h. In the following, a temporal resolution of 1 h is used.

Wind lidar
Wind profiles were measured using a scanning Doppler lidar (Leosphere Windcube WLS100). It operates in the infrared (1.543 µm) with a low pulse energy (0.25 mJ) but a high pulse repetition frequency (20 kHz). The Doppler shift due to the particles' 295 motion along the beam direction (radial wind speed) is determined through heterodyne detection followed by fast Fourier transform analysis. The acquisition time was set to 1 s during the campaign. The pulse duration is 200 ns, corresponding to an axial resolution of 50 m given the pulse shape, with a minimal and maximal range of 100 m and 7.2 km, respectively. The axial resolution can be lowered to 25 m by reducing the pulse duration (100 ns) while increasing the pulse repetition frequency (40 kHz) which in turn reduces the minimal and maximal range to 50 m and 3.3 km, respectively. In practice, the maximum 300 range is limited by the signal level. A minimum Carrier to Noise Ratio (i.e. signal to background noise ratio) of −27 dB is required to keep the radial wind uncertainty (determined from the spectrum peak width) below 0.5 ms -1 . Therefore, observations in the free troposphere are possible only when elevated layers of aerosols are present. Even in the boundary layer, several days of very low aerosol load occurred during the campaign, in which case the Carrier to Noise Ratio threshold was lowered to −30 dB. With such a low Carrier to Noise Ratio, the measurements must be considered with caution.

305
Profiles of the three components of the wind vector were determined using the Doppler Beam Swinging technique originally proposed for Doppler radar (e.g. Koscielny et al., 1984). Here, the measurement cycle includes one vertical beam, for which Page 14 sur 43 with a rather flat bottom, and the closest distance existing between one of the slanted beams and an obstacle was ~1 km.
Therefore, the influence of the orography on the measurement are assumed to be negligible here.

Lake and atmospheric sampling
The shipboard payload is shown in Fig. 6. The lake water surface and subsurface were sampled at the middle of the "Petite Lac" to measure water isotopes and chemicals from the boat.
The lake water thermal stratification was monitored using an EXO sonde, equipped with temperature, pressure, pH, dissolved 320 dioxygen, ion conductivity and chlorophyll sensors. Profiles recorded in the middle of the "Petit Lac" (Fig. 1) once or twice per day (Table A1) and showed that the thermocline (Fig. 7) was below a depth of 10 m, in good agreement with previous studies of the Lake (Danis et al., 2003).
Subsurface water samples were collected at a depth of 2 m in HDPE capped flasks. Surface water samples were collected using a 30 x 30 cm silica glass plate immersed into water for a minute, then gently removed from water vertically (Cunliffe et al., . The water falling from the plate in a continuous flow was not sampled, then, the dropwise water was collected in HDPE capped flaks. The surface microlayer samples were then collected by scraping the remaining water on the glass plates (using a rubber scraper) in amber glassware capped vials. On 22 June, the CRDS isotope analyser taken on board the boat to sample a cross-section of the "Petit Lac" through the location used for in situ sampling on the other days. This made it possible to follow the evolution of isotope concentrations as a function of the distance from the shore and the depth of the lake on that 330 day. Water vapour isotope measurements were taken from an inlet at either ~20 cm or ~2 m above the water surface, while the iMet sonde measured temperature, relative humidity and location. Post-processing and calibration of water vapour measurements were done as for the aircraft data (Sec. 3.1.2).

360
The synoptic flow at 700 hPa over the Annecy lake was dominantly south-westerly from 12 to 16 June, except on 14 June when it was almost southerly as the western flank of the high pressure ridge was seen to straddle the Alps (Fig. 8a). On that day, Saharan dust passed over the experimental supersite (not shown). Between 16 and 17 June the synoptic flow over the Alps was substantially weaker, with jets located north and south of central Europe (Fig. 8b) quite unlike the previous period.
Between 18 and 23 June, the synoptic south-westerly flow over the area of interest intensified again as the high-pressure ridge 365 built up.
During the course of the campaign, the area of interest was alternatively under the influence of warmer temperatures (air masses with high equivalent potential temperature) linked with the high pressure ridge (14-15, 19-23 June, Fig. 9a) and colder temperature (air masses with low equivalent potential temperature) associated with the surface low over the British Isles (12-13, 16-18 June, Fig. 9b). Page 19 sur 43 were gathered during the campaign (8 out of the total 28, Table A1). Several isolated thunderstorms were observed on that day, and on 16 June in the morning. The influence of weather events during the campaign will be further discussed in the following sections.
Rayleigh-Mie lidars are very efficient tools to detect aerosol layers, but also clouds (e.g. Platt, 1977;Berthier et al., 2004), both semi-transparent (ice clouds) or opaque (liquid water clouds) to the laser beam. During L-WAIVE, the temporal evolution 390 of the aerosol burden and clouds above the Annecy lake have been monitored using both the ground-based lidar WALI and the airborne lidar ALiAS. In addition, continuous monitoring of aerosols vertical distribution from a lidar allows to get insight into air masses transport. The identification of their particulate constituents (e.g. dust, pollution aerosols,…) may also provide information on the origin of observed air masses. For instance, such a capability was used to improve our understanding of atmospheric circulation in complex situations such as extreme heat wave phenomena . enhanced VDR values are also evident on 19 and 20 June that are associated with local and/or regional pollution advected in altitude and verticaly mixed by dry convection.
The period was rather cloudy as shown in Fig. 11a (shaded areas) where cloud types are indicated as the raining periods including thunderstorms. It is worth noting large lidar-derived VDR values detected in the presence of dense clouds are related 410 to multiple scattering (areas appearing in pink in Fig. 11b), not to the presence of non-spherical aerosols.
The measurements made with the airborne particle sizer corroborate the nature of the particles identified from the lidar observations. Figure 12 shows

Local wind
The temporal changes in wind intensity and direction observed by wind lidar are given in Fig. 13. Weak winds, generally less than 5 ms -1 , are observed in the valley, below 2 km a.m.s.l. The wind intensity does not show repetitive patterns from one day to the next. In contrast, wind direction shows regular variations, with winds directed towards the exit of the valley (south) 430 during the day, consistent with the flushing of aerosols suggested by the lidar observations. During the night, winds are directed towards the city of Annecy, thereby contributing to the accumulation of aerosols in the Annecy lake catchment. The local wind in the valley is obviously disturbed during stormy periods and during the episode of dust long-range transport. On 14 June 2019, the presence of these aerosols at higher altitudes (up to ~5 km a.m.s.l.) allowed the wind to be retrieved with the WL above 2 km a.m.s.l. Strong southerly winds (in excess of 20 m s -1 above 3.5 km a.m.s.l.) were observed in agreement with the 435 meteorological fields in Fig. 8a. The upper part of the regional pollution plume on 20 June is also seen to be associated with stronger southwesterly winds in excess of 10 m s -1 (Fig. 13a).
https://doi.org/10.5194/acp-2020-1194 Preprint. Discussion started: 1 December 2020 c Author(s) 2020. CC BY 4.0 License.  Some of the marked WVMR features observed in the lidar measurements are reproduced in the WVMR fields of ERA5 with an horizontal resolution of 0.25° (Fig. 14b), especially during the second part of the campaign, starting on 18 June 2019. They differ more in the first part which is influenced by thunderstorms and strong rainy periods, and thus more local processes. The 455 boundaries between the different moist air mass types are similar between lidar measurements and reanalyzes, although the reanalyzes are moister in the planetary boundary layer. There is a strong decrease of water vapor at ~3 km a.m.s.l. which marks the transition to the free troposphere where long-range air mass transport occurs. This altitude corresponds roughly to the average altitude of the mountains at the regional scale around the measurement site.
The WVMR values retrieved from the meteorological probe on board the 2 ULAs match well with those derived from the 460 lidar, as shown in Fig. 15 for the ULA flights listed in Table B1 and B2. Furthermore, the in-situ measurements made with the 2 meteorological probes are in excellent agreement with one another. A drier lower troposphere is indeed found during the first two flights on 23 June and the transition to the part of the troposphere influenced by long-distance transport between ~2.5 and 4 km a.m.s.l. is well represented. Differences in the altitude of the transition layer are seen between the in-situ and lidar observations which are related to the fact that the airborne measurements were sometimes taken over the entire valley and even over the surrounding mountains.

Water isotope survey of the lake d'Annecy region
Stable isotope content of water vapour and in liquid water was quantified regarding H2 18 O and H 2 H 16 O, as well as the deuterium excess (d-excess). Water vapour was sampled in-situ using a CRDS, whereas liquid samples, including precipitation, cloud 475 water and lake water were analysed in the laboratory. Here we report the isotope composition (see interpretative framework in Appendix D) as delta-values relative to a standard (e.g. Gat, 1996).

Atmospheric water vapor sampling
In total, 15 flights with ULA-IC have been performed (see Table B2) including 14 flights where the CRDS allowed a representative sampling of 2 and 18 (saturation on the inlet was encountered throughout flight 11). We provide here an 480 overview of the acquired measurement data, while a more detailed analysis on this dataset is ongoing. The isotope content 2 observed during the flights ranged between about -340 and -80‰ for flight altitudes up to ~3.5 km a.m.s.l. as shown in Fig. 16a. For this figure, we considered only the ascending parts of the flights in order to limit the dispersion of points related to atmospheric heterogeneity on leveled ULA legs. As observed in earlier studies, the dataset is clustered along a typical mixing line in 2 − space (Noone, 2012;Salmon et al., 2019;Sodemann et al., 2017). The end-members of the mixing lines show 485 substantial day-to-day variations, from 2 =-110 to -80‰ (~[-16, -12]‰ for 18 , not shown) for the more humid end member (q > 8 g kg -1 ), and from 2 =-340 to -230‰ (~[-30, -20]‰ for 18 , not shown) for the drier end member (q < 3 g kg -1 ). It is important to keep in mind that both, the variation in maximum flight altitudes and the meteorological situation can contribute to variability.  (Table B2 of Appendix A).
Vertical gradients in isotope ratios changed according to the weather evolution, among other factors by modifying the 495 stratification of the lower troposphere above the valley. After intense thunderstorms in the morning of 16 June, the initial strong vertical gradient in 2 at 2.5 km, also observed during both 16 June afternoon (Flight 5) and 17 June morning (flight 6), reduced substantially during 17 June afternoon (Flights 8-9). Large vertical gradients are also observed at higher altitudes (between 3 and 3.5 km a.m.s.l.). Such a variability is probably specific to the valley being embedded in low mountain ranges.
The main vertical structures are derived from the vertical profiles of potential temperature, aerosols, wind and relative humidity 500 (RH). For the illustration, Fig. 17 shows a typical vertical RH profile that highlights the different layers present between the lake (~0.5 km a.m.s.l.) and about 4 km a.m.s.l.: the lake boundary layer between ground level and 1 km a.m.s.l. (layer 1), an atmospheric layer between about 1 and 2.5 km a.m.s.l. influenced by the area of the lake surrounded by mountains about 2-2.5 km high on its southern part (layer 2), a transition towards the free troposphere between about 2.5 and 3.5 km, influenced by the regional circulation (layer 3) and above 3.5 km a.m.s.l. the free troposphere with a synoptic influence, where long-505 ranged transport of air masses may occur.
The gradient in isotope content was related to a strong inversion between layers 2 and 3 where some clouds can be observed (see Fig. 11a), and was at least partly related to the advection and descent of dry air masses from the Libyan anticyclone during the 13-14 June period (Fig. 8a). The near-complete vertical mixing, and thus the uniform isotope content near the surface are testimony to active turbulence above complex terrains during summer. In contrast, the strong gradients above 2.5 km a.m.s.l. 510 are probably due to regional advection, including the descend of airmasses from higher altitudes, as observed in the airborne observations of the water vapour isotopes in the Mediterranean (Sodemann et al., 2017) during the Hydrological cycle in the Mediterranean experiment (HyMeX). Above 3 km a.m.s.l., a humidity gradient is also observed and the formation of different types of clouds at the interface of the layer 3 and the free troposphere has been observed many times during the field campaign (see Fig. 11a). These gradients are related to strong contrasts in air mass transport. 515

Cloud liquid water sampling
Four relevant cloud water samples have been taken with CASCC during 18, 20 and 21 June (Table B1). The isotope content of the cloud water samples are shown in Fig. 18a (coloured circles). They are close to the global meteoric water line (GMWL) with corresponding d-excess values ranging from 12.1 to 14.8 ‰. The cloud water sample taken in the morning of 20 June (Fig. 18a, orange symbols) carries a weak evaporation signal. Equilibrium condensates were calculated from water vapour isotope measurements during the time when the cloud samples were taken (Fig. 18a, coloured squares) using the fractionation factors of Majoube (1971), and air temperature measurements at cloud level, ranging between -4 and 1°C. It is worth noting that given potential sources of uncertainty, the equilibrium condensate values agree remarkably well with the cloud water samples during 18 June (blue symbols), and the afternoon of 20 June (green symbols). Overall, results here confirm that the cloud water formed from equilibrium fractionation from ambient vapour. It is worth mentioning that such agreement between 525 the completely independent sampling and measurement of vapour and cloud water supports the overall validity of i) the cloud water sampling protocol, ii) the airborne vapour isotope measurements and iii) the consistency of the calibrated water isotope dataset derived from the L-WAIVE campaign.

Precipitation sampling
In total, 28 precipitation samples, of which 22 are unique, have been taken during the campaign (see Table B1), with sampling 530 times lasting from 20 min to several hours, depending on rainfall rate. The isotope composition in rainfall ranges from ~-11.2 to 2.2‰ in 18 and -73.5 to 9.6 ‰ in 2 (Fig. 18a, grey triangles), with corresponding values of d-excess between ~3.5 and 20.8‰ (not shown). Replication of measured isotope values on duplicate samples confirm complete sample preservation.
Notably, there is an overall correspondence between the isotope range observed in cloud water samples and in a majority of precipitation samples. Deviations of precipitation samples from the GMWL indicate potential below-cloud exchange, or post-535 condensational exchange processes leading to an enrichment of the water drop in 18 (Graf et al., 2019;Worden et al., 2007).
Samples taken between 12 to 15 June, and partly on 21 June are most enriched in 18 , exhibiting higher values in 18 and 2 and a calculated smaller d-excess (even negative ~-8‰). These samples are from rainfall events associated with local thunderstorms (see Fig. 11a). The low d-excess of these samples may point out the evaporation of rain droplets during their fall. The less negative isotope values indicate exchange between falling raindrops and ambient water vapor below cloud base.

540
Even if evaporation effects from the sampling setup cannot be fully excluded, in particular for the samples with a sampling duration of more than 3h, relatively high d-excess values during most samples, and the consistency of duplicate samples indicate that the influence of sampling artefacts can overall be considered as secondary.

Lake liquid water sampling
Evaporation from the Annecy lake is expected to be an important source for the water vapour in the Annecy valley. In order 545 to link the atmospheric profiles of water vapour isotopes to the lake as a moisture source, 20 lake water samples (Table B1) were taken throughout the campaign within the lake-atmosphere interface layer (6 samples), as well as between 0.1 and 2 m depth (14 samples), and analysed for their stable water isotope composition. The average isotope contents were -8.3±1.5‰ for 18 , and -63.0±6.0‰ for 2 . Lake water samples taken close to the surface are expected to be most affected by evaporation, causing deviations to the right from the GMWL along evaporation lines (enrichment in 18 ). This is clearly observed in samples taken at a depth of 10-20 cm on one day (Fig. 18b, square symbols), the derived d-excess showed less evaporation influence, with a median of 6.3‰, and a standard deviation of 6.3‰. The fact that there are differences between the isotope contents at the interface and between the uppermost layer and 2 m depth points to incomplete mixing due to surface winds.

555
Indeed, temperature profiles taken within the lake show a typically strong, but variable thermocline at about 5-8 m depth during the campaign (Fig. 7). While some sampling artifacts cannot be fully excluded, the influence of evaporation appears to decrease towards a depth of 2 m.
To assess the coherence between the lake water and the water vapour isotope composition measured by ULA and boat, we calculated the isotope composition of liquid condensate resulting from an equilibrium fractionation process (Fig. 18b,   560 triangles). The range of equilibrium condensate 2 values from the ULA (triangles) matches overall with the range of 2 observed in lake water for 16, 17 and 21 June (dots, squares). The 18 in equilibrium condensate is substantially more depleted, confirming the existence of kinetic (non-equilibrium) fractionation during lake evaporation. During 18 and 22 June, equilibrium condensate is more enriched than lake water, pointing to the influence of other sources that contribute to the water vapour isotope composition. It is worth noting that Lake Annecy is fed by a catchment whose surface is 10 times larger than 565 that of the lake via ten main tributaries located on the lake's periphery. The flows of its tributaries increase significantly in situations of heavy rain, which can substantially influence the isotopic content of the lake during the course of a year. This first-order comparison provides the basis for a more thorough analysis on how the lake evaporation signal is imprinted on the atmosphere above the lake in forthcoming studies.

570
For the sampling environments illustrated in Fig. 17, the overall results for stable isotopes in water are summarized in Fig. 19 on a statistical basis. This synthesis is presented in the form of whisker boxes for d-excess (Fig. 19a) For the atmospheric water vapor, the sample number n is significant (numbers given at the right of the whisker boxes in interval. As they do not overlap from an atmospheric layer to another, we can conclude, with 95% confidence, that the true medians do differ. Hence, we highlight a significant variability in the stable isotope content of water vapor depending on the atmospheric layers previously identified. Page 30 sur 43 Between the surface of the lake and 2.5 km a.m.s.l., we find 18 values similar to those recorded by Craig and Gordon (1965) for evaporation over the Mediterranean [-15,-10] ‰. On the other hand, they report much more dispersed d-excess values 585 ranging from 5 to 35‰. Gat (2000) gives narrower values for marine and European air masses with a d-excess interval of [7-11] ‰. Our observations are mainly outside this interval for the first two layers, they overlap it for the 2.5-3.5 km a.m.s.l. layer more influenced by both regional and long-range transports.
Our liquid water samples are not numerous enough to define a confidence interval. As explained previously, we nevertheless note an enrichment in heavy isotope of the surface layer of the lake which is directly in contact with the atmosphere and 590 therefore directly subject to evaporation. The deep layers are significantly less enriched and their isotope content ( 18~-9‰ and 2~-65‰) is close to that given by Jasechko et al. (2013) for samples from Lake Superior (USA) which is not a highaltitude lake but is close to mean sea level. In addition, measurements of 18 profiles in the Annecy lake have already been carried out between the years 2000 and 2002. They have shown 18 values of the order of -9.5‰ for water above the thermocline of "Petit Lac" during the summer months (Danis, 2009), in agreement with our results. This indicates a year-to-595 year stability of the isotopic content of the lake water below the surface microlayer.
To our knowledge, this is the first time that in-situ samples of cloud water and water vapour measurements have been taken directly in clouds. From a first analysis, cloud water formed in an equilibrium fractionation process for the samples collected here, and the cloud water isotopes were relatively similar to local precipitation (taken on other days

605
The mean mixing lines is reported using a black dotted line. (b) Vertical profiles of (‰). Color indicates flight number. Isotope data are averaged at 10s interval. Only data points during ascent are displayed.  The spatiotemporal evolution of the vertical profiles of WVMR appear coherent between the lidar retrievals and the ERA5 reanalysis, as well as with the measurements of the meteorological probes on board the ULAs. A marked diurnal cycle is 630 observed with maximum WVMRs during the day (between 8 and 11 g kg -1 ) when the valley winds are advected above the lake before reaching the measurement site. The vertical profiles of isotope composition are shown to be mostly uniform below 2 km a.m.s.l. (~1.5 km a.g.l.), and to exhibit strong vertical gradients at higher elevations, with a marked decrease against altitude at the interface between the different layers identified above the valley. We note that the cloud water samples are close to the GMWL, (d-excess between 12.1 and 14.8 ‰), which indicates that the cloud water formed from equilibrium fractionation of 635 ambient atmospheric water vapour. There is an overall correspondence between the isotope range observed in cloud water samples and a majority of precipitation samples (d-excess between -8.6 and 20.8/‰). The deviations observed on precipitation samples that are below the GMWL indicate potential below-cloud exchange, or post-condensational exchange processes. The average isotope composition of the lake water, taken at 2 m depth appears different from the one for the surface microlayer sample, due to evaporation processes. Moreover, the 18 in equilibrium condensate above the lake is generally substantially 640 more depleted, confirming the existence of non-equilibrium fractionation during lake evaporation. It is worth noting that during 18 and 22 June, equilibrium condensate was calculated as more isotope-enriched than Petit Lac water, pointing to the influence of other sources on the water vapour isotope composition above the lake.
The notable atmospheric vertical gradients of stable isotope composition, and day-to-day variation throughout the measurement campaign were clearly related to the current weather conditions, modified by local topography. Beyond the value Appendix B: Ultra-light aircraft flights description