Response of water vapour D-excess to land–atmosphere interactions in a semi-arid environment

The stable isotopic composition of water vapour provides information about moisture sources and processes difficult to obtain with traditional measurement techniques. Recently, it has been proposed that the D-excess of water vapour (dv = δ2H− 8× δ18O) can provide a diagnostic tracer of continental moisture recycling. However, D-excess exhibits a diurnal cycle that has been observed across a variety of ecosystems and may be influenced by a range of processes beyond regional-scale moisture recycling, including local evaporation (ET) fluxes. There is a lack of measurements of D-excess in evaporation (ET) fluxes, which has made it difficult to assess how ET fluxes modify the Dexcess in water vapour (dv). With this in mind, we employed a chamber-based approach to directly measure D-excess in ET (dET) fluxes. We show that ET fluxes imposed a negative forcing on the ambient vapour and could not explain the higher daytime dv values. The low dET observed here was sourced from a soil water pool that had undergone an extended drying period, leading to low D-excess in the soil moisture pool. A strong correlation between daytime dv and locally measured relative humidity was consistent with an oceanic moisture source, suggesting that remote hydrological processes were the major contributor to daytime dv variability. During the early evening, ET fluxes into a shallow nocturnal inversion layer caused a lowering of dv values near the surface. In addition, transient mixing of vapour with a higher D-excess from above the nocturnal inversion modified these values, causing large variability during the night. These results indicate dET can generally be expected to show large spatial and temporal variability and to depend on the soil moisture state. For long periods between rain events, common in semi-arid environments, ET would be expected to impose negative forcing on the surface dv. Spatial and temporal variability of D-excess in ET fluxes therefore needs to be considered when using dv to study moisture recycling and during extended dry periods with weak moisture recycling may act as a tracer of the relative humidity at the oceanic moisture source.


Introduction
Climate change has the potential to significantly impact surface and atmospheric water budgets.Our best understanding of future exchanges between the atmospheric water cycle and the land surface for regional to global scales is likely to be gained through analysis of numerical simulations (Decker et al., 2015;Evans and McCabe, 2010;Harding and Snyder, 2012;Wei et al., 2012).Consequently, continual improvement of available models is essential, but this is contingent upon ongoing validation and evaluation of model performance over a broad range of landscapes and climate types (McCabe et al., 2016).To do this effectively, a diversity of datasets that directly quantify processes represented within these models are required (McCabe et al., 2005).Unfortunately, datasets that directly measure land-atmosphere exchange at the process level are limited (Jana et al., 2016).
Published by Copernicus Publications on behalf of the European Geosciences Union.S. D. Parkes et al.: Response of water vapour D-excess to land-atmosphere interactions Water is composed of a number of stable isotopologues that have sufficient abundance to be measured in atmospheric water vapour ( 1 H 16  2 O, 1 H 2 H 16 O, 1 H 18 2 O and 1 H 17 2 O).Deviations of water isotope ratios are reported as where R is the isotope ratio ( 2 H / 1 H or 18 O / 16 O), and VSMOW (Vienna Standard Mean Ocean Water) is the international standard for reporting water isotope ratios, and these ratios have the potential to evaluate land-atmosphere exchange by discriminating processes based on their isotopic signature (Berkelhammer et al., 2013;Lee et al., 2009;Noone et al., 2013;Risi et al., 2013).Isotopic ratios of water vapour (δ 2 H and δ 18 O) can therefore provide information that is complimentary or even unobtainable when using conventional measurement techniques.
The utility of water isotope ratios for tracing sources of moisture derives from the characteristic equilibrium and kinetic isotopic fractionation that occurs when water undergoes a phase change, causing light water molecules to preferentially accumulate in the vapour phase.Soil moisture is typically enriched in heavy isotopes relative to the ocean (Gat, 1996), so water vapour derived from land surface evaporation is expected to have a different isotopic composition to moisture evaporated from the ocean.This has led to a number of studies using stable isotopes in precipitation to partition oceanic and land-derived sources (Froehlich et al., 2008;Tian et al., 2001).However, land-atmosphere exchange is not restricted to periods of precipitation, and there are relatively few studies examining the role of land-atmosphere exchange on ambient humidity budgets using stable isotope observations of vapour (e.g.Aemisegger et al., 2014;Risi et al., 2013).
In addition to the source of moisture, the magnitude of isotopic fractionation that occurs when water evaporates is related to the liquid surface temperature and humidity gradient between the evaporating surface and atmosphere (Craig and Gordon, 1965).The temperature-dependent equilibrium exchange between liquid and vapour is the largest contributor to isotopic fractionation during evaporation, with the fractionation for δ 2 H approximately a factor of 8 greater than δ 18 O.The effect of kinetic fractionation associated with moisture diffusing from the thin laminar layer of vapour in equilibrium with the water surface to the turbulent atmosphere above is influenced by the relative humidity of the atmosphere and wind speed (Merlivat and Jouzel, 1979).The kinetic fractionation factors for δ 2 H and δ 18 O are similar, causing the ratio of δ 2 H to δ 18 O in the evaporating vapour to decrease as kinetic effects increase with decreasing relative humidity.This phenomenon has been observed for evaporative conditions over the Mediterranean Sea (Gat et al., 2003;Pfahl and Wernli, 2009) and the Great Lakes in northern USA (Gat et al., 1994;Vallet-Coulomb et al., 2008).
The D-excess (D-excess = δ 2 H − 8 × δ 18 O) parameter (Dansgaard, 1964) quantifies the non-equilibrium isotopic fractionation.A reproducible relationship between the Dexcess and relative humidity near the ocean surface has been observed across a wide range of locations (Kurita, 2011;Pfahl and Wernli, 2008;Steen-Larsen et al., 2015;Uemura et al., 2008).Therefore, it has been suggested that for precipitation, D-excess is a good tracer of sea surface evaporative conditions (Masson-Delmotte et al., 2005;Merlivat and Jouzel, 1979).However, this view has recently been challenged due to the role local and regional scale land-atmosphere coupling has in modifying the D-excess of atmospheric humidity over diurnal (Lai and Ehleringer, 2011;Simonin et al., 2014;Welp et al., 2012;Zhao et al., 2014) and synoptic timescales (Aemisegger et al., 2014).As evidence for the role ET plays in modifying the D-excess of water vapour (d v ), a diurnal cycle of d v near the land surface across a range of land surface types has been observed (Berkelhammer et al., 2013;Simonin et al., 2014;Welp et al., 2012).The diurnal cycle shows higher values during the day, which has been proposed to be driven by entrainment (Lai and Ehleringer, 2011;Welp et al., 2012), local evapotranspiration sources (Simonin et al., 2014;Zhao et al., 2014) and meteorological conditions affecting the D-excess of the evaporative fluxes (d ET ) (Welp et al., 2012;Zhao et al., 2014), coupled with low nocturnal values resulting from equilibrium exchange between liquid and vapour pools (Simonin et al., 2014) and dewfall (Berkelhammer et al., 2013).For synoptic scales, Aemisegger et al. (2014) showed that moisture recycling from the land surface had a significant impact on d v for in situ measurements in Switzerland.These studies have largely relied on isotopic models to assess the contribution of ET fluxes, but a lack of d ET measurements makes it difficult to draw robust conclusions.
The evidence provided by these studies suggest d v is a tracer of moisture recycling both on diurnal and synoptic timescales, and is influenced by the dynamics of surface moisture budgets in the atmospheric boundary layer (ABL).However, as noted by Welp et al. (2012), ET and entrainment fluxes both increase as the ABL grows through the previousdays residual layer, which can make interpreting the role of local moisture recycling on d v difficult.To overcome this, Simonin et al. (2014) used a trajectory model to simulate the D-excess of vapour evaporated over the ocean.As the d v was greater than the modelled oceanic moisture source, it was assumed that high daytime values were supported by local ET fluxes.Zhao et al. (2014) suggested that since, on cloudy days, no diurnal cycle was observed for the d v , ET fluxes therefore played a dominant role.Whilst these studies provide compelling evidence for the role of ET driving the diurnal cycle of d v , no measurements of d ET were made.To date the only measurements of d ET have been presented by Huang and Wen (2014) over a maize crop in Northwest China.Interestingly, their direct measurements conflicted with previous interpretations and showed that the d ET invoked a negative The aim of this work is to provide much needed d ET measurements to investigate how ET fluxes modulate the d v diurnal cycle.To do this, chamber-based measurements of the ET flux isotopic compositions were combined with in situ measurements of water vapour isotope ratios, meteorological and radon concentration observations.The data were collected in a region of the semi-arid Murray-Darling basin in southeastern Australia.These data represent the first such collection of the δ 2 H, δ 18 O and D-excess in water vapour from this region of Australia.The augmentation of the chamber-based measurements with in situ observations provides a framework to directly assess the role local ET fluxes have on ambient vapour D-excess.

Site description
During the austral autumn of 2011, a field campaign covering the period 27 April to 11 May was conducted at the Baldry Hydrological Observatory (BHO) (−32.87, 148.54, 460 m a.s.l.-above sea level) located in the central-west of New South Wales, Australia (Fig. 1).The climate of the region is characterised as semi-arid with no clear wet season, a mean annual rainfall of 600 mm, and a mean annual temperature of 24.2 • C (source Australian Bureau of Meteorology, 2016; http://www.bom.gov.au/).The BHO grassland eddy covariance flux tower was the central site of measurements and was located in a natural grassland paddock of dimensions approximately 900 m (north-south) by 300 m (westeast), with a gentle slope decreasing in elevation by approximately 20 m from south-east to north-west.The flux tower was located 650 m from the road to the south and 200 m from a reforested paddock to the west.The forest site to the west and south-west was reforested in 2001 with Eucalyptus camaldulensis, Eucalyptus crebra and Corymbia maculata.At the time of the campaign these trees were approximately 10 m tall.All other adjacent paddocks and most of the surrounding region had similar surface characteristics to the grassland measurement site.

In situ water vapour calibration and sampling
In situ water vapour isotope ratios were monitored using a Wavelength Scanning Cavity Ring Down Spectrometer (WS-CRDS L115-I, Picarro Inc., Sunnyvale, CA, USA), while flux chambers were interfaced to an Off-Axis Integrated Cavity Output Spectrometer (OA-ICOS, DLT100, Los Gatos Research -LGR, Mountain View, CA, USA) to determine the isotopic composition of ET fluxes.Using an automated continuous flow calibration system (built in-house), we simultaneously determined calibration coefficients for both analysers.Calibration experiments were designed to determine the water vapour mixing ratio cross-sensitivity of isotope ratios and linearity of the δ 2 H and δ 18 O measurements.More details on the calibration procedure are found in the Supplement.Due to logistical constraints, the calibration system was not transported into the field, so corrections were determined by compositing multiple calibration experiments run before and after the campaign.During the campaign, a secondary portable calibration system was employed to monitor time-dependent drift of the Picarro analyser (CTC HTC Pal liquid autosampler; LEAP Technologies, Carrboro, NC, USA).Two standards spanning expected water vapour δ 2 H (−49.1 and −221.9 ‰) and δ 18 O (−9.17 and −27.57‰) ranges were injected at approximately 18 mmol mol −1 on three occasions during the campaign.
The uncertainty of measurements from both isotopic analysers was estimated by applying mixing ratio crosssensitivity and linearity corrections to all calibration measurements collected prior to, during and after the campaign.For the Picarro instrument, measurement uncertainty was 0.8, 0.2 and 1.9 ‰ for δ 2 H v , δ 18 O v and d v , respectively.No calibrations were performed for the LGR in field, so the measurement uncertainty was estimated by compositing calibration measurements made before and after the campaign, which were 0.9, 0.4 and 3.3 ‰ for δ 2 H, δ 18 O and d v , respectively.
Although no calibration experiments were run on the LGR during the campaign, simultaneous in situ measurements were made with the Picarro when chamber measurements were not operated.During the day, average differences were −0.06 (±2.0), 0.13 (±0.5) and 0.4 (±3.3) ‰ for δ 2 H, δ 18 O and d v , respectively.A comparison of the analysers is shown in Fig. S1 in the Supplement.At night, while the Picarro was able to maintain a steady cavity and optical housing temperature, the LGR cavity temperature dropped by up to 8 • C. In response to the drop in cavity temperature, nighttime LGR measurements of δ 18 O and d v , and to a lesser extent the δ 2 H, were physically unrealistic and discarded from subsequent analyses.Chamber measurements were therefore restricted to between 09:00 LST (when the LGR cavity temperature had stabilised and in situ measurements were again in agreement with the Picarro) and 17:00 LST (before LGR cavity temperatures began dropping).
A schematic diagram illustrating the sampling design for water vapour is shown in Fig. 2. Half-hourly vertical pro- files of humidity and isotopes were sampled by drawing air to the in situ analyser through 10 mm (outside diameter) PTFE tubing, located at five heights on a 7.5 m tower (0.5, 1, 2, 5 and 7.5 m a.g.l.).The instrument was interfaced to a fiveinlet manifold that enabled sequential sampling of the different heights.Approximately 20 m of tubing was required to connect the tower inlet to the analyser.A vacuum pump (MV 2 NT, Vacuubrand, Wertheim, Germany) was used to draw air through all inlets to the analyser at a flow rate of 10 L min −1 , with the Picarro bleeding off the 0.03 L min −1 through its measurement cavity.To avoid condensation, sample tubes and intakes were wrapped in 15 W m −1 heat tape, insulated by Thermobreak pipe and placed inside 100 mm PVC pipe.The sample tube temperature was controlled using a resistance thermometer detector coupled to a CAL3300 temperature controller (CAL controls Ltd., Grayslake, IL, USA).The inlets at each height were constructed from inverted funnels with mesh filters.In this study we present block hourly averages of all measurements collected at all heights.

Flux chambers
To separate the isotopic signatures of the ET flux components, flux chambers were deployed on both bare soil and vegetated plots to determine the isotopic signature of the evaporative fluxes.An open chamber was designed with a high volume-to-footprint ratio to avoid the chamber mixing ratio rapidly reaching the dew point temperature (causing condensation) and to minimise impacts on the evaporation environment.A schematic of the chamber design is shown in Fig. 3. Four flanged metal collars were inserted ∼ 10 cm into the soil column 2 days before the beginning of the campaign.While this was a short settling time for chamber bases, shallow roots of grass cover within the chamber were largely unaffected.All vegetation was removed from bare soil plots when the metal collars were inserted into the soil.A single chamber cover was constructed out of 4 mm G-UVT Plexiglas (Image Plastics, Padstow, Aus-tralia), selected for its higher transmittance of UV (ultraviolet) and blue light.The dimensions of the chamber were 0.1 × 0.1 × 0.8 m (width × length × height), with the inlets and outlets at 0.1 and 0.7 m above the surface, respectively.All sampling tubes were 10 mm PTFE.The inlet to the chamber was connected to tubing that drew in air from 1.5 m above the ground surface.The outlet was connected to a flowmeter (VFA-25, Dwyers, Michigan City, IN, USA) that regulated the airflow at 10 L min −1 and was driven by a twostage diaphragm pump.A T-piece was connected to the LGR, which bled off approximately 0.8 L min −1 .All tubing between the chamber and the analyser were wrapped in heating tape (15 W m −2 ) and foam insulation.High flow rates were used to combat memory effects modifying the isotopic composition of the vapour within the chamber.Analysis of chamber measurements was conducted on 2-5 min of data, so 2.5-6.25 chamber volumes were exchanged.
To monitor the internal chamber environment, an air temperature and humidity probe (HMP155, Vaisala, Vantaa, Finland) was mounted inside the chamber.To monitor the attenuation of the incoming radiation by the chamber, the photosynthetic flux density was measured (LI-190R, Licor, Lincoln, NE, USA) inside and outside the chamber; 10 s averages of the temperature, relative humidity and photosynthetic flux density were stored in a datalogger (CR1000, Campbell Scientific, Logan, UT, USA).In the Supplement we use these ancillary measurements to assess the impact of observed changes in the chamber environment on the isotopic composition of ET fluxes.The largest contributor to uncertainty caused by changing the evaporative environment was the temperature, although these affects were small compared to the overall variability of the chamber-derived ET isotopic compositions.

Isotopic composition of ET flux from chamber measurements
Mass balance or Keeling-mixing (Keeling, 1958;Wang et al., 2013a) models have been applied to determine the isotopic composition of ET fluxes from chamber measurements (Lu et al., 2017;Wang et al., 2013b).The focus of this work was not to evaluate chamber measurement techniques.Considering that it has been shown that Keeling and mass balance methods give very similar results (Lu et al., 2017;Wang et al., 2013b) we focus on using the Keeling-mixing model, given by where q BG is the water vapour mixing ratio entering the chamber through the inlet and δ BG its isotopic composition, q chamber is the mixing ratio in the chamber and δ ET is the isotopic composition of the ET flux.The δ ET is determined from the intercept of δ chamber against 1/q chamber .A key assumption of the Keeling method is that the isotopic composition of the background vapour and the evaporation flux remain constant during the chamber measurements.For chamber measurements longer than 5 min, non-linear Keeling plots were commonly observed, indicating a change in isotopic composition of one of the sources of vapour.We therefore restricted the Keeling analysis to a maximum of 5 min after an increase in the concentration was observed by the analyser.
Ensuring the linearity of Keeling plots also ensured that the influence of memory effects was minimised.Memory effects would constitute an additional moisture source, violating the two-source assumption of the Keeling methods and reducing Keeling plot linearity.The analysis was also restricted to periods where the H 2 O mixing ratio was increasing, so analysis was generally performed on 2-5 min of data.In addition, only chamber measurements where the correlation between δ chamber and 1/q chamber was significant (p < 0.001) were included in this analysis.A few chamber measurements where obvious non-linearity or very small changes in q chamber occurred were also subjectively removed.Of a total of 105 chamber measurements made from the 4 vegetation plots during the campaign, 99 measurements of the δ 2 H ET , and 97 measurements of δ 18 O ET and d ET were retained.For the bare soil plots, 84 of the 86 chamber measurements were retained for the δ 2 H ET , and 77 of the δ 18 O ET and d ET .The eight plots were sampled 2 to 4 times each day on all days except the first two days of the campaign, and 2 and 5 May.
Sampling was restricted to between 09:00 and 17:00 LST (local solar time) as the large temperature dependence of the LGR at low ambient temperatures limited the accuracy of the chamber measurements.
Results from vegetated plots were used to determine ET flux isotopic compositions and determine how ET influences d v .The bare soil plots were used to determine the isotopic composition of soil evaporation fluxes and to provide an estimate of the isotopic composition of water at the evaporation front.The isotopic composition of the water at the evaporation front (δ L ) was determined by rearranging the Craig and Gordon model: where the isotopic composition of the evaporation flux (δ E ) is taken from the bare soil chamber measurements, relative humidity (RH) normalised to the surface temperature determined from infrared surface temperature measurements (Sect.2.3), and the ambient vapour isotope composition (δ A ) determined from Picarro in situ measurements.Equilibrium fractionation and enrichment factors (α, ε = (α − 1) ‰) were calculated from the surface temperature measurements using the equations of Horita and Wesolowski (1994), while the kinetic enrichment factors (ε k ) were determined as in Gat (1996), but using the parameterisation of the exponent of the diffusion coefficients described by Mathieu and Bariac (1996) and the diffusion coefficients determined by Merlivat (1978).

Isoforcing of ET
The isotopic composition of the near-surface atmospheric water vapour is modified by surface ET fluxes.The impact of ET fluxes on surface vapour isotopes varies over diurnal timescales with the strength of vertical mixing in the ABL or over synoptic timescales as background moisture conditions change.The magnitude and isotopic composition of the ET flux as well as the amount of water vapour in the atmosphere also have an influence.The ET isoforcing (I ET ) represents a useful quantity to study the influence of ET fluxes on the surface vapour and is defined as where F ET is the ET flux in mol m −2 s −1 , H 2 O is the ambient mixing ratio in mol-air mol-H 2 O −1 measured by the local meteorological tower, and δ ET and δ A are the isotopic compositions of the evaporation flux and ambient water vapour, respectively (Lee et al., 2009).
For each chamber measurement, a surface isoforcing was calculated for δ 2 H, δ 18 O and D-excess from the determined ET isotopic composition, as well as the hourly averaged ET flux, mixing ratio and δ A values.The importance of surface fluxes modifying surface vapour isotope composition was investigated for diurnal and synoptic timescales.

Plant and soil sampling
Grass samples were collected three times a day for the duration of the campaign.They were sampled randomly within 100 m of the instrumentation.Each sample consisted of approximately 10 grass leaves, which were placed in 12 mL Exetainer vials (Labco, Ceredigion, UK).The grass samples were assumed to represent bulk leaf water.Soil samples were collected every 2 days throughout the campaign by sampling from the top 5 cm of the soil column.They were collected in 50 mL glass bottles.Soil and plant samples were stored in a fridge (4 • C), before using the distillation method of West et al. (2006) to extract liquid water samples that were analysed on a Delta V Advantage Isotope Ratio Mass Spectrometer (Thermo Fisher Scientific Corporation, Massachusetts, USA).For δ 2 H analysis, water samples were introduced into a H-device containing a chromium reactor, while for the δ 18 O analysis, water samples were equilibrated with CO 2 on a Gas Bench II chromatography column (Thermo Fisher Scientific Corporation, Massachusetts, USA) before being transferred to the Isotope Ratio Mass Spectrometer (IRMS) for analysis.

ET fluxes and meteorological measurements
To measure ET fluxes, an eddy covariance system comprising a Campbell Scientific 3-D sonic anemometer (CSAT-3, Campbell Scientific, Logan, UT, USA) along with a Licor 7500 (Li-7500, Licor Biosciences, Lincoln, NB, USA) analyser was installed at an elevation of 2.5 m.The system was located approximately 10 m from the stable isotope observation tower and sampled at 10 Hz, with flux averages output at 30 min intervals.The ET fluxes from the eddy covariance tower are used to quantify the isoforcing of ET on the overlying atmosphere.
A meteorological tower was co-located with the eddy covariance system, providing complementary data to aid in the interpretation of measurements.The tower comprised a Kipp and Zonen CNR4 radiometer, Apogee infrared surface temperature, RIMCO rain gauge, Vaisala HMP75C temperature and humidity probe, RM Young wind sentry (wind speed and direction), Huskeflux ground heat flux plate and Vaisala BaroCap barometric pressure sensor.Both meteorological tower data and eddy covariance data were inspected visually to detect and remove spikes.The low-frequency eddy covariance data (30 min resolution) were corrected for coordinate rotation (Finnigan et al., 2003) and density effects (Leuning, 2007) using the PyQC software tool (available from http://code.google.com/p/eddy).

Radon-222 measurements
The naturally occurring radioactive gas radon ( 222 Rn) is predominantly of terrestrial origin and its only atmospheric sink is radioactive decay (Zahorowski et al., 2004).The surface flux density of radon is relatively constant in space and time, and since the half-life is much greater than ABLmixing timescales, it is an ideal tracer of vertical mixing strength within the ABL (Chambers et al., 2015c;Griffiths et al., 2013;Williams et al., 2010).Hourly radon concentrations were measured by an Alpha Guard (Saphymo GmbH, Frankfurt, Germany) placed in a ∼ 20 L enclosure.The enclosure was purged at ∼ 15 L min −1 with a vacuum pump (2107 Series, Thomas, Wisconsin, USA) that sampled from a height of 2 m through 10 mm (outside diameter) PTFE tubing.Radon measurements were used to aid the interpretation of the diurnal variations in vertical mixing (see Griffiths et al., 2013).

Meteorological observations
The last rain event was 10 days prior to the campaign, after which clear skies saw the soil dry to a moisture content close to minimum values observed for the site (Fig. 4).The meteorological and radon measurements shown in Fig. 5 indicate the 2-week campaign was conducted under predominantly calm meteorological conditions.In the middle of the campaign (2 May), a cold front moved across south-eastern Australia, producing cloudy conditions and 1.4 mm of precipitation at the site.No change in soil moisture was observed over the 0-10 cm soil layer following the rain event.
Wind directions were variable during the campaign (see Fig. S2a and b). Figure S3 shows that from 27 to 30 April,  dominant daytime wind directions were mainly from the east.After 3 May winds were from the south, except on 7 and 8 May when the wind was from the west and had a fetch from the adjacent forest.At other times the fetch did not overlap the forested site.Daily maximum temperatures on clear days ranged from 16 to 23 • C, whilst nighttime minimum temperatures fell to between 8 and −4 • C. From 7 May onwards nocturnal temperatures fell below zero.On clear nights the surface temperature fell below dew point temperature, indicating dewfall.Apart from the night of 27-28 April and the cloudy nights between 1 and 3 May, the surface tempera-ture fell below dew point temperature and dew or frost was observed in the morning, although heavier from 7 May onwards.
Radon concentrations were low during the day, when the convective boundary layer reached its maximum height, and high at night, when radon emissions were confined within the shallow nocturnal boundary layer.The accumulation at night was variable indicating a varying degree of nocturnal stability, mixing depth and occurrence of transient mixing events (Griffiths et al., 2013).There was general agreement between high nocturnal radon concentrations and low wind speeds, but no direct relationship.The lack of a direct relationship indicates that radon can provide additional information about nocturnal mixing and surface exchange that compliments standard meteorological measurements (Chambers et al., 2015a, b;Williams et al., 2013).
ET fluxes were in general quite low, reflecting the low soil moisture content.The ET flux did show a marked increase the day after the small rain event on 2 May and noticeably smaller fluxes were observed after the first night frost was observed.The health of the grass visibly deteriorated from 7 May, coinciding with frost formation.ET flux isotopic compositions from vegetated chambers were enriched relative to vapour and distributed to the right of the MWL (slope = 3.2).Similar isotopic compositions were measured from bare soil and vegetated chambers.Mean and standard deviations (1σ ) for vegetated and soil chambers were −47.1 (±13) and −50.2 (±11) for δ 2 H, −5.03 (±3.8) and −6.3 (±2.7) ‰ for δ 18 O, and −6.3 (±23) and −0.12 (±15) ‰ for D-excess, respectively.The similar ET isotopic composition from bare soil and vegetated chambers could indicate soil evaporation was the dominant process contributing to total ET.However, as pointed out in the discussion (Sect.4.3), convergence of soil evaporation and transpiration isotope compositions as the soil evaporation source becomes progressively enriched (and D-excess lower), probably makes it difficult to identify the dominant process from these observations.Nevertheless, since the last significant rain event prior to the campaign, progressive reduction of Dexcess of moisture at the evaporation front, and to a lesser extent in the 0-5 cm layer, caused low D-excess of overall ET fluxes compared to d v .This would indicate that ET imposes a negative forcing on d v .
Temporally, a clear trend was not observed for ET isotopic compositions over the measured portion of the diurnal cycle or over the campaign.No measurements were made at night or during the rapidly changing conditions of the morning transition, which may have led to our data missing some observed changes in ET isotope compositions.

In situ water vapour isotopes and ET isoforcing
Observed water vapour mixing ratios and stable isotope compositions are shown in Fig. 7. δ 2 H and δ 18 O variability was similar, reflecting changes in both the synoptic and local meteorology.Prior to the rain event (2 May), relatively moist conditions (higher H 2 O mixing ratios) were observed as air was transported from the warmer ocean off the east coast of Australia (see wind direction in Fig. S3).After 5 May, trans- port of air masses from the colder sea surface south of continental Australia brought drier conditions to the site (lower H 2 O mixing ratios).Moisture source regions were confirmed by backward air trajectories calculated using the Stochastic Time-Inverted Lagrangian Transport Model (STILT; Lin et al., 2003, not shown).These two time periods are hereinafter referred as the "wet period" (before 2 May) and the "dry period" (after 5 May).The wet period coincided with more enriched isotopes and less diurnal variability.In the later part of the campaign, a reproducible diurnal cycle for δ 2 H and δ 18 O was observed (see Fig. 8 for diurnal composites), presenting a sharp increase at sunrise before decreasing from mid-morning (when vertical mixing increased) until the next sunrise.These observations emphasise the complex relationship between stable isotope observations in water vapour and both local-and synoptic-scale meteorology.
The d v dataset showed a robust diurnal cycle of high values during the day and low values at night, consistent with what has been observed across a growing number of locations (Bastrikov et al., 2014;Berkelhammer et al., 2013;Simonin et al., 2014;Welp et al., 2012;Zhao et al., 2014).Wet period daytime d v values were on average lower than those observed for the dry period.Nocturnal d v was consistently lower during the night, but variable from night to night and across individual nights, with no clear difference observed between wet and dry periods.Contrasting daytime measurements of wet and dry periods indicate a role of large-scale processes, whilst the lack of contrast for nocturnal observations shows the importance of local processes.
The I ET was always positive for δ 2 H and δ 18 O and mostly negative for D-excess, but showed large variability across individual days (Fig. 7).I ET was most sensitive to the magnitude of the ET fluxes, producing the greatest forcing on ambient vapour in the middle of the day.The I ET time series did not correspond to temporal variability of vapour δ 2 H, δ 18 O or D-excess.δ 2 H and δ 18 O often decreased during the day while I ET was positive.Whilst the high d v values observed during the day were associated with negative isoforcing, over the course of the campaign the highest daytime d v values did not correspond to the least negative I ET .These observations illustrate that local ET fluxes were not overly important for day-to-day and diurnal d v trends.
The level of agreement between the analysers presented some uncertainty in calculating the D-excess isoforcing.The sign of the isoforcing is dependent on the difference between d v and d ET (Eq.8).In some cases this difference was small and within the range of agreement between the two analysers.While this caused problems for accurate calculation of the absolute values of D-excess isoforcing, for all chamber measurements passing our quality control requirements, D- excess decreased with concentration.This indicates that for all measurements, the D-excess isoforcing was negative.

Relationship between water vapour isotopes and local meteorology
The relationships between local meteorological variables and water vapour isotopes were examined to interpret the role of local processes (Table 1).Regression statistics are shown for both hourly observations and average daytime values (between 11:00 and 16:00 LST).Selecting daytime measurements removes variability associated with transition between the stable nocturnal and daytime convective boundary layer, as well as nocturnal periods when local surface equilibrium exchange and dewfall affect vapour isotope compositions.Correlations determined using only measurements in the middle of the day therefore provide a better indicator of how local meteorology and ET isotopic composition modified ambient water vapour isotope ratios from day to day.Correlations calculated with hourly data were weak for δ 18 O and δ 2 H.Only correlations with air temperature (R 2 = 0.24 and 0.04, respectively) and mixing ratio (R 2 = 0.2 for both isotopes) were significant, and δ 2 H also showed a weak correlation with RH (R 2 = 0.09).For daytime observations, only δ 2 H showed a significant correlation with daytime I ET (R 2 = 0.45, p < 0.05), but the slope was negative in contrast to positive isoforcing.The weak relationships with local meteorology indicate the importance of larger-scale precipitation processes and atmospheric mixing occurring as moisture was transported to the site.
As the diurnal cycle for d v was consistent with growth and decay of the ABL, strong relationships were observed with air temperature and RH for the hourly observations.While the local air temperature and RH could modify d ET on diurnal timescales and in turn local d v , the chamber measurements showed relatively constant d ET .These correlations therefore result from the coincident diurnal variation of the d v , RH and air temperature.
Daytime average d v showed significant correlations with the air temperature, RH, ET flux and mixing ratio.The relationship with ET fluxes was weak (R 2 = 0.3) and positive, but as negative D-excess isoforcing was observed, a negative relationship would be expected.Likewise, the slope between air temperature and d v was negative, counter to what theory would predict for local or remote moisture sources.The strongest relationship was observed with daytime RH (R 2 = 0.74), which had a negative slope (−0.52 ‰ % −1 ) consistent with an inverse relationship between d v and RH for a large unchanging evaporation source.The strong relationship of d v with the daytime RH could indicate an important role for the evaporation conditions at remote moisture sources, as is discussed below in Sect.4.2.

Diurnal variability of vapour isotopes
Diurnal composites were divided into dry and wet periods and are shown in Fig. 8.At sunrise (approximately 06:30 LST) surface heating initiated vertical mixing, shown by the radon concentration maximum, causing temperature and ET flux to increase and RH to decrease.Weak vertical mixing immediately after sunrise and injection of ET into the still shallow surface layer caused near-surface humidity to increase.Similarly, for δ 2 H and δ 18 O, the observed spike immediately after sunrise was likely caused by ET fluxes with an enriched heavy isotope composition, possibly from re-evaporation of dewfall.During the dry period, vapour δ 2 H 12:00 18:00 00:00 06:00 and δ 18 O increased more steeply, caused by the combination of a shallower surface layer observed at the start of the morning transition, shown by higher radon concentrations, and more dewfall on the surface providing a greater initial evaporation source.Rapidly decreasing radon concentrations during this morning ABL transition caused by vigorous vertical mixing entraining air from the residual layer of the previous day diluted ET fluxes and caused the δ 2 H, δ 18 O and the mixing ratio to first stabilise and then decrease.ET fluxes rapidly increased as the ABL grew, but were not large enough to offset the dilution by dry air being mixed down from above or stop depletion of surface δ 2 H and δ 18 O.The d v also increased after sunrise, but aligned more closely to when strong vertical mixing commenced, as shown by the close agreement with radon concentrations.The Dexcess isoforcing was negative, evidence that d v increased from encroachment mixing as the new mixed layer grew in depth and not ET fluxes.The dry period showed a greater increase in d v during the morning transition, likely the result of higher d v in background water vapour and greater differences between the d v of the residual and nocturnal layer.
In the afternoon, d v decreased back to values similar to those observed prior to sunrise, with a simultaneous decrease in solar insolation, ET and a decay of convective mixing.Radon shows how reduction in vertical mixing causes the concentration of tracers emitted from the surface to increase.So while ET decreased, small fluxes were still observed well after 18:00 LST, when large changes in δ 18 O and d v were observed.Hence, as the I ET was positive and negative for δ 18 O and D-excess, respectively, small ET fluxes into a poorly mixed surface layer may have led to observed changes.
During the night, dewfall caused δ 2 H and δ 18 O to decrease as heavy isotopes were removed in condensation, especially during the dry period when greater surface cooling was observed.However, dew formation is an equilibrium processes and therefore did not affect d v .Composites of dry and wet period nocturnal d v measurements do not show clear nocturnal trends, but individual nights showed considerable variability.A regression of nocturnal d v with radon concentrations produced a significant negative relationship (p < 0.001, R 2 = 0.31), indicating that atmospheric stability has some control over nocturnal d v .High radon is associated with the most stable atmospheres, enhancing the effect of surface exchange in the early evening.Low radon, on the other hand, is associated with periods of atmospheric turbulence in which moisture above the nocturnal inversion with a high d v is mixed down towards the surface.
As has been previously observed (Steen-Larsen et al., 2013;Welp et al., 2012) and predicted by isotopic models (Gat, 1996), our observations showed water vapour δ 2 H and δ 18 O are controlled by different atmospheric and hydrological processes than d v .The diurnal cycle was the dominant mode of variability for d v , consistent with previous studies for a range of ecosystems (Simonin et al., 2014;Welp et al., 2012;Zhao et al., 2014).However, results also showed that D-excess variability was controlled by local meteorological conditions and surface exchange at night, ABL growth and decay during transitional periods between the nocturnal and convective ABL, and larger-scale processes in the middle of the day.

Entrainment and the d v diurnal cycle
The radon measurements showed that when the depth of the ABL was rapidly changing through the morning and evening transitions, entrainment from the residual layer and ET fluxes into a rapidly decaying convective boundary layer caused the observed d v diurnal cycle.Between these transitions, when mixing extends to the capping inversion, entrainment fluxes introduce an additional moisture source from the free troposphere that could modify surface vapour isotopic compositions.Air above the ABL is drier and moisture is more depleted than at the surface.Drying and depleting trends for water vapour, δ 2 H and δ 18 O throughout the day, particularly during the dry period (Fig. 8), indicate an important role for entrainment from the free troposphere.Whether this moisture flux impacts on d v is less clear, as it remained reasonably stable once a maximum was reached after the morning transition period.The sign of the isoforcing of moisture entrained from the free troposphere is uncertain, as few free tropospheric d v measurements exist (He and Smith, 1999;Samuels-Crow et al., 2014).Nevertheless, d v values did not show a clear trend until vertical mixing began decaying later in the afternoon; therefore, free tropospheric d v probably has a similar value to moisture already residing in the ABL.

Remote hydrometeorological processes
While the main focus of this study was to examine the role of local land-atmosphere exchange for the diurnal variability of d v , the synoptic context of measurements warrants further examination for comparison against previous studies of d v diurnal cycles.The slope between daytime RH and d v (−0.52 ‰ % −1 , Table 1) was similar to those determined for measurements over the Mediterranean Sea and different ocean basins (between −0.43 and −0.53 ‰ % −1 ) (Kurita, 2011;Pfahl and Wernli, 2008;Steen-Larsen et al., 2014, 2015;Uemura et al., 2008).Aemisegger et al. (2014) showed that this robust relationship is not restricted to coastal locations or measurements over the ocean surface.Using a trajectory model to investigate continental moisture recy-cling in Europe, they found a similar relationship between d v and RH of remote moisture sources during the cold season (−0.57‰ % −1 ), but not for warm season observations.They concluded moisture recycling is weakest during winter, causing d v to retain the signature of the RH of oceanic moisture sources, while in summer moisture recycling increased and attenuated the relationship.Similarities with their winter data indicate that our daytime d v measurements were at least partly determined by RH at the oceanic moisture source.
Along an air masses back trajectory, entrainment fluxes from the free troposphere could be a major driver of daytime d v variability.Mixing of warm dry air down to the surface, presumably with a relatively high D-excess (He and Smith, 1999;Samuels-Crow et al., 2014), would give the same negative relationship between d v and RH observed here.However, for a strong relationship between d v and RH, there must be a dominant moisture source.For the fraction of entrained air in the ABL to cause the strong linear relationship, the D-excess of vapour and RH in both the ABL and free troposphere must be reasonably constant, as in a two-source mixing model.Considering the variability of synoptic-scale weather patterns observed (Sect.3.1), this seems unlikely.Thus, while we cannot definitively rule out the importance of entrainment along back trajectories, it seems more likely that the d v -RH relationship was derived from a large unchanging moisture source such as the ocean.
A practical application of the d v -RH relationship introduced by Aemisegger et al. (2014) was to determine the Dexcess of the liquid moisture source.Based on the closure assumption of Merlivat and Jouzel (1979), it was shown when RH is 100 %, d v is equal to the D-excess of the liquid moisture source.If no further kinetic fractionation or mixing of vapour with a different d v -RH occurred between the point of evaporation and measurement location, extrapolating the regression between d v and RH to 100 % RH gives an estimate of moisture source D-excess.For our measurements, a value of −8 ‰ was determined, remarkably similar to the D-excess determined for ocean water off the east coast of Australia by Xu et al. (2012) using a global ocean model.In contrast to recent literature (Simonin et al., 2014;Welp et al., 2012;Zhao et al., 2014), this suggests that although the common diurnal cycle was observed, daytime observations are potentially a tracer of RH at the oceanic moisture source, but it is likely restricted to periods when moisture recycling is weak.
Whilst we have shown a relationship between the RH and d v consistent with an oceanic vapour source, the consistency of the relationship over longer time periods is uncertain.Indeed, it may be the reason why we show a strong relationship, whereas the study of Welp et al. (2012) did not for six mid-latitude sites in China and the USA, where longer datasets were available.As pointed out earlier, lowers slopes and weaker relationships result from stronger moisture recycling, which indicates moisture recycling and soil moisture state may be the most important variable controlling the relationship between d v and RH.Here we present data from after an extended dry period, where the dominant moisture source is the ocean surrounding the Australian continent.So during wetter periods, an increase in the local and remote moisture recycling probably weaken the relationship between local d v and RH (Aemisegger et al., 2014).However, for locations such as semi-arid Australia where extended dry periods prevail, the relationship between d v and RH may be reasonably robust and prevail as a tracer of oceanic evaporative environments.

Controls of d ET
The chamber d ET measurements showed ET fluxes imposed a negative isoforcing on d v , in contrast to interpretations in previous studies investigating d v variability on diurnal timescales (Simonin et al., 2014;Zhao et al., 2014).However, it is expected that the sign and magnitude of the Dexcess isoforcing would vary both spatially and temporally, in particular with the soil moisture state.After a rain event, soil moisture D-excess would decrease following a pseudo-Rayleigh process (Barnes and Allison, 1988).Therefore, immediately after a rainfall event, d ET would be higher and probably impose a positive isoforcing.Here the negative d ET caused the d v to decrease rapidly as convective mixing shut down.When isoforcing is positive after a rain event, the diurnal cycle observed here and elsewhere may therefore not be observed.Although equilibration between liquid and vapour pools, as eluded to by Simonin et al. (2014), may still help maintain observed trends.As soil dries, a tipping point when the ET fluxes switch from positive to negative isoforcing will be observed.This has implications for studies attempting to use d v as a tracer of continental moisture recycling, as the large spatial variability of rainfall and the associated soil moisture state would lead to large spatial and temporal variability for d ET .Although, the strongest moisture recycling is expected for wet soils when d ET is higher, variability in d ET may still be important.
Relative magnitudes of evaporation and transpiration fluxes are important for d ET , as the two processes could have different D-excess values and could vary strongly between precipitation or irrigation events.The classical view of ET isotope fluxes is that transpiration has an isotopic composition closer to the source moisture than evaporation, and therefore a higher D-excess.However, greater fractionation of the evaporation source pool causes its D-excess value to decrease over time, causing the D-excess of the fluxes to converge.The impact of converging isotopic signatures of ET component fluxes on moisture recycling would depend on the land surface type, but would constitute an important variable influencing the D-excess of local and remote moisture recycling.Further studies investigating how ET partitioning and drying of soil moisture reservoirs following irrigation or precipitation events would lead to a better understanding of how moisture recycling influences the ambient d v on continental and local scales.

Conclusions
To determine how local ET fluxes modified water vapour D-excess, in situ observations were collected in a semi-arid region of south-eastern Australia.The diurnal cycle exhibited high values during the day and low values at night, reflected findings from previous studies.With chamber-based measurements of isotopic compositions in evaporative fluxes, it was shown that local ET fluxes exhibited a negative forcing on the ambient water vapour D-excess that could not explain the high daytime values.A strong negative relationship was observed between the locally measured relative humidity and vapour D-excess during the daytime, consistent with relationships observed for oceanic moisture sources.During the evening transition, collapse of the convective boundary layer and small ET fluxes with negative D-excess isoforcing were responsible for lowering the D-excess of water vapour near the surface.In addition, a negative nocturnal correlation between D-excess in water vapour and radon concentrations indicated transient nocturnal mixing events shifted the D-excess back towards the higher values observed during the day, with the most stable (least turbulent) nights producing the lowest D-excess values.In the morning, encroachment and entrainment of high D-excess air from above caused Dexcess of surface vapour to increase back to the synoptic values.
Overall, it was found that the magnitude of the D-excess diurnal cycle was controlled predominantly by interplay between synoptic forcing and local ABL processes and was modified further by nocturnal surface exchange processes and turbulent mixing.The low D-excess of the ET fluxes determined from flux chambers in this study illustrated that the impact of large-scale moisture recycling may be both spatially and temporally variable, depending on the soil moisture state.This has implications for studies using D-excess to investigate moisture recycling.

Data availability
For access to the data used in this paper contact Stephen Parkes by email (stephen.parkes@kaust.edu.sa).
The Supplement related to this article is available online at doi:10.5194/hess-21-533-2017-supplement.

Figure 1 .
Figure 1.(a) Location of the Baldry Hydrological Observatory, with the heavy black border outlining the extent of the Murray-Darling Basin, (b) location of the field site used for the campaign, illustrating the semi-arid grassland and adjacent reforested site.

Figure 2 .
Figure 2. Sampling system for the automated in situ collection and measurement of water vapour isotopes from the tower.

Figure 3 .
Figure 3. Chamber design used for determining the isotopic compositions of ET fluxes.
S. D.Parkes et al.:  Response of water vapour D-excess to land-atmosphere interactions

Figure 4 .Figure 5 .
Figure 4. Precipitation and 0-10 cm soil moisture for the month leading up to and including the field campaign.

3. 2
Relationship between δ 2 H and δ 18 O of the different water pools A summary of the isotopic composition of all observed and modelled water pools are presented in Fig.6.The local Meteoric Water Line (MWL)(Hughes and Crawford, 2013) is to the left of the global MWL(Craig, 1961), illustrating the characteristically high D-excess of precipitation in the region(Crawford et al., 2013).Ambient vapour observations aligned closely with the local MWL, but with a distribution that fell both to the left and right of the local MWL.Alignment between observations and the MWL show that equilibrium fractionation was the dominant process modifying δ 2 H and δ 18 O in water vapour, while non-equilibrium kinetic processes shift observations away from the MWL and are more easily observed for d v measurements.Plant and soil water pools were enriched relative to the vapour and distributed to the right of the MWL, indicating evaporative enrichment.Soil water isotopes at the evaporation front (δ L ) were very enriched and had lower D-excess values (50 ± 12, 31 ± 3.8 and −131 ± 22 ‰ for δ 2 H, δ 18 O and D-excess) relative to the average soil moisture between 0 and 5 cm (−15 ± 4.2, 2.6 ± 2.5 and −36 ± 17 ‰ for δ 2 H, δ 18 O and D-excess).Low D-excess and enriched isotopes indicated large evaporative enrichment under non-equilibrium conditions consistent with the δ 18 O soil profile measurements ofDubbert et al. (2013) and the δ 2 H profiles ofAllison et al. (1983).The uncertainty of modelled isotope values was the most sensitive to parameterisation of the Craig-Gordon model.Changing the diffusion coefficient exponent (n) had the greatest impact on modelled soil water (n = 0.66, 42.7 ± 12, 21.8 ± 3.8 and −130.8 ± 22 ‰).However, changing parameterisation did not change the conclusion that soil moisture at the evaporation front was heavily enriched with very low D-excess values.

Figure 6 .
Figure 6.Relationship between δ 2 H and δ 18 O for observed and modelled water pools.Linear regressions are shown for local and global meteoric water lines (MWL).Data from Hughes and Crawford (2013) are for monthly cumulative rainfall samples between 2005 and 2008.

Figure 7 .
Figure 7. Time series of hourly water vapour mixing ratio, isotopic composition and ET isoforcing (I ET ).

Figure 8 .
Figure 8. Data plotted by time of day and divided into dry and wet periods (see text in Sect.3.5).Diurnal composites are shown for dry (red) and wet (blue) periods.

Acknowledgements.
Stephen Parkes was supported by the Atmospheric Mixing and Pollution Transport (AMPT) project at the Australian Nuclear Science and Technology Organization (ANSTO) and the King Abdullah University of Science and Technology.The Baldry Hydrological Observatory field campaign was supported by Australian Research Council Discovery grants DP0987478 and DP120104718.Matthew McCabe acknowledges the support of the King Abdullah University of Science and www.hydrol-earth-syst-sci.net/21/533/2017/ Hydrol.Earth Syst.Sci., 21, 533-548, 2017

Table 1 .
Correlation between meteorological variables and the isotopic composition of water vapour.Values outside the brackets are statistics for the hourly observations.Inside the brackets are correlation statistics for average values calculated between 11:00 and 15:00 LST, hence representing activity during a convective boundary layer.Significant correlation are shown in bold; p < 0.001 for hourly observations and p < 0.05 for the daytime averages (due to the smaller number of points).
* Isoforcing correlations were calculated for simultaneous vapour and chamber measurements.Hourly averaged values were used for both.