The role of dew and radiation fog inputs in the local water cycling of a temperate grassland in Central Europe

. In a warmer climate, non-rainfall water (hereafter NRW) formed from dew and fog potentially plays an increasingly important role in temperate grassland ecosystems under the scarcity of precipitation over prolonged periods. Dew and radiation fog occur in combination during clear and calm nights, and both use ambient water vapor as a source. Research on the combined mechanisms involved in NRW inputs to ecosystems are rare, and the condensation of soil-diffusing vapor, as one of the NRW 10 input pathways for dew formation, has hardly been studied at all. The aim of this paper is thus to investigate the different NRW input pathways into a temperate Swiss grassland at Chamau during prolonged dry periods in summer 2018. We measured the isotopic compositions ( δ 18 O, δ 2 H, and d = δ 2 H – 8· δ 18 O) of both ambient water vapor and the NRW droplets on leaf surfaces combined with eddy covariance and meteorological measurements during one dew-only and two combined dew and radiation fog events. We employed a simple two end-member mixing model using δ 18 O and δ 2 H to split the dew input pathways from 15 different sources. Our results showed a decrease of 0.8–5.5 mmol mol -1 in volumetric water vapor mixing ratio and a decrease of 4.8–16.7‰ in ambient water vapor δ 2 H due to dew formation and radiation fog droplet deposition. A nighttime maximum in ambient water vapor δ 18 O (–15.5‰ to –14.3‰) and a 3.4–3.7‰ decrease in ambient water vapor d were observed for dew formation in unsaturated conditions. In conditions of slight super-saturation, a stronger decrease of ambient water vapor δ 18 O (0.3–1.5‰) and a minimum of ambient water vapor d (–6.0‰ to –4.7‰) were observed. The combined foliage NRW and 20 ambient water vapor δ 18 O and δ 2 H suggested two different input pathways: (1) condensation of ambient water vapor and (2) of soil-diffusing vapor. The latter contributed 9–42 % to the total foliage NRW. The dew and radiation fog potentially produced 0.06–0.39 mm night –1 NRW gain on foliage, which was comparable with 2.8 mm day -1 daytime evapotranspiration. The ambient water vapor d was correlated and anti-correlated with ambient temperature and ambient relative humidity respectively, suggesting an only minor influence of large-scale air advection and highlighted the dominant role of local moisture as a source 25 for ambient water vapor. Our results thus highlight the importance of NRW inputs to temperate grasslands during prolonged dry periods and reveal the complexity of the local water cycle in such conditions including different pathways of water deposition. diffusion decreased, whereas the relevance of atmospheric water vapor for dew formation increased. This atmospheric water vapor had a rather local isotopic signature, which suggests that large-scale moisture advection only has a minor influence in the nocturnal NRW gains during dew and radiation fog events. Our results thus highlight the importance of NRW inputs to

complementary to 2 H and 18 O. The d is often used as a tracer for the water vapor source of a given water pool in the water cycle (Aemisegger et al., 2014;Galewsky et al., 2016;Gat, 1996;Welp et al., 2012;Yakir and Sternberg, 2000;Yepez et al., 2003). For example, at the local scale, local evaporation is a vapor source with lower d, while the entrainment from free troposphere is a vapor source with higher d (Delattre et al., 2015;Parkes et al., 2017). The diurnal cycle of deuterium excess in a well-mixed convective boundary layer has been studied previously (e.g., Lai and Ehleringer (2011)), whereas relevant 85 processes affecting d in the NBL are much less well known, in particular for grasslands. Monteith (1957) identified two input pathways for dew formation: 1) the downward pathway through the condensation of ambient water vapor onto foliage, and 2) the upward pathway through the condensation of soil-diffusing vapor onto foliage. Soil vapor diffusion is driven by the temperature gradient between the soil and the atmosphere and between different depths of the soil (Monteith, 1957;Oke, 1970). The temperature gradient generally reaches a maximum at the soil-90 atmosphere interface (2-4 °C warmer than the adjacent air at 1-2.5 cm in height for short grass or foliage surface (Monteith, 1957;Oke, 1970). The diffusing soil vapor can therefore condense onto cooler foliage. After Monteith (1957) had quantified the downward and upward components of dew formation by absorbing the NRW on foliage with filter paper, research has rarely been focusing on distinguishing these two pathways of dew formation. Furthermore, Monteith (1957) distinguished the two pathways by collecting the NRW in separate nights when only one or the other of the two pathways was assumed to occur. 95 In Monteith (1957), the NRW condensing from soil-diffusing vapor was quantified in very calm nights with a 2 m wind speed (hereafter u2m) of less than 0.5 m s -1 , whereas the maximum NRW condensing from ambient water was assumed to occur in slightly windy nights with u2m in the range of 2-3 m s -1 . However, for clear calm nights with u2m between 0.5 and 2 m s -1 , condensation of ambient water vapor and soil-diffusing vapor can occur in combination, with NRW on the foliage being a mix from these two pathways. Stable water isotopes and the "Keeling-plot" approach (Dawson, 1998;Keeling, 1958;Phillips et 100 al., 2005) was used in this study to quantify the individual contributions of these two sources.
Our aim was to (1) investigate the isotopic fractionations during dew-only and dew-fog combined events; (2) estimate the potential gain of NRW from atmospheric vapor and from soil-diffusing vapor; and (3) assess the potential ecological relevance of NRW inputs to a temperate grassland ecosystem. We carried out three 24 h observation campaigns during summer 2018 using stable isotopes combined with EC and meteorological measurements to clarify the meteorological conditions and 105 isotope fractionations for dew and radiation fog formations, to split the dew components from ambient water vapor and soildiffusing vapor, and to explore the potential role of dew and radiation fog in temperate grasslands.

Study site and observation campaigns
The Chamau site (hereafter CH-CHA; 47°12′36.8″ N, 8°24′37.6″ E) is an intensively managed temperate grassland (4-6 cuts 110 per year) at 393 m a.s.l., located in a valley bottom in Switzerland. The EC and meteorological measurement station (Fig. A1 in Appendix A) have been operational since 2005. The precipitation at the CH-CHA site was 870 mm in 2018, which was 297 mm (about 25%) less than the multiyear average over 2006-2017 April to September in 2018, with respect to the corresponding monthly climatological values in the period 2006-2017, the monthly precipitation was on average 81 mm, which was averagely 49 mm (38%) less (Fig. 1a), and the monthly average temperature was on average 17.3 °C, which was 115 1.8 °C higher (Fig. 1b).
Three 24 h observation campaigns were carried out during expected dew/fog events on 25-26 July (event 1), 20-21 August (event 2), and 9-10 September (event 3) 2018. The time series were all recorded in CET (UTC+1). The corresponding consecutive no-rain periods were 23-27 July, 18-21 August, and 8-12 September 2018 respectively. Because of the extreme summer drought in 2018, no harvest of grassland was carried out during the three campaigns, but two harvests were carried 120 out 46 d before event 1 on 9 June 2018, and one day after event 3 on 10 September 2018 respectively. The leaf area index https://doi.org/10.5194/hess-2020-493 Preprint. Discussion started: 13 October 2020 c Author(s) 2020. CC BY 4.0 License.

Eddy covariance and meteorological data and calculations
The EC measurements at 20 Hz were setup at 2.4 m a.g.l. (see Zeeman et al. (2010) for more details), based on measurements 130 with a 3-D sonic anemometer (Gill R3,Gill Instruments Ltd.,Lymington,UK), and an open path Infrared Gas Analyzer (IRGA, Li-7500, Li-Cor, Lincoln, NE, USA). The EC measurements were processed to 30 min averages for evapotranspiration rate (mm h -1 ), horizontal wind speed (hereafter u2m, in m s -1 ), H2O flux (hereafter FH2O, in mmol m -2 s -1 ; minus value means downward flux, whilst positive value means upward flux), atmospheric specific humidity (hereafter qa, in g kg -1 ), and dew point temperature (hereafter Td, in °C) (Buck, 1981;Campbell and Norman, 1998). The meteorological measurements at 0.1 135 Hz for air temperature (hereafter Ta, in °C), relative humidity (hereafter RH, in %), and long-wave outgoing radiation (hereafter LWout, in W m -2 ) were setup at 2.0 m a.g.l. (see Zeeman et al. (2010) and Fuchs et al. (2018) for more details). The horizontal visibility (in km) was measured every 10 s by a fog sensor (MiniOFS, Optical Sensors Inc., Goteborg, Sweden) and a present weather detector (PWD10, Vaisala Oyj, Helsinki, Finland). The meteorological measurements were processed to 30 min averages for Ta, RH, and LWout, and to 1 min averages for visibility. 140 The vegetation surface temperature (T0, in °C) was determined after Stefan-Boltzmann's law (Stull, 1988) as: where an emissivity (hereafter ε) of 0.98 was used to calculate temperatures for wet leaf surfaces (hereafter index w; T0 = T0w), and a value of 0.96 was used for dry leaf surfaces (hereafter index d T0 = T0d) after López et al. (2012); σ is Stefan-Boltzmann constant at 5.67 · 10 -8 W m -2 K -1 . 145 The saturation specific humidity (q0, in g kg -1 ) and the relative humidity (h0) with respect to surface temperature T0 for wet and dry vegetation surfaces was calculated using Tetens formula ( (Buck, 1981;Campbell and Norman, 1998), see the equations in Appendix B).

Sampling of the NRW on foliage & isotope ratio mass spectrometer measurements
To analyze the isotopic compositions of the NRW on foliage, the NRW droplets were taken during dew and radiation fog 150 formations. The sampling of the NRW on foliage (hereafter fNRW) was carried out on a grassland area of 100×130 m 2 around the "EC & meteo" measurements ( Fig. A1 in Appendix A). Nine replicated fNRW samples were absorbed from leaf surfaces with cotton balls at the end of the nights of events 1 and 3 (once sampling per event), but bihourly during the night of event 2 (four times of sampling per event). After collection, the samples were immediately transferred into gas tight 12 ml exetainers (Labco Exetainer® vial, High Wycombe, UK) and stored in a portable cooling box filled with ice blocks. Before extracting 155 the water in a cryogenic vacuum distillation system (Prechsl et al., 2015), the samples were stored at -19°C. The measurements of the isotopic compositions for fNRW (hereafter δfNRW) and in soil moisture (hereafter δs) of extracted water samples were performed using an isotope ratio mass spectrometer (IRMS, DELTAplusXP, Finnigan MAT, Bremen, Germany). The measured uncertainties of δ 18 O and δ 2 H are ±0.1‰ and better than ±1.0‰, respectively (Gehre et al., 2004;Werner and Brand, 2001). 160 https://doi.org/10.5194/hess-2020-493 Preprint. Discussion started: 13 October 2020 c Author(s) 2020. CC BY 4.0 License.

Isotopic compositions and mixing ratio measurements for ambient water vapor
The isotopic compositions and the volumetric mixing ratio for ambient water vapor were measured at 0.5-1 Hz using a cavity ring-down laser spectrometer (L2130-I, Picarro Inc., Santa Clara, CA, USA). The L2130-i was placed in a house 200 m away from the EC & meteo measurements (Fig. A1 in Appendix A). Ambient air was pulled into the L2130-i cavity through a PTFE intake hose, with an inner diameter of 1/4 inch, and a PTFE-filter inlet Solberg International Ltd.,Itasca,165 IL, USA) fixed at 6 m a.g.l.. The intake hose was thermally isolated, heated using a resistive heating wire (Raychem 5BTV2-CT, Von Rotz, Kerns, Switzerland) wrapped around the entire length of the intake tube to prevent condensation, and flushed with an external membrane pump (N022, KNF Neuberger GmbH, Munzingen, Freiburg, Germany) at a rate of 9 L min -1 to minimize memory effects within the inlet system. The isotopic compositions of ambient water vapor (hereafter δa) and the volumetric ambient water vapor mixing ratio (hereafter w) were measured with an instrumental flow rate of 300 mL min -1 . 170 The instrument's response time in this setup was found to be on the order of 10 s Aemisegger et al. (2012).
To correct for instrument drifts and to normalize the data to the international VSMOW-SLAP scale, the raw data were calibrated using a Standard Delivery Module (SDM; A0101, Picarro Inc., Santa Clara, CA, USA) by performing two-point calibrations every 12 h  using two liquid standards (standard 1: δ 18 Os1 = -11.43‰, δ 2 Hs1 = -81.84‰, ds1 = 9.64‰; standard 2: δ 18 Os2 = -40.66‰, δ 2 Hs2 = -325.67‰, ds2 = -0.37‰ measured by an IRMS). The δ 18 O and δ 2 H of 175 the standards thus bracket the range of the measured δ 18 Oa and δ 2 Ha. Laser spectrometric measurements are known to be affected by a water vapor mixing ratio dependent bias due to spectroscopic effects (absorption peak fitting, and baseline effects). In our study, all measurements were performed at w > 12 000 µmol mol -1 , therefore no mixing ratio dependent isotope bias correction was necessary (see more details in Aemisegger et al. (2012)). The L2130-i was calibrated using a dew point generator (LI-COR LI 610, Li-Cor Inc., Lincoln, NE, USA) following the procedure by Thurnherr et al. (2020). 180 The second-order parameter d of ambient water vapor (hereafter da) was calculated with the calibrated δ 18 Oa and δ 2 Ha.
The overall random uncertainties of δ 18 O and δ 2 H measurements were 0.2‰ and 0.8‰ respectively (for more details about the uncertainty quantification, see Aemisegger et al. (2012)). Calibrated δ 18 Oa and δ 2 Ha were then averaged over 30 min intervals.
To compare the ambient water vapor measurements with the fNRW, the NRW equilibrium liquid (aNRW) from this vapor was calculated. Under the assumption of equilibrium fractionation, the isotopic compositions of aNRW (hereafter δaNRW) 185 formed from ambient water vapor (δa) were calculated using the temperature-dependent equilibrium fractionation factors following Horita and Wesolowski (1994) (see details in Appendix C).

Determination of atmospheric layer heights
The isotopic fractionation during phase change at the Earth surface is linked to the micrometeorological layers near the surface (Fig. 2). The inclusion of a zero-plane displacement (hereafter z-plane, Fig. 2) in wind profiles allows us to separate the 190 downward flux from ambient water vapor and the upward flux from soil-diffusing vapor. The height of this z-plane (hereafter zd, Fig. 2) is typically two-thirds of mean vegetation height (hereafter zc, Fig. 2; Stull (1988)). The roughness length (hereafter z0) is a measure of the aerodynamic roughness of the surface, and is around one-tenth of zc (Fig. 2;Stull (1988)). The wind speed is zero at z0 above zd (zd + z0, Fig. 2; Stull (1988)). Therefore, we consider three pathways of NRW inputs onto the foliage of grasslands for dew and radiation fog: 1) the downward component of dew formation condensing from ambient water 195 vapor (hereafter "aDew"), 2) the upward component of dew formation condensing from soil-diffusing vapor (hereafter "dDew"), and 3) radiation fog deposition (hereafter "aFog").
We determined the top of the NBL as the lowest height where the vertical stratification of the atmosphere becomes isothermal (∂T/∂z = 0, Stull (1988); Tombrou et al. (1998)). During the three events in this study, the NBL top at 1:00 CET of the events was at 730 m, 700 m, and 680 m a.g.l., respectively (Fig. 3); the NBL height was obtained from air pressure after 200 Campbell and Norman (1998); the vertical temperature and pressure profiles were extracted from the hourly European Centre for Medium Range Weather Forecast (ECMWF) reanalysis product ERA5 reanalysis dataset within the Copernicus Climate https://doi.org/10.5194/hess-2020-493 Preprint. Discussion started: 13 October 2020 c Author(s) 2020. CC BY 4.0 License.

Partitioning of NRW inputs using a two end-member mixing model
We split the contribution of NRW input pathways into the two main processes described in Sect. 2.3: (1) the downward component of dew formation (aDew) and fog droplet deposition (aFog), and (2) the distillation (dDew) of soil-diffusing vapor on plant leaves. In unsaturated conditions, the NRW on foliage (fNRW) was a mix of aDew and dDew, while in saturated 210 conditions, fNRW was a mix of aDew and aFog. "Unsaturated conditions" in this context refers to the standard 2 m height of meteorological measurements. Both aDew and aFog were condensed from ambient water vapor, thus we used the term "aNRW" if either dew or fog input, or the combination of both, was meant. Dew formed in unsaturated conditions is a mixture of aNRW and dDew but lacks contribution from fog deposition, thus the isotopic signature of the NRW resulting from the isotopic compositions of dDew (hereafter δ 18 OdDew and δ 2 HdDew) and the proportion of dDew (hereafter fdDew) in fNRW was 215 expressed as: (2) where faNRW is the proportion of aNRW in fNRW; δ 18 OdDew, δ 2 HdDew, fdDew, and faNRW are unknown. Therefore, four unknowns 220 with only three equations (Eq. 2-4) required two time points, at 23:00 CET and 1:00 CET in event 2, to obtain empirical estimates for the four unknowns. By doing so, we implicitly assumed that δ 18 OdDew and δ 2 HdDew were constant within this 2 h period, and only fdDew and faNRW were allowed to change between these two sampling times. For δfNRW, the median value for each sampling was taken, and for δaNRW the 2 h average was computed from 30 min data.

Statistics and imaging 225
In unspecified explicit, we reported means ± standard deviation. For δfNRW and dfNRW, we reported median considering the heterogeneous distribution for the sampling of NRW on foliage. The calculating and imaging were processed in R version 3.6.3 (R Core Team, 2020). For linear regression between δ 2 H and δ 18 O the orthogonal regression was used (total least square, Gat (1981)), whereas the ordinary least-squares method was used for the da-RH and linear regression.

Weak turbulence and high relative humidity
Dew and radiation fog generally form during clear-sky nights with a weak large-scale pressure gradient, low wind speeds and weak turbulence. During the three field campaigns presented in this study, u2m and FH2O showed an abrupt weakening from around 17:00 CET onwards (Fig. 4a, b). With nightfall, u2m remained below 0.7 m s -1 (Fig. 4a), and FH2O was at very low (-235 0.4 to 0.3 mmol m -2 s -1 , minus value means downward flux, and positive value means upward flux; Fig. 4b), indicating a vanishing of turbulent fluxes. These are favorable conditions for dew and radiation fog formations.
The three events with dew or radiation fog were characterized by high relative humidity with respect to air temperature (RH) measured at 2 m above ground level. From around 17:00 CET, RH increased rapidly, and reached 100% around 03:00 CET during event 2, and around 20:30 CET during event 3 (Fig. 4c). These saturated conditions led to the formation of fog 240 https://doi.org/10.5194/hess-2020-493 Preprint. Discussion started: 13 October 2020 c Author(s) 2020. CC BY 4.0 License. characterized by a horizontal visibility < 1 km (Fig. 4d). Fog appeared around 05:00 CET during event 2, lasting for less than an hour until sunrise, whilst the onset of fog was much earlier during event 3 (around 23:00 CET), lasting for a longer period until dissipation around sunrise. The visibility was always > 1 km in event 1, indicating that fog was absent during event 1.
Therefore, event 1 can be considered as a dew-only event, whilst events 2 and 3 were characterized by a combination of dew and partial influence of radiation fog. 245

Surface cooling and the sign of condensation
Both grassland surfaces and ambient air started to cool down from around 17:00 CET onwards, due to substantial net longwave radiation loss, which was not compensated by the low remaining incoming short-wave radiation levels. The leaf surfaces of the grassland cooled more rapidly than the near-surface atmosphere, thus with nightfall, T0 remained cooler than Ta, although both of them gradually decreased (Fig. 5a). The first sign of condensation occurred when the leaf surfaces cooled down below 250 dew point temperature (Fig. 5a, T0 < Td). The level of T0 (T0d) became lower than Td at around 0:30 CET in event 1, 21:30 CET in event 2, and 19:00 CET in event 3 (Fig. 5a), determining when the first signs of condensation can be expected. During event 3, the surface already cooled down below the dew point rapidly after sunset (T0 < Td, Fig. 5a), indicating that condensation already started with nightfall.
The specific humidity of the air, qa, steeply increased by 2.0-3.2 g kg -1 from around 17:00 CET until sunset (Fig. 5b), 255 suggesting the inversion of moisture from local evaporation into a shallow inversion layer. The increase of qa over time is enhanced by cold-air drainage down the slopes and along the valley bottom where the CH-CHA site is located as compared to conditions without advection. With nightfall, qa reached a nighttime maximum of 9.6-12.5 g kg -1 (Fig. 5b). Especially, in events 1 and 2, before starting to decrease, qa fluctuated for a short period from sunset until the first sign of condensation (Fig.   5b). When condensation started (T0 < Td, Fig. 5a), qa gradually decreased (Fig. 5b). With q0 falling to values below qa (Fig.  260 5b), super-saturation with respect to the leaf surfaces occurred, thus computed theoretical h0 exceeded 100% (Fig. 4c). The decrease of qa was much faster in event 3 (0.4 g kg -1 h -1 , Fig. 5b) than that in events 1 and 2 (0.2 and 0.3 g kg -1 h -1 , Fig. 5b), indicating stronger condensation of ambient water vapor.

Characteristics of precondensation and condensation periods
According to the temperature and humidity conditions, the periods from 17:00 CET until sunrise were defined as: 1) 265 precondensation period (hereafter P1) with the weakening of turbulence and with T0 > Td; and 2) condensation period (hereafter P2) with T0 < Td. The precondensation period (P1) was further separated into: P1a) starting around 17:00 CET until sunset with the weakening of turbulence and the increase of qa; P1b) from sunset until the first sign of condensation with short-term fluctuations of qa. The condensation period (P2) was further split into: P2a) with dew only under RH < 100%; P2b) with dew and radiation fog occurring in combination under RH = 100%. 270

Splitting the components of dew using a two end-member mixing model
Under unsaturated conditions, with respect to aNRW, δ 18 OfNRW and δ 2 HfNRW deviated to the enriched and depleted sides of δ 18 OaNRW and δ 2 HaNRW, respectively (Fig. 7a, b), suggesting a mix of NRW on foliage (fNRW) from the condensation of soildiffusing vapor (dDew) and the condensation of ambient water vapor (aNRW). Based on the measurements from 23:00 to 1:00 CET in event 2, the averages of δ 18 OdDew, δ 2 HdDew, and ddDew during this 2 h period were estimated as -1.0‰, -71.8‰, and -315 63.4‰ respectively (Fig. 7a, b, c), and the corresponding contributions of dDew in fNRW were 28% and 9% respectively (Fig.   7d). A linear extrapolation from the two hours between 23:00 CET and 1:00 CET to the beginning of dew formation at 21:30 CET of event 2 increased the contribution of dDew to 42% (Fig. 7d). Similarly, when using the values of δdDew from event 2 for estimating the contribution of dDew during event 1, the proportion of dDew was around 18-31% for our sampling at 3:00 CET of event 1 (vertical whiskers in Fig. 7d). 320 https://doi.org/10.5194/hess-2020-493 Preprint. Discussion started: 13 October 2020 c Author(s) 2020. CC BY 4.0 License.

Fractionation during condensation of ambient water vapor
We only considered equilibrium fractionation (shown as δaNRW in Fig. 9) when simulating the isotopic compositions of the NRW component condensing from ambient water vapor. An alternative approach would be to consider both equilibrium and non-equilibrium fractionation factors (see Appendix D; see also Lee et al. (2009) and Wen et al. (2012)) because of the laminar 325 sublayer in the leaf boundary layer. To compare these two methods, we applied the method of Wen et al. (2012) on our data to simulate the isotopic composition of the NRW component condensing from ambient water vapor (shown as δnaNRW in Fig.   9; see also Appendix D). We found that δnaNRW was more depleted as compared to the NRW on foliage (shown as δfNRW in Fig.   9), and the depletion of δnaNRW was more severe with the increase of h0 (Fig. 4c), which is in agreement with Wen et al. (2012).
The depletion of δnaNRW with respect to δfNRW was most likely due to the overestimation of the non-equilibrium fractionation 330 factor when h0 exceeded 100% (going up to 132% in our study, see Fig. 4c), because Jouzel et al. (1987) pointed out that nonequilibrium fractionation is negligible above -10 °C in the process of vapor condensing to liquid. However, non-equilibrium fractionation driven by molecular diffusion might have played an important role in a laminar fog boundary layer (hereafter FBL; (Castillo and Rosner, 1989;Epstein et al., 1992)), which led to more depleted δfNRW than δaNRW at 5:00 CET in event 3 ( Fig. 7a, b) when radiation fog occurred. Heavier isotopologues move more slowly than their lighter counterpart in air 335 Merlivat (1978)), hence the rate at which heavy isotopologues ( 1 H2 18 O and 1 H 2 H 16 O) in ambient air pass through the laminar FBL to be condensed at the liquid-vapor interface is smaller than the rate of condensation of their lighter counterpart. Therefore, δfNRW can become more depleted than δaNRW. Fog lasted as from 23:00 CET until sunrise of event 3, and appeared around 5:00 CET within half an hour before sunrise in event 2 (Fig.   4d). However, we only observed a lower δfNRW than δaNRW in event 3 (Fig. 7a, b), suggesting that the depletion of δfNRW might 340 also be related to the duration of radiation fog. The condensation of ambient water vapor for dew formation can be approximated as an equilibrium fractionation process accordingly, as was also observed by Wen et al. (2012) and Delattre et al. (2015). The condensation of ambient water vapor to form radiation fog can cause slight depletion of the NRW compared to the equilibrium liquid obtained from ambient water vapor.

Potential NRW gain from the condensation of soil-diffusing vapor 345
Splitting the input pathways of dew formation using stable isotopes will allow future studies to quantify dDew gain with the combination of lysimetric or filter paper absorption measurements. In our study, as shown in Sect. 3.4, we estimated that dDew contributed 9-42% of total NRW (Fig. 7d) during our observation periods. Monteith (1957) estimated that the condensation rate of soil-diffusing vapor was 0.01-0.02 mm h -1 (with u2m < 0.5 m s -1 ) using filter paper absorption measurements. In Monteith (1957), the condensation rate of ambient water vapor varied from 0.004 to 0.035 mm h -1 depending on the wind 350 speed (with u2m < 3 m s -1 ) and humidity conditions. Thus, the contribution of dDew in the total NRW was potentially 22-83 % according to the condensation rate of Monteith (1957). Following the condensation rate of Monteith (1957), the potential total NRW gain was 0.06 -0.39 mm night -1 (see details in Table 2). This amount of NRW gain was comparable with the average evapotranspiration rate of 2.8 mm day -1 (daytime) during the continuous no-rain periods of the three events (see details in Table 3). 355 In future research, combining isotopic compositions measurements with lysimetric measurements to quantify dDew gain would provide useful benchmark data to better evaluate the isotope-based estimates of NRW input. The NRW gain can be measured directly by a lysimeter as the net water gain of the soil and plants (Kaseke et al., 2012;Riedl et al., 2020;Ucles et al., 2013), while dDew is an indirect estimate based on stable water isotope data of the transfer of moisture from one part of the surface (soil surface) to another (foliage) within grassland ecosystems. 360 https://doi.org/10.5194/hess-2020-493 Preprint. Discussion started: 13 October 2020 c Author(s) 2020. CC BY 4.0 License.

Diurnal patterns of isotopic compositions in ambient water vapor
The diurnal patterns of da reflected the main drivers of ambient moisture variability. During the daytime 13:00-17:00 CET, da was at a plateau (12.2‰ to 18.0‰, Fig. 6d) compared to condensation periods in the night (P2), when da reached its daily minimum (-11.8‰ to -4.7‰, Fig. 6d). The transition from higher daytime da to lower nighttime da occurred from 17:00 CET until sunset (P1a, Fig. 6d). Entrainment from the free troposphere played a dominant role in daytime atmospheric moisture 365 during 13:00-17:00 CET, and caused a higher da than in the night (Fig. 6d), and a decrease in δa (Fig. 6b, c) (Delattre et al., 2015;Lai et al., 2006;Lee et al., 2006;Parkes et al., 2017;Welp et al., 2012). On the contrary, during P1a, under reduced entrainment from the free troposphere (weakened u2m, and reduced FH2O in Fig. 4a, b) compared to mid-day values, local ET caused a steep decrease of da (Fig. 6d) and increases of δa (Fig. 6b, c), which was in accordance with the previous research by Lai et al. (2006), Huang and Wen (2014), and Parkes et al. (2017). During P1b, the fluctuation of δa (Fig. 6b, c) was due to 370 short-term variability of the isotopic compositions of soil evaporation (within 1 h before sunset, 0-5 cm soil moisture with δ 18 O varying from 5.5‰ to -8.5‰, with δ 2 H varying from -8.5‰ to -72.8‰, and d varying from -5.0‰ to -52.4‰), which was in accordance with the reports by Welp et al. (2012). The decrease of δa during P2b suggested radiation fog with local moisture as a source for ambient water vapor, which was in contrast with Spiegel et al. (2012) in Greenland that found the increase of δa with fog during the passage of a cold front. The correlated da -Ta (Fig. 10a) and anti-correlated da-RH (Fig.  375 10b) in our study suggested an only minor influence of large-scale air advection and highlighted the dominant role of local moisture as a source for ambient water vapor (Aemisegger et al., 2014).
During dew and radiation fog (P2), the condensation of ambient water vapor could essentially be described by an equilibrium fractionation process, with da remained constant at a low nighttime minimum level (Fig. 6d) (Delattre et al., 2015;Huang and Wen, 2014). However, soil evaporation occurred synchronously with condensation. Soil evaporation in saturated 380 ambient air (RH = 100 % at 2 m a.g.l.) is essentially an equilibrium fractionation process (Eichinger et al., 1996;Priestley and Taylor, 1972), which did not affect the variability of da during P2b (Fig. 6d). Whereas, non-equilibrium fractionation is intrinsically dominant in the processes of soil evaporation in unsaturated ambient air (RH < 100 % at 2 m a.g.l.), which induced a slight da decrease during P2a (Fig. 6d). In addition, cold air drainage along the valley to the bottom where the CH-CHA site is located (Eugster and Merbold, 2015), might have enhanced the effect of local soil evaporation on δa variability. 385

Ecological functions of non-rainfall water
From the perspective of ecological functions, dDew might be more important than previously thought, although it has no largescale hydrological significance of moisture transfer from one part of the surface to another (Monteith, 1957). This can be expected if the transfer of moisture is from a hydrological pool that is inaccessible to plants (e.g., soil-diffusing vapor) to another that is accessible to plants (e.g., droplets forming or depositing on leaf surfaces or on the surface soil where it can be 390 accessed by the fine roots). The condensation of soil-diffusing vapor was comparable with the condensation of ambient water vapor in our study (contributing 9-42% during our observation periods, Fig. 7d), and was the dominant pathway of NRW inputs during very calm night (u < 0.7 m s -1 ; see also Monteith (1957)) when the flux from ambient air to the grassland surface was very small. Soil vapor diffusion occurs as long as a temperature gradient exists (see the soil temperature at different depths in Fig. E1 of Appendix E), which results in vapor pressure differences along that gradient. Therefore, soil vapor diffusion 395 transfers the deeper soil vapor to the surface, from where it moistens the air in contact with the soil surface. Subsequently, this moisture condenses onto foliage and becomes available to the plants. Wang et al. (2017) observed that 0.0092 mm of water were transferred from deeper soil layers to the surface by vapor diffusion in a grassland plot, although it was doubtful whether the water went onto foliage or was absorbed by the top soil. The soil diffusion rate increases with the decrease of soil water content (under volumetric soil moisture content higher than 10% (Barnes and Turner, 1998;Philip and De Vries, 1957)), which 400 makes soil vapor diffusion more important in conditions of low soil moisture. The NRW inputs for dew and radiation fog are expected to be taken up by plants through foliar water uptake or be dripping down to wet the soil surface, thereby potentially https://doi.org/10.5194/hess-2020-493 Preprint. Discussion started: 13 October 2020 c Author(s) 2020. CC BY 4.0 License.
preventing permanent damage of the plants by drought stress (e.g., Schreel and Steppe (2020)). The ecological functions of NRW was also reflected in its thermal effect on the plants. The leaf wetting by NRW which potentially cooled the leaf surfaces by 1.5 °C in comparison to dry leaf surfaces (differences between T0w and T0d, Fig. 5a), thereby alleviating potential plant heat 405 stress during the early morning hours when solar radiation quickly increases after sunrise.
Further research should thus focus on the plant water status in response to NRW inputs from dew and radiation fog.
In addition, future research focusing on the continuous measurement of the isotopic compositions (δ 18 O and δ 2 H) of soil vapor would give more quantitative insights on vapor transfer in soils during dew and radiation fog nights. The condensation of soildiffusing vapor is expected to play a more important role in temperate grasslands than in arid grasslands, if soil salinity and 410 canopy resistance are also taken into consideration: Soil salinity reduces the rate of soil vapor diffusion (Gran et al., 2011). In a laminar boundary layer during dew and radiation fog events, dense canopies in temperate grasslands (LAI was 1.5-3.2 m 2 m -2 for summer 2018 at the CH-CHA site, Sect. 2.1) potentially shield the uppermost soil vapor from being exported into the near-surface atmosphere, while sparse canopies in arid grasslands (LAI around 0.5 m 2 m -2 as e.g. in Wen et al. (2012)) should result in most of the soil-diffusing vapor being emitted into the atmosphere. 415

Conclusion
Our results reveal different input pathways for dew and radiation fog in a temperate grassland during three dry intensive observation periods in summer 2018 in Switzerland. Dew and radiation fog occurred in clear calm nights with very low wind speed (u < 0.7 m s -1 ) and weak turbulence with near-zero net water vapor flux at the vegetation surface (FH2O at -0.4 to 0.3 mmol m -2 s -1 ). Condensation of ambient water vapor during dew and radiation fog was found to be predominantly an 420 equilibrium fractionation process, which was deduced from the rather constant da during NRW nights. This caused a decrease of 0.8-1.6‰ δ 2 Ha h -1 in ambient water vapor during dew and radiation fog. In unsaturated conditions (determined at the meteorological 2 m reference height), condensation occurred from ambient air above the canopy as well as soil-diffusing vapor below the canopy, as was indicated by a 3.4-3.7‰ decrease of da. Local evaporation at high relative humidity from 17:00 CET until sunset caused the lowering of da to values in the range of 2.4‰ to 4.8‰ as compared to the higher daytime da (12.2‰ to 425 18.0‰). A further decrease to da values in the range of -11.8‰ to -4.7‰ was observed during the occurrence of dew and radiation fog at night. Dew only formed under unsaturated conditions with a mixed NRW condensing from ambient water vapor and soil-diffusing vapor. The comparison between the foliage NRW δfNRW and the equilibrium NRW δaNRW of ambient water vapor allowed us to trace the source of the NRW input pathways during dew formation. The NRW condensing from soil-diffusing vapor contributed 9-42% of the foliage NRW. The correlated da -Ta and anti-correlated da-RH suggested an 430 only minor influence of large-scale air advection and highlighted the dominant role of local moisture as a source for ambient water vapor.
In future studies, continuous isotope measurements of foliage NRW, ambient water vapor and soil vapor should be combined with direct lysimetric and filter paper absorption measurements, as well as physiological measurements to more precisely quantify the NRW input pathways, and the mechanisms of plant water status responding to NRW input on foliage. 435 Confirmation of dew and radiation fog inputs into temperate ecosystems during summer drought by the isotopic compositions of NRW and ambient water vapor would then allow assessing the potential response of these ecosystems to warming and increased frequency of summer droughts under the global climate changes.
The dew and radiation fog potentially produced 0.06-0.39 mm night -1 NRW gain on foliage, which was comparable with 2.8 mm day -1 daytime evapotranspiration. With increasing relative humidity, the share of vapor originating from soil 440 vapor diffusion decreased, whereas the relevance of atmospheric water vapor for dew formation increased. This atmospheric water vapor had a rather local isotopic signature, which suggests that large-scale moisture advection only has a minor influence in the nocturnal NRW gains during dew and radiation fog events. Our results thus highlight the importance of NRW inputs to https://doi.org/10.5194/hess-2020-493 Preprint. Discussion started: 13 October 2020 c Author(s) 2020. CC BY 4.0 License.  (2002): "aDew" means dew formed from ambient water vapor, "aFog" means fog formed from ambient water vapor; "aDew" and "aFog" are both condensed from ambient water vapor, thus "aNRW" represents the condensation of ambient water vapor if either dew or fog input, or the combination of both was meant. ; "dDew" means dew formed from soil-diffusing vapor. The horizontal wind speed (u) is zero at zd + z0. 670 Figure 3. Nocturnal vertical profiles of air temperature (T) vs. height (z, in m a.g.l.) at 01:00 CET for the three events interpolated to the location of the measurement site based on ERA-5 reanalysis data (Hersbach et al., 2020;Horanyi, 2017). https://doi.org/10.5194/hess-2020-493 Preprint. Discussion started: 13 October 2020 c Author(s) 2020. CC BY 4.0 License. [FH2O], (c) relative humidity at 2 m a.g.l.
[RH], relative humidity with respect to 675 the surface temperature [h0w for wet surface, and h0d for dry surface]; (d) 1 min averages of visibility (< 1 km with fog, and > 1 km with the absence of fog).
https://doi.org/10.5194/hess-2020-493 Preprint. Discussion started: 13 October 2020 c Author(s) 2020. CC BY 4.0 License. Figure 6. The volumetric mixing ratio and isotopic compositions for ambient water vapor. The 30 min averages and standard deviations of (a) volumetric ambient water vapor mixing ratio (w), and (b-d) the isotopic compositions of ambient water vapor (δ 18 Oa, δ 2 Ha, and da). P1a was from 17:00 CET until sunset with the weakened turbulence and increased specific humidity; P1b was a short-term variability of specific 690 humidity; P2a was dew formation in unsaturated ambient air; P2b was dew and radiation fog in combination in saturated ambient air.
https://doi.org/10.5194/hess-2020-493 Preprint. Discussion started: 13 October 2020 c Author(s) 2020. CC BY 4.0 License. Figure 7. (a-c) Isotopic compositions of different non-rainfall water (fNRW, aNRW, and dDew), and (d) the proportions of dDew (fdDew). "fNRW", the NRW on foliage; "aNRW", the NRW condensed from ambient water vapor; "dDew", the dew component condensed from soil-diffusing vapor. P2a was dew formation in unsaturated ambient air; P2b was dew and radiation fog in combination in saturated ambient 695 air. Figure 8. The relationship of δ 2 HfNRW-δ 18 OfNRW with respect to the orthogonal regression of δ 2 HaNRW-δ 18 OaNRW and local meteorological water line (LMWL: δ 2 H = 7.68 × δ 18 O + 6.97, Prechsl et al. (2014)). "fNRW" means non-rainfall water (NRW) on foliage, and "aNRW" means the NRW equilibrium from ambient water vapor. "RH" is relative humidity at 2 m a.g.l.. 700 https://doi.org/10.5194/hess-2020-493 Preprint. Discussion started: 13 October 2020 c Author(s) 2020. CC BY 4.0 License. Figure 9. Comparing the isotopic compositions of the simulated NRW to the isotopic compositions of the NRW on foliage (δfNRW). The δaNRW was calculated from ambient water vapor δa considering equilibrium fractionation, and δnaNRW was calculated from ambient water vapor δa considering both equilibrium and non-equilibrium fractionation). P2a was dew formation in unsaturated ambient air; P2b was dew and radiation fog in combination in saturated ambient air. 705 Figure 10. The relationships of (a) da-Ta and (b) da-RH for the 24 h measurements during the three events. The da is the deuterium excess of ambient water vapor; Ta is the ambient temperature at 2 m a.g.l.; RH is relative humidity at 2 m a.g.l.. https://doi.org/10.5194/hess-2020-493 Preprint. Discussion started: 13 October 2020 c Author(s) 2020. CC BY 4.0 License. Table 1. Partitioning the contribution s of dDew from a mix of dDew and aDew. The fNRW means the non-rainfall water (NRW) on foliage; aNRW represents either dew or radiation fog, or dew and radiation fog in combination condensed from ambient water vapor; 710 dDew means dew condensed from soil-diffusing vapor; fdDew means the proportion of dDew in total foliage NRW.  Table 2. Estimating the potential non-rainfall water (NRW) gain of the three events in our study according to the condensation rate of Monteith (1957). The fNRW means the non-rainfall water (NRW) on foliage; aNRW represents either dew or radiation fog, or dew and radiation fog in combination condensed from ambient water vapor; dDew means dew condensed from soil-diffusing vapor. Event Period (h night -1 ) Condensation rate following Monteith (1957) (