The Alp Weissenstein research site (CH-AWS, at 2000 m.a.s.l.) is part of a managed (grazed) alpine grassland ranging from 1900 to 2500 m.a.s.l. The vegetation composition was classified as Deschampsio cespitosae–Poetum alpinae community with red fescue (Festuca rubra), Alpine cat’s tail (Phleum rhaeticum), white clover (Trifolium repens) and dandelion (Taraxacum officinale) as dominant species (Keller, 2006), complemented by alpine meadow-grass (Poa alpina) and lady’s mantle (Alchemilla vulgaris). The soil types are slightly humous to humous sandy loam (Hiller et al., 2008), hence the permanent wilting point is estimated at around 0.1 m³ m^-3. The mean annual air temperature and precipitation were 1.9 °C (2015–2020; measured all year round at the site between 2015 and 2020; before 2015, data were only collected between May and October at the site) and 1213 mm (2013–2020; measured all year round between 2013 and 2020; before 2012, only the liquid precipitation was measured at the site), respectively. During the main growing season (May to September), monthly mean air temperatures were between 5.0 °C and 10.8 °C (2006–2018) with July as the hottest month ( Figure 1A ), while average monthly precipitation ranged from 87 to 128 mm ( Figure 1B ). During the growing season in 2019, the monthly temperature ranged from 0.8 °C to 11.7 °C with June as the hottest month ( Figure 1A ), whilst the monthly precipitation ranged from 61 mm to 173 mm ( Figure 1B ).

EC measurements for H₂O vapor and net ecosystem CO₂ exchange have been carried out during the growing season since 2006 (tower coordinates: 46°34’59.5” N, 9°47’25.5” E at 1978 m.a.s.l.). In mid-November 2014, the site was equipped with mains power for year-round operation. The EC instruments at CH-AWS in 2019 consisted of a three-dimensional sonic anemometer (model HS-50, Gill Instruments, Solent, UK) and an enclosed-path infrared gas analyzer (IRGA; Li-7200, Li-Cor, Lincoln, NB, USA), installed at 1.4 m agl (above ground level). EC measurements were recorded at 20 Hz and processed to 30 min averages using the EddyPro software Version 7.0.6 (LI-COR, 2019) following established community guidelines (Aubinet et al., 2012) for H₂O (F _H2O in mmol m^-2 s^-1) and CO₂ fluxes (F _CO2 in μmol m^-2 s^-1; net ecosystem exchange NEE). The micro-meteorological sign convention was used, with negative values denoting a downward flux, while positive values stand for upward fluxes. See details of footprint analysis of eddy-covariance measurements in Zeeman et al. (2010). Vapor pressure deficit (VPD) was quantified for 30-min intervals from ancillary air temperature and relative humidity measurements at 1.4 m agl (HygroClip HC2, Rotronic, Bassersdorf, Switzerland). Photosynthetic photon flux density (PPFD in μmol m^-2 s^-1) was measured at 1.3 m agl every 10 s (PARlite, Kipp & Zonen B.V., Delft, The Netherlands) and then averaged to 30-min intervals. Volumetric soil water content (SWC) was measured by two sensors (EC-5, Decagon Devices, Inc., Pullman, WA, USA) at 5 cm depth. NEP was calculated with the opposite sign of NEE (NEP = –NEE).

Diurnal NEP (g C m^-2) was calculated from the CO₂ flux ( F C O 2 ):

where M _c is the molar mas of carbon (12 g C mol^-1), t is measurement intervals (1800 s) of F C O 2 , and a is a unit conversion factor (10^-6 mol μmol^-1).

For a long-term data series of air temperature at standard 2 m agl, an additional meteorological measurement setup was operating since 2006, installed at about 1180 m distance to the east and at approximately 40 m higher elevation compared to the flux tower (Michna et al., 2013). This additional setup provided air temperature (T _a) at 2 m agl (shaded, sheltered HydroClip S3, Rotronic AG, Basserdorf, Switzerland), and precipitation from an unheated pluviometer (LC, Texas Electronics, Dallas, USA). In November 2012, a precipitation gauge (1518H3, LAMBRECHT meteo GmbH, Göttingen, Germany) with a heatable orifice was installed between these two measurement stations (Michna et al., 2013) and provided annual total precipitation, including snowfall. Leaf wetness data were averaged from the measurements of two leaf wetness sensors (BNS, G. Lufft Mess-und Regeltechnik GmbH, Fellbach, Germany) installed since 2005 at 0.1 m agl close to the grassland canopy, using blotting paper inside a clip holder. The leaf wetness data was recorded by the voltage signal resulting from a fixed current applied from the center to the rim of the blotting paper. When the paper got wet by dew, fog, or rain, the blotting paper became conductive, and an increase in voltage signal was observed. We note that the leaf wetness sensors overestimated the leaf wetting duration ( Figure S2 ), because the blotting paper dries out slower than the vegetation. By comparing the BNS sensor with a more accurate leaf wetness sensor (PHYTOS 31, Meter Group AG, Munich, Germany) at a later time of our observation campaigns (5–6 July 2020), the termination of leaf wetting was defined as the point when leaf wetness by BNS steeply and linearly decreased ( Figure S2 ).

All variables were aggregated to 30 min averages or sums. The time series was recorded in CET (UTC+1 hour).

The evapotranspiration rate (ET in mm h^-1) was calculated from the H₂O flux ( F H 2 O ) as (Stull, 1988):

where M H 2 O is the molar mass of H₂O (18 g mol^-1), and b is a unit conversion factor [= (10^-3 mol mmol^-1) · (10^-6 m³ g^-1) · (3600 s h^-1) · (10³ mm m^-1) = 0.0036 mol m² s mm mmol^-1 g^-1 h^-1].

According to our measurements at the CH-AWS site during 2006 to 2018, the hottest three months were typically June, July, and August, with average air temperatures (at standard 2 m agl) of 8.9, 10.8 and 10.3 °C ( Figure 1A ), respectively. As compared to the long-term averages, the respective three months in 2019 were hotter with 11.7, 10.9, 10.8 °C ( Figure 1A ). The precipitation in June 2019 was 61 mm, which was only 51% of the long-term average of 120 mm in June during 2006–2018 ( Figure 1B ).

The 2019 heatwave occurred in Switzerland from 25 June to 1 July in 2019 (MeteoSwiss, 2019). No rain was recorded at the site from 23 June to 30 June 2019, but 0.2 mm rain was collected at 16:00 CET on 1 July 2019. Therefore, in this study, we only considered the 8-day rain-free period between 23 and 30 June 2019, with 23–24 June before the heatwave, and 25–30 June during the heatwave. Sunrise was around 04:30, and sunset was around 20:20 during the heatwave.

To assess the combined effect of a well-developed natural drought during the heatwave in June 2019, we conducted two intensive measurement campaigns as intensive observation periods (IOP) at the end of the heatwave. These campaigns were carried out during two consecutive dew nights on 28–29 (IOP1 from 12:00 to 12:00 the next day) and on 29–30 (IOP2) June 2019.

Tukey’s honest significance test was used to assess differences among averages over sampling times and genera by the R-function agricolae::HSD.test (Steel, 1997) and one-way ANOVA. Reported statistical significance represents p < 0.05 with capital letters indicating temporal differences, and lower-case letters denoting genera or soil-depth differences. The isotopic and LWP results were reported in mean and standard errors of mean (SEM). We note that differences of isotope composition are always reported in absolute terms in per mil (‰). Correlation coefficients of regressions of NEP with T _a, PPFD, and RH were analyzed before and during the heatwave, with “***”, “**”, “*” and “ns” indicating p < 0.001, p < 0.01, p < 0.05, and p ≥ 0.05, respectively. During IOP1 and IOP2 at the end of the heatwave, considering the individual variability of plants, median values of leaf water isotope (δ¹⁸O_lw) and LWP by species at each sampling time were used for analyzing their correlations with environmental conditions (RH, SWC, and δ¹⁸O_soil), with “***”, “**”, “*” and “ns” indicating p < 0.001, p < 0.01, p < 0.05, and p ≥ 0.05, respectively. All analyses were carried out with R version 4.1.2 (R Core Team, 2021).

The diurnal and nocturnal air temperature averaged 14.5 °C and 7.3 °C before the heatwave during 23-24 June, but was 20.1 °C and 11.2 °C on average during the heatwave from 25 to 30 June ( Figure 2A ). The highest temperature of 25.4 °C was observed on 26 June 2019 (15:30; Figure 2A ), indicating the peak of the heatwave, followed by 27 June 2019, the second hottest day. H₂O fluxes varied from –0.3 to 13.7 mmol m^-2 s^-1 ( Figure 2B ), corresponding to 3.0–4.7 mm of diurnal ET before the heatwave, and 4.9–5.7 mm during the heatwave ( Figure 3A ). SWC decreased from 0.32 to 0.15 m³ m^-3 during this rain-free period from 23 to 30 June ( Figure 2C ). NEP varied from –22 to 21 μmol m^-2 s^-1 ( Figure 2D ). The daytime NEP was 4.2–6.3 g C m^-2 before the heatwave, but ranged from 5.4 to –2.9 g C m^-2 during the heatwave ( Figure 3B ). Negative NEP (–2.9 g C m^-2) occurred on 27 June, the day after the hottest day (26 June). Highest VPD was 2.06 kPa and 2.65 kPa before and during the heatwave ( Figure 2E ), respectively. The leaf wetness levels indicated that dew occurred during each night of the rain-free period, being fully evaporated from vegetation surfaces after 07:30 of day ( Figure 2F ). With dew occurrence during each night of the rain-free period (23–30 June), the corresponding nocturnal VPD was as low as 0.14–1.41 kPa ( Figure 2E ).

Before the heatwave, NEP almost linearly increased with PPFD ( Figure 4 ), with higher diurnal NEP (6.3 g C m^-2) on 23 June than that on 24 June (4.2 g C m^-2; Figure 3B ). The suppression of heat-drought stress on NEP was substantial on the hottest day (26 June with 2.0 g C m^-2) during the heatwave ( Figure 3B ), with 63% of reduction in NEP compared to the previous day (25 June with 5.4 g C m^-2 of NEP). Negative diurnal NEP (–2.9 g C m^-2) occurred on the second hottest day (27 June; Figure 3B ), with longer period (09:00–15:00) of net carbon emission (negative NEP) under PPFD > 1300 μmol m^-2 s^-1 and T _a > 22.5 °C ( Figure 4E ), as compared to the hottest day during 14:00–16:30 under PPFD > 1800 μmol m^-2 s^-1 and T _a > 25.3 °C ( Figure 4D ). Daily NEP recovered to the levels of 4.7–5.0 g C m^-2 on 28–30 June after the hottest two days on 26–27 June ( Figure 3B ).

During the early morning hours before 07:30 with leaf wetting by dew, NEP increased by T _a before and during the heatwave ( Figure 5A ; p < 0.01 and p < 0.001 before and during the heatwave, respectively), but the turning T _a from negative (net carbon emission) to positive (net carbon sequestration) NEP was higher during the heatwave period (13.7 °C for 25-30 June) than that before the heatwave (8.2 °C for 23-24 June). With vegetation wetting by dew in the early morning hours, NEP exponentially increased with PPFD (p < 0.01 and p < 0.001 before and during the heatwave, respectively), but NEP was lower during the heatwave than that before the heatwave under same levels of PPFD ( Figure 5B ). In the early morning hours with vegetation wetting, NEP slightly increased by leaf wetness levels before the heatwave ( Figure 5C ; p ≥ 0.05), but significantly decreased by leaf wetness levels at the beginning of the heatwave on 25–26 June ( Figure 5D ; p < 0.01).

LWP of Taraxacum and Trifolium were positively correlated with RH ( Figure 9A ), indicating the controls of atmospheric humidity on their LWP. LWP of Poa was positively correlated with SWC ( Figure 9B ), indicating the controls of soil moisture on their LWP. This corresponded to the significant decline of Poa LWP at the end of the heatwave ( Figure 6 ), under the conditions of low SWC (0.15 m³ m^-3; Figure 2C ) close to wilting point. LWP and δ¹⁸O_lw of Alchemilla, Taraxacum and Trifolium were negatively correlated ( Figure 9C ), indicating the controls of leaf water content on their δ¹⁸O_lw. Stronger drought-stress of Poa might induce partial stomatal closure, and thus result in their more enriched leaf water isotopes (higher δ¹⁸O_lw) not relevant to LWP. The slightly improved Alchemilla LWP at the end of the heatwave might be derived from the accumulated benefits by dew, which contributed more to Alchemilla leaf water (14%; Figure 8D ) compared to the other three genera (≤ 12%). δ¹⁸O_lw of Alchemilla, and Taraxacum was negatively correlated with RH ( Figure 9D ), indicating the control of atmospheric conditions on their δ¹⁸O_lw. δ¹⁸O_lw did not show significant correlation with SWC and δ¹⁸O_soil ( Figures 9E, F ), indicating the minor control of soil water on leaf water isotopes.

The benefits of dew on NEP were observed in the early morning hours of 23–24 June before the heatwave ( Figure 10 ). From sunrise to 06:30 on 23 and 24 June, PPFD was at similar levels ( Figure 10C ), hence higher NEP on 24 June ( Figure 10A ) might be induced by higher temperature ( Figure 10D ) as compared to 23 June. But from 06:30 to 07:30, although higher PPFD and temperature on 24 June, NEP was at the similar levels as that on 23 June ( Figure 10C ), probably induced by higher potential of dew formation as indicated in higher RH on 23 June ( Figure 10B ).

However, with the occurrence of heatwave, dew benefits on NEP were cancelled out, as shown in the reduced NEP with increasing RH ( Figure 5C ). The chamber tracer experiment at the end of the heatwave showed that dew isotope signal was not transferred to leaf sugar ( Figure 8A ), indicating that dew water did not participate in carbon assimilation during the heatwave. The possible reason could be derived from the minor contribution (3–14%) of dew water to plant leaf water ( Figure 8D ), corresponding to previous research that foliar water uptake can only increase leaf water content by 2–11% (Limm et al., 2009). Due to the heat-drought stress, partial stomatal closure and vapor pressure gradient (Li et al., 2023) from leaf to atmosphere (saturated leaf internal environment vs unsaturated atmospheric conditions) might limit the uptake of dew water via leaf, thus most of the dew water during the heatwave could evaporate after sunrise instead of being used for carbon assimilation. Oliveira et al. (2021) showed that dew evaporation processes induced CO₂ loss of a maritime pine forest during the rain-free period after wildfire, but Simonin et al. (2009) reported that the reduction in leaf water deficit by fog water can result in improved carbon gain. Therefore, the effect of dew on ecosystem exchange varied by environmental conditions, e.g., environmental stress (Oliveira et al., 2021), the duration of canopy wetting (Simonin et al., 2009), the wettability of the leaf surface (Brewer and Smith, 1994; Hanba et al., 2004), and the foliar water uptake capacity of plants (Simonin et al., 2009).

Our tracer chamber experiment was carried out in a single chamber, and with only once sampling after sunrise, thus it was not possible to investigate the effect of dew on carbon assimilation after dew totally evaporating from surfaces. Future research on hourly resolution and longer period of after-sunrise isotope measurements is recommended to answer this question.

Despite heat-drought stress, alpine grassland kept high ET during the heatwave ( Figure 3A ), as long as soil moisture was available ( Figure 2C ) to meet its evaporative demand (Teuling et al., 2010; Wolf et al., 2013).

De Boeck et al. (2016) showed that a combined heatwave and drought stress induced a reduction in above-ground biomass of alpine grassland plants. In this study, the reduction of NEP by the heat-drought stress was most pronounced during the two hottest days (26 and 27 June; Figures 4D, E ) of the heatwave period, Compared to 26 June (25.4 °C; Figure 4B ), 27 June showed a longer period of net carbon emission ( Figure 4E ) despite lower maximum T _a (24.8 °C; Figure 5E ) and SWC ( Figure 2C ). This might be explained by an intensifying effect of drought stress, corresponding to previous research by De Boeck et al. (2016), and the influence of severe heat stress on nighttime refilling. Nevertheless, heat stress was a bit relieved after the heatwave peak, when temperatures only reached 21.5 °C, and NEP recovered immediately. The rapid re-growth ability of alpine plants after heat-drought stress (Brilli et al., 2011) might explain the recovery of NEP during 28–30 June ( Figure 3B ) after the hottest two days. This also corresponded to the fact that LWP of plants (except Poa) did not significantly decreased at the end of the heatwave ( Figure 6 ). The refilling of plant tissues during the night is a well-known phenomenon (Schulze et al., 2019), helping to relieve water losses during heat-drought stress, and might contribute to the recovery of diurnal NEP on 28–30 June after the peak of the heatwave to the levels slightly lower than the beginning of the heatwave on 23–25 June. The redistribution of soil water by deeper plant roots to the shallower soil depths might provide sufficient soil moisture sources for plant tissue refilling (Burgess and Bleby, 2006) after the peak of the heatwave. This may also explain why the predawn LWP of Alchemilla, Taraxacum, and Trifolium did not decline at the end of the heatwave. Yet, the nighttime refilling mechanism of plant tissues might not always lead to a full recovery of NEP at daily scales, particularly in the case of extreme heat-drought stress. As a result, the most severe decline of NEP was observed during the peak of the heatwave on 26–27 June. The most severe heat stress on 26 June might cause the suppression of tissue refilling during the following night, and thus more pronounced NEP reduction was observed on the next day (27 June) instead of on the hottest day (26 June). The decline of Poa LWP ( Figure 6 ) might be due to their more severe drought stress (as indicated in their higher δ¹⁸O_lw as compared to the other three genera; Figures 7A , 8A ) and dependence on soil moisture ( Figure 9B ). The decline of Poa LWP had minor effect on NEP, probably due to their small plant size and stress-induced partial stomatal closure. We could not quantify the individual contribution of different genera on ecosystem water and carbon fluxes, and effect of eddy-covariance footprint; thus, we suggest that future research can combine ecosystem-scale eddy-covariance fluxes, footprint models and plant-scale chamber fluxes, complemented by plant community compositions to quantify the effect of different genera on ecosystem exchange.

Dew amount was found to be underestimated by eddy-covariance H₂O fluxes because dew occurs on clear and calm nights with stably stratified nocturnal boundary layer (Jacobs et al., 2006; Li et al., 2021). High accuracy weighing lysimeters can be an option to quantify dew amount into ecosystems (Riedl et al., 2022; Ucles et al., 2013). CO₂ fluxes can be measured by eddy-variance (Eugster and Siegrist, 2000) and laser (Maier et al., 2022) approaches. For alpine ecosystems, a challenge is the topographic variability that induces large uncertainties of CO₂ fluxes (Hammerle et al., 2007). Due to the low quality of CO₂ fluxes during nighttime by eddy-covariance measurements, we could not assess the dew effect in darkness. Furthermore, the benefits of dew on ecosystems can continue after dew drying out on vegetation surfaces, which could be traced by high-resolution measurements of H₂O and CO₂ isotopes, but was not possible in this study based on 3–13 h intervals of destructive isotope sampling. Therefore, additional methods, e.g., synchronized and continuous laser measurements of H₂O and CO₂ fluxes (Li et al., 2021; Maier et al., 2022) and their isotopic fluxes (Siegwolf et al., 2021) at both plant and ecosystem scales need to be explored to assess the long-term benefits of dew at plant (e.g., among species difference) and ecosystem scales.

Plant water stress can be induced by low soil moisture or high atmospheric water demand (Liu et al., 2020). In this study, leaf water status of Poa was dependent on soil moisture ( Figure 9B ), whilst LWP of Alchemilla, Taraxacum, and Trifolium were mainly controlled by atmospheric humidity conditions ( Figure 9C ). Among the four genera, Poa was most substantially affected by heat-drought stress as indicated in their lower LWP ( Figure 6 ), higher δ¹⁸O_lw and δ¹³C_ls ( Figures 7A , 8A, B ). This could be induced by the reliance of Poa on soil moisture ( Figure 9B ), and lower foliar water uptake of Poa ( Figure 8D ). More severe drought stress of Poa induced their lower WUE (higher δ¹³C_ls; Figure 8F ) as compared to Taraxacum, and Trifolium. Palmately-lobed and hairy leaves of Alchemilla might prolong the dew water retention on their leaves, and thus induced stronger foliar water uptake ( Figure 8D ) and slightly increased LWP ( Figure 6 ) compared to the other three genera.

Both Alchemilla and Trifolium depended on shallower soil water depth ( Figure 7B ), but they probably benefited from dew water to maintain their plant water status ( Figure 6 ) in response to heat and drought stress. In the case of Trifolium, the hairy trichomes on the edges of the leaves probably promoted foliar water uptake ( Figure 8D ). Many high-elevation plants have hairy structures to help reduce water loss, reflect excess radiation, and protect plants from pathogens (Zeng et al., 2013; Hamaoka et al., 2017). The control experiment at the same site by Prechsl et al. (2015) found that C₃-grasses did not shift to deeper soil water under drought treatment, indicating that high-elevation plants could benefit from leaf structures regulating their energy and water balances. On the contrary, with deeper soil water sources, the plant water status of Poa strongly declined ( Figure 6 ) in response to heat-drought stress. The maintenance of Taraxacum LWP ( Figure 6 ) in response to heat-drought stress might be beneficial from their waxy leaf surfaces ( Figure S1 ) and deeper soil water uptake ( Figure 7B ). These results indicated that both soil moisture and atmospheric conditions can affect the ecosystem carbon and water exchange under heat-drought stressed conditions through their varied influence on different plant genera. Due to the similar δ¹⁸O values for dew and soil water ( Figure 7B ), we could not split the contributions of dew (foliar water uptake) and nighttime plant tissue refilling (root water uptake; Schulze et al., 2019) on plant water status using the natural-conditioned data, because both root water uptake and foliar water uptake could occur during nighttime, and dew water can also drip off to the soil (Dawson, 1998) and disturb the isotopic signal of root water uptake fluxes. But our chamber-tracer experiment indicated that the contribution of dew on leaf water can vary from 3% to 14% ( Figure 8D ). We did not isolate the soil from the vegetation when amending tracer on the grassland plots, hence the tracer could have been directly applied on the soil, and the tracer sprayed on leaf surfaces could also have dripped into the soil. But according to the slight depletion of δ¹⁸O_lw in our chamber-tracer experiment, the drip-off effect of dew was probably minor compared to direct foliar uptake of dew and atmospheric water vapor. Based on the facts of similar isotopic signal of dew and soil moisture in natural conditions, previous research used excised leaves and isotopically depleted/enriched water to distinguish the two (root and foliar) water sources (Kim and Lee, 2011; Goldsmith et al., 2017). However, these controlled experiments were performed with self-made chambers acting as a heat trap preventing radiative cooling, which is the most important driver of dew formation in natural conditions (Curtis, 1936; Li et al., 2023), and may thus not reflect natural conditions. Future research should therefore apply approaches that allow to estimate the dew influence on plant water under varying soil moisture conditions in the field (Li et al., 2021).