This section summarises the results of the publications by category,
i.e., water, then carbon fluxes and WUE.
1. Water fluxes
The increase in atmospheric evaporative demand, induced by climate warming, compels winegrowers to optimise water resource management, especially in arid regions. In this context, flux station measurements provide a very good understanding of the amount of energy converted into evapotranspiration in irrigated vineyards, measured as latent heat (LE = ET in energy units). The integration of complementary instruments (
e.g., sap flow sensors and lysimeters) also allows the partitioning of ET between vine transpiration in planting rows and soil evaporation in interrow (with or without herbaceous vegetation). This makes possible the identification of the main drivers of these fluxes, the testing of ET spatialization strategies, and the validation of more accessible instrument systems.
1.1. Partitioning of net radiation
Key aspects addressed in “water” studies include how available energy (
i.e., Rn-G) is converted when reaching the vineyard surface. In particular, the fraction transformed into latent heat (LE), which corresponds to ET = LE.λ-1, where λ is the latent heat of vaporisation. The Bowen ratio (β = H/LE; Bowen, 1926) is the most widely used metric to describe this partitioning. Under non–water-limited conditions, the ratio declines linearly as evaporative demand increases (Wilson et al., 2002a), making it a suitable agricultural indicator relative to plant hydric stress. During the vine growth period, three studies of the corpus have reported monthly averaged β close to 1 (
i.e., H = LE) in irrigated vineyards located in the province of Gansu, China (Gao et al., 2020; Li et al., 2009; Zhao et al., 2017). These studies report Bowen ratios ranging from 0.65 to 1.2, with an overall mean of 0.84, indicating that 57 % of the available energy was used for ET. This value is close to the full-annual averages reported for annual crops in the FLUXNET2015 database (β = 0.85), which are among the few ecosystems showing mean values below 1 (
e.g., forests range from 1.42 to 2.25; Lin et al., 2022). However, Chen et al. (2024) reported lower values in another vineyard located in the same region (β = 0.14 and 0.25,
i.e., 89 % and 80 % used for ET), despite similar climatic conditions and irrigation regimes. These differences are attributed to vineyard planting systems,
i.e., VSP versus the single pergola system. The latter likely enhances ET by (i) increasing vine transpiration through greater leaf exposure (as β is negatively correlated with leaf area index: Gao et al., 2020; Zhao et al., 2017), and (ii) increasing soil water availability due to interrow shading under the trellised canopy. There is also a strong seasonal variability in β, driven by water availability, vine phenology and atmospheric evaporative demand (Gao et al., 2020; Li et al., 2009; Zhao et al., 2017). Using 4 years of measurements in a furrow-irrigated vineyard,
Zhao et al. (2017) reported a clear U-shaped pattern, similar to that observed in many northern hemisphere ecosystems, although with a different magnitude (Lin et al., 2022). Prior to budbreak, H dominates (β > 1). After budbreak, β decreases sharply as atmospheric evaporative demand increases, and the vine develops its leaf area. When the canopy achieves its maximum development, β stabilises around 0.5 (= 67 % used for ET), reflecting a balance between available energy and non-limited evaporative processes. The sustained evaporative regime is maintained by irrigation, which plays a key role in controlling energy partitioning. In this context, drip irrigation appears more effective than furrow irrigation in reducing β (Gao et al., 2020).
Advection of heat and energy fluxes can affect the local evaporative demand, with potential impacts on EBR and β. In summer, when a strong thermal gradient exists between two adjacent ecosystems (
e.g., a relatively cool irrigated vineyard and a relatively hot grassland), heat is advected toward the cooler site. These episodes are frequent in vineyards, particularly those located in arid and heterogeneous agricultural regions (Kustas et al., 2022). Heat advection introduces two main issues. First, the supply of hot and dry air increases atmospheric evaporative demand, leading to higher water loss through LE in irrigated vineyards. Chen et al. (2024) showed that nighttime irrigation reduces these losses by limiting thermal gradients compared with morning irrigation. Second, lateral heat advection increases EBR because most methods only measure vertical fluxes. This leads to strong uncertainty in LE calculations (Bambach et al., 2022; Chen et al., 2024) when LE is measured by various non-EC approaches relying on LE as a residual of energy balance. All these results highlight β and EBR as a comprehensive indicator for adaptive strategies applied in vineyards.
1.2. Evapotranspiration partitioning
Partitioning ET into vine and grass transpiration (T) and soil evaporation (E) is another active research focus, but the EC technique cannot directly disentangle these two components. To make this possible, complementary instruments are needed, such as micro-lysimeters for E and sap-flow systems for vine transpiration (T
vine). Generally, a high T/ET ratio is targeted because it indicates that the vine uses most of the available water rather than losing it through E. Studies of the corpus consistently showed that T is the dominant water flux in arid vineyard ecosystems, although not to the same extent. Its contribution to ET ranges from 56 to 86 % during the vegetative period (Campos et al., 2025; Gao et al., 2019; Kool et al., 2016; Zhang et al., 2024; Zhao et al., 2017).
Zhao et al. (2017) further demonstrated that T
vine accounts for most of ET. However, in some cases, E can exceed T when soil water content is high, particularly under generous irrigation practices such as furrow irrigation (Zhao et al., 2018). A strong relationship was also reported between interrow water flux and soil water content under both bare soil conditions (ET = E; Kool et al., 2016) and with cover crops (herbaceous T increases strongly after a rainy event; Campos et al., 2025). In contrast, T
vine is less directly driven by soil water reserves, although prolonged water stress can inhibit vine–atmosphere H
2O exchange, sometimes irreversibly (Shtai et al., 2024). Campos et al. (2024) reported strong diurnal variations in the T
vine/ET ratio under drought, as the T
vine contribution to ET changes from > 75 % to < 25 % between the morning and late afternoon because of stomatal closure.
Overall, these results, combined with the previous findings regarding β, highlight the importance of a precise irrigation system focused on vine rows to avoid water losses. These findings also call for further investigations into the specific impact of the vine's biological mechanisms on vineyard water fluxes, a topic that remains poorly addressed in the corpus. In particular, canopy conductance has been identified as a key driver of T
vine/ET, explaining 50 % of its seasonal variability at the daily timescale and up to 90 % at the monthly timescale (Zhang et al., 2024).
1.3. Toward more accessible methods to estimate ET
Monitoring ET represents a major challenge for viticulture. This is true for both irrigated regions, where optimising water inputs is crucial, and non-irrigated regions, which are becoming increasingly vulnerable to severe droughts due to climate change. Another challenge is the high-resolution spatialization of ET, as it is modulated by factors such as topography, soil properties, and exposure, which can vary greatly across and within vineyards. Several studies in the corpus have used flux station measurements to develop and validate more affordable alternatives for estimating ET.
Commonly used alternatives to the EC-based measurements from flux stations indirectly derive ET as a residual of the energy balance equation by measuring its other major components, mainly Rn, G, and H. Rn and G can be measured with dedicated instruments,
e.g., a radiometer and soil heat flux plates. For H, we tested two methods and compared them with ET measured directly by flux stations: scintillometry and surface renewal. Scintillometry, a relatively expensive method, analyses the scattering of an optical signal by turbulence above the canopy to measure H. It provides reasonable mesoscale ET data over flat terrain (Perelet et al., 2022; Vendrame et al., 2020). The surface-renewal method, more affordable, uses rapid temperature changes at a single point to estimate local H, without needing specific temperature ramp calibration for vineyard conditions (Parry et al., 2019; Shapland et al., 2012; Tosoni et al., 2021). However, both methods remain relatively inaccessible for routine use. Another alternative to flux station measurements is the Micro Bowen Ratio (MBR) system. The MBR system applies the β ratio at the microscale to derive H and ET (Hossein Ashktorab et al., 1989). It is relatively inexpensive, easy to deploy, and well-suited for small plots, but still requires multiple MBR units to capture the spatial heterogeneity of vineyard ecosystems adequately (Holland et al., 2013; Kustas et al., 2019). The main drawbacks to these alternatives compared with EC-based flux stations are: (i) the absence of direct measurements of water vapour transport, (ii) their strong sensitivity to advection, and (iii) the neglect of other energy storage terms (
e.g., canopy heat storage) in the overall energy balance equation. Because ET is estimated from the residual of such an equation, all the closure errors may manifest as errors in ET. This poses a major issue in viticulture, which is commonly subjected to frequent advective episodes in summer.
One other option is to calculate ET using modelling. The FAO Penman-Monteith (FAO-PM; Allen et al., 1998) has been widely used. This model estimates potential evapotranspiration (PET) based on commonly measured climatic variables, such as net radiation, air temperature, and wind speed. PET represents the ET of a well-watered, homogeneous reference grassland. ET for a specific crop can then be approximated by applying a crop coefficient (Kc) to the PET as follows (Equation 2):
E
T
=
P
E
T
×
K
c
E
q
u
a
t
i
o
n
2
Although Kc values for vineyards have been reported in the literature, they were obtained under very specific pedoclimatic and viticultural conditions (Allen et al., 1998). ET measurements from flux stations enable the determination of reliable, site-specific Kc values that winegrowers can use to estimate vineyards' ET easily. For example, this approach has been implemented in Portugal (Ferreira et al., 2012), Mexico (Er-Raki et al., 2013; Zermenio-Gonzalez et al., 2017), and Chile (Villagra et al., 2014).
More complex process-based models can also estimate ET without a flux station. These models account for vineyard ecosystem heterogeneity at two levels of complexity.
Two-source models simulate vine and soil water fluxes (Li et al., 2014; Ortega-Farias et al., 2006), while
three-source models also include under-row fluxes (Montes et al., 2014; Poblete-Echeverría & Ortega-Farias, 2009). 22 studies in the corpus applied spatial upscaling methods with remote imagery (
i.e., satellite, aircraft and drone) to improve ET model prediction compared to flux station measurements (
e.g., in Aboutalebi et al., 2022). These spatialized modelling methods hold considerable potential for the development of future tools, either to assess global viticultural budgets or to investigate within-vineyard processes,
e.g., the impact of different irrigation regimes on ET within a single vineyard (Burchard-Levine et al., 2025). Readers interested in more details on applications in vineyards are referred to the studies conducted as part of the GRAPEX project listed in Kustas et al. (2022).
In addition to process-based models, two recent studies from the corpus used data-driven modelling to estimate ET in vineyards with a high degree of accuracy (Fuentes et al., 2024; Garcia-Rodriguez et al., 2024). This approach relies on large datasets of flux station measurements and environmental variables from which statistical relationships are derived using machine learning.
2. Carbon fluxes
Studies on vineyard carbon fluxes have primarily focused on measuring the net ecosystem exchange (NEE). A negative NEE value indicates that the vineyard acts as a carbon sink (
i.e., GPP > R
eco), while a positive NEE means it is a carbon source (
i.e., GPP < R
eco). These studies aimed to (i) determine whether vineyard ecosystems function as carbon sinks or sources on an annual timescale, (ii) characterise the seasonal dynamics of NEE, and (iii) identify the main drivers of carbon flux.
2.1. NEE annual budget and temporal dynamics
With two exceptions, all sites reported negative annual NEE budgets, giving an average carbon sequestration of –117.6 g C m
-2 year
-1 (Figure 5a). The mean value of Veneto-SB is the lowest in terms of carbon uptake (–46.3 g C m
-2 year
-1). This site and Lombardy-CH also show strong interannual variability, including (i) positive budgets in 2019 (+135.5 g C m
-2 year
-1) and 2022 (+48.5 g C m
-2 year
-1), respectively, and (ii) two successive near-neutral budgets for Veneto-SB (–14.5 and –2.5 g C m
-2 year
-1 in 2016 and 2017). Figure 5b shows that, compared to other ecosystems, the five vineyards sequester carbon at comparable levels (
p-value > 0.05) to grasslands (–22.3 g C m
-2 year
-1), shrublands and savannas (–105.6 g C m
-2 year
-1) or annual crops (–174 g C m
-2 year
-1), but less than forests (–333.4 g C m
-2 year
-1).
Figure 5. Annual NEE budgets from NEE_Y of (a) vineyards of the corpus, and (b) other ecosystems from the FLUXNET2015 database located in the 35–45°N band.
a: The NEE value of Veneto-SB for 2015* was provided by
Tezza et al. (2019), and the others came from the ICOS IT-Lsn flux station, with the 2021 annual budget excluded due to quality issues.
Valenti et al. (2019) provided the NEE value for Lombardy-CH for 2017*, and the others came from the FLUXNET IT-Arz flux station, with the 2020 annual budget excluded due to quality issues. The error bars of both sites show the associated random uncertainty of each annual budget (NEE_RANDUNC). The average NEE annual budget (indicated by a vertical red dashed line) was also computed for the whole period.
b: The FLUXNET2015 data displayed in the boxplot are annual values averaged for the full period of study per site. The result of a Kruskal–Wallis comparison of these values across ecosystems is displayed in bold letters, with similar letters indicating non-significant differences (
p-value < 0.05). The number of sites displayed in the boxplots (n) is indicated in italics.
![]()
Beyond the annual NEE budget, the four studies that provide 12 consecutive monthly values (
i.e., NEE_Y without Mezzolombardo-TR) reveal a marked seasonality (Figure 6a). The pattern is comparable to that of other extratropical ecosystems such as deciduous forests, with a V-shaped pattern (
e.g., Ullah et al., 2025). Vineyards act as carbon sinks in summer due to active photosynthesis (NEE < 0), and as carbon sources in winter during vine dormancy (NEE > 0). The end of the uptake period occurs in mid-autumn across sites, while its onset varies from early spring to early summer depending on local conditions.
b. Data for (i) 2015* were from
<xref ref-type="bibr" rid="ref99">Tezza <italic>et al.</italic> (2019)</xref>, and (ii) 2016-2023 are provided by the ICOS IT-Lsn flux station, with 2021 discarded due to quality issues.
c. Data for (i) 2017* were from
Valenti et al. (2019) and (ii) 2017–2022 are provided by the FLUXNET IT-Arz flux station, with 2020 discarded due to quality issues.
![]()
Figure 6. NEE seasonal cycle from (a) publications of the corpus, (b) Veneto-SB and (c) Lombardy-CH.
b. Data for (i) 2015* were from
<xref ref-type="bibr" rid="ref99">Tezza <italic>et al.</italic> (2019)</xref>, and (ii) 2016-2023 are provided by the ICOS IT-Lsn flux station, with 2021 discarded due to quality issues.
c. Data for (i) 2017* were from
<xref ref-type="bibr" rid="ref99">Valenti <italic>et al.</italic> (2019)</xref> and (ii) 2017–2022 are provided by the FLUXNET IT-Arz flux station, with 2020 discarded due to quality issues.
Only two sites from the corpus, Veneto-SB and Lombardy-CH, allow the examination of NEE interannual variability (Figure 6b and 6c). The onset of carbon uptake differs between the two sites, occurring later for Lombardy-CH (May–June) than Veneto-SB (April–May). This delay does not appear to be driven by varietal phenology, as Chardonnay (Lombardy-CH) is generally documented as an early-budbreak cultivar compared to Sauvignon blanc (Veneto-SB; Zapata et al., 2017). Climate also does not seem to explain this, as air temperatures were comparable between the two sites during the pre-budbreak period. This difference may instead reflect site-specific cultural practices (
e.g., later pruning at Lombardy-CH) or ground cover management (
e.g., inter-row grass cover at Veneto-SB). However, this information was not reported for both sites, preventing a robust conclusion.
Apart from this difference, the overall seasonal cycle of NEE is broadly consistent between the two sites and years. However, interannual variability is significant in summer, with standard deviation of monthly NEE reaching up to ~30 g C m
-2 month
-1 at both sites. Relative to the mean summer carbon uptake (–25.4 and –51 g C m
-2 month
-1 at Veneto-SB and Lombardy-CH, respectively), this variability represents 66 and 59 % of the mean flux magnitude and is thus sufficient to reverse the sign of NEE during some months. These results explain the high variability of the NEE annual budgets previously observed (standard deviation = 87.9 and 82.4 g C m
-2 year
-1 for Veneto-SB and Lombardy-CH, respectively; Figure 5). More multi-year data are required in different contexts to characterise this variability better.
2.2. Climate influence on NEE
Most “carbon” studies in the corpus investigated the influence of meteorological variables on NEE. These studies consistently identified global radiation (
i.e., the primary driver of photosynthesis; Zermeno-Gonzalez et al., 2015), air temperature, vapour pressure deficit (VPD; Badraghi et al., 2023; Guo et al., 2014), and soil water content (Tezza et al., 2019; Vendrame et al., 2019) as the main drivers of the NEE annual cycle in vineyard ecosystems. Among these factors, VPD and soil water content emerged as critical constraints on carbon sequestration during hot and dry compound events. High VPD combined with low soil water content is often associated with stomatal closure, which prevents gaseous exchange (Shtai et al., 2024). Wong et al. (2022) also identified short-term stressful events at their site through abnormal drops in GPP.
To better understand these processes and explore them across a more pedo-climatically diverse dataset, Figure 7 presents relationships between NEE and global radiation, VPD, or soil water content. Incoming solar radiation emerges as the strongest climatic driver, with a near-linear relationship explaining 57 % of the seasonal variability in NEE (Figure 7a). This pattern largely reflects differences among climatically contrasted sites (from humid subtropical,
n = 55, to Mediterranean,
n = 28) and primarily captures large-scale seasonal energy forcing. In contrast, VPD shows a non-linear influence on NEE (adjusted R
2 = 0.50), with carbon sequestration increasing with VPD up to approximately 4 hPa month⁻¹, followed by a decline at higher atmospheric demand (Figure 7b). This threshold-like behaviour is most evident in arid, semi-arid, and Mediterranean contexts, whereas the relationship appears closer to linear in humid subtropical sites, likely due to non-limiting water availability. Soil water content is the least explanatory variable (adjusted R
2 = 0.26; Figure 7c). A second-order polynomial model did not improve the fit compared to the linear model (adjusted R
2 = 0.26; not shown), highlighting the weak explanatory power of this variable in our dataset. This result is also influenced by large site-to-site differences in soil moisture regimes and the heterogeneous SWC gradient across the dataset. The existing literature further supports a complex relationship between soil moisture and NEE that cannot be fully captured by simple regressions applied to modelled climatic variables, which likely explains the limited performance observed here.
Despite the simplicity of the approach and the use of ERA-Land reanalysis, these results highlight the value of an open-access, multi-site, standardised dataset. It should be noted that in situ climatic data are not provided in the "carbon" studies of the corpus, which explains the use of reanalysis products in this context. A larger observational network would enable more detailed analyses, particularly of interannual variability and ecosystem responses to extreme climatic events, which are expected to increase in frequency under climate change (
IPCC, 2023).
Figure 7. Second-order linear polynomial regressions between monthly NEE from NEE_M and (a) cumulative global radiation, (b) VPD, and (c) soil water content of the first layer.
The grey area shows the 95 % confidence interval of the regressions. Colours indicate the membership of a site to a Köppen–Geiger class (see Figure 3 for details). The violin plots in the top part of the figure display the distribution of ERA5-LAND climatic values extracted at each site within a Köppen–Geiger class. The respective number of monthly NEE data (n) accounted for is given on the y-axis. The text in the bottom panels displays the equation of the model together with the adjusted coefficient of determination (R²
adj) and the
p-value (P).
![]()
Vineyard NEE is also indirectly driven by climate through its influence on vine phenology. For example, the onset of vine carbon uptake relies on the bud break date, which depends on the variety-specific chilling requirement and its satisfaction, as well as on average air temperature at the end of dormancy (Zapata et al., 2017). Anomalous cool winter-to-spring transition may delay the carbon uptake period. Conversely, abnormally warm springs accelerate the vegetative processes and increase the risks of late frost, which causes temporary/permanent damage to leaf-related organs. According to
Valenti et al. (2019), the occurrence of such an episode in winter 2017 explains the persistence of a near-neutral NEE monthly budget measured in spring at Lombardy-CH (Figure 6c).
2.3. Influence of viticultural practices on NEE variability
Viticultural practices, such as soil management and organic amendment addition, influence vine phenology (Poni et al., 2022) and the dynamics of soil organic matter (Payen et al., 2021). They also directly and indirectly affect the CO
2 fluxes of the vine (Poni et al., 2008). However, the impact of these practices on carbon fluxes has been little investigated in the literature (
e.g., Tezza et al., 2019). This gap is likely due to the difficulty of disentangling the effects of management from the effects of climate when relying on a single flux station and to the limited temporal coverage of data.
Nevertheless, some studies in the corpus provide initial evidence of such impacts. Marras
et
al. (2015) reported that the strong carbon release observed at the Sardinia-VE site in March and April (Figure 6a) may be related to soil tillage, which (i) removed the interrow vegetation, thereby affecting the vineyard GPP, and (ii) increased R
eco via heterotrophic respiration from soil microorganisms. Maintaining grassed interrows was frequently highlighted in the corpus as a key management practice to enhance carbon sequestration (Scandellari et al., 2016; Tezza et al., 2019), particularly during vine dormancy when photosynthesis by adventive and sowed plants is still active (see winter sequestration at the Lazio-MO site in Figure 6a). Callesen et al. (2023) estimate that grassed interrows accounted for 60 % of the annual carbon uptake of a mountainous vineyard. Additionally, these intermediate crops significantly increased soil organic carbon stocks through root rhizodeposition (
McGourty & Reganold, 2005). This effect can be sustained through conservative soil management practices (Payen et al., 2021).
3. Water use efficiency
Only five studies in the corpus investigated the relationship between water and carbon fluxes using the WUE approach (Campos et al., 2024; Gao et al., 2019, 2021; Li et al., 2015; Zhang et al., 2024). A relatively high WUE indicates an efficient system, in which a large amount of biomass is produced with minimal water use. WUE is sensitive to the spatial scale considered (
e.g., plant or plot; Tallec et al., 2013), which complicates comparisons across sites. Several studies computed an ecosystem-based WUE (
i.e., WUE
eco = NEE/ET or GPP/ET; Campos et al., 2024; Li et al., 2015), but its interpretation is difficult because it integrates diverse physical and biological processes (
e.g., GPP, Reco, vine and grass transpiration, under and inter-row evaporation). WUE is more robust when applied at the plant scale (
i.e., WUE
plant = GPP/T with T: vine transpiration), although it requires complementary instruments such as sap-flow sensors (Gao et al., 2019; Zhang et al., 2024) or advanced water flux partitioning methods (Zahn et al., 2022).
However, the studies cited above identified strong relationships between climatic variables and WUE, which were mediated through their effects on carbon and water fluxes (
e.g., global radiation, air temperature, wind speed, VPD, and soil water content). Water availability was identified as a key driver that affects water and carbon fluxes differently (Gao et al., 2019). When water is supplied through precipitation or irrigation, ET increases mainly due to E (since the latter is more sensitive than T to soil water content; see section 1.2). As a result, WUE
eco decreases, reflecting a significant loss of water that does not contribute to vine productivity. For example, the average WUE
eco during the growth period of the vine in an arid, heavily irrigated vineyard (160 mm of rain and 237 mm of irrigation) was estimated to be between 1.2 and 1.6 g C kg
-1 H
2O
-1 (Gao et al., 2021; Zhang et al., 2024). In contrast, Campos et al. (2025) reported a higher value of 4.8 g C kg
-1 H
2O
-1 in a less-watered mountainous site during a hot climatic event (226 mm of rain and 63 mm of irrigation).
Biotic factors have also been shown to influence WUE
eco and WUE
plant dynamics, as both GPP and T are closely linked to plant stomatal activity, especially leaf area index (LAI; Gao et al., 2019; Li et al., 2015) and canopy conductance (Li et al., 2015; Zhang et al., 2024). Gao et al. (2021) investigated the relative importance of biotic and abiotic factors on the seasonality of WUE
eco and WUE
plant. They found that both were primarily driven by LAI at the beginning and end of the vineyard growing season, and by abiotic variables (net radiation and soil water content) during the rest of the year, emphasising the limiting role of water on carbon yield. In controlled situations, Patono
et
al. (2022) showed that drought-recovered grapevines maximise carbon assimilation and its delivery to sinks, resulting in significantly higher amounts of newly fixed carbon compared to both irrigated and water-stressed ones. In the vineyard dynamics, this mechanism may substantially influence the timing of carbon uptake, calling for further in-situ investigation of how vine physiological mechanisms influence WUE.
In summary, WUE is a relevant variable to study the objective of increasing carbon uptake and reducing water use in viticulture. However, it may be complex to use depending on the application scale,
i.e., vineyard or vine. Despite interesting results, especially regarding biotic factors, reliable conclusions of the studies remain difficult because of the lack of site diversity.