The role of the leaf to air vapor pressure deficit (VPD) is increasingly recognized as a leading limiting factor to determine plant gas exchange (Novick et al., 2016; Grossiord et al., 2020), while the influence of climate change on VPD rise becomes more and more evident. Anthropogenic-related global warming has increased unceasingly by 0.2°C every decade for more than 30 years (IPCC, 2018). A sharp increase of VPD was observed in the last decades (Yuan et al., 2019) and climatic projections show a future scenario where VPD increase is a diffuse phenomenon in many areas of the globe (Ficklin and Novick, 2017; Barkhordarian et al., 2019).

Higher VPD is in general detrimental for plant productivity because of the negative exponential response of stomatal conductance g_s to increasing VPD (Schulze et al., 1972; Jones, 1992; Oren et al., 1999). This response limits plant dehydration, but, at the same time, reduces carbon assimilation once stomata are partially closed. This explains why VPD was identified as a climatic factor related to tree mortality (McDowell et al., 2013) and limits crop productivity (Eamus et al., 2013). Prolonged periods of high VPD have been related to reduced growth (Sanginés de Cárcer et al., 2018) or large-scale forest die-off (Breshears et al., 2013; Will et al., 2013). Similarly, major crops such as maize, wheat, soybean, and fruit trees were reported to have decreased yields in response to high VPD (Challinor and Wheeler, 2008; Lobell et al., 2011; Zhao et al., 2017; Hsiao et al., 2019). Because plants are at the base of human life, the impact of changing climate and VPD rise on food provided by plants are a main issue in the upcoming future.

The response of plants to VPD variation has been extensively studied both at the leaf and at stand level. The latter is more suitable for assessing the effect of different management systems and structures in a specific site/orchard. At stand level, the surface canopy conductance G_surf (mol m^–2s^–1) refers to the capacity of gas exchange of the whole canopy expressed per unit of ground area, which includes a wide variety of heterogeneous leaf conditions (Medrano et al., 2015) and the community structure in terms of leaf area, roughness, soil properties, and species (Grossiord et al., 2020). In stands where the decoupling coefficient Ω is low (Jarvis and McNaughton, 1986), stomata experience a VPD close to the free atmosphere and the total canopy conductance is assumed to be driven largely by stomata aperture (Köstner et al., 1992; Hogg and Hurdle, 1997; Chapin et al., 2002). Thus, canopy conductance G_surf can be estimated from the rate of canopy transpiration (Ec) and VPD simply as: G_surf = k × Ec/VPD (Köstner et al., 1992; Arneth et al., 1996) where k is a temperature-dependent parameter (see below). G_surf includes two components: the stomatal component (average of different canopy leaves) and the aerodynamic component, which is a function of the orchard structure (i.e., density, height, roughness). Thus, when it is considered on a relatively long-term scale (weeks, months), G_surf can be a proxy for carbon uptake of the community, which, in turn, should also determine the carbon pool for reproductive allocation and seed production. In its turn, maximum value of G_surf (i.e., G_max) indicates the potential gas exchange of the whole canopy.

The potential conductance is reached under non-limiting conditions of soil water availability, light, and optimal temperature. In forests, these conditions are rarely met and G_max is generally estimated. Thus, Oren et al. (1999) proposed to use the G_surf reached at 1 kPa (G_surf@1kPa) as reference maximum conductance. G_surf@1kPa is highly variable across ecosystems because it depends both on the stand characteristics and species. In natural forest formations, it ranges from 0.2 to 0.7 mol m^–2s^–1 from savannah to deciduous forests (Grossiord et al., 2020). Tang et al. (2006) reported G_surf@1kPa 0.024 (mol m^–2s^–1) for some broad leaves of the genera Betula, Acer, and Ostrya and Fernández et al. (2009) from 0.28 to 0.44 in Nothofagus antarctica and Diostea juncea. On conifers, Köstner et al. (1996) and Köstner et al. (1996) reported G_surf@1kPa = 0.16–0.4 (mol m^–2s^–1) in Pinus sylvestris.

At the same time, some studies showed that many species, including hazelnut, present G_max occurring before 1 kPa (Tang et al., 2006; Herbst et al., 2008). Some other species reach G_max even above 1 kPa such as for example Pseudotsuga menziesii, where the G_max = 0.64 (mol m^–2s^–1) is reached at about 1.8 kPa (Fernández et al., 2009). Also some studies on leaves of fruit trees as olive (Rodriguez-Dominguez et al., 2019), walnut (Rosati et al., 2006), and different apple cultivars (Massonnet et al., 2007) show that G_max is reached between 1.5 and 2 kPa, probably related to the incidence of the fruit load in the whole tree physiology. Indeed, a correct approach to define the maximum potential gas exchange would be to consider the actual G_max, which may occur at different VPD according to the species or stand management.

Still, the reference G_surf@1kPa is often used to determine the sensitivity of a species. The sensitivity of stomatal response (even at canopy level) to VPD refers to the relative reduction in G with increasing VPD (i.e., -dG/dlnVPD). It is commonly believed that high G_surf@1kPa predicts high sensitivity (Oren et al., 1999). The sensitivity is measured as the slope (m) of the function G_surf = G_surf@1kPa – m lnVPD. G_surf@1kPa and m seem highly correlated (average R² = 0.75) with a slope of approximately 0.6 across species. In other words, the higher the G_surf@1kPa, the faster the stomata close with VPD increase, compared to species or individuals that present a lower G_surf@1kPa. However, it is likely that the sensitivity changes whether we consider G_surf@1kPa or G_max.

In this study, we explored the relationship between VPD and whole tree conductance G_surf as a contribution to understanding the limitations of potential productivity in eight commercial hazelnut orchards distributed in different countries. These sites have been considered as eight representative orchard conditions in a range of bearable climates for hazelnut growth, going from mild temperate to dry warm. Increasing VPD connected to global warming (IPCC, 2018; Barkhordarian et al., 2019) will certainly threaten these areas. Thus, we aimed at: (I) identifying the maximum canopy conductance and its sensitivity to VPD at each site and (II) providing a procedure to quantify the maximum potential gas exchange of a site by knowing the frequency of occurrence of VPD values.

The assessment of critical thresholds of VPD that guarantee optimal G_surf will become of great importance in defining suitability maps to locate new plantations and predict the responses of current crops to future climate change.

The comparative study of a species grown in different sites and under different orchard managements is a useful experimental setting to understand orchard responses under climate warming and VPD rise in the field. These study areas had different mean daily VPD during the growing season and despite irrigation, the soil water content was also affected. Australian sites A0 and A1 showed the most severe evaporative conditions with peaks of 5.2 kPa and daily mean temperature up to 46.6°C. These sites could represent the worst climatic scenario in which hazelnut can grow, together with a soil water content of 0.09 m³ m^–3. Italy was also an area with high VPD together with the Australian site A2 located at a higher altitude with respect to A0 and A1. All these regions combined with high daily global radiation in the southern hemisphere represent a very challenging environment for hazelnut that potentially limits the canopy conductance and growth. At present, France (F1) and Chile (C1 and C2) are in a milder condition, but climatic projections forecast increasing atmospheric water demand at global level (i.e., higher VPD) with extreme cases in the southern hemisphere related to the dry phase of El Niño cycles (Zhang et al., 2015; Barkhordarian et al., 2019). Thus, even milder sites may become challenging places to grow hazelnut and it is crucial to know how orchards will respond to these changes in order to adapt the cultivation systems.

The hazelnut response of G_surf to VPD across sites showed some differences between orchards in terms of both the pattern and G_max values. Differences in G_max are representative of wide range of LAI, which developed in orchards; thanks to different management strategies (spacing and training system). Different orchard managements were reported to have a significant effect also on grapevines gas exchanges (Prieto et al., 2020). Because the G_surf at the leaf area level is similar between sites, orchards with higher LAI have higher total G_max at the community level. Indeed, orchards with lower LAI, as, for example, A0 and A1 (0.99 and 0.70, respectively), responded to the VPD similarly to grassland or savannah species (Grossiord et al., 2020). In other words, in sites A0 and A1, the whole tree stomata are not only more tightly closed at low VPD with respect to other sites, but also the velocity to close further is low per unit of increase in VPD. This effect is further enhanced by the low soil water content in site A1, where despite irrigation being homogeneously distributed, the higher percentage of sandy soil together with high temperatures severely compromised the water uptake capacity of trees. In these areas, the orchard management shall favor higher orchard LAI. Most of the other sites present values that are closer to deciduous forests. Indeed, absolute G_max agrees with findings on other broad leaves reported in the literature (Barradas et al., 2005; Tang et al., 2006; Bourne et al., 2015; Grossiord et al., 2020). Another factor that we find linked to the differences in G_max between sites is the fruit load. Orchards with higher fruit load had also higher G_max. Accordingly, other studies on orange trees showed that low fruit loaded trees show < 40% stomatal conductance respect to fully loaded trees (Syvertsen et al., 2003). However, it has to be considered that fruit load and LAI, i.e., a proxy of the total photosynthetic capacity of the orchard, should be autocorrelated (Wünsche and Lakso, 2000).

The pattern of G_surf variation with VPD agrees with the theoretical framework presented by Oren et al. (1999) who described the response of G_surf to VPD by an exponential response curve, where the sensitivity is expressed by the parameter –m and higher G_max values predict higher sensitivity. Accordingly, Australian sites resulted in the less sensitive to increase in VPD and with very low G_max.(average m = −0.03 mol m^–2s^–1 kPa^–1), while the rest of the sites had an average m = −0.17 mol m^–2s^–1 kPa^–1. This latter is close to what found for broadleaved forests (including hazelnut understory) according to Herbst et al. (2008) and for Quercus alba (Oren et al., 1999). Still, these two clusters of sites (Figure 3) highlight differences in the sensitivity within the same species, suggesting a moderate capacity to acclimate. In this sense, this study represents one of the few examples of conductance sensitivity measurements on the same species (Oren et al., 2001; Addington et al., 2004) and the first to compare a single species across continents.

The maximum canopy conductance G_max is an important parameter to define the optimal gas exchange capacity of a stand. G_max should be defined under favorable conditions for the species, i.e., non-limiting light, water availability, and optimum temperature. In this study, we are analyzing trees under standard irrigation and fertilization regimes, thus we can assume that G_max values are as close as possible to the potential for the specific site. Thus, G_max can be a reference value to determine the species sensitivity to the increasing VPD. Often, the reference G is G_surf at 1 kPa because it is difficult to obtain continuous and reliable measurements for low VPD. However, in our data collection, we could benefit from a long and continuous data series from sap flow measurements. In this study, G_max was observed to range widely across sites from A1 (0.091 mol m^–2s^–1) to G2 (0.58 mol m^–2s^–1), occurring on average at 0.57 kPa, thus at a lower VPD compared to the reference of 1 kPa proposed by Oren et al. (1999). Other authors such as Tang et al. (2006) and Herbst et al. (2008) observed similar values of G_max at VPD < 1 kPa in few temperate broadleaves including forests with hazelnut understory, while many fruit trees as olive, walnut, and apple reach G_max at higher VPD (1.5–2 kPa) (Rosati et al., 2006; Massonnet et al., 2007; Rodriguez-Dominguez et al., 2019). In this perspective, hazelnut can be considered as a species with a water-saving behavior, which typically takes advantage of low VPD to maximize stomata opening and the carbon uptake.

This behavior is also confirmed by the pattern of relative canopy conductance. In this study, not only the slope, but also the value of VPD at which G_max occurs is important to assess the sensitivity of a species in a specific stand structure. Thus, the sensitivity described in this study is not exactly comparable with that calculated with the reference G_@1kPa reported by Oren et al. (1999). Indeed, if G_max occurs at 0.5, we will obtain a less negative slope (i.e., lower sensitivity) respect to the case in which we consider the same G_max value at 1 kPa (Figure 6). We simulated a case in which G_max occurs at 1 kPa and we found m = −0.53, while when G_max occurs at 0.5, m = −0.39. This shows that force comparisons with G_@1kPa may show different results suggesting sensitivity higher than it really is. This might be the reason behind the lower slope value in the linear regression between the parameter –m and reference conductance with respect to found by Oren et al. (1999).

Indeed, the comparison of slopes between G_surf and dG_surf/dln VPD expresses the actual degree of sensitivity of the same species between orchard management. Sites A1 and A0 had the lowest sensitivity (less negative slope). These sites are also those with more extreme events of VPD (up to 8–9 kPa). Thus, we can hypothesize that the lowest sensitivity is related to acclimation in these sites and this acclimation capacity is cultivar independent. Despite the cultivar is different in sites A0 and A1, the pattern of G_surf to VPD has a steep increase at low VPD followed by a reduction suggesting that trees might have acclimate their leaf characteristics (e.g., number of stomata, size of the stomata, or others) across the growing seasons and this would maximize their gas exchange.

The interval of VPD in which trees could maintain over 80% of their maximum tree conductance in most of the sites is below 1.7 kPa. In all the sites, values above 5 kPa reduce the conductance to the minimum. In A0 and A1, the interval is particularly narrow. This, together with the low sensitivity in these sites, is a further clue that suggests an acclimation capacity of trees. In other words, hazelnut might have responded to the high site VPD by reducing stomata opening and improve the water use efficiency. Also, G2 resulted in a site with narrow interval of VPD. This has in common with A0 and A1 very high minimum temperatures of the air in this study period even at low VPD. This might have somehow forced the plants to further reduce their maximum canopy conductance capacity at low VPD. The chance that hazelnut can develop acclimation to extreme climates is of great interest for orchard managers and for the food industry, which relies on world nut supplies. Indeed, if the mechanism of stomatal acclimation would miss, VPD variation due to climate change could reduce stomatal conductance by 10–50% (Ficklin and Novick, 2017).

When we consider the implication of the higher or lower sensitivity to the carbon economy of the plant, it is also important to consider the temporal scale (Martínez-Vilalta and Garcia-Forner, 2017). To test what is the impact of the VPD₈₀ width on the total canopy conductance at the stand level, we estimated the total GEC of each orchard based on the occurrence of VPD₈₀ during the growing season and the corresponding mean G_surf in this interval.

Most of the sites had on average 2 weeks (cumulated hours) per growing season in favorable climatic conditions for conductance maximization, i.e., VPD₈₀ while for the rest of the season they are below the 80% of G_max. F1 and A2 had more than 3 weeks suggesting that from the climatic perspective, these are very promising sites for hazelnut, while A1 has a very limited amount of time during the growing season to optimize the conductance. The situation of site A1 is further worsened by the extremely low G_surf in the interval of VPD₈₀, which makes the total GEC of the site undesirable. While A1 remains an extreme case, the compensation between the occurrences of VPD₈₀ and G_surf supports the hypothesis of some acclimation capacity of the species, which leads to a quite even GEC at site level (about 0.45 Mmol m^–2). This acclimation might have been favored by the orchard management that favored a compensatory LAI through pruning and tree density. The extremely high GEC in site F1 is indeed related to the high LAI, close to a forest one, combined with very frequent VPD₈₀. When we normalize GEC by LAI, we get the leaf area unit gas exchange, independently to the orchard structure. What is clear is that the GEC_leaf is highly related to the frequency of VPD₈₀, but the orchard structures with higher LAI have, as a consequence, higher whole canopy gas exchange.

If the occurrence of VPD₈₀ remains stable in future growing seasons, we can expect most sites to maintain a good level of potential productivity at comparable growing conditions. Still changes in VPD might overturn the suitability of some areas. In this perspective, VPD must be monitored. At the same time, the orchard management shall favor high G_surf. This result can be achieved by increasing the LAI by favoring higher trees or by increasing the density where trees are small or still in the early productive stage. Indeed, the higher the fruit load, the higher the G_surf (Syvertsen et al., 2003). We believe that by collecting longer data series of fruit load combined with physiological and structural parameters of different orchards is a paramount goal to set in the future studies, perhaps toward modeling techniques as proposed by Prieto et al. (2020).

These results on suboptimal interval of VPD combined with the whole tree conductance (potential assimilation capacity) are the basis for a new global perspective on crop management. Indeed, they can be the basis for the construction of physiologically-based suitability maps: by using the projections of VPD conditions (Akpoti et al., 2019), it would be possible to roughly estimate the potential reduction of gas exchange and productivity in a given site.

Nonetheless, this study might provide guidelines for the management of existing orchards. We expect that the more the orchard is structured to maintain a range of favorable VPD conditions (within the VPD₈₀ range), the higher the carbon stocked in trees available for nut production will be. Indeed, the canopy structure variation may largely influence the water use efficiency of orchards (Cohen and Fuchs, 1987; Cohen and Naor, 2002).

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

TA and GP discussed and developed the idea and underlying rationale. GP elaborated the data, proposed the analytical approach, and wrote the first draft. VC was in charge of the technical support of all the devices from the installation phase to the maintenance and dismantling, created the thermal dissipation probes (Lab of the University of Padova, Department of Territorio e Sistemi Agro-Forestali), downloaded, and prepared the raw data. All the authors commented and revised the final version of the manuscript.

This study received funding from Hazelnut Company, Division of Ferrero Group. The funder had the following involvement with this study: the study design and the location of study areas were agreed with the funder, besides the funder partly supported the field data collection. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The handling editor declared a past co-authorship with one of the authors ES.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.