Linking Sap Flow and Trunk Diameter Measurements to Assess Water Dynamics of Touriga-Nacional Grapevines Trained in Cordon and Guyot Systems

: The present research aimed to evaluate the water dynamics of grapevines trained in Cordon and Guyot systems by coupling sap ﬂow and trunk diameter measurements under Mediterranean climate conditions. The study was conducted in a vineyard with Touriga-Nacional located at the Douro Valley, Portugal, during 2017. The results showed daily trunk diameter ﬂuctuations (TDFs), with the contraction, recovery and increment phases and higher sap ﬂow (SF) rates at earlier stages. Under harsh pedoclimatic conditions, SF was reduced and TDF ﬂattened. Rehydration and stomatal mechanisms were mostly associated with these responses. Guyot vines showed higher changes in TDF for the same SF values, whereas the TDFs of Cordon vines remained practically unchanged over maturation. Guyot vines generally showed increased values of cumulative increment and maximum daily trunk shrinkage. Although Guyot vines had a similar leaf area index (LAI), they showed higher SF / LAI ratios than Cordon vines. These results highlight the e ﬀ ect of the shorter length of the hydraulic pathways of the Guyot training system, in contrast to the higher trunk and the permanent horizontal branch (cordon) of the Cordon training system, indicating good adaptation to local pedoclimatic conditions. The study pointed to the complementary use of both techniques in the evaluation of grapevine water dynamics.


Introduction
Vitis vinifera is an environmentally sensitive species, mainly to climate and soil conditions [1]. Radiative, thermal and water stresses, generally combined, are the main causes of reduced growth and yield of vineyards in Mediterranean winegrowing regions, particularly under rain-fed conditions [2,3]. Furthermore, these regions, characterized by long growing seasons, mild-warm temperatures and dry summers [4,5] are likely to face a greater and intensified number of abiotic events in the near future [6,7].
Depending on the crop and variety, plants can control their water status by developing strategies to cope with critical conditions of atmospheric evaporative demand and soil water availability [8,9]. Water redistribution within the plant has been identified as a crucial survival strategy [10,11]. However, the mechanism generally considered most important for plant survival in a hot and dry environment is stomatal control to reduce transpiration [3,4,12,13].

Site Description and Plant Material
The study was carried out in a commercial vineyard, located at the Upper Douro sub-region (the eastern sub-region with harsher climate conditions) of the DDR, Portugal (41°04′18″ N, 7°04′51″ W, 160 m) during the 2017 growing season. The climate of this area is typically Mediterranean (warm temperate climate with dry and hot summer; see [40]) with a long-term average annual rainfall of 580 mm year −1 (mostly concentrated in the winter months) and average annual minimum and maximum temperatures of 10.0 and 19.8 °C respectively [41]. Meteorological variables were monitored by an automatic weather station (IMT280, iMETOS, Pessl Intruments, Weiz, Austria) positioned at the experimental site. From the weather data, reference evapotranspiration (ET0) following the FAO Penman-Monteith method [42] and vapor pressure deficit (VPD) were calculated.
Vitis vinifera L. Touriga-Nacional grafted onto 110 Richter rootstocks were planted in 2011 at a vertical (30% slope) spacing of 2.2 × 1.0 m (4545 vines ha −1 ), with E-W orientation. Two training systems were defined: unilateral (single) Cordon (trunk height of about 0.6 m) and single Guyot (trunk height of about 0.4 m) systems with similar numbers of buds (eight) left at winter pruning. From the selected field area, the closest six training consecutive rows (with just one path between training systems) were chosen, of which only the two center rows were used for measurements. The soil, essentially of schist origin with a loam-dominated texture, is classified as surribi-aric anthrosol [43].
The vineyard was managed (similarly in both training systems, e.g., soil fertilization) according to the grower's commercial cultural practices. Between rows, a cover crop was mowed regularly that became senescent during summer. To prevent plant death, around 6-mm events of water (from end-June to late-August; Figure 1), totaling 50 mm, were provided. For this purpose, a drip irrigation system was positioned in the center of the row, consisting of 2.2 L h −1 self-compensating drippers, spaced 1.0 m apart.

Sap Flow Measurements
The thermal dissipation technique [29,44] was used to continuously monitor SF within the xylem of Cordon and Guyot-trained vines over the growing season. Five vines per training system were chosen and the sensors installed in the north facing side of the grapevine's trunk. The SF system

Sap Flow Measurements
The thermal dissipation technique [29,44] was used to continuously monitor SF within the xylem of Cordon and Guyot-trained vines over the growing season. Five vines per training system were chosen and the sensors installed in the north facing side of the grapevine's trunk. The SF system consisted of a Agriculture 2020, 10, 315 4 of 15 pair of cylindrical probes, one heated and the other unheated, and an electrical current supplier (UP GmbH, Ibbenbüren, Germany). The probes were installed in smooth internode tissue of the trunk, avoiding irregular tissue at the nodes. The trunk was thermally insulated with standard styrofoam pipe, extending approximately 0.15 m above and below the probes. Reflecting material (thick aluminum foil) was applied around the probes and trunk (down to the ground) to thermally insulate the system and to minimize the formation of natural thermal gradients. In addition, four non-heated sensor-pairs per training system were installed in other selected vines (avoiding thermal effects between different sensors) in order to account for the remaining natural gradients. In this way, unheated patterns were applied to the heated sets and corrected temperature differences were then used [5]. A data logger (CR1000 with an AM25T multiplexer, Campbell Scientific Inc., Shepshed, UK) was used for measuring the temperature differential between thermocouples every 30 s, programmed to store mean values every 5 min.

Trunk Diameter Measurements
Trunk diameter fluctuations (TDF) were measured using linear variable differential transformers (LVDT; Solartron Metrology model DF ± 2.5 mm, accuracy ± 10 µm, Bognor Regis, UK). Six sensors were installed, divided by both training systems, which had been previously calibrated. Each sampling grapevine was equipped with a sensor, fixed to the main trunk with a bracket made of aluminum, with a thermal expansion close to zero [45,46]. The sensors were covered with thermo-protected silver sheets to prevent the device from heating up and wetting. Readings were taken every 30 s with a CR1000 Campbell datalogger combined with an AM16/32 Campbell multiplexer, programmed to report values every 5 min. Indices from the TDF daily cycles were derived, such as maximum (MXTD) and minimum (MNTD) daily trunk diameters, cumulative increment (CI) and maximum daily trunk shrinkage (MDS). The CI was calculated as the cumulated difference between MXTD of a given day and MXTD of the previous day and MDS as the difference between MXTD and MNTD of the defined day [47].

Other Measurements
Soil water content was periodically measured within a row of each training system (with five replicates). For this purpose, 0.6-m buriable waveguides were installed and periodically connected to a portable TDR (miniTrase, Soilmoisture Equipment Corp., Santa Barbara, CA, USA). At a given date, the fraction of available water (FASW) within the defined soil profile (0-0.6 m) was calculated from the soil water content on the day of measurement and the soil water content at field capacity and minimum soil water content in situ [4].
Determinations of predawn leaf water potential (Ψ pd ) were performed on cloud free days, between mid-June and late-August (harvest), at 5:30 (completed before sunrise). For this, six fully expanded leaves were collected from the outside of the south-facing middle third of the canopy of distinct vines. Immediately, they were introduced in a Scholander Pressure Chamber (model 1505D-EXP, PMS Instrument, Albany, OR, USA). Measurements were performed according to Scholander et al. [48] and following the recommendations of Turner [49].
The leaf area index (LAI) was computed from measuring transmitted photosynthetically active radiation along all the inter-row space (readings were taken perpendicular to rows) by a ceptometer (AccuPAR model LP-80, Pullman, WA, USA), as described by López-Lozano and Casterad [50]. From early-July to late-August, the measuring process was replicated 10 times in each training system, with a 0.5-m row spacing between readings. Estimated LAI values were then averaged to calculate the integrated canopy LAI.
During the irrigation periods, all these measurements were taken 1-2 days before the water supply.

Statistical Analysis
Statistical analysis of FASW, Ψ pd , LAI and SF/LAI data was performed using the SPSS software, version 24.0 (IBM Corporation, Armonk, NY, USA). Analysis of variance (ANOVA) was performed for Agriculture 2020, 10, 315 5 of 15 each of the dependent variables studied, with a confidence interval of 95%, followed by the Tukey test to compare significantly different means.

Weather Conditions
From budburst to harvest (late-March to end of August), daily maximum temperature ranged from 12.0 • C (DOY 81) to 43.6 • C (DOY 168) and daily minimum temperature varied between 0.5 • C (DOY 84) and 24.1 • C (DOY 170). There were several days, from June to August, with mean temperatures above 30 • C ( Figure 1A, Figure S1). Conversely, precipitation was very low or even zero during these months. Nevertheless, there was an atypical rainy day on August 29 (19 mm) at harvest time. The total annual (November 2016-October 2017, Figure S1) precipitation was 320 mm, of which 25% was observed during the growing season (corresponding to about 70% from November 2016 to March 2017). The reference evapotranspiration (ET 0 ) varied between 1.3 (DOY 83) and 8.2 mm d −1 (DOY 196) from end-March to late-August, with a total amount of 805 mm for this time period ( Figure 1B) and an annual sum of 1140 mm.

Fraction of Available Soil Water and Predawn Leaf Water Potential
The local weather conditions had a clear impact on the fraction of available soil water, FASW ( Figure 2A). In fact, FASW sharply declined from 55% (mid-May) to 30% in about two weeks. Despite the water provided by irrigation and precipitation (e.g., DOY 179) that caused occasional increases in FASW, this variable globally diminished to a minimum of about 20% in August (maturation). No significant FASW differences were found between training systems. Similarly, predawn leaf water potential (Ψ pd ) data revealed no significant differences between training systems ( Figure 2B). The Ψ pd generally decreased from −0.40 MPa (mid-June) to around −0.85 MPa at the beginning of August (maturation), followed by a late recovery at harvest (Ψ pd = −0.35 MPa) due to a rainfall event. However, there were no substantial leaf losses (visual observation) owing to sunburn over the growing cycle.

Statistical Analysis
Statistical analysis of FASW, Ψpd, LAI and SF/LAI data was performed using the SPSS software, version 24.0 (IBM Corporation, Armonk, NY, USA). Analysis of variance (ANOVA) was performed for each of the dependent variables studied, with a confidence interval of 95%, followed by the Tukey test to compare significantly different means.

Weather Conditions
From budburst to harvest (late-March to end of August), daily maximum temperature ranged from 12.0 °C (DOY 81) to 43.6 °C (DOY 168) and daily minimum temperature varied between 0.5 °C (DOY 84) and 24.1 °C (DOY 170). There were several days, from June to August, with mean temperatures above 30 °C ( Figure 1A, Figure S1). Conversely, precipitation was very low or even zero during these months. Nevertheless, there was an atypical rainy day on August 29 (19 mm) at harvest time. The total annual (November 2016-October 2017, Figure S1) precipitation was 320 mm, of which 25% was observed during the growing season (corresponding to about 70% from November 2016 to March 2017). The reference evapotranspiration (ET0) varied between 1.3 (DOY 83) and 8.2 mm d −1 (DOY 196) from end-March to late-August, with a total amount of 805 mm for this time period ( Figure  1B) and an annual sum of 1140 mm.

Fraction of Available Soil Water and Predawn Leaf Water Potential
The local weather conditions had a clear impact on the fraction of available soil water, FASW ( Figure 2A). In fact, FASW sharply declined from 55% (mid-May) to 30% in about two weeks. Despite the water provided by irrigation and precipitation (e.g., DOY 179) that caused occasional increases in FASW, this variable globally diminished to a minimum of about 20% in August (maturation). No significant FASW differences were found between training systems. Similarly, predawn leaf water potential (Ψpd) data revealed no significant differences between training systems ( Figure 2B). The Ψpd generally decreased from −0.40 MPa (mid-June) to around −0.85 MPa at the beginning of August (maturation), followed by a late recovery at harvest (Ψpd = −0.35 MPa) due to a rainfall event. However, there were no substantial leaf losses (visual observation) owing to sunburn over the growing cycle.

Sap Flow and Trunk Diameter Fluctuations
In terms of plant water dynamics, Figure  mid-maturation to ripening ( Figure 3C,F). Generally, both variables exhibited higher values at the earlier phase, decreasing in the following, being clearer in the TDF values (Figures 3 and S2). These patterns were consistent with FASW and Ψ pd trends.
During the first interval ( Figure 3A,D), daily SF displayed a typical bell-shaped curve that increased markedly from sunrise until noon (generally following solar radiation with maximum values reaching 1000 Wm −2 , Figure 3G), falling to negligible values at the end of the day. These SF maximums became lower over the maturation period ( Figure 3B,C,E,F). Although irrigation tended to increase SF, the maximum values were lower in the later periods than in the first interval. This reduction in SF rates was a reflection of the progressive soil water depletion (decreasing FASW) and enhanced atmospheric demand ( Figure 3H,I), as well as the decrease in total leaf area by the onset of senescence towards the harvest period (as shown in Section 3.4). Daily maximum values of vapor pressure deficit (VPD) and reference evapotranspiration (ET 0 ) peaked at 7.6 kPa and around 8.0 mm, respectively, during maturation ( Figure 3H-I).
TDF data showed a clear daily cycle, with increasing values until the early morning, followed by a decreasing period and, lastly, a recovery ( Figure 3A-F). However, TDFs were higher and displayed a positive trend at the early stage ( Figure 3A,D), indicating trunk increment during these days. This trend did not occur, and even became negative (no trunk increment), later on ( Figure 3B,C,E,F). This negative trend was moderately sustained and occasionally reversed due to the water supply by irrigation.   Regarding the training systems, the Guyot-trained vines showed higher maximum SF values compared with the Cordon-trained vines over the season (mean of maximum SF values during post-flowering = 0.24 dm 3 h −1 and 0.16 dm 3 h −1 ; post-veraison = 0.18 dm 3 h −1 and 0.11 dm 3 h −1 ; mid-maturation to ripening = 0.18 dm 3 h −1 and 0.08 dm 3 h −1 , respectively), indicating greater whole-vine transpiration ( Figure S2). Taking into account the respective maximum (MXTD) and minimum (MNTD) values, TDFs were mostly similar between training systems during the first interval (e.g., DOY 147: MXTD/MNTD = 1010/866 µm for Guyot vines and 932/783 µm for Cordon vines). Later in the season, these absolute values diminished, though the differences between training systems enlarged (e.g., DOY 220: 759/702 µm and 767/744 µm, respectively), with effects on the CI and MDS results (Section 3.5). Figure 4 illustrates the daily patterns of SF and TDF on three different days (DOY 147, 199 and 220), each representative of the three phenological intervals defined above for both training systems. On DOY 147 ( Figure 4A,D), corresponding to a period of vegetative growth, and in both training systems, SF increased sharply with sunrise, peaked at around 15:00 and decreased to negligible values at the end of the day. On this day, TDF clearly defined a sinusoidal pattern, peaking a maximum value between 6:00 and 8:00, then decreasing between 15:00 and 17:00, followed by a recovery. During the other two sampling days ( Figure 4B,C,E,F), TDF flattened its pattern, with similar timing of the maximum value and minimum occurring later (around 18:00-19:00). On the other hand, in these two days, SF reached its maximum earlier (8:00 and 10:00) on Guyot-trained vines ( Figure 4E,F) but later on Cordon-trained vines (around 18:00), though a smaller peak was also displayed in the morning ( Figure 4B,C).
These daily courses of SF and TDF permit us to define five phases, as described by Herzog et al. [51]. Phase I corresponds to the increase in trunk diameter during the night, when transpiration is considered negligible and the plant rehydrates. Phase II describes the delay between the increase in SF after sunrise and the shrinkage of the trunk. Phase III spans from the beginning of the trunk shrinkage to the maximum SF value (which occurred later). Phase IV corresponds to the delay between the maximum SF and the minimum trunk diameter, normally reached in the afternoon. Finally, phase V describes the delay between the minimum trunk diameter and the minimum SF value. We verified that phases I and II were similar over the three days presented for both training systems (Figure 4). The same did not happen with phases III, IV and V. On DOY 147, the maximum SF was reached after noon (around 14:00), with phase III lasting around 8 h. However, in the remaining two days (DOY 199 and 220), phase III was shorter (about 2-3 h) whereas phase IV was longer compared with DOY 147 for Guyot-trained vines. Regarding Cordon-trained vines, shifts between phases III and IV were displayed on DOY 199 and 220 due to the fact that the maximum SF was reached later than the minimum TDF value ( Figure 4B,C). Phase V presented longer durations on DOY 147, with its interval depending on previously phases. The pedoclimatic conditions were mostly responsible for these patterns.
The relationships between daily courses of SF and TDF on the same days presented in Figure 4 are plotted in Figure 5. Lower changes in TDF for the same SF values were found on Cordon-trained vines, where the TDF remained practically unchanged over DOY 199 and 220. Furthermore, in both training systems, trunk diameter fluctuations were clearer in the late afternoon than in the early morning, particularly in DOY 147. For each day, SF responded more sharply than TDF just after sunrise. Rises in SF from early morning would be expected in response to solar radiation ( Figure 3G,H,I). However, as was observed previously, FASW and Ψ pd decreased ( Figure 2) and VPD tended to increase over maturation ( Figure 3H,I), explaining the lower SF values. vines, where the TDF remained practically unchanged over DOY 199 and 220. Furthermore, in both training systems, trunk diameter fluctuations were clearer in the late afternoon than in the early morning, particularly in DOY 147. For each day, SF responded more sharply than TDF just after sunrise. Rises in SF from early morning would be expected in response to solar radiation ( Figure  3G,H,I). However, as was observed previously, FASW and Ψpd decreased ( Figure 2) and VPD tended to increase over maturation ( Figure 3H,I), explaining the lower SF values.

Sap Flow and Leaf Area Index
The leaf area index (LAI) and the ratio between SF and LAI (SF normalized by LAI) are presented in Table 1. LAI increased from DOY 185 to DOY 199 (around veraison), then started to decline over maturation. No significant LAI differences were found between training systems. However, when SF was divided by LAI, the Guyot-trained vines presented significant higher values (mean increase of 38%) compared with Cordon-trained vines, pointing to higher transpiration rates on a whole-plant scale. vines, where the TDF remained practically unchanged over DOY 199 and 220. Furthermore, in both training systems, trunk diameter fluctuations were clearer in the late afternoon than in the early morning, particularly in DOY 147. For each day, SF responded more sharply than TDF just after sunrise. Rises in SF from early morning would be expected in response to solar radiation ( Figure  3G,H,I). However, as was observed previously, FASW and Ψpd decreased ( Figure 2) and VPD tended to increase over maturation ( Figure 3H,I), explaining the lower SF values.

Sap Flow and Leaf Area Index
The leaf area index (LAI) and the ratio between SF and LAI (SF normalized by LAI) are presented in Table 1. LAI increased from DOY 185 to DOY 199 (around veraison), then started to decline over maturation. No significant LAI differences were found between training systems. However, when SF was divided by LAI, the Guyot-trained vines presented significant higher values (mean increase of 38%) compared with Cordon-trained vines, pointing to higher transpiration rates on a whole-plant scale.

Sap Flow and Leaf Area Index
The leaf area index (LAI) and the ratio between SF and LAI (SF normalized by LAI) are presented in Table 1. LAI increased from DOY 185 to DOY 199 (around veraison), then started to decline over maturation. No significant LAI differences were found between training systems. However, when SF was divided by LAI, the Guyot-trained vines presented significant higher values (mean increase of 38%) compared with Cordon-trained vines, pointing to higher transpiration rates on a whole-plant scale.

Seasonal Variation of CI and MDS
Seasonal variation of cumulative increment (CI) and maximum daily trunk shrinkage (MDS) for both training systems are shown in Figure 6. Regarding the Cordon-trained vines, CI increased markedly initially, with a maximum peak at DOY 156 (1682 µm), then decreased sharply to around 1200 µm (DOY 172) and lastly was relatively stable afterwards ( Figure 6A). Though the CI pattern was similar for the Guyot-trained vines, they exhibited a higher maximum peak that was reached Agriculture 2020, 10, 315 9 of 15 later (2202 µm on DOY 167). These temporal and absolute differences between training systems were relatively sustained throughout the season ( Figure 6A).
The MDS had seasonal behavior similar to the CI, with an initial increase, then decreasing until it became relatively stable, though fluctuations were shown ( Figure 6B). Likewise, the Cordon-trained vines presented globally lower MDS values compared with the Guyot-trained vines. The Cordon-trained vines reached their MDS maximums on DOY 143 (173 µm), 146 (165 µm) and 152 (157 µm) and then decreased to around 35 µm. On the other hand, the Guyot-trained vines peaked at a higher MDS maximum (297 µm) and later in the season (DOY 168), also decreasing towards the end of the season (August). MDS had a tendency to increase after irrigation events for both training systems.
Seasonal variation of cumulative increment (CI) and maximum daily trunk shrinkage (MDS) for both training systems are shown in Figure 6. Regarding the Cordon-trained vines, CI increased markedly initially, with a maximum peak at DOY 156 (1682 μm), then decreased sharply to around 1200 μm (DOY 172) and lastly was relatively stable afterwards ( Figure 6A). Though the CI pattern was similar for the Guyot-trained vines, they exhibited a higher maximum peak that was reached later (2202 μm on DOY 167). These temporal and absolute differences between training systems were relatively sustained throughout the season ( Figure 6A).
The MDS had seasonal behavior similar to the CI, with an initial increase, then decreasing until it became relatively stable, though fluctuations were shown ( Figure 6B). Likewise, the Cordontrained vines presented globally lower MDS values compared with the Guyot-trained vines. The Cordon-trained vines reached their MDS maximums on DOY 143 (173 μm), 146 (165 μm) and 152 (157 μm) and then decreased to around 35 μm. On the other hand, the Guyot-trained vines peaked at a higher MDS maximum (297 μm) and later in the season (DOY 168), also decreasing towards the end of the season (August). MDS had a tendency to increase after irrigation events for both training systems.

Discussion
The 2017 growing season stood out, with low precipitation and high temperatures ( Figure 1A). The mean temperature for the April-August (April-October) period was 23.4 °C (22.6 °C), clearly above the 30-year average value of 19.0 °C for a similar time period (18.6 °C for April-October; [41]). Annual precipitation was only 55% of the long-term average. A high potential soil water deficit (ET0

Discussion
The 2017 growing season stood out, with low precipitation and high temperatures ( Figure 1A). The mean temperature for the April-August (April-October) period was 23.4 • C (22.6 • C), clearly above the 30-year average value of 19.0 • C for a similar time period (18.6 • C for April-October; [41]). Annual precipitation was only 55% of the long-term average. A high potential soil water deficit (ET 0 minus precipitation) of around 715 mm was calculated from budburst to harvest (Figure 1 and Figure S1). This last value was slightly lower than the 730-750 mm reported by Malheiro [52] for a three-year period in the DDR but with an extended growing season (ca. 20 more days).
Regardless of the water applied by irrigation, a cumulative decrease in FASW was mostly found throughout the season (Figure 2A). This pattern was also reflected in Ψ pd reduction as the summer progressed ( Figure 2B), revealing enhanced plant water stress over the season. Ψ pd reflects the equilibrium between soil and leaf water potentials overnight while no significant transpiration is occurring, and it is therefore considered a good indicator of overall plant water status [53]. According to Ψ pd thresholds reported by Ojeda [54] and Van Leeuwen et al. [15], grapevines were exposed to moderate (Ψ pd = −0.45 MPa) to severe (Ψ pd = −0.85 MPa) water deficit during maturation.
During the post-flowering ( Figure 3A,D), corresponding to a moderate plant water stress period, daily fluctuations of trunk diameter clearly exhibited three phases: recovery, increment and contraction. The recovery phase corresponds to the cycles during which the trunk radius increases until reaching the maximum value that morning (around 6:00); the increment phase represents the period during which the trunk radius continues to increase until the start of the retraction phase of the next day cycle, and the contraction phase is defined as the period during which the trunk radius decreases, starting from the early morning maximum (starting between 6:00 and 8:00 and ending between 15:00 and 17:00). In the following two periods ( Figure 3B,C,E,F), there was a reduction in TDF, with only partial recovery during the night as a reflection of low water availability and high evaporative demand [22,47].
Regarding SF, the curve globally followed the daily course of solar radiation ( Figure 3G), with a sharp increase early in the morning (around 6:00; Figure 3A,D) and peaking at high maximum values after noon. When transpiration starts early in the morning, a tension in the evaporative surface of the leaves is created for all plant organs. The water stored in the plant's tissues during the night is then lost, allowing a quick response of the plant to changes in evaporative demand [22,47,55], without depending on the water intake by the roots, which begins later [47]. This explains the rapid increase in SF and the abrupt decrease in TDF during this period (e.g., DOY 147: Figures 4 and 5).
In Figure 3B,C,E,F, a decrease in maximum SF values was evident. This phenomenon was a result of stomatal control in response to lower water availability [24] and higher irradiance and VPD, revealing a key survival strategy of vines under water stress conditions [3,39]. This mechanism has been attributed to abscisic acid (ABA) signaling that induces stomatal closure before a large water deficit develops [56]. Under these atmospheric conditions, irrigation was able to slightly sustain or even increase SF (Figure 3 and Figure S2) but for a short period. In fact, a stomatal closure may also occur even in well-watered vines (which was not the case in the present study) under high atmospheric demand conditions [3]. TDF was also found to be sensitive to water supply in grapevines [55], revealing a small increase in the increment of the trunk but tending to lose these values between irrigation events in our study. Interestingly, daily TDF responded earlier and more visibly than SF during an irrigation event. This finding is in agreement with the results reported by Fernández et al. [35] in olive trees. These last authors also reported a hysteresis phenomenon indicated by clearer trunk diameter fluctuations in the late afternoon than in the early morning. In the present study, similar responses for both training systems were found ( Figure 5). On the other hand, Guyot-trained vines showed higher changes in TDF for the same SF values, where TDF of Cordon-vines remained practically unchanged over DOY 199 and 220 ( Figure 5). The combined SF and TDF responses allowed us to define the five phases ( Figure 4) described by Herzog et al. [51]. Phase III was the most affected from DOY 147 to DOY 199 and 220, being shortened in Guyot-vines and even reversed with phase IV in Cordon-vines. These patterns were related to an earlier sap flow peak (corresponding to earlier stomatal closure) in both training systems and to a later second (and higher) sap flow peak during late afternoon (indicating stomata openness when VPD was decreasing) in Cordon-vines. This response suggests that the Cordon-vines were particularly sensitive to the atmospheric conditions. In fact, stomata openness is related to environmental conditions [4]. On the other hand, under relatively high soil moisture, a reversion between phase I and II was found in olive trees [35], associated with a total recovery of the trunk diameter during the night.
If the soil water is not limiting, transpiration will be conditioned by the leaf area [57]. However, we found that both training systems had similar low/median values of LAI (0.5 < LAI < 2; [58]). On the other hand, the SF/LAI ratio was significantly higher for Guyot-trained vines ( Table 1), indicating that transpiration per unit area was higher in these vines compared to Cordon-trained vines. In a tall forest species, transpiration per unit leaf area and stomatal conductance per unit leaf area were significantly reduced with increasing crown height [59]. This restriction was a result of the longer length of the hydraulic pathways and enlarged gravitational potential opposing the ascent of water, leading to a reduction in leaf-specific hydraulic conductance [60]. Regarding the grapevines, Favero et al. [61] discussed in their study that the higher trunk height of the Geneva double curtain training system (1.9 m) could have contributed to the difficulty of transporting water to the canopy compared with the 1.0-m vertical shoot positioned trellis. In addition, upward growing of the branch promotes enhanced shoot hydraulic conductivity compared with the downward branch orientation [62]. In this way, our results could be due mainly to the fact that the Guyot-trained vines have lower trunk height and upward growing branches in contrast to the higher trunk and permanent horizontal branch (cordon) of Cordon-trained vines, shortening and facilitating the water pathway along the soil-plant-atmosphere continuum.
During the growing season, CI initially increased and then decreased in both training systems ( Figure 6). With decreasing water availability, trunk contraction increases and trunk increment decreases [22,37]. Interestingly, the separation of the fractions of irreversible stem expansion induced by growth and reversible shrinkage and swelling induced by water deficit has been investigated in forest trees in order to better understand the processes of stem radius fluctuations [63,64]. On the other hand, there is evidence that during the fruiting period, the berries become an important sink that can lead to a reduction in the grapevine trunk growth rate [37]. Nonetheless, no significant differences in cluster number and yield were found between training systems in our study (data not shown). Furthermore, according to the last authors, trunk growth ceases regardless of the water status of the plant after veraison. Though LAI reached its maximum value around veraison in the present study (Table 1), CI started decreasing before this phenophase, particularly for Cordon-trained vines ( Figure 6A). This response may be associated with the different pedoclimatic conditions between studies.
A similar pattern was found for MDS ( Figure 6B). These results are agreement with other studies for grapevines [37] and for other crops, such as olive [27] and apple trees [65]. Furthermore, slight MDS rises were generally found to be due to irrigation ( Figure 6B). Conversely, it has been reported that MDS increases under rain-fed conditions compared to an irrigated treatment [37]. The explanation for this apparent discrepancy lies in the occurrence of stress periods in the present study, which prevents the stored tissues from being recharged during the night [47] and promotes a strong reduction in the swelling of the trunk and, therefore, of the MDS [37,47,65]. Guyot-trained vines peaked at higher maximums of CI and MDS values, which were reached later on the season, compared with the Cordon-trained plants. These responses indicate the higher rehydration capacity of the Guyot-trained vines, which may be also associated with the shorter length of its hydraulic pathways [21].

Conclusions
From a viticultural point of view, the adequate selection of the grapevine training system stands out as one of the most critical adaptation strategies for the present and future times. This study exposed the effect of the shorter length of the hydraulic pathways of the Guyot-trained vines, in contrast to the higher trunk and the permanent horizontal branch (cordon) of the Cordon-trained vines, highlighting the adaptive potential of the Guyot training system to the hot and dry areas. In this way, the results support the selection of the Guyot as a training system that is better adapted to the projected climate change in the DDR as in other Mediterranean wine-growing regions. Stomatal control, in order to reduce transpiration, and rehydration as well as adjust of the total leaf area, were the most important plant survival strategies. The combined use of plant-based measurement techniques, such as sap flow and trunk diameter measurements, revealed sensitivity to variable conditions of atmospheric demand and crop water status. Our results reinforce the usefulness of combining SF and TDF measurements in order to assess grapevine water responses to variable environmental conditions. Further research is needed to comprehend the influence of the hydraulic conductivity of plant segments on grapevine water relations and adaptation measures.  Funding: This study was funded by the "Grupo Operacional -VITISHIDRI-Estratégias para a gestão do stress hídrico da vinha no Douro Superior", financially supported by the European Agricultural Fund for Rural Development and the Rural Development Programme 2020.