A full description of the experimental design appears in Epee et al. (2022a) and Epee et al. (2022b). In summary, vines were pruned in winter 2019 according to a 5 (cane node load) × 3 (spur node number) factorial design, and each set of treatments was replicated 15 times to give a total of 75 experimental units (vines) at Site 1. Total node number was also considered as an overall factor, with the total number of nodes on spurs and canes equating to 15 treatments, for a total vine node number ranging from 12 nodes to 56 nodes at Site 1, and from 14 to 56 nodes at Site 2 and Site 3 (Table S1).

On 11 November 2019, about two weeks following the start of budburst at growth stage EL 9, two to three leaves separated (Modified Eichhorn and Lorenz grapevine growth stages ( Coombe, 1995)), vines were shoot-thinned to their respective cane and spur node numbers by keeping one shoot per count node and removing shoots growing on all non-count nodes (i.e., shoots on vine head, trunk, basal shoots and excess shoots on count nodes). Cane node number was considered a surrogate for total vine node number due to the non-significance of the interaction of cane node load × spur node number, and due to three-node spurs having a blind node appearing at node position one ( Epee et al., 2022b). At harvest (on 20 March 2020), 30 berries per vine were randomly sampled on both sides of the canopy, crushed and the juice analysed for TSS with a portable digital refractometer (Atago Co Ltd, USA). Grape bunches of individual vines were hand-harvested, counted and weighed.

In the following growing season (2020/2021), the same node treatments were applied to the same vines as the previous winter, and shoot thinning was conducted as the previous spring. Contrary to the previous season where TSS was only measured at harvest, in the second season, 30 berries were randomly sampled fortnightly on each vine from EL 33 (berries still hard and green) to EL 38 (berries harvest-ripe) on four dates from 9 February to 22 March 2021. At each sampling time, berries were stored in a cooler bag and taken to the laboratory for analysis within 48 hours. Each sample was weighed, crushed and sieved, and the juice was extracted for analysis. TSS was measured as indicated above in 2019/2020. Titratable acidity was measured using the endpoint titration (tartaric acid equivalents in g/L) using 0.1 mol NaOH to pH 8.4 with an auto-titrator (Metrohm USA), and pH was measured using the same auto-titrator (Metrohm, USA) pH meter. On 22 March 2021, grape bunches were hand-harvested separately from canes, spurs and the vine head, then counted and weighed.

Vine total leaf area (TLA) and exposed leaf area (ELA) were calculated on 22 February, four weeks before harvest. TLA was estimated by randomly sampling one healthy shoot located at the proximal section to the trunk of the bottom cane, measuring the midrib length of all leaves on the shoot (including all leaves on lateral shoots when present) and then multiplying the total shoot leaf area by the total number of shoots on the vine ( Dokoozlian and Kliewer, 1995). In total, 10 leaves from each of the 10 vines were randomly collected at the 2020 harvest. The lamina of each leaf was then scanned using the Epson XP-3100 Series digital scanner, and the area was computed using scanned images and the R package LeafArea ( Katabuchi, 2015). The mid-rib of each leaf was measured and used to develop a regression function (y = 41.8 – 22.7x + 4.9x ² – 0.16x ³; R ² = 0.96, p < 0.001) relating midrib leaf length to leaf area. The same regression function was used the following season.

To estimate ELA, digital photos (with scales embedded) of one side of the grapevine’s canopies were taken with a Nikon D5500 DSLR Camera (Nikon, Japan) on a white background. Photos were transferred into the ImageJ software (The National Institutes of Health, USA) from which the green canopy area was segmented and calculated. The vine’s total ELA was calculated as twice the one-side area as determined by the software, and the same procedure was followed for both seasons.

In winter, vines were pruned, and for each vine, the wood was sorted into cane (the dormant shoot or one-year-old wood) and old cane (or two-year-old wood) and then weighed separately. The average cane mass was calculated by dividing the mass of all canes by the total number of canes pruned off, and the average old cane mass was obtained by dividing the mass of old canes by the number of old canes pruned off. The total cane mass was then adjusted by adding the estimated mass of retained canes left on the vine at pruning from the average cane mass.

In winter 2020, 50 vines were selected within the same vine row and pruned for a 5 cane node load × 2 spur node number factorial designed experiment (referred to as Site 2) with five replicates and with the same five node loads as Site 1 vines, in combination with two node numbers on spurs (two and three node spurs). The same measurements were collected at Site 1 in 2020/2021. Site 2 represented a seasonal repeat of the 2019/2020 experiment at Site 1, which excluded the carry-over effect of the previous year's pruning treatments.

At Site 3, 50 vines along a south/north-oriented row were pruned in winter 2020 with the same node treatments as at Site 2. Measurements were the same as at Site 2 except that berry sampling for TSS, TA, pH and berry mass was done from 2 February to 15 March 2021 (harvest day) and TLA and ELA were measured on 15 February 2021 (four weeks before harvest).

A 3-parameter rise to maximum exponential model ( Fekedulegn et al., 1999) was fitted to TSS data, from which the time from 8 to 21 °Brix was interpolated.

Y = A + Be ^(CX) (1)

where A is the Y asymptote, B is the displacement along the x-axis, C is the rate of increase, X is the time interval (in days) and e is Euler's Number.

The optimal source–sink ratio was determined by plotting ELA/FM and TLA/FM (x-coordinate) with TSS (y-coordinate), then fitting the best curve (least sum of square error) to the data points. The optimum point on the curve was set at 90 % of the Y asymptote value and its projection on the x-coordinate gave the optimal source–sink value.

All dependent variables (TSS, TA, pH, berry mass, yield, yield components, cane mass, old cane mass, average cane mass, average old cane mass and vine balance ratios) were compared with the Analysis of Variance (ANOVA) and Fisher’s Least Significant Difference Test (LSD). The independent variables were the factors: cane node load and spur node number. Before running ANOVA, Levene’s test was conducted to check the homogeneity of variances, and the Shapiro–Wilk test was applied to check whether data were normally distributed. The two ANOVA assumptions were met. Considering that there was no interaction between cane node load and spur node number and that spur node number had no significant effect on all dependent variables, only the effect of the factor cane node load was presented and used as a surrogate for the total vine’s node number. The significance level for all tests was set at p < 0.05. Data were analysed in RStudio Version 4.0.5 (R Core Team, 2021). The following R packages were used: dplyr ( Wickham et al., 2021), ggpubr ( Kassambara, 2020), agricolae ( Mendiburu, 2020), Rmisc ( Hope, 2013) and Car ( Fox and Weisberg, 2019).

In general, bunch mass per vine (vine yield) increased with increasing node numbers whereas a drop in vine yield was observed in the second season of node treatments (Figure 2a).

At Site 1 and in both seasons, the total bunch mass per vine and bunch number per vine increased with increasing node loads (Figure 2a,b). Between 2019/2020 and 2020/2021, the total bunch mass, average bunch mass per vine and bunch mass per shoot dropped respectively by 28, 43 and 33 % on average across all node treatments (Figure 2a,c,d). On all node loads at Site 1, the bunch number per vine rose substantially by 21 % on average across all node loads in 2020/2021, with 10-node vines recording the highest rise (63 %; p < 0.0001) and 50-node vines the lowest (1 %; p < 0.05; Figure 2 b). Similarly, the bunch number per shoot increased in 2020/2021 by 17 % on average across all node treatments. The increase was significant on 30-node (22 %; p < 0.05) and 40-node vines (29 %; p < 0.01) but not statistically significant on other node treatments (Figure 2e). When the four yield components were considered together it appeared that the drop in vine yield between 2019/2020 and 2020/2021 at Site 1 was mainly driven by the diminution of the average bunch mass per vine and bunch mass per shoot (Figure 2c). The relative rise in bunch number per vine and bunch number per shoot was not sufficient to offset the depressing effect of low bunch mass. Overall, yield components were affected by both inter-seasonal weather variations (Figure S1) and node load treatments at Site 1, except for berry mass which remained stable during ripening (Figure 1).

Figure 1. Berry composition and mass on four dates at fortnightly intervals on 9 February (D1), 23 February (D14), 8 March (D28) and 22 March 2021 (D42) at Site 1 and 2, and on 2 February (D1), 16 February (D14), 1 March (D18) and 15 March 2021 (D42) at Site 3. N10, N20, N30, N40 and N50 refer to the total node number on vine’s canes being equal to 10, 20, 30, 40 and 50 nodes, respectively. Points represent mean ± standard error of the mean with n = 15 vines for Site 1 and 10 vines for Site 2 and 3 for each cane node treatment. Means sharing the same letter on a given date are not significantly different (Fisher’s protected LSD test, p ≤ 0.05).

2.2.1. Whole vin

The total bunch mass per vine increased linearly with increasing node loads across all three sites (R ² = 0.9, p < 0.001 for Site 1 and 2, R ² = 0.9, p < 0.05 for Site 3; Figure 2a, Figure S2). The bunch number per vine equally followed a similar pattern of a linear increase at all three sites (R ² = 0.9, p < 0.001 for Site 1 and 2, R ² = 0.9, p < 0.01 for Site 3; Figure S3). Bunch mass per shoot was stable across node treatments at Site 2 and over both seasons at Site 1 (Figure 2d). At all sites a clear trend for 10-node vines having the lowest bunch mass per shoot was visible and this trend was significant at Site 3 ( p < 0.001, Figure 2d). Berry mass did not change significantly across all node treatments at each site in the 2020/2021 growing season (Figure 1).

Figure 2. Effect of node load on yield components. N10, N20, N30, N40 and N50 refer to total node numbers on vine's canes equalling 10, 20, 30, 40 and 50 nodes, respectively. Points and bars represent the mean ± standard error of the mean of 15 vines for Site 1 and 10 vines for Site 2 and Site 3. Means sharing the same letter across different node treatments (node load comparison) are not significantly different with the Fisher's protected LSD test, p ≤ 0.05. For each individual node treatment at Site 1, the significance of the difference between both seasons is indicated by the significance symbols of ANOVA: 'ns' not significant and p > 0.05; ‘*’, p < 0.05; ‘**’ p < 0.01; ‘***’, p < 0.001; ‘****’, p < 0.0001.

2.2.2. Vine parts

The canes contributed the greatest mass to the vine yield with smaller numbers and mass borne on spurs and non-count headshoots (Table 3). The fewer shoots that developed on the vine head bore fewer and lighter grape bunches (Tables 4 and 5). At Site 1, all yield components on cane (except for total bunch mass and shoot number) were significantly higher on N10 and N20 vines (Table 3–5). At Site 2, no significant differences were noted across node treatments on cane yield components, except for shoot number and bunch number, which were significantly higher on N50 vines because they bore more shoots on canes and thus more bunches than N20 to N40 treatments (Table 3–5). When the yield components of the three vine parts were compared (Table 3–5), they were the highest on the cane, high on the spur and the lowest on the vine head.

2.2.3. Whole vine and vine part

Comparing the effect of increased node number on yield components for the whole vine and individual vine parts revealed another aspect of the vine’s physiological response. Overall, the spur and head did not respond to node treatments, all node treatments had statistically similar values whereas the cane responded the most (especially at Site 1). After normalising all cane yield components on a per cane basis (see note Tables 3 and 5), almost all of them (except for cane shoot number at Site 2) decreased in value with increasing node number (Table 3–5); the opposite of what was observed on vine yield components which increased in value with increasing node numbers (Figure 2).

Over both seasons and across all three sites, no clear relationship was found between TLA and ELA (Figure S4). At Site 1, the vine TLA and ELA increased with node load ( p < 0.001) evidently because of the high number of shoots on high-node vines, given that the shoot leaf area was similar across all node loads (Table 6, Figure S5). Between 2019/2020 and 2020/2021, the vine TLA did not vary significantly, whereas ELA increased significantly across all node treatments ( p < 0.001; Table 6) possibly due to seasonal variations.

Contrary to TLA and ELA, TLA/FM and ELA/FM at Site 1 decreased with increasing node load ( p < 0.001); however, with different magnitudes of responses as a function of site and season (Table 6). The high TLA/FM and ELA/FM noted on N10 vines was likely caused by the significantly greater number of head shoots of very low fruitfulness (i.e., carrying smaller and fewer grape bunches) they produced compared with N50 vines ( p < 0.01; Table 4). From the first trial year in 2019/2020 to the second in 2020/2021, TLA/FM and ELA/FM rose significantly on N40 and N50 vines ( p < 0.001) but remained stable on low-node vines (Table 6). Since vine TLA was constant over both seasons, the change in TLA/FM was driven by the variation in the total bunch mass per vine (Figure 2a), whereas the change in the ELA/FM was the result of both ELA and the total bunch mass per vine variations.

The patterns described at Site 1 for TLA, ELA, TLA/FM and ELA/FM were very similar at Site 2 and 3, with vine TLA and ELA increasing with node load ( p < 0.001) whereas TLA/FM and ELA/FM were decreased with node load ( p < 0.01).

When the total bunch mass per vine (vine yield), TSS concentrations at harvest and the source–sink ratios (ELA/FM and TLA/FM) were considered together, harvest TSS was low on high-yield vines (N40 and N50 vines) but high on low-yield vines (N10 vines) with N30 vines having intermediate values (Figures 3 and 4). The optimal ELA/FM across all three sites was 0.75 m ² kg ^-1, and the optimal TLA/FM at Site 1 and Site 2 was 2.0 m ² kg ^-1. The optimal TLA/FM could not be determined at Site 3 as there was not enough variability in TSS to create an inflexion point on the curve (Figures 3 and 4).

Figure 3. Relationship between total soluble solids (TSS), total bunch mass per vine, node load and exposed leaf area to fruit mass ratio (ELA/FM).

Figure 4. Relationship between total soluble solids (TSS), total bunch mass per vine, node load and total leaf area to fruit mass ratio (TLA/FM).

Overall, cane mass, old cane mass and PM significantly increased with increasing node numbers at Site 1 ( p < 0.01), but they did not change significantly over both seasons, although there was a clear trend toward higher mass on low-node vines in the second season (Table 7). On the contrary, the average cane mass and the average old cane mass significantly increased with decreasing node numbers ( p < 0.001) except at Site 1 in 2019/2020, where the increase was not statistically significant. The patterns depicted at Site 1 regarding pruning mass (cane and old cane mass) as well as the average cane mass and old cane mass, were mirrored in the other two sites. For example, at Sites 2 and 3, the average cane mass on 10-node vines was more than twice that of 50-node vines, and at Site 2, the average cane mass varied within the same proportions (Table 7).

In 2019/2020 at Site 1, FM/CM and FM/PM increased with increasing node loads, with 10-node vines having FM/CM values at or below 3.5 kg kg ^-1 and FM/PM at or below 2.5 kg kg ^-1. In 2020/2021, both FM/CM and FM/PM did not change significantly across node treatments but followed a trend of increasing values with increasing node numbers. At Site 2 and 3, FM/CM and FM/PM equally remained below 5.0 kg kg ^-1 despite low-node vines having significantly lower values than high-node vines (Table 8).

When the total bunch mass per vine (vine yield), TSS concentrations at harvest and the FM/PM or FM/CM were plotted against one another, no optimal ratios could be determined for harvest TSS (Figures S6 and S7). However, as FM/PM or FM/CM increased, total bunch mass increased while TSS decreased, although this was a weak relationship (R ² < 0.45; Figures S6 and S7). However, ELA/FM and FM/PM or FM/CM were related (e.g., R ² = 0.69 Site 3 season 2020/2021; Figures 5 and 6) where for example an ELA/FM of 0.75 m ² kg ^-1 corresponded to an FM/CM and FM/PM of, respectively, 5.95 and 4.2 kg kg ^-1 for Site 1 season 2019/2020, 3.9 and 2.9 kg kg ^-1 for Site 1 in season 2020/2021, 4.0 and 3.0 kg kg ^-1 for Site 2 and 2.8 and 2.4 kg kg ^-1 for Site 3.

Figure 5. Relationship between total bunch mass per vine, node load, fruit mass–cane mass ratio (FM/CM) and exposed leaf area-fruit mass ratio (ELA/FM).

Figure 6. Relationship between total bunch mass, node load, fruit mass–pruning mass ratio (FM/PM) and exposed leaf area-fruit mass ratio (ELA/FM).

Time to target TSS concentrations (21 °Brix) was delayed, and TSS concentrations at harvest reduced with increasing node numbers (Table 1; Figure 1; Table 2) which concurs with previous research on Sauvignon blanc ( Greven et al., 2015; Trought et al., 2011) and other cultivars ( Freeman et al., 1980) where sugar accumulation was delayed as the node number increased. Previous studies on New Zealand Sauvignon blanc response to increasing node numbers attributed the low TSS accumulation and delayed ripeness observed on high-node vines to their high crop load (i.e., the amount of fruit bore by the vine) but not to the source–sink ratio as leaf area was not quantified ( Greven et al., 2014; Greven et al., 2015). In the present study, the source size was measured through TLA and exposed area ELA.

The high TSS concentrations measured on 10-node vines were directly related to their high source–sink ratio (Figures 3 and 4). With a significantly much larger TLA/FM and ELA/FM (Table 6), low-node vines potentially supplied more photosynthates to the developing berries. Studies on source–sink manipulations show that leaf removal and shoot trimming (source restriction) for a given crop size are accompanied by a decrease in berry TSS concentrations whereas for a given leaf area, fruit thinning (sink restriction) results in high sugar concentrations ( Bravdo et al., 1985; Etchebarne et al., 2010; Keller, 2015; Parker et al., 2015; Parker et al., 2016; Poni et al., 2006; Poni et al., 2009; Santesteban and Royo, 2006). Therefore, as the source size (leaf area) gets larger relative to the sink size, the TSS concentration and accumulation rate rise up to a maximum and then plateaus. Because they had a much larger leaf area for an equivalent amount of fruit, 10-node vines achieved much higher TSS concentrations than 50-node vines. Kliewer and Dokoozlian (2005) found that on vines trained to single-canopy systems, a total leaf area to fruit mass ratio of 0.8–1.2 m ² kg ^-1 was necessary to achieve optimum TSS concentrations. In the present study, an ELA/FM and TLA/FM of 0.75 and 2.0 m ² kg ^-1, respectively, were necessary for optimum berry ripening (TSS accumulation), other factors being equal. Interestingly, these values were similar across all three sites with the only exception of Site 3, where the critical TLA/FM could not be determined because the variability in TSS was not high enough to create an inflexion point on the curve (Figure 4d). Source–sink ratio values for optimal TSS are particularly useful in guiding in-season canopy management practices such as leaf removal as well as crop load management practices such as inflorescence and fruit thinning.

The general rule of a leaf area to fruit mass ratio of 0.8–1.2 m ² kg ^-1 corresponding to a yield to pruning mass ratio of 5–10 kg kg ^-1 ( Kliewer and Dokoozlian, 2005; Kliewer and Weaver, 1971) was not supported by our data. The authors compared several varieties not including Sauvignon blanc. Site properties, climate, training and pruning systems were different from those encountered and practised in New Zealand. Moreover, there was only a weak linear relationship between TSS and FM/PM or FM/CM, suggesting that both metrics do not have a strong linear relationship (Figures S6 and S7). Nevertheless, based on these weak linear relationships, an ELA/FM of 0.75 m ² kg ^-1 corresponded to 3.0–6.0 kg kg ^-1 FM/CM and 2.5–4.5 kg kg ^-1 FM/PM. The wide range of these values was a clear indication that ratios are site, variety, management and season-dependent. Overall, the FM/CM and FM/PM of source-sufficient vines (N10 vines) were at the lower end or below the range of 3.0–5.0 and 2.0–4.0 kg kg ^-1, respectively, whereas that of source-restricted vines (N50 vines) was at the upper end or above those ranges. FM/CM and FM/PM values at the lower end of the range were an indication of high vigour and high capacity (total cane mass or pruning mass and fruit mass) suggesting that at pruning, the node number of such vines must be raised to meet their capacities. On the contrary, values at the upper end of the range signalled low vine vigour and reduced capacity and consequently decreased node load at pruning.

Greven et al. (2015) noted that throughout four growing seasons, vines pruned to high node numbers accumulated TSS at slower rates in the first two seasons and then progressively reached the same TSS accumulation rate as low node vines by the fourth season. In the second harvest at Site 1, N10 and N50 vines reached comparable TSS concentrations, and this could be interpreted as early signs of these vines reaching the stability observed by Greven et al. (2015). This assumption was considered with care for the following reasons: (1) Greven et al. (2014) did not notice a sharp yield drop across their node treatments the second season as in this trial but rather a moderate and gradual yield rise on low node vines and a gradual yield drop on high node vines; (2) low- and high-node vines in Greven et al. (2015) reached comparable TSS accumulation gradually throughout several seasons and not as abruptly as in the second season in this trial. Therefore, the comparable TSS concentrations observed in this study were mainly driven by the sharp yield drop in the second season. Possible reasons for this yield drop are discussed below. Finally, as the results presented here were only collected over two growing seasons, it was too early to assume any stabilisation of the TSS accumulation rates between the high-node and low-node vines.

Grape TA and pH values were not affected by varying node loads, which is consistent with earlier research on other grapevine cultivars such as Concord, Seyval and Chambourcin grapevines ( Dami et al., 2005; Edson et al., 1995; Keller et al., 2004). Trought et al. (2011) also observed that in two seasons out of three, pH did not change when Sauvignon blanc was pruned from 20 to 40 nodes. Even when the source–sink ratio is altered either with leaf removal or fruit thinning, pH and TA remain similar, confirming the assumption that these two maturity variables are independent of the photosynthesis rate ( Parker et al., 2015; Parker et al., 2016) as well as of the source–sink ratio. The null effect of node load on pH and TA concentration is an indication that these properties are influenced by other factors such as soil physicochemical properties, solute transport, bunch microclimate (temperature), water and mineral nutrition ( Conradie and Saayman, 1989; Kliewer, 1973; Pereira et al., 2006; Spayd et al., 1987) rather than node load. Under some circumstances, however, strong vegetative growth together with the development of a dense canopy may increase shading, resulting in high TA concentrations, as was noted on 10-node vines at Site 1 ( Keller and Koblet, 1995; Keller et al., 1998).