Field trials were conducted in the experimental vineyards of the Institut Viti-vinicole in Remich, Luxembourg (lat. 49.54°N; long. 6.35°E) between 2016 and 2020 on Vitis vinifera L. cultivars Pinot gris (clone Remich 6) and Riesling (clone Remich 10). Both cultivars were planted in 1994, grafted onto SO4 rootstock, and trained to a vertical shoot positioning system (VSP). The vineyard faced south with an inclination of 10 %. The rows were oriented in a north–south direction. The soil between the rows was permanently cover cropped and no irrigation took place. The space per plant was 2.4 m² (2 m between rows, 1.2 m between vines). Fungicide applications against Plasmopara viticola (Berk. & M.A. Curtis) Berl. & De Toni and Erysiphe necator Schwein, following an organic plant protection strategy, were carried out in all seasons at intervals of 10 to 12 days. No products with known activity against Botrytis cinerea were applied.

Experiments were realised using randomised complete block designs with four replicate plots with eight vines per plot. Each block consisted of two treatments. The position of the two treatments in each block was randomly distributed.

Treatments were the same in all years and were defined as it follows:

A) Control: vines pruned in winter to one 10-node fruiting cane per vine;

B) Partial double-pruning after bloom (PDP): vines pruned in winter to two 10-node fruiting canes per vine; one of the two canes were removed at BBCH 73 (2–3 weeks after bloom)

In both treatments, one replacement spur per plant with 1 to 2 nodes remained. In treatment A, the cane was tied downhill, while in treatment B, one cane was bound downhill and the second cane uphill. The uphill-bound cane was removed completely at BBCH 73 with a clean cut at the insertion on the older wood. Dates and phenological plant growth stages at the date of second cane removal in the PDP treatment are shown for both cultivars in Table 1.

Table

titre du tableau
2016				2017				2018				2019				2020
Date	DOY	Rie	Pg	Date	DOY	Rie	Pg	Date	DOY	Rie	Pg	Date	DOY	Rie	Pg	Date	DOY	Rie	Pg
12.07.	194	73	73	26.06.	177	73-75	73-75	18.06.	169	73	73	03.07.	184	73	73-75	29.06.	181	73	73-75

Meteorological data were recorded during the period of examination by a weather station of the national agricultural administration ASTA (Administration des services techniques de l'agriculture) located in direct proximity (distance < 100 m) of the experimental vineyard. Air temperatures were measured at 2 m and precipitation at 1 m above the ground. The phenological stage of full flowering (BBCH 65), according to Lorenz et al. (1995), was recorded based on observations of six plants/organs per cultivar.

Potential differences in the cluster structure were assessed in both cultivars using the cluster density index according to the protocol by Ipach et al. (2005) as previously described (Evers et al., 2010). One hundred clusters were assessed per plot at BBCH 79 (2016: 09 August; 2017: 26 July; 2018: 23 July; 2019: 06 August; 2020: 04 August) following the protocol by Ipach et al. (2005).

Based on the observation that the bunch rot epidemic was delayed in the PDP treatment in the initial years of the present study, we hypothesised that this effect might be caused by changes in berry numbers per cluster or berry sizes. Consequently, in 2020, an additional assessment was carried out: Prior to harvest, ten clusters were randomly selected per plot (n = 4 * 10) in each of the two cultivars and for both treatments in 2020. Rachis lengths and cluster masses were determined. Clusters were manually destemmed and the number of berries per cluster counted. Very small, still hard berries (shot berries) were not considered. The average mass of individual berries was calculated by dividing the cluster mass by the number of berries per cluster. As an indicator of cluster compactness (Tello and Ibanez, 2014), the cluster mass per cm of cluster length was calculated.

The B. cinerea disease progress was followed in weekly to bi-weekly intervals between veraison (BBCH 81) and harvest (BBCH 89) by examining 100 clusters per plot (50 clusters from both sides of the canopy) selected randomly at each assessment date. Disease severity was assessed according to the EPPO guideline PP1/17 classifying visually observed disease severity in seven classes (0 %; 1–5 %; 6–10 %; 11–25 %; 26–50 %; 51–75 %; 76–100 %).

To describe the temporal progress, disease severity in both treatments was plotted over time (expressed as day of the year (DOY)). Disease progress curves were fitted to these data according to the sigmoidal equation (1) as previously described (Molitor et al., 2015a).

y = 100 1 + e - ( ( x - x 0 ) / b )

where y is the disease severity, x corresponds to the assessment date expressed as the day of the year (DOY), x₀ is the inflection point of the curve (disease severity of 50 % reached) and b is the slope factor of the curve at the inflection point.

Solving this equation for x provides the time point at which a specific disease severity value was reached. To quantify differences between the treatments in the temporal position of the annual epidemic, the x_{5 %}-values (DOY corresponding to a disease severity of 5 %) were used according to Beresford et al. (2006).

In all years, grapes were harvested separately per plot. Harvest dates were 6 October 2016 (Pinot), 20 October 2016 (Riesling), 27 September 2017, 25 September 2018, 1 October 2019 (Pinot), 7 October 2019 (Riesling) and 8 October 2020. All grapes from each plot were classified into a marketable (no bunch rot) and non-marketable (botrytised parts of clusters) fraction based on a visual assessment of the harvest team. The average yield per plant of both fractions, average total yield per plant and proportion of the non-marketable fraction in the total yield were calculated. A set of 100 healthy berries per plot (50 on each side of the row) were randomly sampled just prior to harvest. Every sample was pressed, the juice was centrifuged and total soluble solids concentration was measured by FT-IR (FOSS NIRSystems, Laurel, MD, USA).

Data sets were generally assessed on a per plot basis (n = 4).

Differences between treatments in the same year/assessment date as well as within the same cultivar were assessed based on the mean comparison (P < 0.05) using the software R version 4.0.2. Normality and homogeneity of the data were assessed using Shapiro–Wilk and Fisher–Snedecor tests, respectively. When the data followed a Gaussian distribution and had homogenous variances, a comparison of means was made using Student’s t-test; when data followed a heterogenous Gaussian distribution but had variances, a comparison of means was done using a Welch t-test that is robust against heterogeneous variances and when the distributions were not normal, a Wilcoxon–Mann–Whitney test (U-test) was used.

To detect consistent effects over all five years, annual mean values of the (i) density index, (ii) marketable yield per plant, (iii) non-marketable yield per plant, (iv) total yield per plant, (v) proportion of the non-marketable yield on the total yield and (vi) total soluble solids at harvest were normalised by dividing the value of the treatment by the value of the control treatment. Furthermore, the annual deviation (in days) from the control were calculated per each treatment based on the day of the year, reaching 5 % disease severity. Normalised values of both treatments were compared pairwise (P ≤ 0.05) by an independent-sample Welch test (unequal variance t-test).

Key meteorological data for the years 2016 to 2020 are reported in Table 2.

Table

titre du tableau
	Year	Growing Season	BBCH 65 Riesling	BBCH 65 Pinot gris
Year	Mean temperature (°C)	Precipitation sum (mm)	Mean temperature (°C)	Precipitation sum (mm)	Date	Day of the year	Date	Day of the year
2016	10.9	760	15.6	462	26.06.	178	24.06.	176
2017	11.0	725	15.5	413	10.06.	161	10.06.	161
2018	11.8	716	17.0	295	05.06.	156	04.06.	155
2019	11.4	764	15.8	397	21.06.	172	20.06.	171
2020	11.9	774	16.2	380	09.06.	161	03.06.	155
Average	11.4	748	16.0	389	14.06.	165	12.06.	163

The average temperatures of the growing seasons (April–October) were rather similar in 2016, 2017, 2019 and 2020, ranging from 15.5 °C to 16.2 °C. In 2018, an average growing season temperature of 17.0 °C was recorded (Table 2). The lowest precipitation sum within the growing season was observed in 2018 (295 mm) and the highest precipitation sum was recorded in 2016 (462 mm) (Table 2). The phenological stage of full flowering (BBCH 65) was reached in Riesling between June 5 (2018) and June 26 (2016) and between June 3 (2020) and June 24 (2016) in Pinot gris (Table 2).

Cluster density in Pinot gris was significantly higher in 2017 in the PDP treatment than in the control (Table 3). In the other years, no significant differences between the treatments were recorded (Table 3). The average normalised density index in PDP was 1.01 and did not differ from the control (Table 3). In Riesling, significantly lower cluster density was recorded in the PDP treatment compared to the control in 2016 and 2020 (Table 3). The average normalised density index in the PDP treatment was 0.92, while the reduction was not statistically significant according to the Welch t-test (P = 0.08) (Table 3).

Table

titre du tableau
		2016	2017	2018	2019	2020	Average normalised density index
Pinot gris	Control	2.7	3.5	3.2	4.2	3.6	1.00
	PDP	2.8	3.7	3.0	4.32	3.5	1.01
	Sig.	0.245	0.023	0.062	0.724	0.114	0.74
Riesling	Control	3.6	3.6	2.8	3.7	3.5	1.00
	PDP	3.3	3.6	2.7	3.6	2.8	0.92
	Sig.	0.012	0.486	0.380	0.697	0.003	0.08

In Pinot gris, the analyses of cluster structure parameters in 2020 did not show significant effects of the treatments on cluster length, weight, number of berries per cluster and berry weight, while the cluster weight per cm of cluster length was significantly reduced in PDP (Table 4). In Riesling, PDP significantly reduced the cluster weight, berry weight and cluster weight per cm of cluster length in 2020. The cluster length and number of berries per cluster were not significantly affected (Table 4).

Table

titre du tableau
		Cluster length (cm)	Cluster weight (g)	Berries per cluster (n)	Berry weight (g)	Cluster compactness (g/cm)
Pinot Gris	Control	14.0	193.4	166.5	1.20	13.5
	PDP	14.5	174.1	144.9	1.22	11.9
	Sig.	0.374	0.152	0.065	0.657	0.025
Riesling	Control	11.9	178.4	138.6	1.29	14.8
	PDP	12.1	125.5	120.1	1.04	10.3
	Sig.	0.566	< 0.0001	0.059	0.015	< 0.0001

In 4 out of the 5 years of the +study (2017, 2018, 2019 and 2020), disease severity in Pinot gris was significantly lower in PDP treatment than in control. In Riesling, the disease severity at the final assessment date was lower in the PDP treatment for all five years, while differences were significant in 2016, 2017 and 2018.

Table

Marketable yield (healthy grapes) was significantly reduced by PDP in 2 out of the 10 cultivar*year combinations (Pinot gris 2019; Riesling 2020). The average normalised marketable yield in the PDP treatment was 0.95 in Pinot gris (5 % yield reduction) and 0.9 in Riesling (10 % yield reduction), with no significant differences compared to the control (Supplementary Table 3; Figure 3).

The non-marketable yield (rotten grapes) was significantly reduced by PDP in 5 out of the 10 cultivar*year combinations (Pinot gris 2018 and 2019; Riesling 2016, 2018 and 2020). The average normalised non-marketable yield in the PDP treatment was lower than control corresponding to 0.53 in Pinot gris (47 % reduction) and 0.41 in Riesling (59 % reduction) (Supplementary Table 3; Figure 3).

Table

In the present study, PDP was demonstrated to have impressive effects on the development of bunch rot epidemics. In all the 10 cultivar*year combinations, the disease severity at the final assessment date was lower in PDP than in the control; this effect was significant in 7 out of the 10 cultivar*year combinations (Figure 1). PDP delayed the bunch rot epidemic (measured as DOY reaching 5 % disease severity) significantly in both the cultivars and the five years (Figure 2, Supplementary Table 2). The average delay of the epidemic was 10.3 days in Pinot gris and 8.3 days in Riesling (Figure 2; Supplementary Table 2), indicating a high treatment efficiency. In addition, the proportion of non-marketable yield in PDP was reduced by 41 % (Pinot gris) or even 53 % (Riesling) as compared to the control (Figure 3; Supplementary Table 3).

The observed delay of the epidemic is of special interest for relatively cool climate grape growing regions, where the date of harvest is frequently determined by crop health status rather than by optimum grape maturity (Molitor et al., 2012a). Here, the delay of the epidemic creates a temporal as well as a maturation buffer before the grapes reach a disease severity threshold that forces the grower to harvest and thus increases the chances of reaching the status of full grape maturity. Since wines from late-harvested grapes are often perceived as being of higher quality (Spring, 2004), PDP may contribute to increasing the potential wine quality.

In the present study, a reduction of the total yield in PDP compared to the control was observed (on average: –10 % in Pinot gris; –19 % in Riesling), but the effect was not statistically significant. However, this yield reduction as a consequence of PDP is approximately equivalent to yield reductions observed as a consequence of other crop cultural measures reducing cluster compactness (Evers et al., 2010; Molitor et al., 2011a; Molitor et al., 2011b; Molitor et al., 2012a; Molitor et al., 2015a; Molitor et al., 2017). The reduction of the marketable yield caused by PDP (5 % in Pinot gris and 10 % in Riesling), however, is distinctively lower than the reduction of the total yield due to the reduced proportion of clusters or cluster parts affected by rot. This lower proportion of the negative fraction in PDP is supposed to reduce the time needed for negative selection processes during harvest.

The efficiency of PDP against bunch rot infections is likely related to (i) a lower cluster compactness or (ii) a devigorating effect on single shoot growth by the increased bud load, which might improve the canopy microclimate after the removal of the second cane. Both effects potentially have a strong impact on the onset and development of bunch rot epidemics. This might alleviate effects caused by compact bunches and dense canopies such as: (i) increased risk of berry cracking due to high pressure in the interior parts of the cluster (Smart and Robinson, 1991), (ii) low sun exposure of clusters and low air circulation in the cluster zone leading to microclimatic conditions favouring the development of fungal pathogens (Zoecklein et al., 1992), (iii) rapid spread of fungal mycelium from berry to berry (Hed et al., 2009), and (iv) reduced spray penetration and coverage of inner fruit by fungicides (Hed et al., 2009; Molitor et al., 2015a).

Although the effects of lower shoot growth and less compact bunches on the development of botrytis epidemics are difficult to separate, we assume that it is rather the improved cluster structure than the microclimate that determines the effectivity of PDP. In this study, we have only gathered detailed data on cluster structure during the 2020 growing season. These data show that berry number per cluster and cluster weight were reduced under PDP irrespective of cultivar, but that there were cultivar x treatment interactions for berry size, which was not reduced in P. gris, but significantly reduced in Riesling. The latter finding was consistent with lower cluster density index values observed under PDP only in Riesling. This indicates that Pinot gris can compensate for perturbations in source/sink relations with an increase in berry size for a longer time than Riesling, in line with other source/sink manipulation experiments from the same vineyard (Molitor et al., 2012a).

The extent of yield reduction under PDP seemed to be correlated with its efficiency against botrytis infections, giving an indication that, indeed, the modifications of cluster structure (i.e., berry number and size) determine the efficiency of PDP, and not a more general devigorating effect caused by the increase in vine capacity via the pruning level. If a high number of organs compete for reserves, as in the case of PDP relative to the control, then the shoot length and leaf area development decrease as the amount of reserve carbohydrates allocated to an individual shoot decrease (Kliewer and Dokoozlian, 2005). This limits the C supply to the flowers, thus decreasing cell division rates, flower size, fruit set (Keller et al., 2010) and seed number, leading to reduced berry size and numbers (Keller, 2015). Cell division and expansion are still dependent on source/sink relationships after berry set, and thus right up the removal of the second cane in PDP. Such source limitations have been shown to successfully reduce berry size (Hed and Centinari, 2018; Kotseridis et al., 2012) and may play a role in defining cluster morphology under PDP.

Previous studies investigated the potential to delay the bunch rot epidemic using direct or indirect measures. While botryticide application (fenhexamid) was reported to delay the epidemic by 0.1 to 3.8 days (Molitor et al., 2018), the application of bioregulators (prohexadione-Ca; gibberellic acid) by 2.5 (Molitor et al., 2011b) or 0.3 days (Evers et al., 2010), flower debris removal by 3.7 days (Molitor et al., 2015b), cluster-zone leaf removal by 7.3 days (Molitor et al., 2011b), late first shoot topping by 4.3 days (Molitor et al., 2015a) and cluster division by up to 28.2 days (Molitor et al., 2012a), the observed delay of the time taken to reach 5 % disease severity between 8.3 and 10.3 days due to PDP is remarkable. Consequently, PDP might be incorporated as a strong brick in a complex bunch rot control strategy, including different indirect or direct control measures. Consequently, the decision of which and how many measures are integrated into the complex strategy needs to be decided upon depending on the local and annual disease pressure, the specific degree of bunch rot susceptibility of a cultivar and the oenological target (Molitor et al., 2017).

As part of the complex bunch rot control strategy, PDP could provide an additional example for the successful integration of crop cultural measures into Integrated Pest Management.

Via its non-chemical reduction of grape sensitivity to bunch rot, it contributes to the minimisation of pesticide use and its potentially negative effects on the environment.

The time demand for the removal of the second fruiting cane (including the new shoot growth) in PDP was approximately 10 hours per ha (data not shown). Compared to the time demand of other cultural practices, such as manual cluster-zone leaf removal or cluster division (75–100 h/ha (Schultz et al., 2003)) or the costs for the application of botryticides or bioregulators, PDP represents a relatively cost-effective operation with high efficiency

In practice, the workload for removal of the second fruiting cane might be lower in cultivars with a lower adhesion power of the tendrils because here, young shoots are fixed to the trellis less firmly.

Table