Abstract
Background and goals Vineyard management practices can improve production outcomes, including fruit quality and disease prevention. The results of these practices are variable due to vineyard factors such as production capacity, growing environment, and the intensity at which the practice is applied. The meta-analysis reported herein synthesizes data on the effects of vineyard management practices, to better inform vineyard managers of their likely outcomes.
Methods and key findings A meta-analysis was used to investigate vineyard practices including cluster thinning, irrigation, leaf removal, pruning severity, pruning timing, shoot thinning, and shoot trimming, and their effects on yield and on yield components such as berry number per cluster, berry weight, cluster weight, and cluster number. Cluster thinning, leaf removal, shoot thinning, and irrigation reduction significantly reduced yield. Reduced pruning severity increased yield as the bud number increased. All yield components were significantly reduced by reduced irrigation. Cluster thinning and shoot thinning could partially compensate for reduced cluster number with corresponding increases in berry number per cluster, berry weight, and cluster weight. When pruning was less severe, the decreased berry number per cluster, berry weight, and cluster weight partially counteracted the increase in cluster number. In many cases, the timing and severity of vineyard management practices led to different outcomes; this interaction must be considered when making management decisions.
Conclusions and significance This meta-analysis provides important information to better predict yield outcomes after implementation of various vineyard management practices.
Introduction
In horticultural products such as winegrapes, there is often a trade-off between quantity and quality (Gary et al. 1998). The quantity and quality of winegrapes are affected by macro- (regional), meso- (site), and micro- (canopy) climates (Smart 1985). Although wine quality is difficult to define (Jackson and Lombard 1993, Chapman et al. 2005), as it varies by wine style, variety, region, and winemaker, there is a common understanding that yield is fundamental to the economic viability of a vineyard (Ashenfelter and Storchmann 2016). There are a variety of vineyard management practices that manipulate canopy microclimate and thus affect fruit and wine composition and yield (Reynolds and Vanden Heuvel 2009, Bonada et al. 2021). The grower maximizes profitability by increasing yield, quality, or both, over that which would be obtained if the management practice was not implemented. Management practices have different expected outcomes (Smart et al. 1990, Jackson and Lombard 1993) and the effectiveness of a given management practice varies depending on its severity, timing, interaction with the environment, and the grape variety. Shorter harvest periods and increased frequency of extreme weather events such as heat waves, drought, torrential rain, and unseasonal frost due to climate change have stimulated investigations into various management practices (Webb et al. 2010, Petrie and Sadras 2016). For example, using leaf removal to open up the canopy and expose clusters to sunlight to improve flavor, color, and phenolics in the fruit (Arnold and Bledsoe 1990, Hunter et al. 1991, Haselgrove et al. 2000) may also enhance sunburn during a heat wave (Gambetta et al. 2021).
Adequate water availability is crucial to ensure the balance between canopy growth and yield that optimizes fruit maturity and composition (Matthews and Anderson 1988, Chaves et al. 2007, Oliveira and Nieddu 2013, Trigo-Córdoba et al. 2014). In many grapegrowing regions, irrigation is used to modulate production and fruit composition. Irrigation is often supplemental to rainfall, and water stress is allowed during ripening to increase fruit phenolics, at the cost of reduced yield that varies with stress timing and severity (Bonada et al. 2023). Increased temperatures and changing rainfall patterns due to climate change increase demands on water resources and provide incentive to revise water management strategies and their relation to vineyard productivity (Cancela et al. 2016, Bonada et al. 2020). Revised irrigation practices must consider altered rainfall patterns that advance the onset of water stress during the growing season (Sebastian et al. 2015, Mirás-Avalos et al. 2016) and should prioritize using irrigation to manage heat waves (Garcia-Tejera et al. 2023). New techniques focus on efficient water use (De La Hera et al. 2007, Acevedo-Opazo et al. 2010, Pérez-Álvarez et al. 2021) to ensure economic yields (Tangolar et al. 2015).
The most common vineyard canopy intervention is winter pruning, which affects vine size and shape, altering shoot and cluster numbers and, ultimately, yield (Winkler 1931). Pruning can be timed to delay budbreak and reduce the incidence of frost damage, or to delay and spread out ripening to decompress harvest times during hot summers (Petrie et al. 2017). The effect of delayed pruning on yield is varied. Some increases have been reported (Friend and Trought 2007), but more commonly, yield decreases have been found (Poni et al. 2022).
Other management practices used extensively to regulate yield and improve grape composition include leaf removal, cluster thinning, shoot thinning, and shoot trimming (Keller et al. 2005, Poni et al. 2006, Intrieri et al. 2008, Nicolosi et al. 2012, Würz et al. 2017, Kok and Bal 2019). The effects of these practices on yield vary depending on their timing and severity (Poni et al. 2006).
Understanding how management practices affect yield and grape composition is important as grapegrowers adapt to changing environmental conditions or consider planting future vineyards. Knowing which practices give uniform outcomes across a range of sites can inform management decisions that affect vineyard profitability, for example, with cluster thinning (Preszler et al. 2010). This meta-analysis examines a wide variety of published research on vineyard management practices to understand the relationships between the management practices and their effects on yield.
Materials and Methods
Data collection and selection
Meta-analysis principles were used to identify data sources using the Web of Science (WoS) database. The specific selection criteria were published and peer-reviewed literature, with no date boundaries. The search was performed between 22 and 24 April 2022. All languages and document types were considered in the initial selection, but only articles were used in the final search. The goal was to find publications that provided original data (rather than modelled or simulated) on at least one common management practice. The data used covered one or more individual years or were averaged over more than one year. The data had to be presented in either graphs or tables for it to be extracted for re-analysis. Only data on Vitis vinifera vines grown for wine production were included. The initial search terms are listed (Table 1), with the final search terms being “grape* AND (quality OR matur* OR composition OR ripen*) AND (canopy OR irrigat* OR yield OR management)”. Each search was repeated using the Scopus database, but fewer results were obtained, so the results from the WoS database were used. We also included any relevant papers we knew or that were discovered using “snowballing” (e.g., publications identified from citations in selected papers), that were not identified by either search.
Search terms used to identify publications in Web of Science (WoS) and Scopus, and number of results obtained.
A Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) approach (Moher et al. 2009) was used to select appropriate publications from which data was extracted for further analysis (Figure 1). Data from graphs were extracted using the Graphgrabber software tool (Quintessa Ltd.). Data were initially extracted across a broad range of viticultural management practices and then selected based on relevance to vineyard management and the volume of data available. The yield components extracted were yield, cluster weight, cluster number, berry weight, and berry number per cluster. All mentions of berry number are per cluster. Fruit maturity will be the focus of a second manuscript.
Process used to identify sources of data. Figure created with Biorender.com.
Data normalization
Data for all five yield components (yield, cluster weight, cluster number, berry weight, and berry number per cluster) were often given in the publications. Some of these can also be calculated from other components, e.g., berry number per cluster can be calculated from cluster weight and berry weight. Sometimes, there were large discrepancies in the data given. For example, cluster weight measured from a subsample of clusters can provide a significantly different value than if the cluster weight is calculated from the yield and cluster number. A similar discrepancy occurred with berry number per cluster. For this reason, we calculated cluster weights from the data provided and then used the calculated cluster weight to calculate berry number per cluster from the reported berry weight.
Ratios were used to quantify the response of a given management practice (Benayas et al. 2009, Tuomisto et al. 2012).
Eq. 1
Using ratio analysis eliminated the need to convert data obtained from different publications to a common unit. A response ratio of zero indicates no difference between treatment and control; that is, the treatment had no effect. A positive response ratio indicates that for that management practice, the treatment has a larger result than the control: the larger the ratio, the greater the difference. Conversely, a negative response ratio indicates that the treatment had a smaller result than the control. The treatment chosen as the control was the one that best reflected the status quo. For instance, for the leaf, cluster, and shoot thinning experiments, the control had no leaves, clusters, or shoots removed. We did not include data on removal of shoots from the vine trunks (desuckering), although it is likely that some shoots without fruit were removed in the shoot thinning experiments. The control for shoot trimming was that where no shoot trimming occurred. Note that canopy manipulations such as shoot trimming may have been used in a vineyard for both treatment and control groups as part of normal vineyard management, but for experimental purposes, some vines received an additional shoot trimming. For the irrigation experiments, the controls were the samples that received more water, regardless of what was designated as the control in the original publication. Some original publications determined the effect of irrigation on vines in a traditionally non-irrigated vineyard, so their control was the rain-fed, non-irrigated vineyard and the irrigated vines were the treatment. To maintain consistency, we reversed this and used the irrigated vines (those receiving more water) as the control, while the rain-fed vines (or those vines receiving less irrigation water) were considered a treatment. For pruning severity, the control was the vines that had the lowest bud numbers. For pruning timing, the control was the vines that were pruned at the standard winter timing (during dormancy) and the treatments were the later pruning times (approaching, or after, budbreak).
Statistical analysis
To determine whether the response ratios for a given management practice were significantly different from zero, that is, that there was a significant change to the trait with a given management practice, a mixed model was used. In this mixed model, the response was the Response Ratio (for yield or cluster number, etc). The random factors included were Paper and Season, with Season nested in Paper. Where a particular paper described more than one site, each site was treated as a separate paper for the analysis. The inclusion of these random factors and their nesting structure allows for the different studies having different numbers of results and that some addressed multiple seasons of results. This approach avoids potential concerns about false replication. Minitab 20 Statistical Software (Minitab, Inc.) was used for the statistical analysis.
Response ratios for the different treatment levels or timings were obtained using the same mixed model as above and with a fixed factor: Management Practice, Treatment Level, or Timing included. Fisher pairwise comparisons were then used to obtain the differences in the mean response ratio for each treatment level or timing.
Where a given management practice was divided based on timing, the phenological stages and terminology used was that of the modified Eichhorn-Lorenz (E-L) system (Coombe 1995). For example, flowering occurs between E-L stages 19 and 26 and berry development, between E-L stages 27 and 33. These stages were obtained from each publication and, if necessary, converted from other growth stage schemes (Eichhorn and Lorenz 1977, Baillod and Baggiolini 1993, Lorenz et al. 1995) using the equivalent values from Coombe (1995). Groupings were also made based on severity. For irrigation levels, data from publications were extracted or converted to the uniform unit of mm water applied, and groupings were based on the percentage applied to the treatment compared to the control. Different proportions of cluster removal were also investigated. For leaf removal and shoot trimming, the proportion of leaf area removed was obtained where possible, and for shoot thinning, the proportion of shoots removed was used.
Results
Data collection
The search terms identified 4565 articles from the WoS database (Table 1) that were then narrowed down using the PRISMA approach (Figure 1).
The final number of papers from which data were selected for analysis was 413 (Figure 1), 315 of which were used in this publication (Supplemental Table 1). An extensive range of data were then extracted from these publications and tabulated. These data were then refined to the seven most common vineyard management practices with at least one yield component: cluster thinning (58 data sets), irrigation (127), leaf removal (114), pruning timing (13), pruning severity (18), shoot thinning (24), and shoot trimming (29). There were 22 sets of data for cover crops, but we excluded this management practice as it requires more than one season to implement. Cluster cutting, where a portion of the cluster was removed, was the next most common management practice, but with only four sets of data, it was also excluded. Some publications covered more than one management practice and one primary author could have more than one publication addressing a given management practice. For cluster thinning, only manual cluster thinning results were included, because mechanical cluster thinning can remove whole or partial clusters, while manual cluster thinning removes the entire cluster (Petrie and Clingeleffer 2006, Diago et al. 2010). This means mechanical cluster thinning can reduce average cluster weight, while hand thinning has the opposite effect of increasing berry weight (Petrie and Clingeleffer 2006). Mechanical cluster thinning can also damage berries and impede their development, reducing average berry weight, which is unlikely to occur during hand thinning (Petrie and Clingeleffer 2006). We did not expect any difference in vine performance between mechanized and manual leaf removal or shoot thinning (Guidoni et al. 2008), so both types of treatments were included for those management practices.
Irrigation (reduced)
The response ratios for yield and all yield components decreased significantly when irrigation was reduced (Figure 2). The decreased yield as a result of reduced irrigation was due more to reduced berry and cluster weights than to reduced berry number per cluster or cluster number. To further investigate these effects, irrigation was split into four levels based on the percentage of irrigation the treatment received relative to the control: <25%, 25 to <50%, 50 to <75% and ≥75% irrigation. The response ratios for yield decreased as the percentage of irrigation applied decreased (Figure 3). The response ratio for yield was not significantly altered when ≥75% irrigation was applied. Compared to the highest rate of irrigation (≥75%), the response ratio decreased nine-fold at <25% irrigation, five-fold at 25 to <50% irrigation, and four-fold at 50 to <75% irrigation. The response ratios were not significantly different between 25 to <50% and 50 to <75% irrigation (Figure 3). At ≥75% irrigation, only berry weight decreased significantly, but not enough to cause an overall decrease in yield. For irrigation between 25 and 75% irrigation, the yield components berry weight, cluster weight, and cluster number decreased to a similar extent, but berry number per cluster was unchanged. At <25% irrigation, all yield components were reduced significantly more than for the other levels of irrigation (Figure 3).
Yield component response ratios for seven different vineyard management practices using mixed models. 95% confidence intervals are shown. Confidence intervals crossing the zero line indicate that the response ratio is not significantly different from zero. Response ratio = (treatment result / control result) – 1. Note the scale of the y-axis is not consistent between each panel in the Figure.
Yield component response ratios for irrigation with different levels of irrigation applied (Irrig app) using mixed models. The different levels of irrigation were based on the percentage of irrigation the treatment received relative to the control. 95% confidence intervals are shown. Confidence intervals crossing the zero line indicate that the response ratio is not significantly different from zero. Different letters indicate significant differences between the levels for each yield component. Response ratio = (treatment result / control result) – 1.
Cluster thinning
Berry number, berry weight, and cluster weight response ratios showed significant increases for cluster thinning. However, these did not compensate for the greatly reduced response ratio for cluster number, so yield was reduced significantly (Figure 2). Cluster thinning data were divided into two groups based on the phenological stage of thinning: before ripening (E-L ≤ 33) and during ripening (E-L 34 to 39). There was a significant difference in yield response ratios between timings (Figure 4). Thinning at both stages significantly reduced yield, but cluster thinning just prior to veraison and during ripening (E-L 34 to 39) reduced the treatment yield significantly more than earlier cluster thinning (E-L ≤ 33). This was primarily because when cluster thinning occurred earlier, the berry number per cluster increased significantly, increasing cluster weight and partially compensating for the cluster number loss. When cluster thinning occurred later, at E-L 34 to 39, the cluster weight increase was due to increased berry weight. The response ratios for berry weight and cluster number were not significantly different between the two cluster thinning stages (Figure 4).
Yield component response ratios for cluster thinning at different phenological stages using mixed models. 95% confidence intervals are shown. Confidence intervals crossing the zero line indicate that the response ratio is not significantly different from zero. Different letters indicate significant differences in the response ratio between the levels. E-L, Eichhorn Lorenz. Response ratio = (treatment result / control result) – 1. Note the scale of the y-axis is not consistent between each panel in the Figure.
When cluster thinning was divided into two groups based on the proportion of clusters removed, yield was reduced in both cases, but more so when 50% or more of the clusters were removed (Figure 5). When fewer clusters were removed (<50%), berry number, berry weight, and cluster weight were unchanged, while when 50% or more of the clusters were removed, berry number and cluster weight increased, compensating to some extent for the reduced cluster number (Figure 5). Berry weight was unchanged regardless of the proportion of clusters removed.
Yield component response ratios for cluster thinning with different proportions of clusters removed using mixed models. 95% confidence intervals are shown. Confidence intervals crossing the zero line indicate that the response ratio is not significantly different from zero. Different letters indicate significant differences in the response ratio between the levels. Response ratio = (treatment result / control result) – 1. Note the scale of the y-axis is not consistent between each panel in the Figure.
Leaf removal
The decreased yield due to leaf removal was driven by significant decreases in berry number, berry weight, and cluster weight. There was no change in cluster number (Figure 2). We split the leaf removal data into four different levels based on the phenological stage: shoot and inflorescence development (E-L ≤ 18; this was generally E-L 16 to 18, but occasionally leaf removal timing was described as “preflowering,” so the designation of an E-L stage was not possible), flowering (E-L 19 to 26), berry development (E-L 27 to 33), and ripening (E-L 34 to 39). Yield response ratios differed significantly between these levels (Figure 6). Leaf removal during shoot and inflorescence development (E-L ≤ 18) and during flowering (E-L 19 to 26) reduced yield more than during either berry development (E-L 27 to 33) or ripening (E-L 34 to 39) (Figure 6). Yield was not significantly reduced when leaf removal occurred during berry ripening (E-L 34 to 39), nor were any yield components altered significantly. There was no significant difference in yield between leaf removal before flowering (E-L ≤ 18) or during flowering (E-L 19 to 26), nor when leaf removal was conducted during berry development (E-L 27 to 33) or berry ripening (E-L 34 to 39) (Figure 6). The yield reductions with leaf removal prior to berry ripening (E-L 34 to 39) were due to reductions in berry number per cluster, berry weight (not significant for E-L ≤ 18), and cluster weight. There were no changes in cluster number attributed to leaf removal at any stage (Figure 6).
Yield component response ratios for leaf removal at different phenological stages using mixed models. 95% confidence intervals are shown. Confidence intervals crossing the zero line indicate that the response ratio is not significantly different from zero. Different letters indicate significant differences in the response ratio between the levels. E-L, Eichhorn-Lorenz. Response ratio = (treatment result / control result) – 1.
Where the information was available, the leaf removal treatments were divided based on the proportion of leaf area removed. The yield response ratios were significantly less than zero when >25% of leaf area was removed. The largest drop in yield occurred when >50% of leaf area was removed, and this yield decrease was significantly greater than the decreased yield when less leaf area was removed (Figure 7). Cluster number was not significantly altered at any level of leaf removal. Berry number per cluster and cluster weight were reduced more when over half of the leaf area was removed than when less than half of the leaf area was removed, and these components were primarily responsible for the decreased yield. When >25 to 50% of leaf area was removed, yield reduction was attributed to reduced berry and cluster weight, as berry number and cluster number were unchanged. No yield components were reduced significantly when ≤25% leaf area was removed (Figure 7).
Yield component response ratios for leaf removal with different proportions of leaf area (LA) removed using mixed models. 95% confidence intervals are shown. Confidence intervals crossing the zero line indicate that the response ratio is not significantly different from zero. Different letters indicate significant differences in the response ratio between the levels. Response ratio = (treatment result / control result) – 1. Note the scale of the y-axis is not consistent between each panel in the Figure.
Pruning severity
When more buds were present, yield increased. The increased cluster number more than compensated for the decreased berry number, berry weight, and cluster weight (Figure 2). Yield increased significantly for all levels of additional buds left on the vine. Yield was significantly greater when over five-fold more buds remained than with either two- to five-fold more or ≤ double the bud number remained. Similarly, cluster number was significantly greater when more buds remained, with more clusters at over five-fold bud number than at two-to five-fold bud numbers, which in turn had more clusters than when only twice or fewer bud numbers remained than in the control (Figure 8). These increased cluster numbers more than compensated for the lower cluster weight, the result of fewer berries per cluster and lower berry weight at all levels of pruning severity. Generally, when bud number increased, the yield increased due to increased cluster number and despite reduced cluster weights.
Yield component response ratios for pruning with different proportions of buds remaining after pruning using mixed models. 95% confidence intervals are shown. Confidence intervals crossing the zero line indicate that the response ratio is not significantly different from zero. Different letters indicate significant differences in the response ratio between the levels. Response ratio = (treatment result / control result) – 1. Note the scale of the y-axis is not consistent between each panel in the Figure.
Pruning timing
No yield components were altered significantly by delayed pruning (Figure 2). The available data were split into two pruning timings: E-L stages 2 to 10 and E-L stages 11 and later. The control was E-L 1, dormant pruning. The yield response ratio for pruning at E-L 2 to 10 was not significantly different from zero and no yield components were altered when pruning occurred between E-L 2 to 10 (Figure 9). However, yield was dramatically reduced when pruning occurred at or after E-L 11 (Figure 9), due to decreased berry number per cluster, cluster weight, and cluster number. Berry weight was not altered in either pruning stage.
Yield component response ratios for pruning at different phenological stages using mixed models. 95% confidence intervals are shown. Confidence intervals crossing the zero line indicate that the response ratio is not significantly different from zero. Different letters indicate significant differences in the response ratio between the levels. E-L, Eichhorn-Lorenz. Response ratio = (treatment result / control result) – 1.
The later the pruning, the lower the yield (Figure 10). Although these relationships were not strong, the R2 for the quadratic response (R2 = 0.44) is an improvement on the linear response (R2 = 0.36), explaining eight of the remaining 64% of unexplained variation. We could surmise that the effect on yield was less severe when pruning was performed up until E-L 8 to 10 (Figure 10).
Linear (solid line) and quadratic (dashed line) relationships between yield response ratio and the Eichhorn-Lorenz (E-L) stage at which pruning occurred. The p value for both relationships was p < 0.001. Inflexion point is E-L 3.85.
Shoot thinning
Although the response ratios for berry number, berry weight, and cluster weight increased significantly with shoot thinning, these increases did not compensate for the significant decrease in cluster number response ratio, so the response ratio for overall yield decreased (Figure 2). To further investigate, shoot thinning was divided into two timings: that occurring during the shoot and inflorescence stages (E-L ≤ 18), and that occurring during flowering, berry development, or ripening (E-L 19 to 39). Shoot thinning took place as early as E-L 9 and as late as veraison, E-L 35. The yield response ratios for shoot thinning performed at both stages were significantly less than zero, although there was no significant difference between these two response ratios (Figure 11). The decreased yield when shoot thinning occurred at E-L ≤ 18 was attributed to decreased cluster number, which was partially compensated by increased berry number per cluster, berry weight, and cluster weight. With later shoot thinning at E-L 19 to 39, the decreased yield was attributed to the decrease in cluster number, which was not offset by an increase in berry weight (Figure 11).
Yield component response ratios for shoot thinning at different phenological stages using mixed models. 95% confidence intervals are shown. Confidence intervals crossing the zero line indicate that the response ratio is not significantly different from zero. Different letters indicate significant differences in the response ratio between the levels. E-L, Eichhorn-Lorenz. Response ratio = (treatment result / control result) – 1. Note the scale of the y-axis is not consistent between each panel in the Figure.
Yield differed based on the proportion of shoots removed by shoot thinning (Figure 12). Where >25% of shoots were removed, the yield response ratios were reduced significantly, but when ≤25% of shoots were removed, the yield response ratio was not significantly reduced (Figure 12). Cluster number was reduced significantly at all levels of shoot thinning, but less so when ≤25% of shoots were removed. When >50% of shoots were removed, the reduced yield was due to reduced cluster number, since berry number, berry weight, and cluster weight were unchanged. When >25 to 50% of shoots were removed, there was some compensation for the decreased cluster number with increased berry number, berry weight, and thus, cluster weight (Figure 12). When ≤25% of shoots were removed, although cluster number was reduced, berry number, berry weight, cluster weight, and yield were unchanged (Figure 12).
Yield component response ratios for shoot thinning with different proportions of shoots removed using mixed models. 95% confidence intervals are shown. Confidence intervals crossing the zero line indicate that the response ratio is not significantly different from zero. Different letters indicate significant differences in the response ratio between the levels. Response ratio = (treatment result / control result) – 1. Note the scale of the y-axis is not consistent between each panel in the Figure.
Shoot trimming
There were no changes in yield or yield components due to shoot trimming (Figure 2). Shoot trimming was divided into three levels based on the phenological stage at shoot trimming. These were before and during flowering (E-L ≤ 26; as early as E-L 15), berry development (E-L 27 to 33), and ripening (E-L 34 to 39). There were no overall yield changes when shoot trimming varied based on phenological stage (Figure 13). However, berry number per cluster increased when shoot trimming occurred early, at or before E-L 26, and decreased when trimming occurred during berry ripening (E-L 34 to 39). Berry number was unchanged when shoot trimming occurred during berry development (E-L 27 to 33). The reduced berry number from this later shoot trimming was reflected in a reduced cluster weight, but the reduction in overall yield was not significant. The significant increase in berry number when shoot trimming occurred during E-L 1 to 26 did not change cluster weight or overall yield (Figure 13).
Yield component response ratios for shoot trimming at different phenological stages using mixed models. 95% confidence intervals are shown. Confidence intervals crossing the zero line indicate that the response ratio is not significantly different from zero. Different letters indicate significant differences in the response ratio between the levels. E-L, Eichhorn-Lorenz. Response ratio = (treatment result / control result) – 1. Note the scale of the y-axis is not consistent between each panel in the Figure.
Despite a decrease in berry number when >25 to 50% leaf area was removed by shoot trimming and a decrease in cluster number when ≤25% leaf area was removed, these changes were not reflected in overall yield at any level of shoot trimming (Figure 14).
Yield component response ratios for shoot trimming with different proportions of leaf area (LA) removed using mixed models. 95% confidence intervals are shown. Confidence intervals crossing the zero line indicate that the response ratio is not significantly different from zero. Different letters indicate significant differences in the response ratio between the levels. Response ratio = (treatment result / control result) – 1. Note the scale of the y-axis is not consistent between each panel in the Figure.
Discussion
The vineyard practices in the articles we reviewed, from most to least common, were irrigation (33.2%), leaf removal (29.8%), cluster thinning (15.1%), shoot trimming (7.6%), shoot thinning (6.3%), pruning severity (4.7%), and pruning timing (3.4%). The research was performed in a variety of countries and the earliest was published in 1976. Italy, Spain, and the United States contributed ~20% of the publications each, with Australia and Canada each contributing 6.5%, and the remaining 27% from another 20 countries. Irrigation was examined in 15 countries, with most publications coming from Spain (43). The 12 articles focusing on pruning timing came from Italy, Spain, New Zealand, and Australia.
We used response ratios to increase our understanding of how these management practices affect yield and yield components (Figure 2), to inform decisions on when and where they are best used, and to more precisely predict the effect that a given management practice will have on yield.
Irrigation (reduced)
Irrigation can be necessary to supplement rainfall and ensure grape quality and quantity (Jackson and Lombard 1993, Gamero et al. 2014). Too much or too little water can be detrimental to wine quality (Balint and Reynolds 2017). Mild water stress may reduce vine vigor, resulting in reduced berry size and improved skin-to-pulp ratio, which increases color and aroma (Bravdo et al. 1985). The reduced vigor also shifts the partitioning of carbohydrates from shoot tips to fruit, improving fruit microclimate (Chaves et al. 2007) and upregulating the synthesis and accumulation of phenolic substances (Castellarin et al. 2007) that characterize fruit and wines from water-stressed vines (Casassa et al. 2015, Bonada et al. 2021). The two strategies most often used to reduce irrigation are sustained deficit irrigation (SDI), where the deficit occurs throughout the season, and regulated deficit irrigation (RDI), where the irrigation deficit is targeted to a specific growth stage (Munitz et al. 2017) to better control canopy growth and reduce berry size with only minor yield loss (Keller et al. 2008, Drori et al. 2022). Our selected publications included some that compared irrigation to rainfed vineyards and others comparing a standard irrigation strategy to lower irrigation volumes or altered application patterns. We did not divide our data based on irrigation timing or method, only based on the reduction in overall irrigation volume. However, the articles covered a range of irrigation strategies, timings, and deficits and that could mask some interactions between timing and irrigation strategy. The control was always those vines where the most irrigation was applied, often 100% of the vine’s full evapotranspiration requirement. Overall, reducing irrigation also significantly reduced all yield components; that is, the treatment yield component was always lower than the control yield component (Figure 2). The effect of water stress can depend on its timing during the growth cycle (Chorti et al. 2016). For example, water stress between budbreak and flowering can reduce shoot growth and fruit set (Mirás-Avalos et al. 2016). Whether due to reduced fruit set (Hardie and Considine 1976) or reduced berry weight and cluster number per vine (Alexander 1965, Levin et al. 2020), yield is particularly reduced when the water deficit occurs in the first few weeks after flowering (McCarthy 1997). Yield components in subsequent seasons are also affected, as flowering is a critical period of sensitivity of cluster number (bud fertility) and berry number per cluster for the following year (Guilpart et al. 2014). Yield loss due to reduced berry size is thought to result from a reduced rate of cell division in the pericarp cells (McCarthy 1997, Ojeda et al. 1999). Preveraison water stress also results in reduced berry weight and thus yield (Mirás-Avalos et al. 2016), but this is attributed to reduced cell expansion at this stage (McCarthy 1997, Ojeda et al. 1999). The effect of water stress on berry weight between flowering and veraison is irreversible (Van Zyl 1984). However, postveraison water deficit does not always affect yield (McCarthy 1997, Levin et al. 2020).
The reduced response ratio for berry number per cluster and cluster number was small; however, with many data points and small 95% confidence intervals, these small decreases were significant (Figure 2). There is inconsistency in the literature as to whether cluster number or berry number is more sensitive to water deficit (Levin 2020). Our meta-analysis showed that cluster number decreased significantly at <75% irrigation, but that berry number only decreased at the lowest level of irrigation, at <25%. This suggests that cluster number is more sensitive to water deficit. This is likely to have a carryover effect, with water stress reducing shoot number and cluster number per shoot in the following season (Petrie et al. 2004, Guilpart et al. 2014). The greatest effect of reduced irrigation was on berry weight and therefore, cluster weight (Figures 2 and 3). Berry weight was reduced at all levels of reduced irrigation (Figure 3), presumably due to reduced cell division and cell expansion (McCarthy 1997, Ojeda et al. 1999). Berry number was only reduced at the lowest level of irrigation, when <25% irrigation was applied due to reduced fruit set (Hardie and Considine 1976), contributing to the larger decrease in cluster weight at this irrigation level.
At 75% or more irrigation, berry weight was reduced, but this did not translate to a lower cluster weight. Cluster number and overall yield were also unchanged. Conversely, reductions in berry number, berry weight, cluster weight, and cluster number were greatest at <25% irrigation. Between 25 and 75% irrigation reduction, the effect on yield components was similar, with reduced cluster number, berry weight, and cluster weight contributing to decreasing yield. Because increased concentrations of anthocyanins and phenolics, and therefore potential improved grape quality, can occur with irrigation deficits (Matthews and Anderson 1988, Matthews et al. 1990, Bucchetti et al. 2011), our findings suggest that growers could reduce irrigation by up to 25% (i.e., ≥75% irrigation applied) to save water, without suffering a substantial yield penalty. Whether this level of irrigation reduction improves quality will be investigated in a subsequent publication (Cameron et al. in preparation).
Cluster thinning
An extreme method of reducing yield, cluster thinning is used primarily to prevent overcropping and ensure adequate maturity (Jackson and Lombard 1993), to reduce cluster zone congestion and the incidence of botrytis disease, or where only a target volume of fruit can be delivered based on winery and/or sales capacity (Keller et al. 2005). Unsurprisingly, both the timing and amount of crop removed affect final yield (Gatti et al. 2012). The developing fruit is a strong sink for carbohydrates (Hale and Weaver 1962), and its partial removal can alter the source-to-sink ratio (Pastore et al. 2011) and thus the balance between fruit growth and canopy development, allowing the canopy to grow larger or later into the season and degrading canopy microclimate (Smart et al. 1990, Jackson and Lombard 1993). Early cluster removal allows the vines to compensate and replace some yield loss by improving set (berry number), as shown here, or by increasing berry size (Weaver and Pool 1973). In general, cluster thinning after shoot growth has slowed may allow assimilates to be diverted to the remaining fruit (Keller et al. 2005), rather than growing unwanted excess canopy.
When clusters were thinned before ripening, there was less yield loss than when clusters were thinned during ripening, E-L 34 to 39 (Figure 4). In both instances, the cluster number was similar and largely accounted for the yield decrease. However, with early cluster thinning, the berry number per cluster and cluster weight increased, partially compensating for the decreased cluster number and mitigating some yield loss (Figure 4). The trend toward increased berry weight (Figure 2) was not significant when clusters were removed pre-ripening. Because we calculated the berry number per cluster, we obtained berry number data from papers where this response was unreported and seldom remarked upon. We propose two reasons for this increased berry number per cluster with pre-ripening cluster thinning. The first is reduced berry abscission. A large proportion (70 to 80%) of flowers do not develop into fruit (Dry et al. 2010) and are shed as the berries develop during the few weeks of fruit set (E-L 27), a process called abscission. The remaining berries would normally continue development to mature fruit. The increased berry number found with cluster thinning before ripening suggests that there was less abscission (or berry abortion) and thus, more berries per cluster, because cluster removal increased carbohydrate availability to the remaining clusters. Cluster thinning during ripening did not alter berry number, consistent with the lack of significant abscission after fruit set (Bessis and Fournioux 1992). Secondly, cluster removal in a given year results in an increased leaf area-to-fruit ratio, which could lead to increased stored carbohydrates that facilitate improved fruit set, and thus berry number per cluster, in the subsequent year (Smith and Holzapfel 2009).
Berry weight increased when clusters were thinned later in the season, but there was no change when clusters were removed earlier. Berry size/weight is limited by the number of pericarp cells, so is determined before flowering and during Stage 1 of berry development (Harris et al. 1968). However, subsequent factors such as temperature, water status, seed number, and competing “sinks” will contribute to the final berry weight by changing cell size (Harris et al. 1968). Late cluster thinning reduced competing sinks and allowed more carbohydrates to be allocated to the remaining fruit, increasing berry weight.
There were conflicting findings regarding the effect of cluster thinning on berry weight. One study reported that clusters were heavier due to increased berry number, not increased berry weight, regardless of the timing of cluster thinning (Gatti et al. 2012). This study’s early and late thinning corresponded to “flowers separated” (E-L 17) and “berry lag phase” (E-L 32), both classified as early thinning in our study. Another study found no difference in berry weight with either early (one month postbloom) or late (veraison) cluster thinning (Keller et al. 2005). Because one reason for cluster thinning is to improve grape quality, understanding how it changes berry size is important, as increased berry size can decrease color in red wine grapes (Gil et al. 2015), which may be undesirable for red wine quality.
The proportion of cluster removal varied among studies. When 50% or more of the clusters were removed, unsurprisingly, both yield and cluster number were significantly less than when fewer clusters were removed (Figure 5). With <50% of clusters removed, there were no significant compensation responses observed through increased berry number, berry weight, or cluster weight. However, when half or more of the clusters was removed, there was some compensation for yield loss through increased berry number per cluster and increased cluster weight. Lighter cluster thinning did not have as great an effect on the crop load (leaf area-to-fruit ratio) as that of heavier cluster thinning; therefore, the relative amount of additional carbohydrate supply from the leaves was not enough to support compensatory growth after lighter cluster thinning. With a higher proportion of cluster thinning, there was a greater effect on the leaf area-to-fruit ratio, with resulting benefits to subsequent-season flowering and berry number, as discussed above. Maintaining berry number and cluster weight (rather than increasing) through lighter cluster thinning could benefit wine quality, because a less compact cluster has less potential for botrytis (Sternad Lemut et al. 2015). In contrast, the higher level of cluster thinning and resulting higher berry number (Figure 5) carries the potential for tighter clusters and greater disease incidence. Clearly, the timing and proportion of clusters removed act in concert and final results depend on their relative impact. Berry number only increased when cluster thinning occurred early, between E-L ≤ 33, and when half or more of clusters were removed (response ratio 0.151, p = 0.001). When <50% clusters were removed during early cluster thinning, there was no significant increase in berry number (response ratio = 0.082, p = 0.17). This suggests that the increased berry number (and potential for tighter clusters and increased disease risk) from early cluster thinning requires a greater proportion of cluster removal and will not occur when fewer than half of clusters are removed.
The loss of yield due to cluster thinning was the greatest significant result for any practice examined. Any improvements in grape and wine quality and grape price must be substantial to offset this yield penalty. Since cluster thinning is expensive (Preszler et al. 2013), it may be preferable to use other techniques such as pruning, competitive cover crops, or decreased irrigation or fertilization to obtain the desired crop load. However, in marginal climates where spring frosts or cool and windy flowering conditions make the relationship between bud number and yield more uncertain, cluster thinning provides more precise control of yield (Clingeleffer 2010), and therefore, insurance against adverse events early in the season.
Cluster thinning is not necessarily applied randomly. Early cluster thinning may target the smaller upper clusters, while later cluster thinning can target partially unripe (green) clusters or clusters that are too compact or too large. Such details were not commonly described in the articles and so were not part of our meta-analysis. Removing the smaller upper clusters early in the season could have contributed to the increased average cluster weight. The systematic removal of clusters later in the season could also have biased our results.
Leaf removal
Leaves on main shoots can start to export carbohydrates when they are ~30% of their final size (Hale and Weaver 1962) but continue to import some carbohydrates until they reach 50% of final size (Koblet 1969). Photosynthetic rates increase as the leaves grow, peak when the leaf is full-sized (Alleweldt et al. 1982), then decline as the leaf ages (Kriedemann et al. 1970, Poni et al. 1994) or there is an increase in the leaf area-to-fruit weight ratio (Petrie et al. 2000). The effect of leaf removal thus depends on which, when, and how many leaves are removed (Caccavello et al. 2017).
Traditionally, basal leaves were removed between fruit set and veraison (E-L 27 to 34) to improve exposure of the fruit zone in denser canopies (Guidoni et al. 2008, Poni et al. 2018), thus improving grape composition (Sivilotti et al. 2016) and reducing the severity of cluster disease (Poni et al. 2006). In this scenario, basal leaves removed later in the growth cycle may be exporting less carbohydrates than apical leaves, so the impact on yield components is minimal (Bledsoe et al. 1988, Vasconcelos and Castagnoli 2000). This conclusion has been questioned, as some of the largest leaves are located in the zone that is targeted for leaf removal and may be contributing more to canopy photosynthesis than is assumed (Poni et al. 1994). Loss of photosynthetic activity due to basal leaf removal may also stimulate photosynthesis in apical leaves to compensate (Petrie et al. 2003).
Early leaf removal during flowering has been used specifically to reduce yield by temporarily reducing the availability of carbohydrates for the setting berries (Poni et al. 2006, 2023). Yield is reduced through reduced fruit set (Candolfi-Vasconcelos and Koblet 1990) and the more open, loose clusters are more resistant to disease (Tardaguila et al. 2010, Bubola et al. 2017). The reduced leaf area generated by early removal is often temporary, as the growth of laterals is stimulated and compensates for the reduced leaf area (Vasconcelos and Castagnoli 2000, Bubola et al. 2017).
Another, more recent approach is removal of apical leaves after veraison (Palliotti et al. 2013). At this stage, apical leaves are highly functional and their removal reduces the availability of carbohydrates. This reduction delays sugar accumulation, addressing the issue of progressively earlier fruit maturation and compression of the harvest period due to climate change and warming temperatures (Cameron et al. 2020). The effect of this practice on yield is likely minimal, even if much of the canopy is removed, since fruit growth has mostly ceased (Palliotti et al. 2013).
In our analysis, cluster number was the only yield component where leaf removal did not change the response ratio (Figure 2) and this was true regardless of the timing (Figure 6) or proportion of leaves removed (Figure 7). The unchanged cluster number in the same year as leaf removal is unsurprising, as it would need to stimulate cluster abortion for this to occur. A reduction in cluster numbers in subsequent years could theoretically occur, because reduced leaf area, particularly from early leaf removal (four weeks postbloom), lowers the leaf area-to-fruit ratio and thus carbohydrate reserves, reducing inflorescence number and cluster number per vine (Bennett et al. 2005, Caccavello et al. 2019). We found no evidence for this. When leaf removal occurred before veraison, yield was reduced primarily due to lower berry number and berry weight, both reflected in cluster weight (Figure 6). These effects are consistent with a “looser cluster” (Poni et al. 2006, Diago et al. 2010, Komm and Moyer 2015, Moreno et al. 2015), and overall, this loss of yield by leaf removal could be offset by the improved quality and tolerance to rot (Poni et al. 2006, Preszler et al. 2013, Hannam et al. 2015). This effect was magnified by the earlier leaf removal, with removal at or before E-L 26 (i.e., flowering and earlier) producing the largest decrease in the yield and berry number response ratio. We attributed this to decreased availability of carbohydrates, which increased berry abortion (Lebon et al. 2008). Later leaf removal during E-L 27 to 33 (berry development) produced a smaller reduction in yield and berry number. Leaf removal between E-L 34 and 39 did not affect yield or yield components. The reduced effect of later leaf removal on yield suggests that the mostly basal, fruiting zone leaves are less photosynthetically active at this time (Kriedemann et al. 1970, Alleweldt et al. 1982), so there was less impact on carbohydrate supply to the fruit.
The proportion of leaf removal also influences yield: yield decreases were found as more leaf area was removed (Acimovic et al. 2016). A greater proportion of leaf removal reduced fruit set because the shoot apex is a strong sink for carbohydrates, at the expense of the cluster (Frioni et al. 2018). Our meta-analysis found that the yield reduction was particularly large when over half the leaf area was removed (Figure 7). Contributing to this yield loss were reduced berry number and weight, and thus, cluster weight. This is consistent with findings of reduced berry set and berry weight due to defoliation (Coombe 1959). When a large proportion of canopy was removed at flowering, the carbohydrate supply was severely reduced, strongly reducing berry number due to reduced berry set (berry abortion) (Candolfi-Vasconcelos and Koblet 1990). When fewer leaves were removed, this trend was not as severe, and there was no change in yield or yield components when ≤25% of the leaf area was removed (Acimovic et al. 2016).
The effects of timing and severity of leaf removal do not act in isolation. Our data set is skewed, so when leaf removal was conducted early, prior to, and during flowering (E-L ≤ 26), only five of the 73 data points were for lighter leaf removal, i.e., ≤25% leaf area removed. For late leaf removal (E-L 34 to 39), only three of 56 data points represented more severe leaf removal, where >50% leaf area was removed. The unchanged yield response due to late leaf removal (E-L 34 to 39) may be because most of the leaf removal done at this stage was lighter (with 50% or less leaf area removed). A light early leaf removal may not provide the expected or desired lower berry number, looser cluster, and improved disease resistance, for example (Figures 6 and 7). When we further segmented our data, we found that leaf removal early in the season (E-L ≤ 18) significantly reduced berry number at >50% and at >25 to 50% leaf area removed (response ratio = −0.162, p < 0.001 and response ratio −0.143, p = 0.023, respectively). There were no data for ≤25% leaf area removed. When leaf removal occurred at E-L 19 to 26, the reduction in response ratio for berry number per cluster was only significant with heavier leaf removal, i.e., when >50% leaf area was removed (response ratio = −0.326, p = 0.001). When leaves were removed during E-L 27 to 33 and E-L 34 to 39, berry number was unchanged at all levels of leaf removal. Therefore, if the aim of leaf removal is to make the clusters looser, then leaf removal must occur at E-L stages ≤26 and be more severe, with >50% leaf area removed.
Although few studies removed apical leaves, in this treatment, the yield response ratio was significantly less than zero (results not shown). The effect on yield from apical leaf removal has been found to vary depending on the stage when apical leaves were removed and whether they were net importers or exporters of carbohydrates (Buesa et al. 2019, De Bei et al. 2019, Gatti et al. 2019, Lu et al. 2022). In our examples the removed apical leaves provided an important source of carbohydrates for the developing fruit.
Pruning severity
Pruning determines bud number, which will affect canopy size and yield (Winkler 1931). More severe pruning restricts cluster number and allocates carbohydrates to shoot growth and development; conversely, light pruning that leaves many buds favors fruit growth and can result in many, smaller shoots (Winkler 1931). There was a linear relationship between node number and yield per bearer, although cluster weight decreased as node number per bearer increased (Clingeleffer 1989). Unsurprisingly, we found that when bud numbers increased, so did yield, particularly when the increase was five times or greater (Figure 8). Cluster number increased similarly, and offset the significant decreases in berry number, berry weight, and cluster weight that were found at all levels when bud number increased. The decreases in berry and cluster weights were greatest when the greatest number of buds remained. Much of this research was performed to understand the effects of converting vineyards from manual to mechanical pruning (Clingeleffer 2010), and whether the corresponding change to more and smaller clusters resulting from more buds improves fruit composition. This will be addressed in a subsequent publication.
Pruning timing
Late pruning has been proposed as a method to avoid early spring frosts by delaying budbreak (Friend and Trought 2007), and to expand the window of fruit ripening and thus combat the advancement and compression of grape maturity due to climate change (Petrie et al. 2017). The effects of later pruning on yield have varied from large increases (Friend and Trought 2007) to large decreases (Petrie et al. 2017, Zheng et al. 2017); however, most studies report decreased yield (Poni et al. 2022). Reduced yield may be due to loss of flowers with later pruning (Frioni et al. 2016, Zheng et al. 2017) and/or a high incidence of blind nodes and lower shoot fruitfulness (Palliotti et al. 2017). One study found that decreased yield with late pruning (E-L 15 and 22) was due primarily to fewer clusters, particularly on basal shoots (Frioni et al. 2016). The authors attributed this to source limitation during cluster development, due to fewer mature leaves that would have provided carbohydrates and to elimination of growth during the late pruning in which the vine had already invested considerable carbohydrate reserves. Another study found that cluster architecture was altered by late pruning, with shorter, less compact clusters contributing to lower yields (Silvestroni et al. 2018). Where increased yield with later pruning was reported, it was attributed to clusters and shoots developing later in the season, under warmer, more favorable conditions, and to reduced berry abortion (Friend and Trought 2007). Warmer temperatures also enhanced flower fertilization in this study: the mix of berry types changed with later pruning, with a greater proportion of large, seeded berries and thus greater average berry and cluster weights.
We found that yield was heavily reduced when vines were pruned at or after E-L 11, due to reductions in cluster number, berry number, and cluster weight (Figure 9). Berry weight was unchanged, suggesting that late pruning reduced shoot fruitfulness, possibly due to a carbohydrate deficit as the vine uses carbohydrate reserves to re-establish its canopy (Frioni et al. 2016, Petrie et al. 2017). Flowering in grapevines is sensitive to carbohydrate availability and reduced levels stimulate inflorescence abortion and/or reduce set (Caspari et al. 1998). Late pruning can also disrupt budbreak and result in fewer and less fruitful shoots (Petrie et al. 2017). Lower yields often result in earlier ripening (Frioni et al. 2016); therefore, yield reduction as a consequence of later pruning could counteract any phenological delay (which is often the aim of the process) (Petrie et al. 2017). Most articles found that the delay in budbreak due to late pruning reduced or was not maintained at subsequent phenological stages (Palliotti et al. 2017, Allegro et al. 2020), which counters the use of late pruning to delay ripening.
The potential for a quadratic relationship between yield response ratio and the E-L stage of pruning suggests that time of pruning could be used with minimum yield penalty if pruning was completed by E-L 8 to 10, after which time the yield is likely to be severely reduced (Figure 10). Time of pruning may be more cost effective at substantially reducing yield than other techniques (Silvestroni et al. 2018); however, the importance of yield to vineyard economic viability and the variation in response suggest that more research is required to fully understand and predict the yield outcome of delayed pruning.
Shoot thinning
The primary aims of shoot thinning are to fine-tune yield (Bernizzoni et al. 2011), produce higher-quality fruit (Tardaguila et al. 2008), and optimize canopy density (Poni et al. 2018). The outcome of shoot thinning depends on its timing and severity (Reynolds et al. 1994, Kok and Bal 2019). Growing shoots can compete for resources, which is especially apparent early in the season before they begin exporting carbohydrates back to the rest of the vine. Therefore, if shoots are removed early, there are extra resources available for the remaining shoots, which could compensate with increase shoot growth and/or yield (Naor et al. 2002, Bernizzoni et al. 2011). Later shoot thinning has less effect on yield because older shoots are more self-sufficient and do not compete for resources, so there is less chance for yield compensation (Miller et al. 1996).
The reduced yield response ratio due to early shoot thinning (E-L ≤ 18), i.e., during shoot and inflorescence development, was due to fewer clusters, which was not fully compensated for by increased berry number and weight, and thus cluster weight (Figure 11). Although there may have been extra carbohydrates partitioned to the remaining shoots to bolster yield through berry number, berry weight, and cluster weight, this did not compensate for the reduced cluster number. When shoot thinning occurred during or after flowering, E-L 19 to 39, greater effect on yield was expected, as the vines had less opportunity to compensate for fewer clusters. We found no difference in yield decrease, and the reduced cluster number was only compensated by increased berry weight. However, we had relatively little data for shoot thinning at this later phenological stage, as it is a difficult task from a management perspective. It is also possible that more non-fruitful shoots were removed as part of these late shoot thinning operations. Later shoot removal decreased cluster number and any extra partitioning of carbohydrates to the remaining shoots bolstered berry weight, although this did not translate to a cluster weight increase and the shoot removal was too late to influence berry number.
When ≤25% of shoots were removed, yield was not significantly affected (Figure 12), despite fewer clusters. Yield was not significantly decreased until >25% of shoots were removed (Figure 12), due primarily to fewer clusters and despite a compensatory increase in berry number per cluster, berry weight, and cluster weight when >25 to 50% of shoots were removed (Figure 12). We expect that greater shoot removal decreases yield more because clusters are also being removed. Shoot thinning removes leaves, clusters, and growing shoot tips and is also likely to remove shoots with no clusters, or with smaller clusters, so the effect on leaf area-to-fruit weight ratio and the vines’ ability to compensate is potentially complex. With more severe shoot removal (over half of shoots removed), the vine could not fully compensate for the loss of clusters, with no changes in berry number, berry weight, or cluster weight. An interaction between timing and percentage of shoot removal is expected (Kok and Bal 2019). When shoot thinning occurred early, E-L ≤ 18, yield decreased significantly only when >25% of shoots were removed (>25 to 50% shoots removed: response ratio = −0.189, p = 0.001; >50% shoots removed: response ratio = −0.296, p = 0.001; ≤25% shoots removed: response ratio = −0.106, p = 0.08). For later shoot thinning (E-L 19 to 39), yield reductions were not significant, although we found little data because shoot thinning at these later stages is operationally difficult. This suggests that early shoot thinning will reduce yield only if >25% shoots are removed and that a light early shoot thinning will not significantly reduce yield.
Shoot trimming
Shoot trimming has generally been regarded as having little effect on yield or grape composition, but is useful to facilitate movement around the vineyard (Poni et al. 2023), with the additional advantage of a potential for mechanization (Parker et al. 2016, Poni et al. 2023). Shoot trimming has been used to reduce berry sugar accumulation (Filippetti et al. 2015, Herrera et al. 2015, Parker et al. 2016), offsetting increased sugar accumulation and advanced grape maturity due to climate effects (Jones and Davis 2000, Cameron et al. 2020). To delay maturity, shoot trimming targets removal of some of the leaves that provide a carbohydrate source (Valentini et al. 2019), especially later in the season. Shoot trimming earlier in the season, especially in VSP systems, when shoot growth is still occurring and many expanding leaves are still importing carbohydrates, can eliminate a sink that competes with fruit set (Collins and Dry 2009). This is reflected in increased berry number with early shoot trimming (E-L ≤ 26) (Figure 13). Shoot trimming also promotes lateral shoot growth, contributing to the increased photosynthetic demand which can redirect carbohydrates and delay sugar accumulation in the fruit (Vasconcelos and Castagnoli 2000). The interaction between the growing shoot tip, a carbohydrate sink, and stage of the season (especially the self-sufficiency of the canopy for carbohydrates) determines whether shoot trimming affects yield.
Our results reflected this dependence, and we found that the phenological stage at shoot trimming did not affect the response ratio for yield, despite some effects on yield components (Figure 13). Berry number increased significantly with early shoot trimming (E-L ≤ 26) due to increased carbohydrate supply. The significant decrease in berry number with later shoot trimming (E-L 34 to 39) and resulting cluster weight decrease was surprising. As mentioned, berry number per cluster was not always reported (or calculated) and this particular result has rarely been examined by researchers. It is possible that the late trimming reduced carbohydrate accumulation and reserves (Holzapfel et al. 2006), leading to fewer berries per cluster in subsequent seasons (Filippetti et al. 2015). An increased yield response due to shoot topping (trimming) was especially pronounced when it occurred during flowering (Collins and Dry 2009). Even when we further segmented our phenological stage groups to include flowering (E-L 19 to 26), we found no significant influence of shoot trimming on yield, although berry number increased (results not shown). The proportion of leaves removed by shoot trimming did not affect yield or yield components, other than decreased berry number when >25 to 50% of leaf area was removed, and decreased cluster number when ≤25% leaf area was removed (Figure 14).
Shoot trimming later in the season may be used to slow berry ripening, without imposing a yield penalty (Valentini et al. 2019); however, results vary. Given the widespread use of shoot trimming and the importance of yield, this relationship requires further investigation.
Conclusion
Some vineyard management practices seek to influence yield directly, while others may affect yield even when that is not the primary intention. Given the importance of yield to vineyard economic viability, we used meta-analysis principles to analyze previously published peer reviewed research articles to gain further insight into the effect of these practices on yield. Both timing and severity of the various practices gave varying yield responses and must be considered together to understand and predict the outcome of applying these practices.
Supplemental Data
The following supplemental materials are available for this article in the Supplemental tab above:
Supplemental Table 1 List of publications and vineyard management practices. CT, cluster thinning; IRR, irrigation; LR, leaf removal; PS, pruning severity; PT, pruning timing; STH, shoot thinning; STR, shoot trimming.
Footnotes
We appreciate the input from Ms. Lucy Kendall with a preliminary iteration of this analysis, Professor Victor Sadras for discussion on the implementation of the meta-data analysis, Dr. Mike Trought for discussion on analysis methods, and Dr. Keren Bindon for providing additional background data. Acknowledgement also goes to Associate Professor Graham Hepworth of the Statistical Consulting Centre at The University of Melbourne for valuable statistical input and advice. Thank you to the anonymous reviewers of the manuscript who provided valuable feedback.
Cameron W, Petrie PR and Bonada M. 2024. Effects of vineyard management practices on winegrape yield components. A review using meta-analysis. Am J Enol Vitic 75:0750007. DOI: 10.5344/ajev.2024.23046
By downloading and/or receiving this article, you agree to the Disclaimer of Warranties and Liability. If you do not agree to the Disclaimers, do not download and/or accept this article.
- Received June 2023.
- Accepted January 2024.
- Published online April 2024
- Copyright © 2024 by the American Society for Enology and Viticulture. All rights reserved.
References
Vol 75 Issue 1
Jump to section
Related Articles
Cited By...
More from this TOC section
Similar Articles
