Practical Winery
58-D Paul Drive, San Rafael, CA 94903-2054
phone:415/479-5819 · fax:415/492-9325
Winegrowing -

September/October 1999 

Evaluation of Cabernet Sauvignon: three vine spacings, two trellis systems; Oakville district, Napa Valley

By Don Williams and Richard Arnold,
Robert Mondavi Inc., Oakville, CA

High grapevine planting density, common in many parts of Europe, is a relatively new concept in California viticulture. Planting costs, equipment availability, the average price per ton of grapes, and California’s soil and climate, which allow vines to grow fairly large in most valley floor locations, were all factors favoring fairly wide vine spacing. In Napa Valley prior to the late 1980s, it was rare to find vine densities exceeding 600 vines per acre, and 8x12 foot spacing (vine x row, totaling 454 vines per acre) was one of the most common plantings.

Emergence of phylloxera in Napa Valley in the 1980s made it clear that a large percentage of valley floor vineyards would need to be replanted before the turn of the century. While economically challenging, this created the opportunity to evaluate not only different rootstock selections, but also new clones, trellis systems, and vine planting densities. The possible impact of high-density plantings in Napa Valley vineyards on vine growth, crop yield, and wine style was largely unknown.

In 1985, one vineyard in the Oakville district of Napa Valley was planted with 1x1.8 meter vine spacing, a density of 2,178 vines per acre. This was the first large-scale planting (35 acres) at this density in Napa Valley. When the vines matured, the vineyard demonstrated that it was possible to produce high quality wines from closely spaced vines. No alternate spacings were planted, however, so relative differences could not be compared.

In 1989 Robert Mondavi Winery established a trial on six acres of the To-Kalon estate in Oakville, CA, to assess the effects of yield and wine style from different vine spacings. Crop yield and shoot growth characteristics have been recorded every year since 1992, and wine quality differences have been evaluated on a production scale (4 to 8 tons) since 1993.

Materials and methods:

There are three vine spacings in the trial: 1x1 meter (3.1 x 3.1 feet), referred to as close treatment, 1 x 1.8 meter (3.1 x 3.6 feet), moderate treatment, and 1.5 x 2.7 meters (5 x 9 feet), wide treatment. A vertical trellis with shoot-positioning wires (VSP) is used in the close and moderate spacings. In the 1.5 x 2.7 meter spacing, two different trellis systems were compared: VSP trellis (wide treatment) and a divided-canopy Lyre trellis (Lyre treatment). These four treatments are described in Table I. Each treatment has two replicates, totaling 1.5 acres per treatment.

Two rootstocks, 110R and 5C, were planted in each vine spacing to observe potential differences between rootstocks. Approximately half the rows in each treatment and block were planted to each of the two rootstocks.

The rootstocks were planted in the spring of 1989 and field-budded with Cabernet Sauvignon FPMS Clone 8 that fall. The vines were trained out into bilateral cordons in 1990 and 1991. Quadrilateral cordons were established in the Lyre trellis treatment in 1991 and 1992.

Soil in the trial area is described as Bale series clay-loam. The soil is deep, with no apparent restriction zones down to at least three meters. Drainage is fairly good. Row orientation is northeast-southwest, with a gentle slope of less than 3%.

All treatments use vertical shoot positioning wires, but the cordon and top wire heights vary. For the close spacing, the cordon wire was established 51 cm (20 inches) above ground level, and the top wire positioned at 122 cm (48 inches). For both the moderate and wide treatments, the cordon wire was at 76 cm (30 inches) and the top wire at 173 cm (68 inches). The Lyre trellis had the highest cordon development, with the cordons positioned at 112 cm (44 inches) and the top wire at 198 (78 inches).

The vines were balance-pruned each year to 15 or 16 nodes per kilogram of pruning weight. Because of vine vigor differences, cordon development time differed between the treatments. Cordons in the moderate and wide spacings filled out the trellis within three to five years. The close-spacing cordons took longer to establish, and some vines in the Lyre trellis still do not have cordons fully developed. Shoot spacing was kept similar in all treatments, averaging 16 centimeters between two-bud spurs.

There were two replications of each treatment. Data vine sets used six vines per rootstock per block in the moderate, wide, and Lyre treatments (n=24), and 10 vines per rootstock per block in the close spacing (n=40). Since this report focuses on vine spacing, the averages for the two rootstocks in each treatment have been combined and are presented together, though a few comments on major differences found between the rootstocks are mentioned in the results.

Because the trial involved production-scale lots (4 to 8 tons), with the goal of maximizing wine quality for each treatment, some cultural practices differed between treatments. These included irrigation amounts and timing of tillage.


For most parameters presented, values represent seven-year averages (1992 to 1998). Leaf-area estimation data was not collected in 1996 or 1997, so all leaf-area parameters represent five-year averages (1992 to 1995 and 1998). Single-factor statistical analysis of variance was run on the viticultural data, evaluating treatment by year. Statistical differences reported are all at .05 significance level.

Light readings were taken in the fruiting zone with a Decagon Corp. 80-sensor ceptometer. Readings were usually taken in August, shortly after veraison.

Leaf-size determination from 1992 to 1995 was made by a weight/size calculation. In each treatment, two 20-leaf samples per rootstock and block were taken from data vines in early August (after hedging). Main shoot leaves (taken equally from the basal, mid-, and upper portions of the shoot) and lateral leaves were sampled separately.

After the petiole stem was excised, 20 leaves from each sample were weighed as a group and an average leaf weight was recorded. A cork borer of known area was then used to cut out discs from the mid-portion of each leaf, and the 20 discs were weighed. Leaf size was calculated by comparing disc size/disc weight against the leaf weight.

In 1998, the leaf size was determined by a leaf area measuring device (Delta T Manufacturing Co.) at the University of California Oakville Experimental Vineyard. Some leaves were subsequently sized using the earlier size/weight ratio method for comparison, and differences between the two techniques were found to be less than 10%.

Node number was determined by node counts of data vine shoots after leaf fall, usually in December. While some lateral nodes had broken off, particularly in years of early rain or frost, it was presumed that the lateral leaf area not accounted for was small.

Grapes were hand-harvested, crushed in an Amos crusher/destemmer, and fermented in 10-ton stainless steel tanks. Skin contact time was four to five weeks, depending on the vintage. The wines were pressed in a membrane press, and 10-gallon samples were taken after draining and pressing. The wine samples completed malolactic fermentation, were put through a racking and SO2 addition program, and were bottled without wood ageing for analytical and sensory evaluation.


Canopy growth — Vine growth data are listed in Table II and leaf canopy data in Table III. Shoot number increased with wider vine spacing. The wide spacing, with 50% more space between vines, averaged 91% more shoots per vine than the close spacing and 42% more shoots per vine than the moderate spacing.

The close and moderate spacing treatments both have one meter separation within the vine row, but due to stronger shoot vigor and earlier cordon development, the moderate spacing averaged more shoots per vine every year until 1998. Shoot number per vine differences between the close and moderate vine spacing were not significant, however.

At the 1.5 x 2.7 meter spacing, the divided-canopy Lyre trellis averaged 59% more shoots than the single- canopy wide treatment. Since the Lyre trellis has twice the available cordon wire length, this indicates that the Lyre cordons have not completely filled the available trellis wire.

On a per vine basis, as vine spacing increased, the pruning weight increased by a higher percentage than the number of shoots. Therefore the average shoot weight increased with vine spacing, with the wide-spacing shoots significantly heavier than the close-spacing shoots. On a per-square-meter-of-soil basis, however, the ratio is reversed, with the close spacing having the highest pruning-weight and shoot-weight values and the wide spacing having the lowest of these values of the three vine spacings.

Due primarily to a lower hedging height, the close spacing averaged three fewer nodes per main shoot than the other three treatments. The moderate, wide, and Lyre treatments had similar average node numbers per shoot after top hedging. Main shoot leaf size, total leaf area per shoot, and percent lateral leaf area did not show significant differences.

The close spacing and Lyre treatments represent the extremes in per-vine values for number of shoots, cluster number, crop yield, pruning weight, and percent lateral leaf area (though differences in canopy hedging height probably affected lateral shoot production). For a number of other parameters, the Lyre trellis values were closer to the close spacing than either the moderate or wide spacings, including average shoot weight, cluster weight, leaf size, and total leaf area per shoot.

Crop yield parameters — Table IV shows seven-year averages of crop yield parameters. Cluster number and crop weight per vine increased as vine spacing increased, primarily due to increased shoot number per vine. Cluster weight also increased, with a 49% increase from the close to the wide spacings.

The increase in cluster weight was due primarily to significantly more berries per cluster with each increase in vine spacing. There was a trend towards increased berry size with wider vine spacing, but the differences were not significant. The increase in both cluster weight and berry number per cluster as vine spacing increased occurred in all seven vintages.

Comparing trellis systems at the 1.5 x 2.7 meter spacing, the Lyre trellis had significantly more clusters and greater crop weight per vine, due to more shoots, but significantly lower cluster weight and number of berries per cluster than the single-canopy, wide-spacing vines.

Total yields over the seven years averaged between 6.8 to 7.0 tons per acre for the close, moderate, and Lyre treatments. In no individual vintage were the values that close, however, and the close, moderate, and Lyre treatments each had years where they had the highest calculated tons per acre. The wide spacing averaged 5.6 tons per acre, approximately 20% lower than the other three treatments, but this difference was not significant at the .05 level.

Canopy/crop ratios — Table V lists two common ratios that relate canopy growth to crop yield. Differences between values were not significant, but the two 1.5 x 2.7 meter treatments averaged approximately 20% higher crop-weight/pruning-weight ratios and 20% lower leaf-area/crop-weight ratios than the two closer vine spacing treatments.

Rootstock differences — The vines on 110R rootstock established themselves earlier and have averaged slightly higher vine-vigor values (shoot number, shoot weight, pruning weight per vine, cluster number per vine, and leaf-area/crop-weight ratio) than the vines on 5C rootstock. The 5C vines averaged slightly higher crop-yield values (berry weight, cluster weight, crop weight per vine, tons per acre, and crop-weight/pruning-weight ratio). Differences between rootstocks that exceeded 10% are listed in Table VI.

Must and wine analysis —Malic acid, tartaric acid, and arginine content of berries sampled just prior to harvest are presented in Table VII, and wine analysis data are in Table VIII. The goal was to harvest the treatments between 23.0° and 24.5° Brix, depending on the vintage, and to have no more than 0.5° Brix separation between treatments in any one year. This was accomplished in most, but not all vintages, as other berry maturity characteristics were also used as criteria for harvest. Since the wines were not always picked at similar Brix values, statistics have not been run on wine analysis.

In most vintages, harvest of the four treatments happened within one week of each other. The close spacing was usually harvested first, but often had a lower alcohol content. Averaged across all vintages, the wine alcohol levels were all between 13.2% and 13.4%.

The wide spacing had the highest average color intensity (sum of absorbance values from the 420, 520, and 620 wavelengths). The close spacing had the highest color intensity in several vintages, but over-cropping in 1993 resulted in much lower color intensity and total phenolic (A280) values that particular year.

Most other berry and wine parameters showed few consistent trends from vintage to vintage. These included wine potassium content and malic acid, tartaric acid, and arginine values from berry samples taken just prior to harvest.

Sensory evaluation — Duo-trio difference tastings (in dark glasses) and preference-ranking tastings (blind, in clear glasses) were performed on pre-aged wine samples in all vintages from 1992 through 1997. Only preference-ranking tasting was performed on the 1998 wines. In 1992, separate wines were made from the two rootstocks in all spacings, and tasting evaluations were made within each rootstock. In 1997, the Lyre trellis was not evaluated due to a very sluggish alcoholic fermentation.

No significant differences were found between any wines in 1996 or 1997 or within the 110R rootstock wines in 1992. In other vintages significant differences at the .05 or higher level were found in several treatment comparisons.

Difference tastings between the close and moderate spacing wines showed significant differences in three of the six vintages they were compared. Comparisons between both close-to-wide and moderate-to-wide spacing wines were significantly different in two of five vintages in which they were evaluated. The Lyre-trellis wine was significantly different from the wide-spacing wine in only one of five vintages.

In the preference-ranking tastings, the tasters were asked to rank the wines for overall preference in all vintages, rather than being directed towards any particular attribute. Significant preference between wines was only found in two vintages, 1995 and 1998. In 1995, both the wide and Lyre wines were significantly preferred to the moderate-spacing wine. In 1998, the close spacing was significantly preferred to the moderate- and wide-spacing wines, and the Lyre trellis was significantly preferred to the moderate-spacing wine.


When this trial was being set up, it was apparent that not all cultural practices could be identical in the growth and development of the vines. Different trellis and canopy heights, variable harvest maturity, amounts of irrigation and thinning, and choosing what criteria to normalize for comparison purposes were factors that needed consideration.

Differences in vine size ruled out establishing any per vine parameters to normalize. Having similar tons per acre yield was considered, and multi-year averages of three of the four treatments have very close values. While economic comparisons on a per acre basis may be useful, optimum wine quality may result from different tons per acre yields for the different vine spacing.

It was decided that the most valid way to compare the different vine spacings was to use the same shoot spacing and balanced pruning formula, and aim for similar canopy-growth/crop-yield ratios (crop weight/pruning weight or leaf area/crop weight). Accurately predicting these ratios prior to harvest is difficult, however. While differences between treatment ratios were not significant in this trial, they did exceed 20%.

Harvest timing varied because these were production-scale wine lots, and the winemakers were looking at berry flavor development and phenolic changes in skins and seeds as guides for determining harvest timing. Unfortunately, these did not always occur at similar Brix levels between treatments. If the Brix differences were too large, however, phenolic extraction differences from differential alcohol levels would occur. In some vintages a compromise was reached in an attempt to keep treatment alcohol levels within 0.5% by volume.

Concerns of excessive fruit shading with closer vine spacing have not been realized in this trial site. Although the canopy height exceeds the row width in the meter-by-meter spacing, and lateral leaf area is highest in this spacing, a desirable light environment still existed (light readings in the fruiting zone averaged over 15% of ambient in all treatments). The VSP trellis pulls most of the leaves out of the fruiting zone, allowing adequate light penetration to the fruit. In addition, at Napa Valley’s latitude (37 North), sunburn damage is a concern during summer heat spikes, and partial fruit shading is desirable.

Based on our winemaking experience over the seven years of this trial, keeping crop-weight/pruning-weight ratios in the 3.5 to 4.5 range and keeping yields under seven tons per acre are desirable for optimal Cabernet Sauvignon from this site. Use of cover crops in high-rainfall years and veraison crop thinning in higher-yielding years are tools that can be used to keep canopy and crop levels from getting too high.


As vine density increased, there were decreases in per-vine values for both canopy and crop size. There were also significant decreases in individual shoot weight, cluster weight, and number of berries per cluster values. The consistently smaller and looser cluster size associated with closer vine spacing, particularly in the meter-by-meter, implies that root competition for water and/or nutrients occurs early enough in the season to affect berry set. Smaller cluster size can be desirable, depending on the site, variety, clone, and wine style for which the grapes are intended.

Seven-year averages showed very similar per acre yields from the two closest vine spacings, 1 x 1 and 1 x 1.8 meters. Both yielded over one ton per acre more than the 1.5 x 2.7 meter spacing on the same trellis system. The Lyre trellis treatment has averaged similar yield to the two closer spacing treatments, but to date has not fully filled out its cordons.

Within the same vine spacing of 1.5 x 2.7 meters, adding shoots and clusters via a quadrilateral cordon Lyre trellis system resulted in consistently lower cluster weight and shoot weight compared to a bilateral cordon, single vertical trellis. The wines were found to be significantly different from each other in only one vintage, however.

Wine "quality" conclusions are difficult to make, and blind comparison tastings did not always show significant differences. The preferred Cabernet Sauvignon wine style at this site has tended to come from a combination of factors, rather than one particular parameter of measurement. These include small cluster size, low to moderate shoot vigor and crop weight per vine, a fairly low crop-weight to pruning-weight ratio, and limiting maximum per acre yields.

Edited from presentation to ASEV Vine Spacing Symposium, June 29, 1999