Abstract

Introduction

Yield monitors for winegrapes have only been commercially available in Australia for the last three vintages. Nevertheless, early work on precision viticulture (Bramley and Proffitt, 1999, 2000, Bramley et al., 2000, Bramley 2001a,b) has demonstrated that, along with differential global positioning systems (dGPS) and geographical information systems (GIS), this technology enables grapegrowers and winemakers to acquire detailed georeferenced information about vineyard performance and to start using this to tailor production of both grapes and wine. However, the early work has shown that vineyards are highly variable. For example, yield mapping in the Coonawarra, Padthaway, Clare Valley, Riverland and Sunraysia districts shows that the range of yield variation is typically 8 to 10-fold and can be expected to be at least 4-fold, with fruit quality attributes also varying significantly (Bramley 2001a,b, Figure 1, Table 1). These early results suggest that for targeted management to be successfully implemented, an improved understanding of the winegrape production system will be required. In particular, the relationships between the inputs to winegrape production and the outputs (grapes and wine) in terms of both yield and quality need to be better understood

Figure 1. Variation in yield and selected vine and fruit indices over two years in a 7.3 ha Coonawarra vineyard under Cabernet Sauvignon (Bramley, 2001b). These maps show the sum of data normalised (μ = 0, σ = 1) for each year (1999 and 2000) and therefore indicate zones of consistent performance over the two years. Table 1 provides further statistics.

if the effects of an inherently spatially variable resource (ie the land supporting the vineyard) are to be managed in such a way that offers improvements over the current system of uniform management. Figure 1 suggests that such understanding will not be gained unless the required research is conducted in a way that recognises that vineyards are highly variable. In other words, relationships between yield, fruit and vine indices, soil properties and viticultural management need to be investigated on the basis of measurements made at common locations within blocks rather than using block (or trial plot) mean estimates which form the basis for much current knowledge. Similarly,

Table 1. Summary statistics for yield and selected fruit and vine indices (1999-2000) in a 7.3 ha Coonawarra vineyard under Cabernet Sauvignon (see Figure 1)

vineyard sampling undertaken as the basis for operational decision making will need to account for sample location in addition to number, and may need to be done more intensively than it is currently. For example, whilst irrigation scheduling decisions may involve sophisticated soil moisture monitoring equipment, it is uncommon for more than one instrument to be used in a single block or management unit, and often these are shared between several blocks or a single instrument used for management of whole properties (ie many hectares). Similarly, assessments of crop load, vigour and berry maturity made on a small number of vines are used to assist in decisions as to the use of crop thinning, leaf plucking and the timing of harvesting in blocks of several ha. Crop forecasting and the payment of premiums for fruit meeting certain quality specifications also depends on an ability to representatively sample vineyard performance. However, there is current disquiet in the industry amongst growers who are cynical about the ability of wineries to assess fruit quality in a robust manner and ensure that it is being harvested under optimal conditions. This paper aims to demonstrate that it is incumbent on both the grower and winemaker to ensure that vineyard sampling strategies are appropriate to both the specific objective of the sampling and also the inherent characteristics of the vineyard being sampled.

Assessment of the vineyard soil resource

The effects of variation in soil characteristics on yield variation in contrasting vineyards in Sunraysia (Figure 2) and Coonawarra (Figure 3) has previously been illustrated by Bramley et al. (2000) and Bramley (2001a,b). At both sites, harvesting was carried out each year using a Gregoire G120 grape harvester fitted with a differential GPS and HarvestMaster™ grape yield monitor. Each site was also surveyed for bulk electrical soil conductivity using an EM38 sensor mounted on a plastic sled and towed behind a 4-wheel motorbike (Bramley et al. 2000). At the 7.3 ha Coonawarra site (Figure 3), soil depth was measured in pits dug at 190 points, whilst at the 4.5 ha Sunraysia site, soil cores were collected from 130 points. The depth to the clayey B horizon was noted at the time of sampling this site and clay content in the 5-15 and 55-65 cm depth increments was analysed subsequently. Yield maps were produced following the protocol of Bramley and Williams (2001), with local (yield, EM38) and global (all other properties) kriging carried out using Vesper (Minasny et al., 2000). In the case of Figure 2, the yield map shown is the sum of individual maps of normalised (μ = 0, σ =

Figure 2. Variation in yield (1999-2000), the amount and position of clay in the soil profile and bulk electrical soil conductivity as measured by EM38 in a 4.5 ha

Sunraysia vineyard under Ruby Cabernet.

Figure 3. Variation in yield (1999-2001), soil depth and bulk electrical soil conductivity as measured by EM38 in a 7.3 ha Coonawarra vineyard under Cabernet Sauvignon.

1) yield produced separately for 1999 and 2000 whilst in Figure 3, the map shown was constructed as the average of normalised yield maps produced separately for 1999-2001.

It is suggested on the basis of Figure 3 that on the terra rossa soils of Coonawarra, variable soil depth leads to variation in the volume of soil that the roots can explore and thus the volume of plant available water, which in turn exerts a substantial effect on yield. In contrast, at the Sunraysia site (Figure 2), the presence of a poorly drained clayey B horizon can lead to waterlogging problems in winter and spring in those parts of the block where this horizon occurs closest to the surface. Whatever, both Figures 2 and 3 suggest that there would be much value in being able to assess the soil at high spatial resolution. Current practice in vineyard soil survey is for the soil to be sampled on 75 m grids. However, there is a considerable risk that such a strategy may lead to potentially important soil variation going un-noticed. For example, as Figure 4 shows, imposition of a 75 m grid onto the Coonawarra site, would result in only 16 observations being made in the block and a serious reduction in the utility of the map produced from these – in this case, of soil depth – compared to one based on more intensive sampling (190 points were used in this case in order to be consistent with the vine analysis shown in Figure 2). Note that since global kriging using only 16 points is not justified, Figure 4 also provides a comparison of the results obtained using a simpler interpolation methodology (inverse distance squared (IDW)). As can be seen, a similar loss of map resolution results when a 75 m sampling grid is used compared to more intensive sampling, irrespective of the method used for map interpolation.

It should not be inferred from Figure 4 that somewhere in the vicinity of 190 observations would be essential for delineation of the variation in soil depth at a resolution appropriate to targeted management. Rather, it is suggested that a tool such as EM38 (Figures 2 and 3) could be used as an aid to more detailed, but cost-effective vineyard soil survey. Thus, the results of an EM38 survey, which for the 7.3 ha Coonawarra block, would take approximately 4 hours if the sensor were to be used in both the horizontal (sensing to 60 cm depth) and vertical (to 120 cm) positions with the survey being carried out in every third row, could be used as an index of inherent soil variation and therefore, a basis for a better targeted soil sampling program. McBratney and Pringle (1999) have recently suggested that even in the case of broadacre applications of precision agriculture, where site-specific management to a resolution of (400 m²) is envisaged, sampling grids of 20-30 m would be required. Discussion with vineyard managers suggests that the minimum area for which they might consider differential management is three row widths; ie approximately 10 m. Clearly, vineyard soil survey at the required resolution in the absence of tools like EM38, would be prohibitively expensive; on the other hand, as Figure 4 suggests, the 75 m grid offers poor value for money in terms of the information it provides.

Crop assessment and forecasting

Wine companies depend on crop forecasting for operational planning and the contracting of fruit supply from non-company vineyards. An estimate of crop size is

Figure 4. Effect of sampling intensity on resolution of soil depth in a 7.3 ha Coonawarra vineyard.

generally made in early December for the purposes of assisting with negotiations with contract wineries for crushing requirements and wine storage. Another estimate is often made in early January, both as confirmation of the earlier estimate, and also as a means of assessing the amount of fruit available and options, if required, for sourcing additional fruit on the open market from non-contracted growers. The estimates are made on the basis of bunch counts soon after flowering (ie December in many Australian regions) followed by the number of berries per bunch when pea-sized (ie January). Knowledge of the number and size of bunches, together with the long-term average berry weight at harvest enables calculation of the expected yield. The problem faced by those charged with predicting yield is the number and location of vines to be sampled. Whilst there are long-established methods for estimating the required sample number for an estimate with a specified confidence interval (eg Cline, 1944), few wineries have information that would assist in the choice of location of vines to be sampled.

For vintage 2000 in Australia, the industry as a whole prepared for a crush of approximately 150,000 t more than was actually achieved. If it is assumed that 100,000 t of this can be ascribed to poor forecasting (the remainder might be due to crop damage from birds or storms or the incidence of disease), and that the costs associated with processing fruit and storing bulk wine are around $200 t^-1, the cost of poor crop forecasting industry-wide can be seen to be of the order of $20M. This analysis does not include the costs incurred by growers who may have been left without fruit contracts prior to vintage as companies predicted a lack of capacity to deal with it, nor by wineries who, during the disappointing vintage, found either that they had insufficient fruit of desired specification to meet the production targets for selected wines, or spare capacity for further production. As Figures 2 and 3 suggest, a basis for identification of zones of consistent performance and consequent knowledge of vineyard variability, should lead to an improved capacity for accurate crop forecasting. The use of normalisation (μ = 0, σ = 1) of data obtained in any single year has proved useful in the work conducted at Sunraysia and Coonawarra for which the sum of maps has worked well for datasets covering two years; the addition of a mixture of positive and negative numbers means that following summation, anything strongly negative or positive indicates consistently low or high values (eg Figure 2). For datasets covering more than two years, taking the average of the normalised maps (eg Figure 3) gives a less “spiky” result than that obtained by summation.

Crop assessment is also critical to the use and timing of a number of viticultural practices including harvesting. In the case of winegrapes, a primary determinant of the decision of whether or not to harvest is the ripeness or maturity, as measured in terms of baumé (°Bé). In essence, this is a measure of berry sugar content; for red grapes a target of 13.5 °Bé at harvest is typical. Viticulturists at Agriculture Victoria (Irymple, Sunraysia) have recently developed a recommendation for baumé assessment in vineyards (Dr Mark Krstic – pers. comm.) which advises growers that to get a measure of baumé that is representative of the maturity of fruit in the whole vineyard, the juice of berries sampled from 5 bunches on each of 4 vines should be analysed. The suggestion that 5 bunches be sampled is designed to account for within vine-variation, whilst the recommendation to sample 4 vines in a vineyard is designed to account for within-vineyard variability; for the purposes of this recommendation, “vineyard” is deemed to indicate an area of not more than 5 ha. Figure 5 examines the implications of this recommendation using the data from the target vines assessed at the Sunraysia site, and in light of the knowledge (Figure 2) that the vineyard could usefully be divided into western and eastern halves to facilitate more targeted management. Random sampling of the whole block (dashed circles in Figure 5) results in a baumé estimation of 13.4°. However, targeting the sampling according to prior knowledge of relative performance in the block gives baumé estimates of 13.7° in the better eastern portion of the block compared to 12.5° in the poorer western side (Figure 5); the mean of all 120 sampled vines is 13.1° with a range of 10.9-15.7°. The difference between 13.1 and 13.4° is not significant (p<0.05) and presumably reflects the effects of the different sample sizes (120 compared to 4) and a slightly skewed distribution of baumé data. Similarly, there is no significant (p<0.05) difference between the whole block estimate and that for the better eastern half. However, the difference between the estimates of 12.5 and 13.7° for the western and eastern halves of the block would certainly be sufficient to impact (by approximately one week) on decisions relating to the timing of harvest; the fruit in the eastern portion may be ready for picking whereas that in the western portion may not be. In the absence of any knowledge of variability in block performance, the grower has no choice but to believe the whole-block estimate and hope that post-harvest assessment of baumé at the winery weighbridge delivers a similar verdict, rather than resulting in a penalty being incurred for fruit assessed as being under-ripe. However, knowing that the maturity in the block follows a somewhat similar pattern to yield variation (Figures 2 and 5), the grower is in position to make a choice: He/she could decide that the season is such that the fruit in the western portion is unlikely to ripen further and to proceed with the harvest. In so doing, the grower could further choose to either harvest the block as a single unit (and perhaps risk a

Figure 5. Implications of vineyard variability for analysis of baumé in a 4.5 ha Sunraysia vineyard, vintage 2000. Dashed circles indicate vines selected for sampling the whole block using a completely random strategy. The black and white circles indicate vines randomly selected for sampling in the lower and higher yielding halves of the vineyard respectively (see Figure 2).

price penalty) or keep the fruit from the two sections of the block separate, in the knowledge that by accepting a lower price for one part of the crop, a higher price is almost guaranteed for the remainder. Alternatively, the grower could decide that there would be no penalty incurred by delaying the harvest for a week to allow the less ripe fruit to “catch up”. Clearly, prior knowledge of the variability of this vineyard could have a beneficial impact on its profitability especially if price premiums were payable for delivery of fruit to the winery of required specifications. A capacity for targeted harvesting could deliver similar benefits.

Conclusions

Vineyard productivity in terms of both yield and quality may be highly variable and associated with variation in the underlying soil resource. If targeted management is to be used as a means of managing and controlling variation in the grape and wine production process, information about the soil resource will be required at a much greater resolution than that provided by the industry standard 75 m grid surveys. Furthermore, since many viticultural management decisions are dependent on vine and berry sampling, knowledge of vineyard variation will lead to improved crop assessment and consequently, improved management. Yield mapping, and tools such as EM38 soil survey can provide useful information to assist in identifying zones within vineyards for which differential management might be appropriate. They should also lead to an improved capability for crop forecasting.

Acknowledgments

This work has been variously funded by CSIRO Land and Water, Southcorp Wines Pty Ltd and the Commonwealth Cooperative Research Centres Program under the aegis of the Cooperative Research Centre for Viticulture. It has benefited considerably from the input and assistance of Dr Tony Proffitt and his colleagues at Southcorp Wines (Coonawarra site), from the cooperation of Peter Walmsley and Englefield Vineyard Contracting (Sunraysia Site), and from the excellent technical support of Susie Williams. I am also most grateful to Terry Evans (Southcorp Wines) for carrying out EM38 surveys of the Coonawarra site and to staff of the Farrer Centre, Charles Sturt University, for their EM38 survey of the Sunraysia site.

References

Bramley, R.G.V. (2001a). Progress in the development of precision viticulture - Variation in yield, quality and soil properties in contrasting Australian vineyards. In: Precision tools for improving land management. (Eds L D Currie and P Loganathan). Occasional report No. 14. Fertilizer and Lime Research Centre, Massey University, Palmerston North. In press.

Bramley, R.G.V. (2001b). Variation in the yield and quality of winegrapes and the effect of soil property variation in two contrasting Australian vineyards. Proceedings of the 3^rd European Conference on Precision Agriculture, Montpellier, France. In Press.

Bramley, R.G.V. and Proffitt, A.P.B. (1999). Managing variability in viticultural production. The Australian Grapegrower and Winemaker 427 11-16. July 1999.

Bramley, R.G.V. and Proffitt, A.P.B. (2000). Managing variability in agricultural production: Opportunities for precision viticulture. Proceedings of the 5^th International Symposium on Cool Climate Viticulture and Oenology (Workshop 12 – Precision Management), Melbourne, 16-20 January, 2000.

Bramley, R.G.V., Proffitt, A.P.B., Corner, R.J. and Evans, T.D. (2000). Variation in grape yield and soil depth in two contrasting Australian vineyards. In Soil 2000: New Horizons for a New Century. Australian and New Zealand Second Joint Soils Conference. Volume 2: Oral papers (Eds. Adams, J.A. and Metherell, A.K.) 3-8 December 2000, Lincoln University. New Zealand Society of Soil Science. 29-30.

Bramley, R.G.V. and Williams, S.K. (2001). A protocol for winegrape yield maps. Proceedings of the 3^rd European Conference on Precision Agriculture, Montpellier, France. In Press.

Cline, M.G. (1944). Principles of soil sampling. Soil Science, 58, 275-288.

McBratney, A.B. and Pringle, M.J. (1999). Estimating average and proportional variograms of soil properties and their potential use in precision agriculture. Precision Agriculture, 1, 125-152.

Minasny, B., McBratney, A.B. and Whelan, B.M. (1999). VESPER version 1.0. Australian Centre for Precision Agriculture, The University of Sydney. (www.usyd.edu.au/su/agric.acpa).