Effects of soil type on vineyard performance and berry composition in the Río de la Plata coast ( Uruguay )

*Corresponding author : gecheverria@fagro.edu.uy Aims: Vineyards in Uruguay are concentrated over soils formed from Quaternary sediments; however, in recent years vineyard surface over soils formed from metamorphic rocks has increased. In this context, this study assessed the relationships between soil physical properties and grapevine vegetative development, yield and berry composition, in order to understand how vine response is affected and characterize the viticultural production potential for different regions.

The volume of soil potentially explored by roots depends on its physical properties.Texture is widely used for soil characterization and as root-zone volume indicator (Oliver et al., 2013).In fact, soil depth and clay fraction exert a strong influence on grapevine development and yield (Bodin and Morlat, 2006;Trought et al., 2006), maybe because they determine water supply and aeration (Lanyon et al., 2004;Tardáguila et al., 2011).However, the apparent influence of soil properties on berry composition is indirect (Zerihunet al., 2015), being mediated by their effects on canopy size, which is negatively associated with berry quality.On shallow soils, grape soluble solids content and pH are higher and titratable acidity is lower (Trought et al., 2008).Moreover, soils with less water and organic matter lead to higher anthocyanin concentrations in grape skins (Cheng et al., 2014).
Usually, soils devoted to quality-grape production show lower yields (Van Leeuwen et al., 2008;Renouf et al., 2010).However, if yield is to be used as a management goal in viticulture, optimal yield ranges for different cultivars, soil types and climatic conditions need to be developed, and so soil quality must be monitored (Lanyon et al., 2004).This is justified by the fact that high yields can be obtained in soils with large ranges in coarse elements or clays (Seguin, 1986); therefore, it is not possible to define an ideal soil using texture only (Seguin, 1983).
According to Morlat et al., 1999), it is possible to estimate the terroir-induced vigor and precocity of the vines by considering soil depth, water availability, natural drainage, root depth and slope, among other factors.This approach has been validated by Bodin and Morlat, 2006) and Morlat and Bodin, 2006).
A number of studies have reported that the best grape quality is obtained under moderate water stress conditions during the maturation period (Matthews and Anderson, 1988;van Leeuwen and Seguin, 1994;Choné et al., 2001;Ojeda et al., 2002;Roby et al., 2004).The absence of water deficit in vines during ripening determines a low quality potential due to increased competition between reproductive and vegetative sinks, as well as disorders in secondary metabolism (Koundouras et al., 1999).Similarly, severe water stress has negative consequences for grape and wine quality (Morlat et al., 1992).
Soil characterization is time-consuming and costly.Recently, electromagnetic sensors have been used for assessing soil apparent electrical conductivity (ECa), making it possible to take a large number of measurements in a short time and to differentiate soil areas according to water availability (Sudduth et al., 2005;Goulet and Barbeau, 2006).ECa is strongly correlated with soil texture (Bramley et al., 2011;Rodríguez-Pérez et al., 2011) and has proven to be a reliable predictor of within-field spatial variability, since it is significantly correlated with plant growth and yield (Bramley et al., 2011;Rossi et al., 2013).This technique allows for a rapid soil characterization that would facilitate vineyard management decision making.
In Uruguay, the viticultural areas are concentrated on the coastline of the Río de la Plata, mainly in the southern and southwestern regions of the country, over soils formed from Quaternary sediments (QS).In the last decade, vineyard area has increased in eastern and other regions, over soils formed from metamorphic rocks (MR) (Celio, 2016).
In this context, the current study, conducted across four vintages, aimed to relate grapevine vegetative development, yield, and berry composition with soil texture classes and water contents in order to characterize the viticultural potential of soils from six regions along the Río de la Plata coast of Uruguay.

Location of the studied sites
Nine plots distributed over 300 km along the Uruguayan coast of Río de la Plata, on different types of soils and geological parent materials, were selected for the current study.The plots are numbered 1 to 9 from west to east, with plots 1 and 2 in Colonia del Sacramento (Colonia Department) and plots 8 and 9 in Laguna del Sauce and Pueblo Edén (Maldonado Department) (Supplementary Figure 1).This region is characterized by high annual rainfall.According to the Instituto Nacional de Meteorología of Uruguay (http://www.meteorologia.com.uy/ServCli/tablasEstadisticas), accumulated rainfall over the growing season (September 1st to March 31st) is 721 mm in the west, 662 mm in the central region, and 645 mm in the east of the Uruguyan coast of Río de la Plata.The accumulated monthly average is 103±17 mm in the west, 95±10 mm in the central region, and 92±16 mm in the east.

Description of the experimental plots
The study was conducted from 2011 to 2014 in nine commercial, non-irrigated vineyards, where experimental plots were defined.In each plot, 30 Tannat (Vitis vinifera L.) vines from three rows with ten vines each were randomly selected.Plants were trellised to Vertical Shoot Positioning and pruned using a Guyot system; rows were oriented northsouth in all sites.Location, rootstock, year of plantation and vine spacing are reported in Table 1.

Vine vigor measurements
Exposed leaf surface (SFEp) was estimated at veraison as proposed by Carbonneau, 1995.At harvest, one shoot with clusters was collected from 10 vines, and leaves, clusters and wood were separated.Fresh weight per organ was measured, as well as shoot length.Then, samples were dried at 50°C in an oven and expressed as dry weight per plant organ.Dry weight per linear meter was estimated using the average number of shoots per linear meter from all the vines in the plot and the average dry weight of 10 shoots.A relative indicator of vigor was expressed as dry weight per cm of shoot.

Root characterization
Roots were studied by excavating trial pits in the row from soil surface to the parent material.The amount, diameter and distribution of roots were determined at different distances from the vine axis and in parallel to the row.Roots were painted, photographed, mapped over a 100-cm 2 grid, and classified according to their diameter as < 3 mm, 3 to 5 mm, and > 5 mm.
Root depth was defined as that where 90% of active roots (diameter < 3 mm) could be found.This depth was used for estimating Dryness index (DI) and soil textural class for the root zone (TCra).

Soil characterization
The soil of each plot was described according to FAO (2006) guidelines and classified following USDA Soil Taxonomy (Soil Survey Staff, 1999).Two samples were taken from each horizon in the excavated trial pits for physical and chemical analysis.Moreover, at least five more samples were collected on each plot using a manual drill to complement the observations of structure, texture, depth and presence of active roots.
Soil texture was determined by the method of Bouyoucos (1962) and soil organic carbon (SOC) by that of Walkley and Black (1934).The relative clay, sand and silt contents were calculated in all horizons down to the active-root depth, thereby assigning a new TCra.
During the 2012-2013 growing cycle, soil ECa was measured in the inter-row with an EM38 electromagnetic equipment (EM38-Geonics, Mississauga, Ontario, Canada), using the vertical   dipole, taking at least one measurement every 7 plants.These measurements were performed on the same dates as pre-dawn leaf water potential readings.
The DI was used to determine the available soil water and was calculated according to Riou and Lebon (2000), as adapted by Ferrer et al. (2007).The DI was estimated from September to harvest.The available water capacity (AWC) within the soil profile occupied by roots was estimated following the method proposed by Fernández (1979).This measure was considered the starting point in the water balance analysis.

Leaf water potential
Pre-dawn leaf water potential (LWPpd) was determined with a pressure chamber (Soil moisture equipment, Santa Barbara, CA, USA).Measurements were made before dawn in 20 adult, healthy leaves per plot at four different stages of grapevine development: fruit-set, veraison, pre-harvest and harvest.

Yield components and bunch rot incidence
At harvest, the yield of the 30 plants per plot was individually weighed, the number of clusters was counted and the average weight per bunch was calculated by dividing yield per vine by the number of clusters.Rot incidence was estimated by weighing bunches with at least 5% of berries affected and was expressed as percentage of total yield per vine.Berry weight was measured in samples of 250 randomly collected berries.

Grape samples and analysis
Harvest was carried out at "technological maturity" for each plot.The criteria for determining the harvest date were: maximal sugar content (g/L), total acidity (g H 2 SO 4 ) between 4.0 and 5.5, pH between 3.2 and 3.5, and onset of berry weight decrease; the appearance of characteristic symptoms of bunch rot and/or pH greater than 3.5 prevailed over other parameters.These parameters were analyzed periodically according to OIV (2009) methods.For this purpose, replicated 250-berry samples from all vines in each plot were collected weekly from veraison to harvest.Berry composition was determined after manually destemming the berries and obtaining the juice by crushing the pulp with an electric blender (HR2290, Phillips, The Netherlands).Soluble solids contents (SS) were measured using a refractometer (Atago N1, Atago, Tokyo, Japan); pH was determined with a pH meter (HI8521, Hanna Instruments, Villafranca Padovana, Italy); and total acidity (TA) was measured by titration and expressed as g sulfuric acid/L juice.
Total anthocyanins (ApH1), extractable anthocyanins (ApH 3.2), phenolic richness (A280) and cell maturity index (EA) were determined in the grape samples according to Glories and Augustin (1993).All the measurements were carried out in duplicate with a Shimadzu UV-1240 Mini spectrophotometer (Shimadzu, Japan), using glass (for anthocyanins) and quartz (for absorbance at 280 nm) cells with 1cm path length.The indexes were calculated considering the respective dilution of the grape extracts, according to González-Neves et al. (2004).

Statistical analysis
Multivariate techniques, such as Principal Component Analysis (PCA) and Hierarchical Clustering (HC), were used to determine the relationships between vine variables (vigor, yield and berry composition) and soil classes.Variables were standardized by vintage year, as follows: where V x is the value of the original variable at position x (mean by plot and year), and V and SD v are, respectively, the average and the standard deviation of the original variable by vintage year.
Plant response to TCra was analyzed by MANOVA and means were separated using the LSD Fisher test (p<0.10).Discriminant analysis was used to characterize the relationships between soil (depth, AWC, ECa, DI) and vine (LWPpd) water-related variables and soil profile (textural class and geological material).Moreover, Pearson's correlation coefficient was employed to assess significant relationships among variables.All statistical analyses were carried out using the Info Stat software.

Results
According to USDA classification, three soil groups were identified on the studied vineyards (Table 2), and they belonged to three textural classes: silty clay loam, silty clay, and clay loam.Parent materials were either QS or MS ("average degree" and "quaternary sediments").In plot 9, the presence of limestone over the MR justified the sub-classification.Great differences among sites were observed for root zone depth, varying from 36 to 70 cm, with soils formed over MR being the shallowest.Depth, soil textural class and SOC determine AWC, which ranged from 57 to 123 mm.According to the monthly rainfall regime in the region (96.7 mm per month), soils would be at field capacity in the springtime.
Significant differences in vine vigor were detected among soil TCra classes (Table 3).Clay loam soils showed lower SFEp values, whereas shoot length was lower under silty clay loam soils.No significant differences among soil classes were detected for total shoot and linear dry weights.However, vines grown on clay loam soils presented the highest dry weight per cm of shoot.In summary, biomass production was lower in clay loam soils.
Yield components showed significant differences among soil classes (Table 3), except for berry weight.Vines on clay loam soils showed lower yield, bunch dry weight and rot incidence, whereas vines on silty clay soils showed the highest yield and bunch rot incidence.
Berry composition was mainly unaffected by soil textural class, except for phenolic and anthocyanin concentrations (Table 3).Vineyards on clay loam soils produced berries with the highest concentrations in total and extractable anthocyanins and phenolic compounds, whereas the lowest values were observed for berries from vineyards on silty clay soils.No significant differences were detected for SS, TA, pH and EA.
Vine variables (vigor, yield components, sanitary status and berry composition) were separated according to plot and soil TCra classes by PCA (Figure 1).Principal component (PC) 1 explained 37.6% of the variance and PC2 explained 17.1%, together explaining 54.7% of the total variance in the dataset.
Those traits associated to berry quality (SS, EA, ApH1, ApH3.2 and A280) were located on the negative side of PC1, close to plots 6, 7 and 8, with clay loam soils.In addition, plot 9 (clay loam soils) was associated to berry quality, although with higher   pH values and lower yields than the former plots.Plot 3, also on clay loam soils, was not related to quality attributes but to vigor variables.
Yield components were located close to plots 4 and 5, on silty clay soils, and plots 1 and 2, on silty clay loam, on the positive side of PC1.This group of plots was associated to less negative LWPpd values.
In contrast, biomass production (dry weight of shoot per linear meter, total shoot dry weight, bunch dry weight and SFEp per vine) had a more heterogeneous distribution.Shoot length, SFEp per vine and dry weight of shoot per m were located close to plots 5 (silty clay) and 3 (clay loam), on the negative side of PC2.In contrast, total shoot dry weight and yield components (berry weight and bunch dry weight) were located on the positive side of PC2.Berry composition variables were negatively correlated with yield and vigor, except for shoot length and dry weight per cm of shoot.
In addition, a negative correlation between SFEp per vine and EA was observed.A positive correlation between dry weight per cm of shoot and anthocyanins, and a negative correlation with bunch rot weight were also shown by PCA.Moreover, a negative correlation between dry weight of shoot per m and berry sugar and phenolic concentrations was also detected.Berry pH showed a negative correlation with yield components.
The LWPpd measured between veraison and harvest showed less negative values for vines over silty clay and silty clay loam soils and more negative values for those grown on clay loam soils.The LWPpd was negatively correlated with TA.
The parent rock of each plot revealed a relationship between quality attributes and yield over MR, whereas vigor and yield were related to QS.The subclasses established for MR, such as MRad and MR/QS, showed an intermediate behavior.
When using yield components, HC grouped plots according to soil parent material (Figure 2B).As an example, plots 3, 7 and 8 had the highest coarse particle content in the soil profile and MR rock at low depth, whereas plot 6 had the same parent material but at greater depth.
The relationship between TCra and those soil properties related with water availability for plants (ECa, depth, AWC, DI and LWPpd) was studied over the 2013 growing cycle.Discriminant analysis (Figure 3A) showed that most of the variability among plots was distributed along the first canonical axis.Clay loam and silty clay soils were clearly separated by the prediction ellipses, whereas silty  clay loam soils appeared to have a behavior similar to that of the silty clay soils.The most important variable for this separation was AWC.When parent material was included in the analysis (Figure 3B), five independent groups were observed, except for a slight interception area between the ellipses for silty clay/QS and silty clay loam/QS soils.
Pearson correlation coefficients (Supplementary Table 1) for the aforementioned variables proved a strong correlation between depth and AWC, suggesting that soil depth is a determinant factor of soil water availability.A significant correlation was also observed between AWC and DI (r=0.44,p<0.05).Moreover, a high correlation was observed between ECa and depth (r=0.83,p<0.001) and ECa and AWC (r=0.81,p<0.001).A lower correlation coefficient was observed between ECa and DI (r=0.37,p<0.05).Finally, LWPpd measurements were not significantly related to any of the soil water availability variables studied.
Significant correlations between AWC and vigor, yield and berry composition variables for the growing seasons 2011-2014 were observed.Significant positive correlations were detected between AWC and SFEp per vine (r=0.53,p<0.001), yield per vine

Discussion
In this study, the potential for viticultural production of soils from six regions along the coast of Río de la Plata was characterized for the first time.As reported by a number of authors (Tisseyre et al., 2007;Trought et al., 2008;van Leeuwen, 2010), the influence of soil on vegetative expression, yield, berry composition and vine sanitary status was verified.Soil physical properties and root zone depth were highly correlated with vine performance, as previously indicated by Bodin and Morlat (2006), Trought et al. (2006) and Oliver et al. (2013).
Apart from texture, soil structure plays a relevant role in vineyard performance.In some plots with high fine particule contents, roots were not able to explore deep soil layers.oxygen.Nevertheless, Typic and Vertic Argiudoll soils associated with QS showed greater root zone depths than Hapludoll, Hapludert and Abruptic Argiudoll soils when they are formed from MR (MR or MRad).In fact, our results showed that the presence of a silty layer between the soil and MR in the Abruptic Argiudoll soil from plot 9 created deeper soils.
Independently of the assigned textural class (TCra) and parent material, all vineyards except plot 1 had a Bt horizon with more than 40% clay (data not shown).This is a common and distinctive characteristic of the soils in this region.
In accordance with the reports of Van Leeuwen et al., (2008) and Renouf et al. (2010), a negative correlation was observed between two groups of variables: one group related to berry quality and the other related to vigor and yield.Even though the concentration and the evolution of compounds derived from the primary metabolism (SS, TA and pH) is part of the criteria used for deciding harvest date, similar results between different regions were obtained on the same season; however, they did not reflect the velocity in the accumulation of these compounds.Moreover, when the comparison of these variables accounted for the year effect, as expressed by the averages and standard deviations as a function of soil class in the ANOVA, differences were not significant.In contrast, secondary metabolites were different among soil classes, even when including the year effect.
Vines grown on deep soils formed on QS and classified as silty clay or silty clay loam showed greater vegetative expression than those grown on clay loam soils on MR.ANOVA results for vigor variables confirmed this trend, even though no significant differences were detected for total shoot dry weight and total linear dry weight.The HC analysis for vigor variables grouped the plots according to TCra, except for plots 3 and 8, which belonged to the clay loam class and were classified with other plots on silty clay soils because of their high dry weight per cm of shoot and shoot length.It is possible that in those plots, the relative importance of SFEp would be small due to a larger distance between shoots, which might promote an increase in shoot diameter and length without significantly increasing leaf surface.Nevertheless, the different analyses performed did not provide contradictory results, thus an association among soil physical properties and the potential for biomass production can be established, according to the model proposed by Morlat et al. (1999), Bodin and Morlat (2006) and Morlat and Bodin (2006).
Silty clay and silty clay loam soils, deeper and lighter textured, determined higher values for yield components.According to Wang et al., 2001), the size of the root system has a positive effect on shoot growth and, consequently, on yield per vine and berry and bunch dry weights.
HC for yield components showed plots ordered by geological material and TCra almost perfectly, from AWC, to plot 5 with a high AWC to plot 7 with the lowest AWC.In summary, yield components were greater on soils over QS, with lighter texture and higher AWC, as previously reported by Lanyon et al., 2004) and Tardáguila et al., 2011).In fact, yield from vineyards on clay loam soils was 35.4% and 40.1% that of vineyards on silty clay and silty clay loam soils, respectively.
Better results for berry quality traits were obtained under shallow soils with coarse texture.Bunch sanitary status and the concentrations of total and extractable anthocyanins, tannins and soluble solids were higher in vineyards grown on these soils.
According to the Morlat et al. (1999) approach, soil properties in plots 3, 6, 7, 8 and 9 would generate a moderate water stress during maturation, thus favoring berry quality (Matthews and Anderson, 1988;van Leeuwen and Seguin, 1994;Choné et al., 2001;Ojeda et al., 2002;Roby et al., 2004;Mirás-Avalos et al., 2013).However, the weak correlation between LWPpd at maturation and some berry traits such as SS and TA identified by PCA proves that vine water status at maturation is not enough for explaining the plant qualitative response to water availability.In fact, when LWPpd measurements were initiated, vines had already attained the maximum vegetative development for the season.Therefore, in case no cultural practices (defoliation, cluster thinning) are performed, the source:sink ratio would be determined at veraison, as well as yield potential and the synthesis of berry compounds.The modulation of this potential would be dependent on the severity of the water stress at maturation, determined by DI, other environmental conditions (such as solar irradiation) and management practices (e.g.plant architecture).This implies interactions within the production system and assigns a negative correlation between vigor and berry composition, as expressed by Zerihun et al. (2015).Similarly to the results from the discriminant analysis, HC gathered plots in two groups as a function of berry composition attributes, depending on TCra and parent material.As in the former cases, plot 3 (clay loam and MRad) was included into the group of deep soils with light textures developed on QS.The hypotheses for explaining this behavior would be that: i) the higher AWC in relation to plots 6, 7 and 8 (also classified as clay loam) causes less water stress, thus affecting berry composition; and ii) since this soil presented a complex structure, and due to plant age, a second level of roots within the sub-soil may exist, allowing the vines to take water from this subsoil when fine roots from upper layers would have limited ability to do so.The latter hypothesis represents an additional buffer capacity, as mentioned by Hunter et al. (2010).
Another possible factor is the percentage of soil organic matter (SOM; data not shown) because of its negative correlation with the synthesis of ApH1 (r=-0.34;p=0.08) and ApH3.2 (r=-0.37;p=0.06).SOM depends on soil type and its management.In the root zone, the lowest SOM values were detected in plots 6, 7, 8 and 9 (clay loam), then plots 1 and 2 (silty clay loam) and plots 4 and 5 (silty clay), as expected.However, plot 3 (clay loam) on MR showed the highest SOM, even higher than that of soils formed on QS.This fact is justified by the conversion of natural prairies for grazing to vineyards.The high SOM levels explain the high AWC of plot 3 (the proportion of SOM is used for its estimation), the high vigor (SFEp, vine and shoot length) and, as a consequence, the lower berry quality.With a normal precipitation regime during springtime, plot 3 would not present water deficits during this stage, which, in combination with N abundance, would favor vigorous canopy and root development.
As suggested by Cheng et al. (2014), plots 6, 7 and 8, with less water and organic matter, were related to the highest anthocyanin concentrations in grape skins.Soil conditions in these plots would create a balance in the vines biased to sinks (low Ravaz index) causing shoots with greater diameter and longer internodes, thus reducing yield and increasing the concentration of quality traits in the berries.
Sanitary problems observed in vineyards grown on silty clay soils would be associated to the greater vine vigor, which is related to vine microclimate and increasing humidity within the canopy (Ferrer et al., 2011).
Soil depth is the main factor determining AWC (r=0.93),along with texture and SOC.The measurements of ECa were strongly correlated with AWC, as pointed out by Sudduth et al. (2005) and Goulet and Barbeau (2006), and proved to be a reliable predictor of within-field spatial variability of AWC, which significantly affected plant growth and yield (Bramley et al., 2011;Rossi et al., 2013).
From the perspective of a given yield level allowing the expression of the terroir and the economic sustainability of the wineries, optimal source:sink ratios should be defined for each soil in order to maximize yield without compromising berry quality.

Conclusions
The current study proved that soil water availability is variable and depends on depth and texture, which are correlated with soil geological origin.Deep, lighttextured soils formed over QS induced greater vegetative growth, higher yields and more sanitary problems.In contrast, coarse-textured soils formed over MR favored the production of better quality musts.Moreover, this study proved the need for further investigation on soil fertility since this is linked to physical conditions and soil use history.
According to edaphic conditions, it was possible to define the viticultural interest of different areas within the sub-regions of Río de la Plata coast.From the central region to the west, on a 40-km long strip of coastline, it is possible to achieve high yields, whereas to the east, lower yields but better quality are expected.

Figure 1 .
Figure 1.PCA of vine vigor, yield component and berry composition variables at harvest, as a function of plot, textural class in the active root zone (TCra) and geology.Variables were standardized per vintage year.