The study site was a commercial vineyard located 556 meters above sea level in the municipality of Cacabelos (León; Spain) with geographic coordinates of 42°36’N latitude and 6°45’W longitude. This area can be considered as representative of vineyards on acid soils under a Mediterranean climate. This type of vineyards occupies over 210,000 ha in Mediterranean Europe (ESD, 2004).

From a climatic point of view, the research area would be classified as Region I (≤ 1,390 Celsius degree-days) based on the system devised by Amerine & Winkler (Jackson, 2014). The mean reference evapotranspiration (FAO Penman-Monteith) and mean rainfall were 922 mm/year and 616 mm/year for the period 2010/11, respectively (SIAR, 2012). The soil under study is an Inceptisol, Suborder Xerept, Great group Haploxerept according to Soil Survey Staff (2010) and Cambisol Dystric according to IUSS Working Group WRB (2006). The soils in the study area are developed on Tertiary sediments from which calcium minerals are almost completely absent (IDEE, 2012).

The research was conducted on Vitis vinifera cv. ‘Mencía’ grafted on a 60-year-old ‘Rupestris du Lot’ rootstock, which has been classified as highly sensitive to soil acidity (Fráguas, 1999). Planting lines displayed a north-south orientation. The conduction system involved a head trained spur pruned vines, with 4-5 arms per plant. Winter pruning left a thumb-sized arm with two buds. Vineyard had no irrigation system support. No fertilizers or extra amendments other than those used in this research were applied during the study period.

The composition of the two liming materials used in this study is shown in Table 1. As expected, the dolomitic limestone exhibited a higher Mg content than the sugar foam, whereas this latter presented a higher organic matter content. Espejo (2001) stated that the high Ca content of the sugar foam is due mainly to the presence of Ca in the form of slaked lime (Ca(OH)₂) and, to a lesser extent, as carbonate (CaCO₃). The slaked lime progressively reacts with atmospheric CO₂ to produce CaCO₃. This carbonation occurs at a rate which depends on the aggregate size, porosity and water content of the sugar foam.

The experimental design was based on three treatments (control, dolomitic limestone and sugar foam) with three replications each. The whole study plot was therefore split into nine 450 m² subplots in which approximately 180 vines were grown each one at 1.5 × 1.7 m spacing in a completely randomized design. The lime requirement was calculated using the formula proposed by Cochrane et al. (1980) to achieve 80% base saturation. Therefore, 2720 kg/ha of CCE, which corresponded to 2800 kg/ha of dolomitic lime and 3900 kg/ha of sugar foam were applied. The dolomitic limestone was in a powdery state, whereas the sugar foam consisted of aggregates of variable size, which were manually disaggregated before addition to the soil. The liming materials were uniformly spread onto the entire surface of the treatment subplots, and then incorporated with one-pass tillage down to a depth of 20-30 cm in late November 2008.

Before the amendments were added, the soil was sampled and ground to pass a 2-mm mesh sieve. Textural classes according to USDA were determined by the Bouyoucos´s (1962) hydrometer method. Official methods of analysis (MAPA, 1993) were used for the determination of soil organic matter, pH in water (pH_w), electrical conductivity (EC), and the exchangeable contents of Ca, Mg, and K by atomic absorption spectrometry. The exchangeable content of Al was determined by inductively coupled plasma atomic emission spectroscopy (ICP-AES) using 1 M KCl as extractant (Little, 1964). The ECEC was obtained by addition of the exchangeable Ca, Mg, K, and Al contents. CaECEC, MgECEC, KECEC and AlECEC were therefore calculated as the respective quotients: Ca/ECEC, Mg/ECEC, K/ECEC and Al/ECEC, and expressed in %. The available P was determined by visible molecular absorption spectroscopy after extraction with 0.5 M sodium bicarbonate (Olsen et al., 1954).

After liming, the following soil properties were monitored for three years (2009, 2010 and 2011): pH_w, P, CaECEC, MgECEC, KECEC and AlECEC. This monitoring was conducted by sampling the soil in all subplots at 0-30 and 30-60 cm depths, and at the fruit set phenological stage, which occurs in late June in the study area. These depths were chosen because of the depth down to which lime had been incorporated. This was 20-30 cm, and it gave rise to a soil divided into two layers: one directly and another indirectly affected by liming. The soil samples were collected using an auger, sealed in plastic bags, transported to the laboratory and air-dried at room temperature.

The calcium, magnesium, potassium and phosphorus contents in blades and petioles were annually monitored along with the soil properties at the fruit set phenological stage. Specifically, 30 grape basal leaves from opposite bunches were randomly collected per subplot each year in mid-June. They were sealed in plastic bags, and transported to the laboratory. Once in the laboratory the leaves were carefully rinsed with deionized water, and then dried for three days at 70ºC (Bavaresco et al., 2010). Next, they were wet digested with an acid mixture of perchloric, sulphuric and nitric acid at 420ºC for 20 min (Calleja, 1978). The cation contents in the extracts were determined by ICP-AES with quartz torch using a Perkin Elmer Plasma 1000 (Perkin Elmer, Waltham, MA, USA), whereas P content was determined by visible molecular absorption spectroscopy.

The grapes were sampled at harvest every year, specifically on 12^th September 2009, 20^th September 2010 and 11^th September 2011. Those harvested grapes from each subplot were weighed to determine yield. Next, 300 grape berries per subplot were randomly chosen to determine the berry weight parameter, and to analyze must quality. The berry weight parameter was obtained by simply weighing 100 of these berries. The grape must was manually obtained from the 300 berries by gently pressing the grapes, using rubber gloves to avoid sample contamination. In the must thus obtained, the following properties were determined: actual acidity (pH), total soluble solids, total acidity, extractable anthocyanins, and finally, seed maturity. Total soluble solids were measured using a refractometer. Total acidity was determined by titration of the grape must with 0.1 M NaOH to an endpoint of pH 7, and expressed as the equivalent content of tartaric acid in g/L (OIV, 2014). Extractable anthocyanins and seed maturity were determined by the Glories method (Saint-Cricq et al., 1998) and expressed in %. To determine grape nutrient contents, the seeds, skins and flesh from 100 grapes were manually separated and immediately dried at 60ºC to constant weight before wet digestion (Calleja, 1978). Calcium, magnesium, potassium and phosphorus were determined in these extracts by ICP-AES. Calcium, magnesium, potassium and phosphorus in dried grape berries were subsequently assessed as the sum of their contents in skins, seeds and flesh.

Statistical analyses were performed using the R software (R Core Team, 2013). Several analysis of variance (ANOVA) were carried out to study the effect of liming on soil, leaf, grape yield contents, grape must quality properties and grape nutrient contents. In the ANOVAs of the soil chemical properties, the soil depth was also taken into account as a block factor with two levels, i.e. shallow soil (0-30 cm), and deep soil (30-60 cm). In all the ANOVAs the year of sampling with three levels (2009, 2010 and 2011) was also included as a block factor. Therefore, a three-way ANOVA (liming, soil depth and sampling year) was used for each one of the soil properties, whereas a two-way ANOVA (liming and sampling year) was used for the leaf, grape yield, grape must quality properties, and grape nutrient contents.

As long as the liming factor presented a significant effect on any of the properties, the variance of the ANOVA between groups, i.e. due to the liming factor, was split into two independent (orthogonal) variance contributions (summands) or main effects. These are the effect of just liming, and the effect of the specific material used to lime. Provided the experimental design, these are the only two possible contributions to the variance due to the liming factor. This variance decomposition gives rise to two orthogonal contrasts. In the first contrast, liming is compared against no-liming (C1), i.e., the control is compared against the dolomite and sugar foam treatments lumped together. In the second contrast, the liming materials are compared against each other (C2). The orthogonal contrasts allow measuring the effect sizes in a standardized way and thus, they are more rigorous and statistically efficient than the post hoc contrasts used in common practice (Field, 2012).

Before doing the ANOVA the univariate normality hypotheses for every variable were tested using the Kolmogorov-Smirnov test. All soil, leaf (blades and petioles), yield and quality must properties presented univariate normality. Additionally the null hypothesis that the variances of the groups are not different was tested using Levene’s tests. According to this test the hypothesis of homoscedasticity could be accepted for every variable at the 95% confidence level.

Table 2 shows the baseline characteristics of the two soil depths before liming. The AlECEC was 37 ± 7% (95% CI) in the 0-30 cm layer, and 42 ± 19% (95% CI) in the 30-60 cm layer. In both depth intervals the exchangeable Al content is over the 20% limit, which is considered the highest Al saturation that most plants can tolerate (Fageria & Baligar, 2008). Additionally, the important exchangeable soil acidity was revealed by the near-one differences between pH_w and pH_kc1 in both soil layers. Low P, and low exchangeable Ca, K and Mg contents were found too.

The evolution throughout the three years of monitoring of pH_w, CaECEC, MgECEC, KECEC and AlECEC and P for the treatment and control subplots, are shown in Figs. S1 and S2 [supplementary]. As can be seen, sugar foam seems to be more efficient than dolomitic lime to decrease soil acidity, especially to reduce AlECEC, and increase CaECEC and pH_w at both study depths. This effect was particularly marked the first year, while differences between sugar foam and dolomite tended to disappear the second year. Interestingly, available P tended to decrease as a consequence of liming, particularly with sugar foam (Fig. 1). Three-way ANOVAs were used to find out if the differences between liming treatments were statistically significant, and furthermore, if they depended on the soil depth and on the year of sampling, and the interactions between both. According to the results of the three-way ANOVAs, significant effects of liming on all soil properties were revealed (Table S1 [supplementary]). Besides, the effect of soil depth was revealed as significant just on P levels, whereas the year of sampling was revealed as significant only on exchangeable K. On the contrary, for the other chemical properties the effect of liming does not significantly change either with the soil depth, or with the year of sampling, and neither there are significant interactions between the three factors. Thus, whereas KECEC and P must be studied as, respectively, a function of liming and year of sampling, and a function of liming and soil depth, pH_w, CaECEC, MgECEC and AlECEC can be studied as a function of just liming (Fig. 1).

Therefore, one-way ANOVAs were carried out for each soil property (Table 3), and again significant differences of pH_w, CaECEC, MgECEC and AlECEC were observed between treatments. Specifically, pH_w, CaECEC and MgECEC increased while AlECEC decreased in soils in which liming materials were used. However, in order to find out if differences in the soil properties between treatments arose as a consequence of i) just liming (C1), or ii) liming with sugar foam or dolomitic limestone (C2), or iii) both (C1 + C2), the variance between treatments obtained from the one-way ANOVAs was decomposed according to both planned contrasts (C1 and C2). The first contrast (C1) revealed that just liming significantly increased pH_w, CaECEC and MgECEC, whereas significantly decreased KECEC and AlECEC. The second contrast (C2) revealed that sugar foam significantly increased pH_w, KECEC (in 2011) and CaECEC, whereas significantly decreased P, in the subsoil layer, and AlECEC more than dolomitic limestone (Table 3). After three years the effects of liming with both materials, and particularly with sugar foam, on all six soil properties were still remarkable (Table 4).

Means and standard errors of Ca, Mg, K and P content in blades and petioles for the dolomitic limestone and sugar foam treatment, and control subplots are shown in Fig. 2. The shaded areas in Fig. 2 illustrate the concentration ranges that are optimal at fruit set for petioles (Ca (0.6 – 1.4%), Mg (0.3 – 0.8%), K (1.0 – 2.9%) and P (0.1 – 0.6%)) according to Wooldridge et al. (2010), and blades (Ca (1.4 – 2.6%), Mg (0.2 – 0.4%), K (0.8 – 1.4%) and P (0.1 – 0.3%)) according to Fregoni (2005).

According to the ANOVAs there were no significant effects (p>0.05) of liming on any of the leaf nutrient contents in both blades and petioles. Interestingly, there was a significant effect (p<0.05) of year of sampling on Mg, K and P in blades, as well as on Ca, Mg and P in petioles, whereas the interaction between liming and sampling year was not significant (p>0.05) on any of the leaf nutrient contents. Correlation coefficients (not shown) between magnesium, calcium, potassium and phosphorus in leaf and in soil were very low.

There were no significant effects (p>0.05) of liming on berry weight, grape yield, total soluble solids, pH and extractable anthocyanins. On the contrary, there were significant effects of liming on total acidity just in 2011, and seed maturity just in 2009 and 2010 (Table 5). Interestingly, there was a significant effect (p<0.05) of year of sampling on berry weight, total acidity, extractable anthocyanins and seed maturity (Table S2 [supplementary]). It is also interesting the absence of significant interactions between liming and sampling year in all parameters except seed maturity. Therefore, means along with standard errors are shown in Fig. 3 for grape yield, total soluble solids and pH by liming material, while the same statistics are shown for berry weight, total acidity, extractable anthocyanins and seed maturity by both liming material and year.

According to the ANOVA, liming had no significant effects (p>0.05) on any of the nutrient contents. Interestingly, there was a significant effect (p<0.05) of year of sampling on K and P contents in grapes. Additionally, the effect of the sampling year on the grape nutrient contents did not change with liming. This was revealed by the non-significant interaction (p>0.05) between liming and sampling year. Thus, mean and standard errors of Ca and Mg content in grapes by liming material, and the same statistics of K and P contents in grapes by liming material and year are shown in Fig. 4.