2.1. Environmental Conditions in Valtellina
This study evaluates data collected by the technical assistance centre of Fondazione Fojanini in 15 commercial vineyards (V1 to V15), representative of the viticultural belt along the Rhaetian mountainside of the valley (
Figure 1).
Figure 2 shows the last 41 years of data for the weather station of Sondrio from the network of ARPA Lombardia (the Regional Environmental Agency of Lombardy), in the middle of the wine belt of Valtellina, close to vineyards V8–V11.
The yearly average temperature course follows the typical behaviour of Western Europe [
44]. It experienced a gradual rise, starting in the late 1980s and reaching a stable level in the early 2000s, marking the onset of the current warm phase. To account for this change, the 20-year climate normal of 1981–2000 and 2001–2020 have been chosen as references instead of the 30-year standard suggested by the WMO (2022).
The increase in temperatures had obvious effects on the occurrence of heat waves (
Figure 2), which moved from an average yearly value of 13 d y
−1 during the past climate phase to 21 d y
−1 in the current one.
Regarding precipitation, the two phases show a yearly average value that increased from 876 mm in the 1981–2000 period to a 2001–2020 value of 1013 mm (+15%), with increased interannual variability (
Figure 2).
Geographical and climatic features of the vineyards.
Based on DTM and meteorological data, the following geographical and climatic features (
Table 1) were derived for the 15 vineyards:
Elevation [m]
Aspect [°]
PPAR—Potential Photosynthetically Active Radiation [MJ]
ATY—2001–2020 average of yearly mean temperature Tmean [°C]
PrecY—2001–2020 accumulation of yearly precipitation [mm]
HWY—2001–2020 sum of yearly days with maximum daily temperature Tmax above 32 °C [n days]
Most of the vineyards’ aspect ranges from south–east to south–west with PPAR values in the 3300 to 3500 MJ/m2 range. The only exception is V14, East-exposed, with the lowest PPAR of 2794 MJ/m2. Yearly average temperature (ATY) ranges between 12.1 °C (V7) and 14.2 °C (V5 and V2), while yearly precipitation (PrecY) is between 1022 mm (V6) and 1255 mm (V13). The most noticeable difference among the vineyards is provided by the yearly number of heat waves (HWY), from 15 in V14 to 38 in V5.
2.2. The Ripening Progress
The first analysis performed focuses on the ripening indices (
Table 2).
Considering DOY15 (the day of achievement of the grape sugar content of 15 °Brix, assumed as an early ripening stage), the average value is 237 (25 August) with a standard deviation of 8.9 days. The earliest value is 221 (9 August) and the latest is 254 (11 September).
In the case of DOY20 (the day of achievement of the grape sugar content of 20 °Brix, representing technical maturity), the average value is 259 (16 September) with a standard deviation of 9.6 days. The earliest value was 235 (22 August), and the latest was 287 (10 October).
Ripening Length (RL—as the difference between DOY20 and DOY15) is on average equal to 22 days, with a standard deviation of 4.7 days. The shortest period is 14 days, and the longest is 33.
Total acidity at technical maturity (TA) is, on average, equal to 11.01 g L−1 with a standard deviation of 1.66 g L−1, with the lowest value of 7.95 g L−1 and the highest value of 14.33 g L−1.
The variability of each vineyard for 21 years and each year for 15 vineyards is analysed using box plots, as shown in
Figure 3 and
Figure 4.
Regarding the vineyard variability (
Figure 3), the vineyards show consistent behaviour across all four indices examined. There is a strong correlation between the average values of DOY15 and DOY20 (R
2 = 0.85), as well as between average DOY15 and average TA (R
2 = 0.52). However, the relationship between DOY20 and TA displays less consistency.
The vineyard variability of DOY15 is quite similar in all the vineyards, with V4 and V10 being the ones that exhibit a higher variation. Similarly, in the case of DOY20, the highest variability is shown by V2, V4, V10, V13, and V15. The ripening length is particularly stable among the vineyards, except for V10 and V14, which exhibit a longer duration.
In the case of V10, the vineyard shows average late DOY15 and DOY20. V10 is the second highest for elevation among the studied vineyards (613 m a.s.l.), and the late and longer ripening can be explained by a lower thermal regime (12.7 °C of yearly average temperature).
In the case of V14, the vineyard position on the South–Eastern side of the Valley determines a lower thermal regime despite the lower altitude (474 m a.s.l. and 12.9 °C of average yearly temperature).
V15, the highest in altitude at 623 m, is characterised by more favourable thermal conditions, with a yearly average temperature of 13.1 °C. V15 shows a late average DOY15, but this is not translated into a late DOY20 or a longer RL.
In the case of seasonal variability (
Figure 4), the four indices are again consistent, with significant correlations between DOY15 and DOY20, DOY15 and TA, and DOY2 and TA with R
2 values of 0.88, 0.48, and 0.37, respectively. For DOY15 (the average day of the 15 vineyards), the earliest seasons are 2003, 2007, 2009, 2011, 2017, and 2018, with days of occurrence between 12 and 18 August. The latest seasons are 2001, 2002, 2008, 2013, and 2016, all dropping in the first week of September.
For DOY20, the earliest average dates are in 2003, 2007, 2009, 2011, 2017, and 2018 and are comprised of between the end of August and the first 10 days of September. The latest seasons are 2001, 2008, 2013, 2014, and 2016, all in the last week of September.
In the case of total acidity, values below 10 g L−1 were obtained in 2003, 2011, 2015, 2017, and 2020, most of which were years of early ripening (DOY15 and DOY20). Values of TA above 12 g L−1 were obtained in 2001, 2002, 2008, and 2014, all seasons with late DOY20.
Results highlighted that viticultural activity in Valtellina is strongly affected by the interannual variability of environmental conditions [
45,
46,
47]. Additionally, this should be considered to define effective agronomical management of the vineyard [
48,
49].
To assess a possible trend in the timing of ripening during the period of study, linear regressions between year and DOY15, DOY20, RL, and TA were set up (
Table 3).
Even with very low values of R
2, significant advances in DOY20 (−0.345 d y
−1), decreases in TA (−0.037 g L
−1 y
−1), and shortenings in RL (−0.107 d y
−1) were found when considering the whole ensemble of data from the 15 vineyards. This trend agrees with similar results detected in other wine regions in recent years [
50]. However, analysing the data separately for each vineyard, no significant regression between year and DOY15 or between year and TA was found. A significant advance in DOY20 was found only in 3 vineyards (V6 R
2 = 0.148, −0.571 d y
−1; V10 R
2 = 0.187, −0.675 d y
−1; and V13 R
2 = 0.187, −0.686 d y
−1), while a significant shortening of ripening length was found only in 2 (V6 R
2 = 0.277, −0.464 d y
−1; V16 R
2 = 0.316, −0.401 d y
−1).
2.3. Ripening and Environmental Conditions
2.3.1. Influence of Geographical and Climatic Features
For each vineyard, a linear regression was set up between the values of DOY20, DOY15, RL, and TA averaged for the 2001–2021 period and the geographical and climatic features (Elevation, Aspect, PPAR, PrecY, ATY, and HWY).
Figure 5 shows only the significant regressions.
Elevation shows a high significant (p-value ≤ 0,001) positive regression with DOY15 (R2 = 0.717; +0.028 d m−1) and TA (R2 = 0.608; +0.0073 g L−1 m−1) and at a lower level (0.01 ≥ p value > 0.001) with DOY20 (R2 = 0.554; +0.0264 d m−1). In other words, as elevation increases, the ripening process tends to occur later in the season. Additionally, at the point of technical maturity, the TA of the grapes tends to be higher.
The yearly average air temperature (ATY) has a highly significant (p-value ≤ 0,001) negative regression with DOY15 (R2 = 0.731; −5.304 d °C−1) and DOY20 (R2 = 0.709; −5.6081 d °C−1) and at a lower level (0.01 ≥ p-value > 0,001) with TA (R2 = 0.456; −1.1891 g L−1 °C−1). In this case, with increasing temperatures, ripening occurs earlier in the season while TA at technical maturity is lower.
Air temperature is strongly related to altitude and, to a lesser extent, to PPAR and exposure, resulting from the surface energy balance [
11]. In light of this, a strict relation between altitude and ripening was generally found [
7,
8,
9], and in Valtellina, Failla et al. [
43] reported that elevation mainly affected technological maturity.
The number of heatwaves HWY showed a significant negative relationship with DOY20 (R2 = 0.3415; −0.3415 d d−1). This means that hot conditions determine early technical maturity.
RL does not show any significant regression with the geographical and climatic indices. A possible reason is that inter-year variability is prevalent compared to the geographical and climatic components.
Furthermore, no significant regression was found between PPAR, aspect, YTP and DOY15, DOY20, and TA. This is probably due to the homogeneity of Aspect and PPAR along most of the vineyards. While the role of PPAR and Aspect are partially covered by the ATY index, the lack of correlation with PrecY can be attributed to the fact that in the Valtellina region, where annual precipitation is usually sufficient for the growth requirements of grapevines, the timing of grapevine events is more influenced by the patterns and dynamics of precipitation events [
28,
51].
2.3.2. Ripening and Environmental Limiting Factors
The four ripening indices were put in relation to environmental indicators through linear regression, considering the whole body of data and the single vineyards.
The environmental indices adopted are:
HHH15—accumulation of high heat hours (representing thermal excess) from the beginning of the season to the day of achievement of 15 °Brix.
HHH20—accumulation of high heat hours from the beginning of the season to the day of achievement of 20 °Brix.
ΔHHH—accumulation of high heat hours from the day of achievement of 15 °Brix to the day of achievement of 20 °Brix.
HW15—number of heat waves from the beginning of the season to the day of achievement of 15 °Brix.
HW20—number of heat waves from the beginning of the season to the day of achievement of 20 °Brix.
ΔHW—number of heat waves from the day of achievement of 15 °Brix to the day of achievement of 20 °Brix.
Prec15—total precipitation from the beginning of the season up to the day of achievement of 15 °Brix.
Prec20—total precipitation from the beginning of the season up to the day of achievement of 20 °Brix.
ΔPrec—total precipitation from the day of achievement of 15 °Brix to the day of achievement of 20 °Brix.
When considering all the vineyards collectively, it is observed that DOY15 shows a significant positive regression with Prec15 (R2 = 0.237, +0.021 d mm−1), as well as a significant negative regression with HHH15 (−0.019 d °C−1) and HW15 (−0.482 d d−1). However, the R2 values for the latter two variables, HHH15 and HW15, are relatively low (0.078 and 0.089 respectively), with the significance of the regression analysis being enhanced by the large dataset used for analysis.
At the single vineyard level, it is observed that Prec15 shows significant regression in 11 vineyards (R2 from 0.217 to 0.516). On the other hand, HHH15 shows significant regression in only one vineyard (V11 with R2 = 0.216), and HW15 shows significant regression in five vineyards (R2 from 0.212 to 0.260).
It is interesting to note that with the exclusion of V11, the significant negative effect of HW15 is related to the vineyards at the highest altitudes (V6, V7, V10, and V15), where thermal resources are lower. The results indicate that, in general, DOY15 is more influenced by Prec15 (cumulated precipitation) than by thermal excess. Higher levels of cumulated precipitation are associated with an earlier start of maturation (DOY15). This is probably caused by a higher plant vigour related to a better plant water status [
52].
DOY20, considering the whole dataset, shows significant negative regression with ΔHHH (R2 = 0.36, −0.16 d °C−1) and ΔHW (R2 = 0.25, −1.26 d d−1) and positive regression with Prec20 (R2 = 0.21, +0.02 d mm−1). Significant negative regressions with HHH20 (−0.03 d °C−1) and HW20 (−0.22 d d−1) were found as well, but characterised by lower R2 (0.15 and 0.13, respectively). Finally, a positive regression, but at a lower significant level (0.05 > p value > 0.01), was found with Δprec (R2 = 0.02, +0.03 d mm−1).
At the vineyard level, ΔHHH, ΔHW, and Prec20 have significant regression in most of the vineyards (14, 9, and 11, respectively). The result shows that DOY20 is generally more affected by thermal excesses that occur during ripening, causing an advance in the achievement of 20° brix. In the case of precipitation, higher regressions are obtained when considering the total amount of cumulated precipitation (Prec20) instead of the portion cumulated during ripening (Δprec). This leads to the idea that the timing of ripening is more related to the plant’s water status throughout the whole growing season.
Considering the whole data set, TA shows significant negative regression with ΔHW (R2 = 0.51, −0.24 g L−1 d−1), ΔHHH (R2 = 0.34, −003 g L−1 °C−1), HHH20 (R2 = 0.23. −0.01 g L−1 °C−1), and HW (R2 = 0.21. −0.05 d d−1). A positive and significant regression was found between Prec20 and the parameter under consideration (+0.01 g L−1 mm−1). However, the R2 value for this regression was relatively low, at 0.10.
Analysing the vineyard-specific results, TA is more linked to thermal excesses during ripening. ΔHW and ΔHHH have the highest R
2, and their regression is significant in most vineyards (10 and 15, respectively). Indeed, high temperatures during ripening could increase the rate of acid degradation in berries [
22]. A positive influence of Prec20 is also observed, even if weak. This is probably linked to acid accumulation at the beginning of the ripening process since the regression between TA and Δprec shows lower significance (0.05 ≤
p value > 0.01) and R
2 (0.02).
Considering the whole dataset, RL shows a significant positive regression with Δprec (R
2 = 0.18. 0.05 d mm
−1) and a less significant but still present regression with ΔHHH (R
2 = 0.03. 0.03 day °C
−1) and ΔHW (R
2 = 0.03. 0.21 d d
−1). Focusing on the singular vineyards, ΔHHH, and ΔHW shows a significant positive regression in very few vineyards (3 and 1, respectively) with low R
2 value. In six vineyards, Δprec has a significant positive regression. Thermal excess during ripening may, in some cases, have a negative influence on the physiology of maturation, as indicated by the modest correlations found with RL. However, the relationships observed in this case are not very strong. Indeed, Δprec has the largest influence on RL. This could be related to the phenomenon of sugar dilution in berries related to precipitation that occurs during ripening time [
28].
In conclusion, the increase in Prec induces a higher TA and a delay in reaching DOY15 and DOY20, while the increase in Δprec is related only to higher values of RL. Conversely, thermal excesses that occur during ripening (ΔHHH and ΔHW) time have more influence on DOY20 and TA than the total amount (HHH and HW). Indeed, ripening period temperatures are important for quality wine production [
53]. Finally, thermal excesses show a lower effect on DOY15 and RL than precipitation.
2.3.3. Modelling the Timing of Ripening
The predictive models for the achievement of 15 and 20 °Brix are based on the accumulation of NHH (normal heat hours, representing hourly thermal resources useful for grapevine development) and GDD (growing degree days with a base temperature of 10 °C) on the day when the two sugar levels are achieved. The variables used in the models are ∑GDD15 and ∑NHH15 for the day of achieving 15 °Brix DOY15, and ∑GDD20 and ∑NHH20 for the day of achieving 20 °Brix DOY20, respectively.
The average values of accumulations over the 21 years for the 15 vineyards, were adopted as thresholds for the achievement of the two sugar contents. The performance of the models was tested by comparing the day of occurrence of the chosen thresholds with the observed day of occurrence of the sugar level. Statistical indices MAE, EF, and R
2 [
54] were used to assess the models’ performance. The obtained results are as follows:
∑GDD15 = 1456
∑NHH20 = 1763
∑GDD20 = 1683
∑NHH20 = 2037
To test the models’ performances in the calibration process, for each vineyard and each year, the day of occurrence of the thresholds was compared with the observed day of occurrence of sugar levels of 15 °Brix and 20 °Brix. The statistical indices MAE, EF, and R
2 are reported in
Table 8.
Regarding the estimation of DOY15 (
Table 8), the NHH model shows better performances in terms of MAE, with an overall lower error of 8.03 days compared to the GDD model’s 10.17 days. Moreover, the NHH model provides better estimation values in 13 out of the 15 vineyards. However, the error is lower than one week only in five vineyards. Regarding EF, the NHH model performs better than GDD, but EF values are positive only for V1 and V15. This means that only for those two vineyards does the model provide a better estimation of DOY15 compared to the average of the observed DOY15 in the specific vineyard. In the case of R
2, the GDD model shows better results in terms of significant regressions, but R
2 values are always quite low.
The simulation of DOY20 showed worse but similar results. The global MAE increased to 10.40 and 14.56 for NHH and GDD, respectively. No vineyard shows MAE values below one week, and EF is positive only for vineyard V1.
However, analysing the averaged data for each vineyard (
Figure 6), it is worth noting that the regressions are robust, with R
2 values of 0.55 for GDD and 0.69 for NHH in the simulation of DOY 15. Similarly, for the estimation of DOY20, the regressions yield R2 values of 0.67 for GDD and 0.76 for NHH.
To provide a long-term view of the ripening timing in Valtellina, the NHH models for DOY15 and DOY20 have been applied to the Sondrio time series of temperatures for the period 1981–2021 (
Figure 7). Ten-year averages of the simulated DOY15 and DOY20 are also provided. Data from Vineyard V6 are shown for the period covered by monitoring activities to provide a qualitative idea of the model simulation.
The simulation of DOY15 is more consistent than the simulation of DOY20 and shows weaker results for the first three years (2001, 2002, and 2003).
Regarding the simulation of DOY 20, the model performances are particularly weak when ripening is late, as in the cases of 2006, 2008, and 2010. This could be explained by the fact that Nebbiolo ripens quite late in the season and the thermal-resources-based models are not able to accumulate enough resources at that time, especially during cold days. This confirms the idea that ripening models should not exclusively rely on temperature as the driving variable of the process.
In more general terms, and focusing on the more reliable simulation of DOY15, the 10-year averages of the last four decades are 254 (1981–1990), 229 (1991–2000), 238 (2001–2010), and 236 (2011–2020), corresponding to 11 September, 17 August, 26 August, and 24 August. This confirms what was observed in the analysis of temperature: a warming phase in the 1990s that stabilised in the last 20 years.