The Variability and Trend of Harvest Dates of Table and Pisco Grapes in Northern Chile Are Independently Influenced by Bioclimatic Indices ^†

Abstract

(1) Background: The variability and trend in harvest dates of table and Pisco grapes have been scarcely studied. This can be closely influenced by bioclimatic indices since they account for the interactions between climatic factors and vine phenology. Understanding the environmental factors influencing harvest timing has become increasingly critical to perform specific viticultural practices. (2) Methods: The aim of this research was to evaluate the influence of bioclimatic indices on variability and trend of harvest date from the 2002–2003 to 2017–2018 seasons in Flame Seedless, Thompson Seedless, Muscat of Alexandria, and Moscatel Rosada growing in Northern Chile. (3) Results: The harvest date of Flame Seedless advanced significantly with an increasing Growing Season Temperature (GST) (from 1 October to 31 December), while Thompson Seedless showed a significant advancement in harvest date with rising the Maximum Springtime Temperature Summation SON_max (from 1 September to 30 November) values. Similarly, the harvest date of Muscat of Alexandria was significantly earlier with higher Heliothermal Index (HI) (from 1 July to 31 January and from 1 August to 30 April) values, whereas Moscatel Rosada exhibited a significant advancement in harvest date as the GST (from 1 July to 31 December and from 1 July to 31 January) increased. The trend in the harvest date of Thompson Seedless was statistically significant, reaching a coefficient of determination of 0.42. (4) Conclusions: Understanding the influence of bioclimatic indices on harvest date in long-term periods is critical in the context of climatic variability since producers can make more informed decisions to optimize grape quality and maintain sustainability in production systems.

1. Introduction

Chile has long been recognized as a leading global producer and exporter of table grapes and Pisco due to its diversity of climates and advanced agricultural practices that have allowed it to position itself as a key player in international markets [1,2]. Chile’s geography allows for the cultivation of a wide range of grape varieties from the arid north to the central valleys, including important table grapes for market, such as Thompson Seedless and Flame Seedless [1,3]. Since agriculture worldwide faces the impacts of climate variability and changing market demands, understanding the environmental and managerial factors influencing harvest timing has become increasingly critical [4].

Grapevines (Vitis vinifera L.) are highly receptive to environmental conditions, in which temperature and radiation are the most critical factors influencing the date of phenological stages such as budburst, flowering, veraison, maturity, and harvest [5,6]. The seasonal growth of grapevines is strongly related to weather conditions [7]. The grapevine varieties have different thermal requirements to achieve full berry ripeness [8]. The technological maturity of grapes, which is essential for achieving the desired wine style, is determined early in the season by the enologist based on key parameters such as sugar accumulation, acidity levels, and phenolic composition [8]. Grape composition at harvest, and subsequent wine quality, has been linked to growing season temperatures [4,6,9]. The high thermal accumulation in a viticultural location during the growing season provides grapes with high sugars, proline, and anthocyanins contents, leading to the production of alcoholic wines with ripe fruit aroma, low acidity, and a marked astringency [10,11,12]. The low thermal accumulation mostly the month before harvest resulted in grapes with a high content of amino acids, organic acids, flavanols, and acidity and lower sugar accumulation, leading to the production of balanced wines with a characterized fruity aroma [9]. The calculation of temperature-based indices has become important for understanding and optimizing vineyard performance, especially in the context of climate variability [13,14,15,16,17,18].

Many bioclimatic indices have been proposed since the last century to differentiate, describe, and delimit viticultural regions [16,19]. The growing season temperature (GST) was proposed by Jones [16], which calculated the average temperature within the seven-month growing season and classified thresholds suitable for different grapevine varieties. The growing degree days (GDD) index is one of the most effective, simple, and commonly used agroclimatic indices [17]. It is closely linked to vine growth cycles, as it incorporates air temperature, a widely available and highly reliable atmospheric parameter. The GDD index is based on the principle that vine development occurs when air temperature surpasses a specific base threshold (usually 10 °C) over a given period of time. The Heliothermal Index (HI), also known as the Huglin Index, accounts for daily temperatures during the grapevine growing season and includes an adjustment factor for day length as a function of latitude [20]. The HI index shows a moderate association with the potential sugar concentration in grapes [21]. Thermal night conditions related to grape maturity can be assessed using the Cool Night Index (CI), which considers the average minimum temperature during the late maturation period [22]. This index aims to enhance the evaluation of a site’s quality potential by accounting for the accumulation of secondary metabolites, such as phenolics, amino acids, and volatile compounds [9,22]. Springtime heat accumulation has been correlated with the timing of maturity in Australian vineyards [4]. The mean and maximum springtime temperature bioclimatic indices provide additional information regarding the climatic characteristics of a site [4].

Traditional bioclimatic indices such as the GST, GDD, and HI are widely used in viticulture, particularly for dry wine production. However, these indices are typically calculated based on fixed seasonal or calendar-based periods, which may not accurately reflect the phenological stages of different grapevine varieties, especially those cultivated for other grape-derived products, such as table grapes, raisins, and fortified or sparkling wines. The development cycles and optimal climatic conditions for these products can differ significantly from those of wine grapes, highlighting the need for more flexible and variety-specific bioclimatic assessments.

The variability and trend in harvest dates of Chilean table and Pisco grapes over consecutive seasons have been scarcely explored. Morales-Castilla et al. [23] reported that long-term warming trends have compressed the phenological calendar in grapevines, leading to earlier harvests and challenges in maintaining fruit quality. Different cultivars respond differently to temperature accumulation, and seasonal climate variations and phenological variability among grapevine varieties suggest that a one-size-fits-all approach to bioclimatic index calculation may not be appropriate. The variability and trend in harvest date could be also affected by interannual climatic anomalies, such as El Niño–Southern Oscillation (ENSO), which can significantly impact regional temperature and precipitation patterns [24]. Variety-specific responses can impact season-to-season harvest dates since grapevines can exhibit different sensitivity to bioclimatic factors [10,25]. Thus, standard calendar-based indices do not account for climate anomalies, which may result in misleading assessments of grapevine suitability.

This study focuses on specific bioclimatic indices such as the GDD, HI, GST, CI, and spring temperature summation. These indices were selected because they capture key climate–vine interactions that are strongly correlated to technological parameters and the quality of grapes and wines [21]. The research goal was (i) to evaluate the influence of bioclimatic indices on the variability and trend of harvest dates of Chilean table and Pisco grapes growing in Northern Chile and (ii) to refine bioclimatic indices by aligning them with actual biological growth cycles for table and Pisco production rather than calendar seasonal definitions.

2. Materials and Methods

2.1. Study Sites, Plant Material, and Experimental Design

This study was conducted at the Vicuña Experimental Center, part of the Instituto de Investigaciones Agropecuarias (INIA), located in Vicuña, Coquimbo Region, Chile (30°02′ S, 70°41′ W, 630 m.a.s.l.). This area is within the Elqui Valley, characterized by a semi-arid subtropical Mediterranean climate. The soil is classified as alluvial clayey (Entisol order), commonly found in viticultural regions with limited water retention capacity. Four Vitis vinifera L. grape varieties were analyzed, including two table grape varieties (Flame Seedless and Thompson Seedless) and two Pisco grape varieties: Muscat of Alexandria and Moscatel Rosada. These varieties were selected due to their economic importance in viticulture in Northern Chile.

The vineyards were established between 1995 and 1998 on their own rootstocks under a pergola training system with a planting distance of 2.0 × 3.5 m. The vineyards were irrigated by a drip irrigation system, ensuring consistent water supply throughout the season. Harvest date records were collected from the 2002–2003 to 2017–2018 growing seasons, covering up to 16 years of data collection.

Table 1. Field characteristics, data available, and differences between the production system.

		Table Grape Production		Pisco Production
		Flame Seedless	Thompson Seedless	Muscat of Alexandria	Moscatel Rosada
General characteristics	Vineyard surface (ha)	1.2	1.0	1.3	1.3
	Year of plantation	1998	1998	1995	1995
	Trellis system	Pergola	Pergola	Pergola	Pergola
	Spacing distance	2.0 × 3.5	2.0 × 3.5	2.0 × 3.5	2.0 × 3.5
Data of harvest date	Seasons	2002 to 2018	2002–2003 to 2012–2013	2002–2003 to 2017–2018	2002–2003 to 2017–2018
	Number of seasons	16	11	16	16
Differences between production system	Yield adjustment	Yes	Yes	No	No
	Use of growth regulators	Yes	Yes	No	No
Phenology	Average growth cycle duration (days) (budburst to harvest)	140	150	210	200

shows the characteristics of the production units from which the information was obtained and the differences between the 2 production systems considered. Briefly, in table grape production, crop load adjustment is carried out to obtain fruit that meets fresh market standards. This involves the removal of entire clusters, typically adjusting to 30 to 36 clusters per vine. Additionally, a berry thinning process was performed on the retained clusters, leaving approximately 90 to 110 berries per cluster. Complementary to these practices, plant growth regulators, such as gibberellic acid and cytokinins, among other plant hormones, are applied to promote berry enlargement. None of these practices were conducted in the production of grapes destined for Pisco production.

2.2. Harvest Criteria for Table and Pisco Grapes

Harvest timing was determined based on minimum industry standards for commercialization. Harvest was defined when berries reached 17.0 °Brix for table grapes (Flame Seedless and Thompson Seedless), meeting the market requirements for commercial quality. Harvest was performed at 22.0 °Brix in Pisco grapes (Muscat of Alexandria and Moscatel Rosada), the minimum value of soluble solids required sugar content for Pisco production.

2.3. Climate Data Collection and Bioclimatic Indices Calculation

Climate data (daily maximum and minimum temperatures) were recorded using an automated meteorological station located 300 m from the vineyards. Data gaps were filled using linear regression models based on nearby weather stations when necessary, ensuring at least 80% complete records across all seasons [13]. To assess the influence of climate on harvest variability and trends, seven bioclimatic indices were calculated, commonly used in viticulture. Each index was calculated for multiple periods, including standard and adjusted periods to better reflect the specific phenological cycle of each grape type (

Table 2. Bioclimatic indices and their calculation periods.

Bioclimatic Index	Equation	Standard Period 1	Adjusted Periods (for Early Ripening Varieties in Northern Chile)
Huglin Index (HI)	$HI={\displaystyle \sum _{d=1}^n}max\left({\displaystyle \frac{(TMean-10+TMax-10)}{2}}\right)k$	1 October–31 March	1 July–31 December; 1 July–31 January; 1 August–30 April; 1 September–30 April; 1 September–31 March
Growing Degree Days (GDD)	$GDD={\displaystyle \sum _{d=1}^n}max{\displaystyle \frac{(TMax+TMin)}{2}}-10$	1 October–30 April	1 July–31 December; 1 July–31 January; 1 August–30 April; 1 September–30 April; 1 September–31 March
Growing Season Temperature (GST)	$GST={\displaystyle \frac{{\sum }_{d=1}^n\frac{TMax+TMin}{2}}{n}}$	1 October–30 April	1 July–31 December; 1 July–31 January; 1 August–30 April; 1 September–30 April; 1 September–31 March
Springtime Temperature Summation (SONmean, SONmax)	$SONMean={\displaystyle \sum _{d=1}^n}Tmean$ $SONMax={\displaystyle \sum _{d=1}^n}Tmax$	1 September–30 November	1 October–31 December
Cool Night Index (CI)	$CI={\displaystyle \frac{{\sum }_{d=1}^{n}TMin}{n}}$	March	December; January; February; April
Mean January Temperature (MJT)	$MJT={\displaystyle \sum _{d=1}^n}Tmean$	January	No adjustments needed

¹ Standard periods for the southern hemisphere according to the exposure by Verdugo-Vásquez et al. [13].

). In cases where missing data were detected, typically one or two days, these were completed using the methodology described by Verdugo-Vásquez et al. [13]. This approach ensured that all bioclimatic indices were calculated using complete datasets, in accordance with their respective definitions. These adjustments were made considering the earlier phenological stages of grapevine development in the study area compared to traditional wine-growing regions [13]. In this fashion, for table grapes, the inclusion of July and August accounts for the earlier onset of vegetative growth in Northern Chile, while the end of the calculation period in December or January corresponds to the typical harvest window. In contrast, for Pisco grape varieties, budburst usually begins in September, with harvest occurring from March to April. Based on this, the indices were computed using periods that better reflect the actual growth cycle of each production system.

2.4. Statistical Analysis

Descriptive statistics, including the mean, standard deviation (SD), and coefficient of variation (CV), were calculated for each bioclimatic index. To analyze long-term trends in harvest dates, linear regression models were applied, with the coefficient of determination (R²) used to assess the goodness of fit. Additionally, Pearson’s and Spearman’s correlation coefficients were computed to examine the relationships between harvest dates and bioclimatic indices as previously reported [13]. All statistical analyses were performed using XLStat software version 2020.3.1 (Addinsoft, Paris, France).

3. Results

3.1. Descriptive Analysis of the Influence of Bioclimatic Indices on Harvest Date

The descriptive analysis and linear trends of the effects of bioclimatic indices on harvest date are shown in

Table 3. Descriptive analysis and linear trend for the bioclimatic indices (Elqui Valley, Chile).

		Descriptive Statistics			Linear Trend
Bioclimatic Indices	Period	Mean	SD	CV	Trend yr−1	R2
Huglin Index (HI: Heat Units)	1 October–31 March	2434.9	84.3	3.5	6.39	0.13
	1 July–31 December	1841.1	137.8	7.5	0.75	0.0007
	1 July–31 January	2308.9	146.4	6.3	3.07	0.01
	1 August–30 April	3258.3	143.4	4.4	7.99	0.07
	1 September–30 April	3025.7	114.4	3.8	7.18	0.09
	1 September–31 March	2708.1	100.2	3.7	5.78	0.08
Cool Night Index (CI: °C)	December	9.8	0.8	8.2	0.006	0.001
	January	11.8	0.8	7.1	0.02	0.02
	February	11.3	0.8	13.4	0.035	0.03
	March	9.4	0.9	9.9	−0.0007	0.00001
	April	6.5	1.1	17.6	0.07	0.09
MJT 1 (°C)	January	20.5	0.7	3.6	0.04	0.06
SON 2 Mean (Heat Units)	1 September–30 November	1435.9	72.6	5.1	−5.65	0.14
	1 October–31 December	1601.8	58.1	3.6	−3.08	0.06
SON 2 Max (Heat Units)	1 September–30 November	2339.7	85.3	3.6	2.20	0.02
	1 October–31 December	2504.4	79.1	3.2	4.36	0.07
Growing Degree Days (GDD: Heat Units)	1 October–30 April	1702.5	95.7	5.6	0.60	0.0009
	1 July–31 December	960.0	129.2	13.5	−9.36	0.12
	1 July–31 January	1284.1	137.3	10.7	−8.28	0.08
	1 August–30 April	1916.9	140.5	7.3	−3.38	0.01
	1 September–30 April	1828.2	115.3	6.3	−1.49	0.004
	1 September–31 March	1671.9	101.5	6.1	−2.42	0.01
Growing Season Temperature (GST: °C)	1 October–30 April	18.0	0.5	2.5	0.002	0.0004
	1 July–31 December	15.1	0.8	5.4	−0.05	0.09
	1 July–31 January	15.8	0.7	4.7	−0.04	0.07
	1 August–30 April	17.0	0.5	3.2	−0.01	0.01
	1 September–30 April	17.5	0.5	2.7	−0.006	0.004
	1 September–31 March	17.9	0.5	2.7	−0.011	0.01

¹ Mean January Temperature. ² Spring Temperature Summation. SD: standard deviation. CV: coefficient of variation. R²: coefficient of determination.

. The Huglin Index varied from 1841 to 3258 heat units (from 1 July to 31 December and from 1 August to 30 April, respectively) with a coefficient of variation (CV) that ranged from 3.5 to 7.5% (from 1 October to 31 March and 1 July to 31 December, respectively). The linear trend of the Huglin Index varied from 0.75 to 7.99 heat units per year (from 1 October to 31 March and 1 August to 30 April, respectively) with a coefficient of determination (R²) lower than 0.13. This value shows a low linear adjustment along the seasons. The Cool Night Index (CI) varied from 6.5 to 11.8 °C (from April to January, respectively) with a CV that ranged from 7.1 to 17.6% (from January to April, respectively). The linear trend of CI was lower than 0.07 °C per year (in April) with an R² lower than 0.09 (in April). The Mean January Temperature (MJT) reached a mean of 20.5 °C with a CV of 3.6%. The linear trend of the MJT was 0.04 °C per year with an R² of 0.06. The Daily Mean Springtime Temperature (SON_mean) ranged from 1436 to 1602 heat units with a CV that varied from 3.6 to 5.1%. The linear trend of SON_mean varied from −5.65 to −3.08 heat units per year with an R² that varied from 0.06 to 0.14. The daily maximum springtime temperature (SON_max) ranged from 2340 to 2504 heat units with a CV that varied from 3.2 to 3.6%. The linear trend of SON_max varied from 2.2 to 4.4 heat units per year with an R² that varied from 0.02 to 0.07. The accumulation of growing degree days (GDD) ranged from 960 to 1917 heat units (from 1 July to 31 December and from 1 August to 30 April, respectively) with a CV that varied from 5.6 to 13.5% (from 1 October to 30 April and from 1 July to 31 December, respectively). The linear trend of GDD varied from −9.36 to 0.60 (from 1 July to 31 December and 1 October to 30 April, respectively) heat units per year with an R² that reached a maximum value of 0.12. The growing season temperature (GST) ranged from 15.1 to 18.0 °C (from 1 July to 31 December and from 1 October to 30 April, respectively) with a CV that varied from 2.5 to 5.4% (from 1 October to 30 April and from 1 July to 31 December, respectively). The linear trend of the GST varied from −0.006 to 0.002 (from 1 September to 30 April and 1 October to 30 April, respectively) °C per year with an R² that reached a maximum value of 0.09.

Figure 1 shows the boxplot of the harvest date in different table and Pisco grape varieties. The earliest harvest date was observed in Flame Seedless reaching the harvest at the −29 day of the year (DOY), followed by Thompson Seedless (at −4 DOY), Moscatel Rosada (at 46 DOY), and Muscat of Alexandria (at 75 DOY). The latest harvest date was registered at 124 DOY in Muscat of Alexandria, at 106 DOY in Moscatel Rosada, at 23 DOY in Thompson Seedless, and at −1 DOY in Flame Seedless. The median harvest date in Flame Seedless occurred at −14 DOY, in Thompson Seedless at 12 DOY, in Moscatel Rosada at 69 DOY, and in Muscat of Alexandria at 101 DOY.

Figure 2 shows the trends of harvest dates in different table and Pisco grape varieties during the period of the 2002–2003 and 2017–2018 seasons. The coefficient of determination (R²) of the harvest date in Flame Seedless and Moscatel Rosada was 0.09, whereas, in Muscat of Alexandria, it was 0.13, which was not statistically significant. The R² of the harvest date in Thomson Seedless was 0.46, showing a statistical significance (p-value: 0.05).

The Spearman correlation coefficient of the harvest dates between the varieties is shown in

Table 4. Spearman rank correlations between the harvest date of the studied table and Pisco grape varieties.

	Flame Seedless	Thompson Seedless	Muscat of Alexandria	Moscatel Rosada
Flame Seedless	-	0.81 *	0.35	0.79 **
Thompson Seedless		-	0.48	0.87 **
Muscat of Alexandria			-	0.51 *
Moscatel Rosada				-

* p-value at 0.05. ** p-value at 0.01.

. The Spearman correlation between the harvest date of Thompson Seedless and Flame Seedless was 0.81 (p-value: 0.05), indicating a strong positive correlation between the harvest dates of these two varieties. The Spearman correlation coefficient of the harvest dates between Moscatel Rosada and Flame Seedless, Thompson Seedless, and Muscat of Alexandria was 0.79 (p-value: 0.01), 0.87 (p-value: 0.01), and 0.51 (p-value: 0.05), respectively, indicating a strong positive correlation of the harvest date of Moscatel Rosada with the rest of the studied table and Pisco grape varieties. The harvest date of Muscat of Alexandria was not statistically related to the harvest date of Flame Seedless and Thompson Seedless.

3.2. Correlations Between Harvest Date and Bioclimatic Indices

The correlation coefficient (R²) and levels of significance for the effects of bioclimatic indices on the harvest date are shown in

Table 5. Coefficient of correlation (R²) and levels of significance of the influence of bioclimatic indices on the harvest date in different table and Pisco grape varieties.

Bioclimatic Indices	Period	Flame Seedless	Thompson Seedless	Muscat of Alexandria	Moscatel Rosada
Huglin Index (HI)	(1 October–31 March)	0.38 *	0.11	0.40 **	0.26 *
	1 July–31 December	0.67 **†	0.55 **	0.38 *	0.59 **†
	1 July–31 January	0.63 **	0.58 **†	0.42 **†	0.53 **
	1 August–30 April	0.55 **	0.24	0.42 **†	0.36 *
	1 September–30 April	0.43 **	0.15	0.35 *	0.32 *
	1 September–31 March	0.46 **	0.18	0.35 *	0.35 *
Cool Night Index (CI)	December	0.11	0.01	0.17	0.23
	January	0.23	0.02	0.35 *†	0.12
	February	0.03	0.02	0.001	0.018
	(March)	0.02	0.06	0.001	0.0002
	April	0.02	0.02	0.15	0.04
MJT 1	(January)	0.12	0.06	0.26 *	0.04
SON 2 Mean	(1 September–30 November)	0.60 **	0.44 *	0.09	0.64 **
	1 October–31 December	0.68 **†	0.46 *†	0.23	0.67 **†
SON 2 Max	(1 September–30 November)	0.47 **	0.68 **	0.24	0.44 **
	1 October–31 December	0.34 *	0.46 *	0.28 *†	0.32 *
Growing Degree Days (GDD)	(1 October–30 April)	0.45 **	0.21	0.31 *	0.33 *
	1 July–31 December	0.65 **	0.45 *	0.18	0.66 **†
	1 July–31 January	0.68 **†	0.47 *†	0.23	0.64 **
	1 August–30 April	0.58 **	0.25	0.26 *	0.44 **
	1 September–30 April	0.46 **	0.21	0.22	0.39 *
	1 September–31 March	0.51 **	0.21	0.2	0.43 **
Growing Season Temperature (GST)	(1 October–30 April)	0.48 **	0.27	0.33 *	0.36 *
	1 July–31 December	0.68 **	0.48 *	0.24	0.71 **†
	1 July–31 January	0.70 **†	0.50 *†	0.29 *†	0.71 **†
	1 August–30 April	0.61 **	0.28	0.28 *	0.46 **
	1 September–30 April	0.48 **	0.2	0.23	0.41 **
	1 September–31 March	0.53 **	0.23	0.2	0.44 **

¹ Mean January Temperature. ² Spring Temperature Summation. The “best” correlations are shown in gray, and the “well” correlations are in dark gray. * p-value at 0.05. ** p-value at 0.01. Standard periods for each index are indicated in parentheses. The † symbol highlights cases where the use of an adjusted period improved the model’s explanatory performance compared to the standard definition *.

. The harvest date for each grapevine variety was compared to the bioclimatic indices using Pearson’s correlation. The bioclimatic indices related to heat accumulation, such as Huglin’s Heliothermal Index (HI) (in 83%), growing degree days (GDD) (in 67%), Mean Spring Temperature Summation (SON_mean) (in 75%), and Maximum Spring Temperature Summation (SON_max) (in 88%), including the growing season temperature (GST) (in 71%) and Mean January Temperature (MJT) (in 25%), showed a significant statistical correlation with the harvest date (p-value < 0.005). These aforementioned bioclimatic indices mostly affected the harvest date of Flame Seedless (in 96%) and Moscatel Rosada (in 96%), followed by Muscat of Alexandria (in 57%) and Thompson Seedless (in 43%). The Cool Night Index (CI) (in 5%) was scarcely correlated to the harvest date in most of the studied table and Pisco grapes, except for those calculated in January for Muscat of Alexandria.

The CI had the highest number of “least” correlated results (in 25%) with harvest date, especially when calculated in March (in 75%) for the majority of the studied table and Pisco grapes, except for Thompson Seedless. The SON_mean reached the highest percentage of the “best” and “well” correlated results (in 25%) with the harvest date, followed by the GST (in 17%), HI (in 17%), SON_max (in 13%), and GDD (in 4%). Among the bioclimatic indices most correlated to the harvest date, Flame Seedless showed the “best” and “well” correlated results in 17% of the studied relations, followed by Moscatel Rosada (in 13% of the relations), Muscat of Alexandria (in 13% of the relations), and Thompson Seedless (in 9% of the relations).

The harvest date in Flame Seedless was “well” correlated to the SON_mean calculated from 1 October to 31 December (R²: 0.68, p-value: <0.01), the GDD calculated from 1 July to 31 January (R²: 0.68, p-value: <0.01), and the GST calculated from 1 July to 31 December (R²: 0.68, p-value: <0.01). In addition, the harvest date in Flame Seedless was “best” correlated to the GST calculated from 1 July to 31 January (R²: 0.70, p-value: <0.01). The harvest date in Thompson Seedless was “well” correlated to the HI calculated from 1 July to 31 January (R²: 0.58, p-value: <0.01) and “best” correlated to the SON_mean calculated from 1 September to 30 November (R²: 0.68, p-value: <0.01). The harvest date in Muscat of Alexandria was “well” correlated to the HI calculated from 1 October to 31 March (R²: 0.40, p-value: <0.01). In addition, the harvest date in Muscat of Alexandria was “best” correlated to the HI calculated from 1 July to 31 January (R²: 0.42, p-value: <0.01) and from 1 August to 30 April (R²: 0.42, p-value: <0.01). The harvest date in Moscatel Rosada was “well” correlated to the SON_mean calculated from 1 October to 31 December (R²: 0.67, p-value: <0.01). In addition, the harvest date in Moscatel Rosada was “best” correlated to the GST calculated from 1 July to 31 December (R²: 0.71, p-value: <0.01) and 1 July to 31 January (R²: 0.71, p-value: <0.01).

Linear regression of the harvest date of the GST calculated from 1 July to 31 January for Flame Seedless and SON_max calculated from 1 September to 30 November for Thompson Seedless are shown in Figure 3. The harvest date of Flame Seedless was significantly earlier as the GST increased, whereas the harvest date was significantly earlier when the SON_max increased in Thompson Seedless.

Linear regression of the harvest date of the HI from 1 August to 31 April and from 1 July to 31 January in Muscat of Alexandria and the GST from 1 July to 31 December and from 1 July to 31 January for Moscatel Rosada are shown in Figure 4. The harvest date of Muscat of Alexandria was significantly earlier as the HI increased, whereas the harvest date was significantly earlier when the GST increased in Moscatel Rosada.

4. Discussion

The results exposed in this trial confirmed the influence of bioclimatic indices on the harvest date in a period that corresponded to more than a decade. As was previously exposed, the harvest date of Flame Seedless and Moscatel Rosada advanced significantly with the increase in the growing season temperature (GST) in an independent way (Table 3). The GST calculates the average temperature within the seven months of the vine growing season and classifies thresholds suitable for the cultivation of different grapevine varieties [16]. This bioclimatic index is one of the most commonly used in broad-scale suitability analysis, probably due to its ease of calculation and minimal data requirements [17,18]. The GST provides a general indicator of the heat available for vines during the key ripening stages, and high values of this index indicate an increase in vine metabolic processes, leading to earlier phenological events. However, the GST does not account for day-to-day variability, extreme temperatures, or other factors, such as day length or rainfall, which could affect the harvest date of grapevine varieties. The harvest date of Muscat of Alexandria was significantly earlier when the Huglin Index (HI) increased (Table 3). A refinement of growing degree days (GDD) is the HI that incorporates the maximum and mean daily temperatures during the growing season and the day length at different latitudes [20]. The HI emphasizes heat accumulation during the grapevine’s metabolically active hours, and the reason for this is that it is closely related to the rate of sugar accumulation and acid degradation [20,26]. Thompson Seedless showed a significant advancement in the harvest date with the rising of the maximum average spring temperature (SON_max) (Table 3). Springtime temperatures are related to maturity timing and have been found to influence phenological timing throughout the growing season, including harvest [4]. Sadras and Petrie [27] reported that warmer spring temperatures are leading to an earlier onset of grapevine development, which in turn results in an earlier maturity date. However, this shift is primarily due to an earlier start of the vine growth cycle rather than an acceleration of the vine developmental rate caused by higher temperatures.

The studied grapevine varieties in this trial responded differentially to bioclimatic indices, probably due to the distinct physiological and phenological characteristics related to their genetic background and environmental adaptability. The GST showed consistent associations with the harvest date, particularly in early-ripening table grape varieties, such as Flame Seedless [13], and also influenced ripening behavior in Pisco varieties, like Moscatel Rosada. This index affects key metabolic processes that drive vine growth and berry ripening, including sugar accumulation, acid degradation, and stress tolerance [28]. Improvements were also observed when using adjusted phenological periods, reinforcing the value of period customization for cultivar-specific climate responsiveness.

Flame Seedless is sensitive to high temperatures, which accelerate its ripening by hastening sugar accumulation compared to other grapevine varieties [29,30]. Zhong et al. [29] reported that Flame Seedless accounted for 61.8% more soluble solids than Victoria due to the upregulation of the VvSWEET15, VvHXK, and MYB44 genes and the downregulation of bHLH14, which is linked to the glucose metabolism. For Muscat of Alexandria, the harvest date was mostly affected by the HI, which incorporates not just temperature but also day length in its calculation. This index could be relevant for varieties whose phenological responses depend on heat accumulation and solar radiation exposure. The HI accounts for extended heat exposure, and it could be representative of varieties with longer ripening periods, such as Muscat of Alexandria. This grapevine variety is commonly known to be sensitive to sunburn and is usually attributed to a higher frequency and intensity of heat waves that increase radiation and temperature in the exposed clusters [31]. Sunburn incidence could be reduced by diffusing light regime, and decreasing radiation may delay the harvesting date of sensitive varieties, such as Muscat of Alexandria [32,33]. Despite this, Muscat of Alexandria is a variety that tolerates heat stress, achieving high stomatal conductance and CO₂ assimilation under these conditions [33]; thereby, radiation and temperature play a key factor in the berry ripening in this variety. Interestingly, the sums of daily degree days (DDD) and radiation were positively correlated to the degradation of total carotenoids, whereas this parameter in Muscat of Alexandria was related to DDD that accounts for the degree days accumulation from veraison to harvest [34]. In this report, Muscat of Alexandria was mostly sensitive to topo-climatic variations in other varieties such as Chardonnay, Chenin Blanc, Gewürztraminer, Colombard, and Pinot Gris [34]. Thompson Seedless showed a significant advancement in the harvest date with the increase in the maximum average spring temperature (Table 3). The Thompson Seedless harvest date is usually defined as when soluble solids reach 17 °Brix. Given that springtime maximum temperatures primarily affect early phenological stages, such as budburst and flowering, the harvest in Thompson Seedless may be more influenced by these temperatures since it is a short-ripening grapevine variety. Matsui et al. [35] reported that Thompson Seedless berries exposed to 40 °C for four days during Stage I of berry development were delayed in ripening and contained lower total soluble solids and titratable acidity at harvest compared to the non-stressed control. In this sense, temperature at early stages in Thompson Seedless is a factor that considerably influences its harvest date.

Bioclimatic index calculation should be adjusted to the biological growth of each variety and not by calendar since there are varieties that ripen extremely early in terms of technological maturation and others much later. Moreover, it is important not only to consider the heat accumulation over the whole vegetative cycle of the vine but also for each phenological stage of vine development. This approach is crucial because grapevine phenology and ripening behavior vary significantly among varieties due to genetic, environmental, and management factors. Thus, calendar-based calculations for bioclimatic indices, such as the GST, fail to account for differences in vine phenology, which can result in inaccurate assessments of environmental suitability for both early- and late-ripening grapevine varieties. The early-ripening grapevine varieties often complete phenological stages outside the standard index calculation windows, leading to the underestimation or misrepresentation of heat requirements [36], whereas late-ripening grapevine varieties may accumulate heat beyond the defined calendar period, leading to inaccuracies in indices, like the HI, which assumes phenology ends with a calendar calculation. In addition, adjusting the bioclimatic indices to match the phenological stages of each variety allows for more accurate assessments of heat accumulation and its impact on ripening and better prediction of phenological events, including floraison, veraison, and harvest [36].

The application of temperature-based models for predicting grape ripeness has gained increasing attention in viticultural research [37,38]. The Grapevine Sugar Ripeness (GSR) model offers a temperature-driven approach to estimate the day of the year when a grape variety reaches a specific sugar concentration based on accumulated thermal time [39,40]. This model has been applied to wine grape varieties [41], but it could be adapted to table and Pisco grapes, presenting an opportunity to refine harvest timing assessments. The results presented in this study have focused on the most used bioclimatic indices in viticulture such as the GST, GDD, HI, and CI to evaluate their influence on the variability and trends of harvest dates. However, integrating a sugar ripeness model could provide a more physiological perspective, directly linking climate variability with grape maturity and sugar accumulation dynamics. Since table and Pisco grapes have distinct sugar thresholds for harvest (17 °Brix for table grapes and 22 °Brix for Pisco grapes in our study), further research could explore whether modifying the GSR model parameters to accommodate these thresholds improves the accuracy of harvest date predictions. Additionally, future comparisons between standard bioclimatic indices and GSR-based predictions could assess their respective strengths in explaining interannual variability and long-term trends in grape maturation under climate change scenarios.

Since several bioclimatic indices showed statistically significant correlations with harvest date, the relatively low R² values in many of the linear regression models suggest that additional factors contribute to the observed variability. These include microclimatic variation within vineyards, differences in soil characteristics, genetic variability among cultivars, and management practices, such as irrigation strategies, pruning intensity, or canopy architecture, all of which can influence phenological development. Moreover, the assumption of linearity may not fully capture the complex, and potentially nonlinear, relationship between heat accumulation and vine physiological responses. Future studies could enhance predictive accuracy by applying nonlinear regression, generalized additive models (GAMs), or machine learning algorithms capable of incorporating and analyzing multivariate datasets. However, despite its simplicity, the current approach based on bioclimatic indices and linear models offers a practical tool for growers, especially in contexts where data availability or technical capacity is limited. For instance, identifying that the GST is a consistent predictor of harvest timing for certain varieties enables producers to plan labor, irrigation termination, and harvest logistics with greater anticipation. Similarly, knowing that spring maximum temperatures strongly influence Thompson Seedless can inform early-season crop management decisions. Therefore, although the models presented here do not explain all sources of variation, they provide actionable insights that can support short- and medium-term adaptation strategies in the face of climate variability. Their ease of use and reliance on readily available climatic data make them especially valuable in operational viticulture, serving as a starting point for precision agriculture tools and more complex phenological modeling in the future.

5. Conclusions

This study confirmed the critical influence of bioclimatic indices on the harvest dates of table and Pisco grape varieties over a period of close to two decades in Northern Chile. Flame Seedless and Moscatel Rosada exhibited significant advancements in harvest dates with an increased growing season temperature (GST), reflecting their sensitivity to cumulative heat during the growing period. Muscat of Alexandria was more influenced by the Huglin Index (HI), indicating its dependence on both heat accumulation and day length, particularly for varieties with extended ripening periods. Thompson Seedless showed earlier harvest dates with rising maximum spring temperatures (SONmax), emphasizing the impact of early-season heat on this short-cycle variety. Despite that the explanatory power of the linear regression models was limited by relatively low R² values, the approach provides a practical and accessible framework for anticipating harvest timing using readily available climatic data. These models can assist growers in planning harvest logistics, adjusting irrigation schedules, or coordinating labor availability, which are all critical aspects of modern viticulture. The simplicity of the method facilitates its use in data-limited or resource-constrained settings, making it especially valuable for operational decision making under conditions of climatic uncertainty. Future research should incorporate nonlinear models and additional agronomic and environmental variables to improve predictive capacity, but the current findings offer a solid starting point for climate-informed vineyard management.