Abstract
Background and goals Temperate woody perennials, including grapes, require chilling and heat accumulation during dormancy to initiate growth in the following spring. The quantification of chilling and heat requirements is critical for a more accurate prediction of endodormancy breaking or budbreak. We aimed to estimate the chilling and heat requirements of Delaware grape in Osaka, Japan, by using partial least squares (PLS) regression to correlate the budbreak dates of Delaware with temperature variations from 1963 to 2022.
Methods and key findings PLS was effective in delineating chilling and forcing periods, and estimating the chilling and heat requirements of Delaware. Three chilling models (0 to 7.2°C, Utah, and Dynamic) and one forcing model (growing degree hour [GDH]) were applied to calculate daily chilling and heat accumulation. According to the PLS regression output, the chilling period for budbreak began in mid-October and extended to early February, whereas the forcing period began in mid-January and extended to the budbreak for each year. The Dynamic model was the most accurate, with a coefficient of variation of 5.3% and the least annual variation in accumulated chill during the chilling period.
Conclusions and significance The chilling (in the Dynamic model) and heat (in the GDH model) requirements of Delaware were determined as 70 ± 4 chill portions (mean ± standard deviation) and 5068 ± 721 GDHs, respectively. These findings can contribute to improving the accuracy of the budbreak prediction model or can help guide physiological and genetic research through identification of different dormancy phases.
Introduction
Temperate woody perennials, including grapes, undergo a dormancy phase during the fall and winter, allowing them to endure low temperatures. Generally, there are two phases of dormancy: endodormancy and ecodormancy (Lang et al. 1987). Endodormancy is regulated by factors within the plants, such as the inhibition or retardation of bud growth, even under favorable environmental conditions. Ecodormancy is regulated by external factors (primarily temperature) that limit growth. The buds are sensitive to chill accumulation during endodormancy. After the chilling requirements are met, the buds enter the ecodormancy phase, during which they are sensitive to heat accumulation. Budbreak occurs after the heat requirements are met. Thus, chilling and heat requirements are the primary determinants of dormancy breaking and growth initiation in spring (Luedeling 2012). Both chilling and heat requirements are specific to species and varieties (Sugiura et al. 2009).
Delaware grape (Vitis vinifera × [Vitis labrusca × Vitis aestivalis]) has its origin in the United States (Maul et al. 2023), where it is cultivated primarily as a winegrape variety. In Japan, Delaware is one of the main table grape varieties that produce seedless berries when treated by the plant hormone gibberellic acid. Determining the chilling and heat requirements of Delaware is critical for a more accurate prediction of the endodormancy breaking and budbreak dates. Endodormancy breaking date is an indicator of the start time of heating during heated cultivation. Heated cultivation inside greenhouses is widely used in Japan to promote budbreak, which allows for the advancement of cultivation management timing earlier than under field conditions. This provides both a more efficient use of labor throughout the season, as well as high-price sales through early shipping. Because abnormal (decreased, delayed, and non-uniform) budbreak is caused when heating is initiated before chilling requirements are fulfilled (Hirose et al. 2000, Sugiura et al. 2009), understanding the chilling requirements to break endodormancy is useful for the management of temperature control in heated cultivation. In addition, budbreak date prediction can be utilized in grapevines growing under field conditions to plan disease and pest control, and to estimate the risk period for frost damage. It is also important in predicting future changes in budbreak dates caused by climate change. The budbreak of Delaware in Osaka, Japan, advanced at 1.5 days per decade between 1963 and 2010 (Figure 1). This advance has been highly correlated with the rising trends in air temperature during the same years (Kamimori et al. 2019). This finding is consistent with previous research that reported significantly advanced spring events (budbreak, leaf emergence, and bloom) of temperate woody perennials, owing to an increase in air temperature (Chmielewski et al. 2004, Wolfe et al. 2005, Hasegawa and Ogata 2007, Guédon and Legave 2008, Grab and Craparo 2011, Guo et al. 2013). However, future air temperature increases may not necessarily result in budbreak-advancing effects. This is because temperature increases in fall and winter are expected to decrease chilling accumulation, thereby delaying the endodormancy breaking process (Kamimori and Hiramatsu 2022). Thus, accurate estimation in both chilling and heat requirements of Delaware is critical to develop effective adaptation strategies for climate change based on Delaware response prediction.
Distribution of budbreak dates for Delaware grapes from 1963 to 2010 in Osaka, Japan (modified from Kamimori et al. 2019 with permission from the Japanese Society for Horticultural Science). *** indicates that the trend is significant at p < 0.001. DOY, day of the year.
Several models have been developed to quantify chilling and heat requirements. Three chilling models are widely used: 0 to 7.2°C (Weinberger 1950), Utah (Richardson et al. 1974), and Dynamic (Fishman et al. 1987a, 1987b). The 0 to 7.2°C model is primarily used for grapevine cultivation in Japan because it permits easy calculation of chill metrics. However, only a few studies have compared the 0 to 7.2°C model with alternative chilling models to assess the accuracy for chill quantification in grapevines. The growing degree hour (GDH) model is widely used to quantify heat requirements (Anderson et al. 1986). To obtain chilling and heat requirements using these models, it is necessary to identify endo- and ecodormancy. However, in experimental methods, whether a bud is endo- or ecodormant is mainly determined by evaluating growth ability under favorable conditions, a method that has not been standardized among researchers. Therefore, it remains challenging to experimentally determine robust and comparable chilling and heat requirements (Fadón et al. 2020).
Partial least squares (PLS) regression is a statistical method that is applied when independent variables outnumber dependent variables (Luedeling and Gassner 2012). Recently, PLS analysis has been successfully used to identify chilling and forcing periods, and to estimate chilling and heat requirements based on the long-term phenology and temperature records of many temperate woody perennial species (Luedeling et al. 2013b, Guo et al. 2014, 2015, Benmoussa et al. 2017, Martínez-Lüscher et al. 2017, Pertille et al. 2022). However, obtaining reliable results from PLS analysis requires long-term temperature data, including appropriate levels of interannual variability, to identify the temperature response signals (Luedeling and Gassner 2012). Thus, PLS analysis is limited to studies with a long-term data set, which is why there are currently only a few studies in grapevines (Martínez-Lüscher et al. 2016). Moreover, relevant reports from East Asia are scarce.
In this study, PLS regression was applied to correlate the budbreak dates of Delaware with temperature data in Osaka over a span of >50 years. The objective of this study was to identify the chilling and forcing periods and to quantify the corresponding chilling and heat requirements of Delaware using three chilling models and one forcing model via PLS analysis. The efficacies of the three chilling models were compared under the climatic conditions specific to Osaka, Japan.
Materials and Methods
Experimental site
Budbreak data were collected from a vineyard of the Research Institute of Environment, Agriculture and Fisheries, Osaka Prefecture (34°31´N; 135°35´E, 70 m asl). The planting density was ~230 vines/ha, with a vine-by-row spacing of 6.0 m × 7.0 m in a north-south orientation. The Delaware grapevines were trained on a horizontal trellis (1.8 m aboveground) with long canes (Supplemental Figure 1). The trellis covered the entire surface of the vineyard, and the shoots of the vines were trained on a net-like trellis top. Horizontal trellises offer the advantages of easy vigor management through long-cane pruning to adjust the spread of the canopy, and easy cluster management, such as gibberellic acid application. The horizontal trellis is the most commonly used training system for table grape cultivation in Japan. The vineyard was maintained using conventional fertilization and disease and pest control; weak vines were replanted accordingly. The mean annual temperature and precipitation in the area from 1981 to 2010 were 16.9°C and 1279 mm, respectively. Temperatures rarely fell below freezing in winter. The climate of the area is humid subtropical (Cfa), according to the Köppen classification (Kottek et al. 2006).
Data collection
Three to four Delaware grapevines (more than five-years-old) were chosen for experiments. Budbreak was determined by visual observation, and the budbreak date was recorded as the average of three to four grapevines. Budbreak date was defined as the date when the buds were swollen and >80% of the total buds were exposed from the scale, which corresponded to stage 5 on the BBCH scale, or to stage 3 on the Eichhorn-Lorenz (E-L) scale for grapevine growth (Lorenz et al. 1995). Observational data spanning 53 years, from 1963 to 2022, were available (no observations were available from 2011 to 2017) (Kamimori et al. 2019).
Daily minimum and maximum temperature data between 1980 and 2022 for the experimental site were obtained from the Agro-Meteorological Grid Square Data (AMGSD) of the National Agricultural Research Organization (Ohno et al. 2016), which provides daily data with 1 km resolution. Monthly transfer functions were defined for temperature data from 1963 to 1979 using temperature data from the Osaka District Meteorological Observatory (located ~17 km from the experimental site; temperature data from 1883 were available) and from the experimental site (retrieved from the AMGSD system). Regarding overlapping days, which were covered by the two recordings, the daily minimum and maximum temperatures at the experimental site were correlated with the daily temperature extremes from the Meteorological Observatory. The relationships that emerged (Supplemental Table 1) were used to estimate temperature data for the experimental site from 1963 to 1979. Daily mean temperatures were calculated by averaging the minimum and maximum values for each day. Hourly temperature data required for the chilling and forcing models were constructed using the make_hourly_temps function in the ‘chillR’ package (ver. 0.72.8; Luedeling 2022) from the R programming language (ver. 4.2.1; R Core Team 2022). The methodology for this function was developed by Luedeling (2018), who used the equations proposed by Spencer (1971), Linvill (1989, 1990), and Almorox et al. (2005).
PLS regression analysis
For the PLS analysis methodology, we closely followed the approach developed by Luedeling and Gassner (2012) and Luedeling et al. (2013a, 2013b), with minor modifications. All analyses were performed with R (ver. 4.2.1), using procedures of the ‘chillR’ package (ver. 0.72.8).
PLS regression analysis was used to analyze the response of Delaware budbreak dates (expressed in Julian days) to variations in mean daily temperatures during 1963 to 2022, using the PLS_pheno function in ‘chillR’. When raw daily temperature data were used for PLS analysis, we obtained an unidentifiable temperature response pattern (Supplemental Figure 2). From this result, it is difficult to recognize distinct periods related to the temperature response. To improve the clarity of the outputs, the application of an 11-day running mean filtering procedure (the mean values of the period starting five days before, and ending five days after, the corresponding date) was applied to the data set, similar to previous work (Luedeling and Gassner 2012, Guo et al. 2013, Martínez-Lüscher et al. 2016). Daily mean temperatures were used as independent variables (357 records; from 1 May of the year preceding the recorded budbreak dates to 22 April, which was the latest budbreak date recorded), whereas the budbreak data were used as dependent variables. The variable importance in projection (VIP) and standardized model coefficients are the two main outputs of PLS analysis (Guo et al. 2015). The VIP values indicate the importance of all independent variables in explaining variation in the dependent variables. We adopted the commonly used VIP threshold value of 0.8 (Wold et al. 2001). The standardized model coefficients indicate the strength and direction of the effect of each variable in the PLS model. From the output of the PLS analysis, two different periods were identified, which were marked by VIP scores greater than 0.8, and a predominance of either negative or positive model coefficients for a given time period.
Chilling and forcing models
In the present study, we used three common chilling models (0 to 7.2°C, Utah, and Dynamic). The most widely used 0 to 7.2°C model defines one chilling hour (CH) as one hour with temperature between the thresholds (0 to 7.2°C) (Weinberger 1950). The Utah model defines chill in chill units (CU) and contains a weighted function based on different temperature ranges, including the cancellation of previously accumulated chill by high temperatures (Richardson et al. 1974). The Dynamic model considers chilling accumulation to be a two-step process based on hypothetical physiological reactions, and does more than simply assigning a weighted function to temperatures (Fishman et al. 1987a, 1987b). In the first step, which is reversible, an intermediate product of the dormancy breaking factor is formed when the bud is exposed to low temperatures; however, higher temperatures can nullify it. In the second step, which is irreversible, once a critical amount of the intermediate product has accumulated, it is transformed into a stable product for the dormancy breaking factor, expressed as the chill portion (CP). CP is accumulated until the end of the chilling period.
We used the GDH model (Anderson et al. 1986) as the forcing model. It assumes that heat accumulates when hourly temperatures are between three threshold values (base temperature [below which no heat accumulation will occur; 4°C], optimal temperature [at which heat accumulation is maximal; 25°C], and critical temperature [above which no heat accumulation will occur again; 36°C]).
Three chilling and GDH models were used to determine the daily chill and heat accumulation from 1963 to 2022 using hourly temperature data. These calculations were implemented using the daily_chill function in ‘chillR’, based on equations by Luedeling et al. (2009a, 2009b). PLS regression analysis was used to determine the chilling and forcing periods by correlating the budbreak dates of Delaware with daily chill and heat accumulation by the PLS_chill_force function in ‘chillR’. Following the previous PLS analysis, an 11-day running mean was applied to the daily chill and heat accumulation. For each year, daily chill and heat values were independent variables (714 records; from 1 May of the year preceding the recorded budbreak dates to 22 April, which was the latest budbreak date recorded), whereas budbreak dates were dependent variables. The PLS outputs were interpreted based on dormancy theory. That is, greater chill accumulation during the chilling period should accelerate the fulfillment of chilling requirements, resulting in early budbreak. Similarly, greater heat accumulation during the forcing period should advance budbreak. Thus, the chilling and forcing periods in the PLS outputs should be interpreted as periods during which the VIP is greater than 0.8 and the model coefficients are negative (Luedeling et al. 2013a, 2013b, Guo et al. 2015, Fadón et al. 2023). Based on these expectations, likely chilling and forcing periods with predominant phases of negative model coefficients were extracted from the PLS outputs.
It should be noted that exact estimates of chilling and heat requirements cannot be determined using the statistical approach alone. However, we can obtain a reasonable approximation of these requirements using PLS analysis (Luedeling et al. 2013b). Assuming that chilling accumulation occurs over the chilling periods defined by the PLS analysis, chilling requirements can be calculated as the total chilling accumulation (including some periods with positive model coefficients or low VIP scores during the chilling periods). Heat requirements for the forcing period were calculated similarly, and the budbreak date was regarded as the end of the forcing period because heat accumulation after budbreak had no effect. These data were determined for each year of observation, and the final chill and heat requirements were calculated as the mean of these values for the period 1963 to 2022. The performance of the chilling models was evaluated by analyzing the variation in chill accumulation across years during the chilling period (Luedeling et al. 2013b, Guo et al. 2015). The standard deviations of the chilling estimates for each model were calculated. The model with the lowest standard deviation was considered to be the most accurate among the tested models. In addition, when several candidate chilling and forcing periods were identified, their respective standard deviations were calculated for comparison.
Results
Relevant periods influencing Delaware budbreak
During 1963 to 2022, the average, earliest, and latest budbreak dates of Delaware in Osaka were 9 April, 30 March, and 22 April, respectively.
The results of PLS analysis between budbreak dates and daily mean temperatures are shown (Figure 2). We identified two periods during which warm conditions had antagonistic effects on the budbreak dates of Delaware. High temperatures in early fall (11 Sept to 31 Oct) were related to delayed budbreak (independent variables with high VIP scores and positive model coefficients), whereas warm conditions during the subsequent phase (10 Nov to 19 April) led to advanced budbreak (independent variables with high VIP scores and negative model coefficients). Moreover, this phase was divided into two subphases based on the strengths of the negative model coefficients and VIP scores. The first stage was from 10 Nov to 23 Dec, with relatively weak model coefficients and VIP scores. Following a brief interruption, the second stage was from 12 Jan to 19 April, with strong model coefficients and VIP scores; the model coefficients tended to increase close to the budbreak dates.
Results of the partial least squares (PLS) regression of budbreak dates for Delaware grapes in Osaka with 11-day running means of daily mean temperatures, from May in the previous year to April in the year of recorded budbreak. Blue bars in the top panel indicate variable importance in projection (VIP) values greater than 0.8, the threshold for variable importance. In the middle and bottom panels, red indicates that the model coefficients are negative (and VIP values greater than 0.8), while green indicates positive (and VIP values greater than 0.8) relationships between budbreak and temperature. The black line in the bottom panel represents the mean temperatures, while the gray, green, and red areas represent the standard deviation of daily mean temperatures for each day of the year. In all panels, the dashed line indicates average budbreak date (9 April), and the shaded areas in the right-side margins represent the range of budbreak dates for the evaluated years.
Chilling and forcing periods for Delaware
The results of the PLS analyses of budbreak dates and CH, CU, and CP are shown in Figures 3, 4, and 5, respectively. While using the 0 to 7.2°C model for chilling calculation, we interpreted the period during 16 Oct to 6 Feb as the chilling period, with negative model coefficients and VIP scores greater than 0.8. When using the Utah model for chilling calculation, we interpreted the period from 10 Sept to 16 Feb as the chilling period. Although the start of the chilling period in the Utah model was earlier than that in the 0 to 7.2°C model by approximately one month, PLS results in both models produced similar results. When using the Dynamic model for the chilling calculation, the PLS regression results showed similar characteristics in terms of changes in model coefficients and VIP score as those of the previous two analyses. We interpreted the period from 16 Oct to 12 Feb as the chilling period. However, in all three chilling models, some periods with positive model coefficients or low VIP scores (early November to mid-December) occurred during chilling periods.
Results of the partial least squares (PLS) regression of budbreak dates for Delaware grapes in Osaka using the 0 to 7.2°C model for chilling calculations, from May in the previous year to April in the year of recorded budbreak. Blue bars in the top panel indicate variable importance in projection (VIP) values greater than 0.8, the threshold for variable importance. In the middle and bottom panels, red indicates that the model coefficients are negative (and VIP values greater than 0.8), while green indicates positive (and VIP values greater than 0.8) relationships between budbreak and daily chill accumulation. The black line in the bottom panel represents the daily chill accumulation, while the gray, green, and red areas represent the standard deviation of the daily chill accumulation for each day of the year. In all panels, the dashed line indicates average budbreak date (9 April), and the shaded areas in the right-side margins represent the range of budbreak dates for the evaluated years.
Results of the partial least squares (PLS) regression of budbreak dates for Delaware grapes in Osaka using the Utah model for chilling calculations, from May in the previous year to April in the year of recorded budbreak. Blue bars in the top panel indicate variable importance in projection (VIP) values greater than 0.8, the threshold for variable importance. In the middle and bottom panels, red indicates that the model coefficients are negative (and VIP values greater than 0.8), while green indicates positive (and VIP values greater than 0.8) relationships between budbreak and daily chill accumulation. The black line in the bottom panel represents the daily chill accumulation, while the gray, green, and red areas represent the standard deviation of the daily chill accumulation for each day of the year. In all panels, the dashed line indicates average budbreak date (9 April), and the shaded areas in the right-side margins represent the range of budbreak dates for the evaluated years.
Results of the partial least squares (PLS) regression of budbreak dates for Delaware grapes in Osaka using the Dynamic model for chilling calculations, from May in the previous year to April in the year of recorded budbreak. Blue bars in the top panel indicate variable importance in projection (VIP) values greater than 0.8, the threshold for variable importance. In the middle and bottom panels, red indicates that the model coefficients are negative (and VIP values greater than 0.8), while green indicates positive (and VIP values greater than 0.8) relationships between budbreak and daily chill accumulation. The black line in the bottom panel represents the daily chill accumulation, while the gray, green, and red areas represent the standard deviation of the daily chill accumulation for each day of the year. In all panels, the dashed line indicates average budbreak date (9 April), and the shaded areas in the right-side margins represent the range of budbreak dates for the evaluated years.
The results of the PLS analysis of budbreak dates and GDH (using the Dynamic model for chill accumulation) are shown (Figure 6). Important negative coefficients for heat accumulation appeared from 20 Nov to 23 Dec, and from 12 Jan to 22 April. Additionally, substituting the Dynamic model with the 0 to 7.2°C or Utah models in PLS regression of GDH produced similar results, with the forcing periods of the 0 to 7.2°C model from 20 Nov to 22 Dec and 12 Jan to 20 April, and of the Utah model from 21 Nov to 17 Dec and 13 Jan to 19 April (Supplemental Figures 3 and 4, respectively).
Results of the partial least squares (PLS) regression of budbreak dates for Delaware grapes in Osaka using the growing degree hour (GDH) model for heat calculations (using the Dynamic model for chill accumulation), from May in the previous year to April in the year of recorded budbreak. Blue bars in the top panel indicate variable importance in projection (VIP) values greater than 0.8, the threshold for variable importance. In the middle and bottom panels, red indicates that the model coefficients are negative (and VIP values greater than 0.8), while green indicates positive (and VIP values greater than 0.8) relationships between budbreak and daily heat accumulation. The black line in the bottom panel represents the daily heat accumulation, while the gray, green, and red areas represent the standard deviation of the daily heat accumulation for each day of the year. In all panels, the dashed line indicates average budbreak date (9 April), and the shaded areas in the right-side margins represent the range of budbreak dates for the evaluated years.
Chilling and heat requirements for Delaware
Estimates of the chilling and heat requirements of Delaware using the 0 to 7.2°C, Utah, Dynamic, and GDH models are shown (Table 1). The chill requirements for Delaware buds in the 0 to 7.2°C, Utah, and Dynamic models were 1182 ± 93 CH (mean ± standard deviation), 688 ± 173 CU, and 70 ± 4 CP, respectively. The Dynamic model provided the most accurate estimates of chilling requirements, with a coefficient of variation (CV) of 5.3% compared to 7.8% and 25.1% for the 0 to 7.2°C and Utah models, respectively.
Estimates of the chilling and heat requirements for budbreak of Delaware grapes in Osaka, Japan, using the 0 to 7.2°C, Utah, and Dynamic chilling models, and the growing degree hour (GDH) forcing model. Units for the chilling models are in chilling hours (CH), chill units (CU), and chill portions (CP), respectively, and units for the forcing models are in GDH. CV, coefficient of variation.
Candidate forcing periods in the 0 to 7.2°C, Utah, and Dynamic models were 20 Nov or 12 Jan to budbreak, 21 Nov or 13 Jan to budbreak, and 20 Nov or 12 Jan to budbreak, respectively. Considering heat accumulation in the three chilling models during each candidate forcing period in all years, the heat requirements (mid-November to budbreak) in the 0 to 7.2°C, Utah, and Dynamic models were 7847 ± 1295 GDH, 7716 ± 1267 GDH, and 7847 ± 1295 GDH, with CVs of 16.5%, 16.4%, and 16.5%, respectively. The heat requirements (mid-January to budbreak) in the 0 to 7.2°C, Utah, and Dynamic models were 5068 ± 721 GDH, 5050 ± 723 GDH, and 5068 ± 721 GDH, with CVs of 14.2%, 14.3%, and 14.2%, respectively (Table 1).
The chilling and forcing periods for the 0 to 7.2°C, Utah, and Dynamic models overlapped by 26, 35, and 32 days, respectively. In the 0 to 7.2°C and Dynamic models, the accumulated chilling requirements (the ratio to total chill requirement expressed as a percentage) by the beginning of the forcing period (12 Jan) were 787 ± 103 CH (66.6%) and 48 ± 4 CP (68.2%), respectively, whereas in the Utah model, the chill accumulation by the beginning of the forcing period (13 Jan) was 208 ± 181 CU (30.2%), and exhibited greater variation than the other models (Table 2).
Estimates of the chilling requirements until chilling and forcing periods began to overlap for budbreak of Delaware grapes in Osaka, Japan, using the 0 to 7.2°C, Utah, and Dynamic models. Units for the chilling models are in chilling hours (CH), chill units (CU), and chill portions (CP), respectively. CV, coefficient of variation.
Discussion
Temperature responses of Delaware budbreak
The PLS regression analysis proved to be a valid method for correlating the temperature effects with the budbreak dates of Delaware (Figure 2). The PLS results revealed that budbreak dates of Delaware were delayed by warm temperatures during 11 Sept to 31 Oct, whereas warm conditions during 10 Nov to 23 Dec and 12 Jan to 19 April advanced budbreak dates. Controlled experiments using detached twigs revealed that the endodormancy of Delaware in Osaka was induced primarily by low temperatures (10 to 18°C) (Tohbe et al. 1998) and was in its deepest phase from late September to late October. Thereafter, endodormancy breaking began, and transitioned to ecodormancy during late January to early February (Horiuchi et al. 1981). According to the PLS results and the previous experimental works (Horiuchi et al. 1981, Tohbe et al. 1998), the high temperatures in September and October appear to delay budbreak by inhibiting dormancy induction and breaking. Conversely, high temperatures from 10 Nov had the effect of advancing budbreak, suggesting that the buds were responsive to heat at that time. Grapevines have the capability to fulfill their critical chilling requirements earlier than most woody perennial species (Horiuchi et al. 1981, Mullins et al. 1992, Martínez-Lüscher et al. 2016). However, the VIP values and negative coefficients for November to December were considerably lower than those for the subsequent periods, suggesting that the budbreak-advancing effect was weak in most years. Subsequent periods (12 Jan to 10 April) with additional chilling were correlated with high VIP values and negative coefficients, indicating that dormancy breaking was in progress due to additional chilling. This finding is consistent with the results of previous studies on temperate woody perennials, which showed that additional chilling exposure reduced a quantity of heat requirement and increased the percentage of budbreak (Takagi and Tamura 1987, Dokoozlian 1999, Luedeling et al. 2013b, Guo et al. 2015). Thus, the temperature response patterns estimated by PLS regression were consistent with those of previous studies on the physiology of dormancy in grapevines and fruit trees.
Chilling and forcing periods for budbreak of Delaware
The chilling periods identified by PLS regression occurred around mid-September or mid-October to early February, which is consistent with the previously mentioned dormancy processes (Horiuchi et al. 1981). Notably, in all three chilling models, some periods with positive model coefficients or low VIP scores occurred during the chilling periods, which is consistent with the results of several studies (Luedeling 2013a, Guo et al. 2015, 2019, Martínez-Lüscher et al. 2017). As previously stated, Delaware buds became responsive to heat in early November, suggesting that chill accumulation at that time may be related to the late occurrence of budbreak. However, as chill accumulation also has a budbreak-advancing effect by breaking endodormancy, we considered that these two opposing effects of chill accumulation should be expressed as discontinuities within the chilling period. As a result, we considered the entire period from mid-September or mid-October to early February to be part of the chilling stage. Two candidate forcing periods occurred around mid-November and mid-January until budbreak (mid-April). The CV for GDH from mid-January to budbreak was lower than that from mid-November to budbreak (Table 1). Thus, the period from mid-January to budbreak was considered to be the forcing period. This result is in line with the previous suggestion that November to December is less effective for heat accumulation. The chilling and forcing periods were similar to the PLS regression analysis for Riesling Italico (V. vinifera) in Mandicevac, Croatia, which was reported to have the chilling and forcing periods from 23 Sept to 27 Feb, and from 22 Jan to 1 May, respectively (Martínez-Lüscher et al. 2016). Considering the differences in climate of Mandicevac and Osaka and the varieties used in the two studies, this was a noteworthy match.
Chilling and heat requirements for budbreak of Delaware
The 0 to 7.2°C model is widely used in Japan as an indicator for endodormancy breaking, or the initiation of heating in heated cultivation. The chilling requirement for Delaware to satisfy endodormancy has been estimated at 600 (Hirose et al. 2000) to 1000 CH by experimental or field observations, close to the estimated 1182 ± 93 CH in this study (Table 1). Although there are recently published reports of chill requirement expressed as CU and CP for grapevines (Londo and Kovaleski 2019, Prats-Llinàs et al. 2019), we could not compare them with our results, owing to different factors such as the criteria used to define the durations of endo- and ecodormancy. The heat requirement of Delaware was 5050 to 5068 GDH, which was close to the ~5000 GDH of V. vinifera var. Ciliegiolo, Sangiovese, and Vermentino, which are classified as early budbreak varieties in Italy (Andreini et al. 2009; calculation of GDH began on 30 Jan). Delaware is also classified as an early budbreak variety (Ministry of Agriculture, Forestry and Fisheries 2015), consistent with the previous study. If long-term budbreak and temperature data for at least 15 years (including meaningful variation in the chill and heat accumulation necessary for PLS to work) from the same field or fields with very similar climatic conditions are available, PLS can be used to quickly estimate the initial chilling and heat requirements (Luedeling et al. 2013b). Therefore, it is expected that the chilling and heat requirements for grapevines in other varieties and climates can be investigated using PLS analysis.
In our analysis, the chilling and forcing periods overlapped for approximately one month, indicating that the chill and heat accumulation in Delaware overlapped and were not sequential. This finding suggests that an overlapping approach to grapevine dormancy modeling is appropriate. Except for the Utah model, the overlap occurred after 12 Jan and continued until the end of the chilling period, suggesting that heat accumulation began to show effectiveness when ~70% of the chilling requirement had been fulfilled (Table 2). These results can guide future research to understand the overlapping period, which is important to improve model fitting in the process of dormancy.
Comparing the CV in three chilling models, the CV of the Dynamic model was lower (5.3%) than that of the 0 to 7.2°C and Utah models (7.8 and 25.1%, respectively; Table 1). This finding is consistent with previous reports indicating that the Dynamic model performs better than other chilling models under various climatic conditions (Campoy et al. 2012, Luedeling et al. 2013b, Guo et al. 2015, Benmoussa et al. 2017, Abou-Saaid et al. 2022). However, the parameters of the Dynamic and GDH models used in this study were not optimized for Delaware. Although the use of fixed parameter values is convenient for comparing the chilling and heat requirements of different species or varieties, the values are not optimal for evaluating specific varieties (Egea et al. 2021). Further research is needed to optimize the parameter values for the Dynamic and GDH models for Delaware based on the results of this study.
It should be noted that outputs derived from PLS analysis are only valid for the sites where observed data were obtained and may not be directly useful under very different climatic conditions. Indeed, PLS analysis of identical varieties of sweet cherry grown in Bonn, Germany (Köppen classification: Cfb) and Zaragoza, Spain (Köppen classification: BSk) exhibited large differences in the estimated chilling and heat requirements (Fadón et al. 2023), implying that this is due to inaccuracies within the chill models in approximating the actual dormancy progression of fruit trees (Luedeling 2012), or to a lack of modeling for chilling and heat interaction (Fernandez et al. 2020). Even for grapevines, using PLS analysis to investigate how different climatic conditions affect the agroclimatic requirements might help evaluate and improve the chilling and heat model accuracy. Future studies are needed to develop a process-based model that includes an understanding of dormancy processes in grapevines by integrating findings from experimental and statistical approaches. The recently developed “PhenoFlex” modeling framework implements the interaction of chilling and forcing using a sigmoid function, which could cover these needs to some extent (Luedeling et al. 2021).
Conclusion
PLS analysis using long-term budbreak and temperature records was effective in determining the chilling and forcing periods and estimating the chilling and heat requirements for budbreak of the Delaware grape in Osaka, Japan. The chilling periods were identified from mid-October to early February, whereas the forcing periods were identified from mid-January to the budbreak date for each year, which were consistent with previous studies on physiology of dormancy in grapevine. The chilling and forcing periods overlapped for approximately one month, indicating that the chill and heat accumulation in Delaware overlapped. The Dynamic model provided the most accurate estimate of chilling requirements, suggesting that CP is a better indicator for winter chill than the conventional measurements of CH by the 0 to 7.2°C model. The chilling and heat requirements of Delaware in Osaka were 70 ± 4 CP and 5068 ± 721 GDH, respectively. Chilling requirements can be a good indicator of the start time of heating during heated cultivation. In addition, chilling and heat requirements provide useful information for improving the accuracy of budbreak prediction models. Furthermore, the findings of chilling and forcing periods could guide physiological or genetic research by specifying the different dormancy phases, allowing investigation into the internal processes occurring inside the grapevine during the dormancy phase changes.
Supplemental Data
The following supplemental materials are available for this article in the Supplemental tab above:
Supplemental Table 1 Slope and coefficient of determination of the linear regression equation calculated from temperature data from the experimental site and from the Osaka District Meteorological Observatory (ODMO) for the period 1980 to 2022. Tmax, maximum temperature; Tmin, minimum temperature.
Supplemental Figure 1 Delaware vineyard of the experimental site in Osaka, Japan.
Supplemental Figure 2 Results of the partial least squares (PLS) regression of budbreak dates for Delaware grapes in Osaka with raw daily mean temperature data, from May in the previous year to April in the year of recorded budbreak. Blue bars in the top panel indicate variable importance in projection (VIP) values greater than 0.8, the threshold for variable importance. In the middle and bottom panels, red indicates that the model coefficients are negative (and VIP values greater than 0.8), while green indicates positive (and VIP values greater than 0.8) relationships between budbreak and temperature. The black line in the bottom panel represents the mean temperatures, while the gray, green, and red areas represent the standard deviation of the daily mean temperatures for each day of the year. In all panels, the dashed line indicates average budbreak date (9 April), and the shaded areas in the right-side margins represent the range of budbreak dates for the evaluated years.
Supplemental Figure 3 Results of the partial least squares (PLS) regression of budbreak dates for Delaware grapes in Osaka using the growing degree hour (GDH) model for heat calculation (using the 0 to 7.2°C model for chill accumulation), from May in the previous year to April in the year of recorded budbreak. Blue bars in the top panel indicate variable importance in projection (VIP) values greater than 0.8, the threshold for variable importance. In the middle and bottom panels, red indicates that the model coefficients are negative (and VIP values greater than 0.8), while green indicates positive (and VIP values greater than 0.8) relationships between budbreak and daily heat accumulation. The black line in the bottom panel represents the daily heat accumulation, while the gray, green, and red areas represent the standard deviation of the daily heat accumulation for each day of the year. In all panels, the dashed line indicates average budbreak date (9 April), and the shaded areas in the right-side margins represent the range of budbreak dates for the evaluated years.
Supplemental Figure 4 Results of the partial least squares (PLS) regression of budbreak dates for Delaware grapes in Osaka using the growing degree hour (GDH) model for heat calculation (using the Utah model for chill accumulation), from May in the previous year to April in the year of recorded budbreak. Blue bars in the top panel indicate variable importance in projection (VIP) values greater than 0.8, the threshold for variable importance. In the middle and bottom panels, red indicates that the model coefficients are negative (and VIP values greater than 0.8), while green indicates positive (and VIP values greater than 0.8) relationships between budbreak and daily heat accumulation. The black line in the bottom panel represents the daily heat accumulation, while the gray, green, and red areas represent the standard deviation of the daily heat accumulation for each day of the year. In all panels, the dashed line indicates average budbreak date (9 April), and the shaded areas in the right-side margins represent the range of budbreak dates for the evaluated years.
Footnotes
We would like to thank Dr. Hisayo Yamane (Kyoto University) for providing valuable comments on this research. We would also like to thank the staff of the Research Institute of Environment, Agriculture and Fisheries, Osaka Prefecture, who collected consistent budbreak data of Delaware grapes over several decades.
Kamimori M and Hosomi A. 2024. Evaluation of chilling and heat requirements for the budbreak of Delaware grape in Osaka, Japan. Am J Enol Vitic 75:0750011. DOI: 10.5344/ajev.2024.23045
By downloading and/or receiving this article, you agree to the Disclaimer of Warranties and Liability. If you do not agree to the Disclaimers, do not download and/or accept this article.
- Received June 2023.
- Accepted February 2024.
- Published online May 2024
This is an open access article distributed under the CC BY 4.0 license.
References
Vol 75 Issue 1
Jump to section
Related Articles
Cited By...
More from this TOC section
Similar Articles
