Climatic Change

, Volume 137, Issue 3, pp 541–556

A crop and cultivar-specific approach to assess future winter chill risk for fruit and nut trees

Article

DOI: 10.1007/s10584-016-1692-3

Cite this article as:
Darbyshire, R., Measham, P. & Goodwin, I. Climatic Change (2016) 137: 541. doi:10.1007/s10584-016-1692-3

Abstract

Anthropogenic climate change will influence winter chill accumulation, with future declines likely in temperate locations. However, these declines only translate as impacts when cultivar winter chilling requirements are not satisfied. This study presents a methodology to evaluate future impacts of declining winter chill through a cultivarspecific approach which is useful for growers, industry and policy-makers to develop adaptation strategies. A risk based system was applied to represent the likelihood of meeting cultivar chilling requirements using low, medium, medium-high and high risk ratings based on percentiles. This was combined with climate projection uncertainty graphically at 16 Australian growing districts historically (1981–2010) and for 2030, 2050 and 2090. The results demonstrated that impacts and likely adaptation options differed between cultivars, some recording limited risk at all sites out to 2090 ('Nonpareil' almond) whilst others recorded greater risk both historically and into the future ('Chandler' walnut). Notably, risk differed across sites and with the future time period. These results highlight which cultivars are susceptible to low winter chill conditions, where this risk does and does not manifest and the different time horizons at which the risk will materialise across Australia's main growing districts. Using this approach, changes in winter chill conditions are presented in a useable form which allows for appropriate climate adaptation strategies to be developed, securing the industries into the future.

1 Introduction

The potential impact of climate change on the accumulation of dormancy breaking temperatures (winter chill) is an important scientific question and has been the subject of several research reviews (Atkinson et al. 2013; Campoy et al. 2011; Luedeling 2012). Each of these reviews highlight that temperate perennial fruit and nut production may be at risk, particularly in more mild production zones, due to warming winters, and that adaptation to climate change will likely be necessary.

Declines in winter chill accumulation are only economically important when cultivar-specific winter chill requirements, which are considered to be threshold limits, are not satisfied. Insufficient winter chill accumulation leads to light, delayed and variable bud break (Oukabli et al. 2003; Petri and Leite 2004; Saure 1985; Voller 1986) and for some cultivars, low fruit set and a low quality of fruit set (Mahmood et al. 2000). These influences on flowering and subsequent fruit and nut production negatively influence yield.

Several investigations into winter chill accumulation under climate change conditions have been conducted including for Australia (Darbyshire et al. 2013; Hennessy and Clayton-Greene 1995), California (Baldocchi and Wong 2008; Luedeling et al. 2009c), the United Kingdom (Sunley et al. 2006) and Germany (Chmielewski et al. 2012) as well as globally (Luedeling et al. 2011). These assessments generally illustrate future potential risk to temperate perennial fruit and nut production in relation to insufficient chill accumulation. However, interpretation of these changes in winter chill accumulation into impacts for particular crops, and cultivars within crops, is often lacking.

A key barrier to crop- and cultivar-specific assessments of winter chill is a lack of information on chilling requirements. This paucity of knowledge is compounded by the use of different, non-convertible (Luedeling and Brown 2011) measurements of chill (Measham et al. 2014). Hence, previous assessments of chilling requirements using chill models that have now been superseded (e.g. Ghariani and Stebbins 1994) cannot be used for new evaluations. Finally, a tendency to conduct research within particular tree crops, such as apple (Allderman et al. 2011), apricot (Viti et al. 2010) and walnut (Charrier et al. 2011), limits the understanding of the impact of declining winter chill across the gamut of crop types and cultivars often grown in the same district, further limiting the ability to assess relative risk between crop types and cultivars to inform climate adaptation strategies.

An important aspect of climate impact assessments is the incorporation and description of climate projection uncertainty. This uncertainty, or reporting on the range of likely future climates, is often explicitly considered and described within climate science studies (Alexander and Arblaster 2009; Smith and Chandler 2010) but tends to receive less focus in climate impact analyses. This divergence of climate science and impact studies, and the concomitant lack of practicable recommendations, has been noted (Luedeling 2012; Measham et al. 2014). The uncertainty described in climate impact studies needs to be understandable and provide information for management action in line with tree phenology, rather than introduce confusion which can hinder action. Smith and Chandler (2010) provide a succinct statement regarding this need; “… we are not concerned so much with being proved “right” or “wrong” with regard to climate change projections … as [we are] with providing expert advice that is both transparent, and can be can be acted on now”. To generate such actionable knowledge, assessments of future winter chill conditions need to be cultivar-specific and include climate projection uncertainty in a useable format.

The first modern assessment of the impact of climate change on winter chill was conducted in Australia and used two general chill thresholds (‘low’ and ‘high’ chill) to make results relevant to tree physiology (Hennessy and Clayton-Greene 1995). Chmielewski et al. (2012) used a range of fixed chill thresholds to provide an indication of the impact of declining winter chill in Germany, without directly linking results to particular cultivar requirements. Sunley et al. (2006) assessed flowering duration of blackcurrant and raspberry against chill accumulation of several chill models to investigate any potential relationship. Regional studies for a few cultivars have also been conducted. Luedeling et al. (2009a) evaluated future winter chill conditions, measuring chill hours, for the Arabian Peninsula and included an assessment of the likelihood of various crops in meeting chilling requirements. Similarly, Measham et al. (2014) inferred future chill suitability of two cherry cultivars at six Australian sites.

Many of these studies tend to have different primary focuses. For instance, the assessment of cultivar chilling requirements (Measham et al. 2014), an evaluation of different chilling models (Chmielewski et al. 2012; Sunley et al. 2006) and the construction of temperature records (Luedeling et al. 2009a). Notwithstanding, these aims have genuine scientific merit, studies which focus on the delivery of cultivar-specific winter chill information across a range of crop types under climate change, including appropriate projection uncertainty, are not available in the literature.

This study aims to provide an example of the construction of a cultivar-specific climate impact assessment of future winter chill. To do so, a methodology was outlined to evaluate climate projection information concisely but include appropriate climate projection uncertainty. Building on this, a cultivar-specific method incorporating four levels of risk to account for different grower risk appetites was developed and combined with the climate projection method to demonstrate where and when future risk will likely appear. Site-based assessments of winter chill accumulation were conducted to provide baseline information and to illustrate the juxtaposition to a cultivar-specific approach. These aims were achieved using seven fruit and nut cultivars and at 16 Australian growing districts.

By applying a cultivar-specific approach, actionable outcomes, such as climate adaptation planning to change cultivar, crop type or location, can be more easily determined. Equally important is information regarding limited or nil impact to encourage investment and take advantage of possible opportunities. For instance, a single site may have one crop type which will require adaptive management to remain profitable yet may also cultivate another which will not be impacted by climate change.

In order to conduct these analyses, a chill model needed to be selected and a set of estimated chill requirements compiled using the units of the chill model. The Dynamic model (Erez et al. 1990; Fishman et al. 1987a) has gained significant support through both its mechanistic design (Darbyshire et al. 2011; Luedeling and Brown 2011) and stable performance in experiments across a range of crop types (Miranda et al. 2013; Perez et al. 2008; Ruiz et al. 2007; Zhang and Taylor 2011). As such, the Dynamic model, which calculates winter chill in units of chill portions, was selected for use. A review of literature was conducted to identify chilling requirements reported in chill portions and seven cultivars representing seven different crop types which are currently grown in Australia were selected.

Results were interpreted using three future time periods. These represent adaptation relevant for on-farm management (2030), a more strategic adaptation horizon for growers, industry and policy makers (2050) and a longer term adaptation horizon for transformational change and longer term industry and policy directions (2090).

2 Methods & materials

Based on a review of the literature, chilling requirements measured in chill portions for seven cultivars which are currently grown in Australia were identified (Table 1). For cultivars with a range of chilling requirements reported, the maximum was selected.
Table 1

Crop, cultivar and associated chilling requirement (CR) measured in chill portions used in this study

Crop

Cultivar

CR range

CR selected

Source of CR

Almond

‘Nonpareil’

22–23

23

(Ramírez et al. 2010)

Apple

‘Golden delicious’

50

 

(Erez 2000)

52–61

61

(Finetto 2014)

Apricot

‘Orange red’

55–64

 

(Campoy et al. 2012)

55–67

67

(Viti et al. 2010)

Cherry

‘Lapins’

35

 

(Erez 2000)

66

66

(Palasciano and Gaeta 2012)

Peach

‘O’Henry’

63

63

(Miranda et al. 2013)

Pistachio

‘Sirora’

59–62

62

(Zhang and Taylor 2011)

Walnut

‘Chandler’

45–50

 

(Pope 2015)

70–72

72

(Luedeling et al. 2009b)

16 sites representing Australia’s main nut, pome and stone fruit growing districts (Fig. 1 and Table 2) were used to evaluate historical and future chill conditions for the seven cultivars. Table 2 includes a list of major crop types grown in each district.
Fig. 1

Sites used to evaluate historical and projected winter chill

Table 2

Sites used to evaluate historical and projected winter chill and main crops currently cultivated. Elevation is metres above sea level

Location

Latitude

Longitude

Elevation

Crops cultivated

Batlow

−35.52

148.14

826

Apple

Griffith

−34.28

146.07

133

Walnut

Orange

−33.28

149.1

869

Apple, cherry

Young

−34.28

148.32

487

Apple, cherry

Applethorpe

−28.62

151.95

875

Apple

Lenswood

−34.94

138.79

553

Apple, cherry

Renmark

−34.13

140.72

24

Apricot, peach, almond

Huonville

−43.01

147.06

39

Apple, cherry

Spreyton

−41.22

146.35

14

Apple, cherry

Swansea

−42.06

148.06

7

Walnut

Mildura

−34.2

142.14

57

Almond, pistachio

Swan Hill

−35.27

143.46

67

Apricot, peach, almond

Tatura

−36.44

145.27

116

Apple, cherry, walnut, apricot, peach

Yarra Valley

−37.84

145.68

177

Apple, cherry

Donnybrook

−33.58

115.83

70

Apple, cherry, apricot, peach

Manjimup

−34.18

116.07

232

Apple, cherry, apricot, peach, walnut

2.1 Historical and projected winter chill

Winter chill was calculated in chill portions using the Dynamic chill model (Erez et al. 1990; Fishman et al. 1987a) following equations in Darbyshire et al. (2011). Historical daily maximum and minimum temperatures were sourced from gridded data (0.05° by 0.05°) produced from the Australian Bureau of Meteorology’s high quality weather station network (Jones et al. 2009) as used in similar studies (Darbyshire et al. 2011; Webb et al. 2012). 30 years centred on 1995 (1981–2010) were used to represent historical conditions.

Projections of future temperature conditions were sourced from CSIRO and Bureau of Meteorology (2015) using data from Coupled Model Intercomparison Project 5. 20-year periods centred on 2030, 2050 and 2090 were used. In constructing future projections of climate, two Representative Concentration Pathways (RCPs) (Moss et al. 2010) were selected, RCP4.5 and RCP8.5.

The RCP4.5 scenario represents an intermediate greenhouse gas emissions pathway with strong mitigation of emissions in the second half of the century. RCP8.5 is a high emissions scenario and includes little mitigation of emissions.

Of the 40 global climate models (GCMs) available, 14 were rejected as they have been found to perform poorly for Australia (Moise et al. 2015). The remaining 26 GCMs were considered for inclusion. A minimum set of GCMs which account for the majority of variability among all GCMs was selected using the Australian Climate Futures approach (Clarke et al. 2011; Whetton et al. 2012) which categorises and then ranks GCM output for climate variables of interest. Two models, MIROC and CanESM2, were found to be representative of the range of GCMs (cool and warm, respectively) for temperature in the regions included in this study (‘Southern Australia’ and ‘Central Slopes’ in CSIRO and Bureau of Meteorology (2015)).

‘Best’ and ‘worse’ case scenario were defined with the best-case option resulting from data produced from MIROC (‘cool’ GCM) forced with RCP4.5 (‘medium’ emissions) and the worse-case from CanESM2 (‘warm’ model) forced with RCP8.5 (‘high’ emissions). Results are discussed in terms of these best and worse-case scenarios, noting that the range across these two scenarios is used to represent the range of uncertainty in the results.

To downscale the projection data both spatially (1.5° to 0.05° grids) and temporally (monthly to daily), the projected maximum and minimum temperature change for each grid cell were applied to the 30 year historical dataset centred on 1995 (Jones et al. 2009).

The Dynamic chill model requires hourly temperature data input. Both the historical and projected daily maximum and minimum datasets were interpolated into hourly values using a sine-logarithmic algorithm (Darbyshire et al. 2011; Linvill 1990). Chill portions were accumulated for each of the 30 years in the historical and projected datasets from 1 March (Chmielewski et al. 2012; Miranda et al. 2013; Perez et al. 2008; Zhang and Taylor 2011) up until 31 August (Darbyshire et al. 2011).

2.2 Site evaluation of winter chill

Accumulated winter chill was evaluated at each of the 16 sites for the historical data and for 2030, 2050 and 2090, using the best- and worse-case scenarios. The range of historical and projected data was represented by the 10th and 90th percentiles of the data (30 years). Accumulated winter chill was presented as percentage anomaly changes from the site median of the historical dataset. This allowed for consideration of the shift from historical conditions and to compare sites on the same scale.

2.3 Evaluation of cultivar-specific winter chill suitability

A statistical approach regarding risk was applied to evaluate future suitability of winter chill for each cultivar. For the historical data and each projected year within the best and worse-case scenarios, 30 years of chill accumulation representing natural variability were produced. Some of these years may meet required cultivar chilling requirements and others may not. To interpret this risk, a grading system was developed (Table 3). This was based on the proportion of the 30 years which reached or superseded the cultivar threshold chilling requirement as detailed in Table 3.
Table 3

Grading system to evaluate risk of not meeting chilling requirements

Risk category

Statistic

Risk interpretation

Low

<10th percentile

90–100 % chance chill will be satisfied (9 to 10 out of 10 years chance chill threshold will be met )

Medium

10-50th percentile

50–90 % chance chill will be satisfied (5 to 9 out of 10 years chance chill threshold will be met )

Medium-high

50-90th percentile

10–50 % chance chill will be will be satisfied (1 to 5 out of 10 years chill threshold will be me)

High

>90 percentile

0–10 % chance chill will be satisfied (0 to 1 out of 10 years chill threshold will be met )

Each of the four risk categories were allocated a colour (Fig. 2). The results from the historical and projected best- and worse-case scenarios were assigned a risk rating and colour according to Table 3 and Fig. 2. For the projection data, the range in the results between the best and worse-case scenarios for each projection year was represented by a colour-and-hash system. The worse-case scenario determined the background risk category colour and the best-case scenario determined the hashing colour (Fig. 3). A solid colour indicates that both best- and worse-case scenarios fall in the same risk category.
Fig. 2

Colours used to represent the four risk categories (low, medium, medium-high and high) as described statistically in Table 3

Fig. 3

Key to interpret projection results. The worse-case scenario risk category determined the background colour and the best-case scenario determined the colour of the hashing

3 Results

3.1 Site evaluation of winter chill

For each of the 16 sites, historical and projected changes to winter chill were compiled. All sites showed a decrease in winter chill accumulation in the future, with greater declines likely further into the future (Table 4). Four sites are displayed in Fig. 4 to illustrate the results with the remaining sites included in the Supplementary Materials.
Table 4

Site based accumulated winter chill measured in chill portions. The range represents the 10th to 90th percentile of the data

Site

Historical

2030

2050

2090

Worse-case

Best-case

Worse-case

Best-case

Worse-case

Best-case

Applethorpe

66.5–82.3

55.8–65.1

60.9–71

45–54.6

55.7–67

19.4–32.3

52.3–63.2

Batlow

98.2–110.2

90.9–105.1

93.4–105.5

85.9–97.1

93–104.4

70.6–81.6

86.3–99.8

Young

79.6–89.3

70.8–81.8

71.6–84.3

63–73.5

69.4–82.7

41.3–51.8

64.8–77.2

Donnybrook

52.9–66.6

38.4–56.1

44.5–60.6

30.6–47.4

41.7–59.5

11.9–23.9

37.1–56.3

Lenswood

86.9–102.9

77.5–91.3

79.9–96.8

63.9–78.7

79.2–96.1

41.6–55.1

74.6–89.2

Manjimup

63.7–79.6

54.2–69.5

57.5–73.6

42.8–62

55.8–73.8

19.9–35.9

52.2–68.1

Huonville

102.5–115.7

92–105.9

96.1–112.8

86.3–100.7

94–109.7

51.3–67.7

90.4–105.8

Spreyton

91.1–105.3

80.3–93.5

83.1–97.7

71.6–85.2

82.5–97.9

41.7–56.9

74.3–90.7

Tatura

76.8–90.1

70.2–80.2

70.2–82.4

61.5–72.1

68.3–81.9

38.1–50.9

63.2–74.9

Yarra Valley

89.7–103.8

81–94.2

81.7–96.8

76–85.9

80–96.5

50.1–62.9

77.9–89.9

Mildura

56.6–70.0

42.9–56.6

46.4–60.9

34.6–46.6

42.4–57.7

12.8–24.8

39.5–52.7

Griffith

62.1–75.3

53.6–67.3

54.7–67.7

45.8–58.2

51.5–65.5

21.2–34

47.7–60.3

Swan Hill

63.7–76.6

53.4–65.9

58.1–69.9

44.6–58.3

53.8–67.2

20.4–34.5

47.3–62.3

Renmark

51.9–68.3

40–53.6

44.3–58.6

29.8–42.5

41.8–55.6

12.3–22

36.6–50.3

Orange

87.3–99.3

81.7–92.5

82.9–93.8

73.5–85.3

82.2–93.8

52.6–65.7

76.8–88

Swansea

87.0–101.7

73.4–87

77.9–90.6

66.5–81.3

75.6–88.3

26.7–42.6

69.5–84.9

Fig. 4

Projected chill conditions for four sites. The black bars and grey shading represent the historical range, blue and red bars represent best and worse-case scenarios for each projected time period, respectively. Numbers at the range of the projection is the chill portion accumulation at the 10th and 90th percentiles

Considering the sites in Fig. 4, Swansea and Applethorpe show the greatest percentage decline, relative to each site’s median. However, the actual chill accumulation at Swansea is much higher than at Applethorpe. For instance according to the worse-case scenario and by 2090, Swansea is expected to accumulate 26.7–46.2 chill portions compared with 19.4–32.3 chill portions at Applethorpe. Equally, Batlow and Donnybrook demonstrate lower percentage decline of chill into the future, relative to each site’s median. Again, considering the actual chill accumulation, a much higher accumulation is expected for Batlow in 2090 than for Donnybrook; 70.6–81.6 compared with 11.9–23.9 chill portions according to the worse-case scenario.

Although the data presented in Fig. 4 and in the Supplementary Materials is informative regarding changes in total winter chill accumulation, interpretation into on-the-ground action is difficult. The changes in chill accumulation are not related to cultivar threshold chilling requirements, with the key question remaining, do these declines have potential production impacts and what action, if any, should be taken?

3.2 Evaluation of cultivar-specific winter chill suitability

Each of the seven cultivars were assessed in terms of ability to meet chilling requirements historically and into the future at each of the 16 sites (Fig. 5). Using this approach, the risk of meeting chilling requirements, the uncertainty associated with climate projections and the relative risk between sites and cultivars can be viewed.
Fig. 5

Cultivar-specific risk assessment of meeting chilling requirements at each of the 16 sites. The background colour represents the risk category according to the worse-case scenario and hashing relates to the risk category according to the best-case scenario. Green is low risk, yellow is medium risk, orange is high-medium risk and red is high risk (also see Table 3, Figs. 2 and 3)

Using ‘Nonpareil’ almond as an example to interpret the results, it is likely (90–100 % chance) that this cultivar will satisfy chilling requirements at all sites until 2050. By 2090, Swan Hill, a current almond production area, recorded low (90 to 100 % of meeting chilling requirements; green hashing in Fig. 5) to medium risk (50–90 % chance of meeting chilling requirements; yellow background in Fig. 5). Another almond production area, Renmark, showed a greater range of potential risk by 2090 ranging from low to high risk (red background with green hashing in Fig. 5).

Using another example, ‘Golden Delicious’ apple indicated some risk of insufficient chill accumulation at several sites. For instance, Donnybrook, a current apple growing region, historically had a 10–50 % chance of meeting chilling requirements. By 2030 and through to 2090, both the best- and worse-case scenarios indicated a 0–10 % chance of meeting chilling requirements. However, at another current growing region, Batlow, there was a 90–100 % chance of meeting chilling requirements historically and through to 2090.

‘Sirora’ pistachio provides a final example. Currently, most Australian production occurs in and around Mildura. The results indicated that historically there is already some risk with a 50–90 % chance of chill being satisfied. By 2030 this risk increased to 0–10 % chance of chill being satisfied for the worse-case scenario and 10–50 % chance according to the best-case scenario. For 2050 and 2090 both best and worse-case scenarios indicated a high risk (0–10 % chance of chill being satisfied).

4 Discussion

In this study, the climate projection method used minimised the number of future climate scenarios and reduced results to be described in terms of best- and worse-case scenarios and the range between these scenarios. This was applied for two reasons, to simplify interpretation of results and to reduce computation load. Such an approach has been recognised within the climate projection science community as important for impact assessments (Whetton et al. 2012). Using the simplified projection information, the site-based results (Table 4 and Fig. 4) still lacked relevance regarding management or policy decisions. Indeed, considering the results, it may be tempting to follow either the best or worse-case scenario, however use of the range of results is required to include appropriate projection uncertainty (Jun et al. 2008). Of course, site-based information can be presented differently, for instance as a range across potential future pathways (e.g. Darbyshire et al. 2013) but interpretation in terms of potential production impacts is still missing.

Here, analyses were conducted as cultivar-specific to create more actionable results. Within the approach, climate projection uncertainty was represented via a colour-and-hash system with the worse-case scenario represented by the background colour and the best-case scenario dictating the colour of hashing. This allowed for an easy to interpret view of the range of uncertainty with solid colours showing high confidence and greater colour divergence showing lower confidence.

The presentation of cultivar-specific results illustrated where and when potential risk will arise which was informative in terms of planning and instigating climate adaptation strategies. For instance, even though apple production areas are likely to face substantial declines in chill accumulation (Table 4 and Supplementary Materials) nil impact is expected for ‘Golden Delicious’ for many production areas up until 2090 (Young, Yarra Valley, Lenswood, Spreyton, Orange, Huonville), while Tatura recorded some risk by 2050 and low risk was found for Batlow out to 2090 (Fig. 5). Higher risk was found at other apple production areas (Donnybrook, Manjimup and Applethorpe). This highlights that for a single cultivar, adaptation options and timelines will differ between sites and a single country-wide strategy will not suffice.

‘Nonpareil’ almond recorded low risk at all sites, including current production areas, up until 2090. Conversely, ‘Chandler’ walnut indicated that historically some current production sites have a medium-high risk (10–50 % chance chill will be will be satisfied) with risk increasing into the future (Griffith, Manjimup). Clearly, adaptation strategies for these two crop types need to be different.

In this study we introduced a risk classification (Fig. 2 and Table 3) to allow growers, industry and policy makers to determine their own risk appetite. Low risk was set as per Luedeling et al. (2009c). In describing this metric the authors noted that “We expect this metric to be more useful for tree crop growers … since it incorporates the economic need for an orchard operation to produce good yields in most years (90%)…”. As some managers may have different risk appetites, in this study three additional risk levels were used; medium (50–90 %), medium-high (10–50 %) and high (10 %).

The results for ‘Sirora’ pistachio provide a good example of different risk capacity. Historically, the major pistachio growing region in Australia (Mildura) recorded a medium risk of insufficient chill accumulation (50–90 % chance chill will be satisfied). Despite this level of risk, the pistachio industry has expanded in recent years (Pistachio Growers’ Association 2015) and issue alerts to manage low winter chill conditions (Joyce 2015), namely through the application of winter oil (Seif El-Yazal and Rady 2012). Hence, for this industry, a medium level of risk appears to be manageable using incremental adaptation strategies. However, greater risk in the future needs to be appreciated, especially given the long production horizon of pistachio trees (20–25 years) and may require more transformational adaptation such as moving to a lower chill cultivar or a different crop type.

In developing adaptation strategies, growers, industry and policy-makers require information at different time-scales. Considering growers and industry, 2030 is a suitable time horizon for management purposes as typically most modern orchards switch to newer cultivars every 10–20 years. Using the methodology outlined in this study, growers could ascertain the likelihood of chilling requirements being met over this time-frame and incorporate personal risk appetite which may include incremental adaptation strategies like the application of dormancy breaking products or more strategic adaptation such as transitioning to lower chill cultivars or crop types. 2050 represents a more strategic timeframe. The results for 2050 can be used to direct growth industries into particular regions, for instance transition of walnut production away from Manjimup and concentration in the low risk production zones of Swansea, Griffith and Tatura. Equally, through provision of cross-industry results, growers in Manjimup can identify other cultivars or crop types to grow which may be lower risk as the industry shifts. Finally, for policy purposes, 2090 provides for long-term positioning of industries. Many of the traditional growing regions will likely have a high risk of insufficient chill for several cultivars by 2090. Repositioning the industries towards more reliable locations which currently cultivate perennial fruit and nuts, or to other non-traditional growing regions, may be required. This could mean delivering policy options that encourage grower investment at low risk sites, development of appropriate infrastructure (e.g. processing plants and cool stores), upskilling work-forces in non-traditional growing districts and/or investing in breeding programs to develop low chill cultivars. Crop production is intrinsically linked to climate, and presenting climate information with the associated policy support for change presents opportunities rather than limitations (Selvaraju et al. 2011). Of course sufficient winter chill is only one climate component required for successful production and other site suitability criteria also need to be considered in adaptation strategies.

A single cultivar from seven crop types was selected for analyses to demonstrate the methodology of creating cultivar-specific climate impact estimates. The crops and cultivars selected were restricted by the availability of chilling requirement information in chill portions. For instance, Cripps Pink apple is economically important for Australia, however an evaluation of chilling requirements in chill portions has not been conducted and it was therefore not included. Furthermore, individual cultivar results should not be considered representative of whole crop types. Some crops have a wide range of chilling requirements, for instance Viti et al. (2010) reported a range of chilling requirements, 37–79 chill portions, for apricot in Italy.

Additionally, often the chilling requirement is reported as a range or the results from different studies disagree (Table 1). Across these divergences, the maximum reported value was used in this analysis leading to likely conservative results. The results are additionally sensitive to the threshold requirements used. This was demonstrated at Applethorpe for ‘Lapins’ cherry (66CP) and ‘Orange Red’ apricot (67CP). The historical results found low risk for ‘Lapins’ cherry and medium risk for ‘Orange Red’ apricot despite the small difference in chilling requirement. This is a result of these chilling requirements falling around the 10th percentile of the historical data (66.5 CP; Table 4). Thus, when making management and policy decisions, chilling requirement thresholds and site results should be evaluated against these potential sensitivities.

In this study it was assumed that the Dynamic chill model and the reported chilling requirements adequately represented the physiological process. Some of the studies used in this analysis (e.g. Ramírez et al. 2010) determined chilling requirements statistically from flowering records and may not represent biological thresholds. More robust analyses (e.g. Zhang and Taylor 2011) which use progressive sampling of bud wood which are forced with warm temperatures in controlled environments likely provide more reliable results. However, these controlled environment studies still face challenges in determining chilling requirements, as demonstrated by Campoy et al. (2012), who used the same experimental protocols in Spain and South Africa and found different chilling requirements for the same apricot cultivar. For example, ‘Orange Red’ was found to require approximately 64 chill portions in Spain and 55 chill portions in South Africa. Many issues with determining chilling requirements experimentally have been summarised by Dennis (2003) with knowledge gaps in understanding the dormancy breaking process and potential phenotypic plasticity (Campoy et al. 2011) meaning some uncertainty in precise cultivar chilling requirements persists.

Compared with other models of chill, the Dynamic chill model has been found to best represent flowering conditions by a number of studies (Miranda et al. 2013; Perez et al. 2008; Ruiz et al. 2007; Zhang and Taylor 2011). However, this model is not mechanistic and the varied chilling requirements reported in Table 1 and by Campoy et al. (2012) demonstrate there may be deficiencies in the model. Typically, as applied in this study, the beginning of chill accumulation is determined by the model, dependent on weather conditions and not cultivar characteristics. The model should be re-parameterised for new cultivars as suggested by the authors (Fishman et al. 1987b) however this has not been successfully accomplished to date. This inability to modify the start of chill accumulation may also contribute to the range of chilling requirements reported. Another aspect of the model form is the continuous chill accumulation through the season. Statistical analyses of flowering in relation to winter chill and spring heat indicate that the windows of chill accumulation and periods of meaningful chill accumulation are different to the continuous chill accumulation according to the Dynamic model (Guo et al. 2013, 2015; Luedeling et al. 2013). Without better physiological understanding it is unlikely this misalignment between chill model structure and plant response will be resolved.

This study presented predictions of the risk of meeting chilling requirements for seven different crops across nuts and pome and stone fruits. By focusing on cultivar chilling requirements rather than potential shifts at sites, the relevance of declining winter chill was interpreted as well as the suitability of crop and cultivar mixes into the future which provides information for adaptation actions. Risk was explicitly included to allow growers, industry and policy-makers to determine their own risk appetite. The approach used was dependent on the determination of chilling requirements in chill portions and the reliability of these requirements to represent yield-relevant thresholds. Continued efforts to establish chilling requirements using existing methods will allow for industry planning whilst greater research into the dormancy breaking process will provide greater certainty of potential impacts.

Acknowledgments

The authors thank Mark O’Connell , Jennifer Whitney and Walnuts Australia for advice on cultivar selection and Australian growing regions. Funding for this research was provided by the Australian Department of Agriculture and Water Resources.

Supplementary material

10584_2016_1692_Fig6_ESM.gif (108 kb)
ESM 1

Fig. 1 Projected chill conditions for 12 sites not included in Fig. 4. Black bars represent the historical range, blue and red bars represent best and worse-case scenarios for each project time period, respectively. Numbers across the range is the chill portion accumulation for the 10th and 90th percentiles. (GIF 108 kb)

10584_2016_1692_MOESM1_ESM.tiff (35 mb)
High Resolution Image (TIFF 35888 kb)
10584_2016_1692_Fig7_ESM.gif (102 kb)
ESM 2

(GIF 102 kb)

10584_2016_1692_MOESM2_ESM.tiff (35 mb)
High Resolution Image (TIFF 35888 kb)

Copyright information

© Springer Science+Business Media Dordrecht 2016

Authors and Affiliations

  1. 1.Faculty of Veterinary and Agricultural SciencesUniversity of MelbourneMelbourneAustralia
  2. 2.Tasmanian Institute of AgricultureUniversity of TasmaniaTasmaniaAustralia
  3. 3.Department of Economic Development, Jobs, Transport and ResourcesVictorian GovernmentTaturaAustralia

Personalised recommendations