The fundamental flaws of chill models
Why they don’t work well in South Africa. By Anna Mouton.
Chill models matter — their outputs have real-life implications. Growers in areas with mild winters depend on chill models for the management of rest-breaking treatments. An incorrect estimate of chill can lead to either insufficient or overzealous rest-breaking treatment. Chill estimates also inform cultivar selection when establishing new or replacing old orchards. Planting an unsuitable cultivar is one of the more expensive mistakes a grower can make.
The problem South African growers face is that chill models don’t work so well in our climate. Local growers mostly rely on the Utah chill-unit model, or the daily-positive Utah chill model, better known as the Infruitec model, which is an adaptation of the original Utah model. But our orchards are not in Utah. How does this impact the accuracy of the models? Dr Nigel Cook, plant physiologist and horticultural consultant, explains.
Start at the beginning
The Utah chill-unit model was published by Richardson and co-workers in 1974. Many South Africans call it the Richardson model, and refer to Utah chill units as Richardson chill units. The Utah model assigns chill units based on average hourly temperatures according to the values in table 1.
Table 1: Chill-unit values for different temperature ranges
|Temperature range in °C
|Number of chill units per hour
Units are added up for every hour in the day. The total for each day is added to a cumulative total. Figure 1 shows the cumulative number of chill units over a four-month period, from 1 May to 31 August, for two different years. The solid lines are chill units calculated using the Utah model, and the dotted lines are chill units calculated using the Infruitec model.
When days are warm, the total number of units per day will be negative. A succession of warm days results in an increasingly negative cumulative number of Utah chill units. This is typical of late autumn or early winter in areas with mild winters. When the weather eventually cools, the total number of Utah chill units per day becomes positive — this is an inflection point. This can be seen at the end of May for year 2 in figure 1.
“What Richardson said, is that the turning point becomes your start date,” says Cook. “That’s where the plant starts recording the chilling. But the start date varies from year to year. Richardson says you have to calibrate every weather station every year to find its start date.”
There is a convention in South Africa to start counting chill units on 1 May. This usually results in a cumulative total that is lower than the true total, because of the initial accumulation of negative chill. “You can’t arbitrarily make the start day 1 May,” stresses Cook. “By reporting the data like that, you can be out by as much as 200 units.”
The risk of a false start
The Utah model works well in cold climates, where the onset of dormancy is rapid. “Dormancy has an entrance and an exit,” says Cook. “In cold climates, the entrance is rapid, but in warmer areas, the entrance process is problematic. There is no correlation between chill units and the plant’s dormancy level in the entrance phase.”
Stone- and pome-fruit trees undergo a series of changes to become dormant, but there is no practical way of telling when the process is complete. In warm areas, the entrance to dormancy can take many weeks, even most of winter. Meanwhile, the trees don’t register chill — they only start to accumulate chill once they are fully dormant.
The Utah model works in cold climates because it happens quickly — it’s safe to estimate when to start counting chill units based on the inflection point of a chart. Cook doesn’t think this assumption necessarily holds for South African conditions. “It looks like the Utah and the daily-positive models are very poor at quantifying the accumulation of chilling in the entrance phase, especially in warmer areas.”
Not knowing the correct date on which to start counting chill units can lead to either an over- or an underestimation of the amount of chill accumulated, depending on the region and the particular season.
Chill units will be inflated if counting continues after the trees have broken their rest. Chill is frequently reported as a cumulative total for the period 1 May to 31 August. This is clearly not appropriate for early-flowering trees, some of which may already be in active growth in July.
Factoring in heat
One of the improvements in the Utah chill-unit model over earlier models was that it provided for negative chill. However, later research suggests that negative chill has a greater impact at temperatures above 21 °C, and that the Utah model should be adjusted to reflect this.
“That’s not included in any of the models,” says Cook. “We don’t have Utah version 2, we only have Utah version 1.”
Cook speculates that short periods of very high temperatures in the middle of winter may lift plants out of dormancy. Affected trees may have to start their dormancy cycle from scratch, thereby losing any accumulated chill. The lack of an easy way to measure the progression of dormancy makes it hard to identify trees that have spent part of winter tossing and turning.
Accounting for heat will become increasingly important thanks to climate change. “Everyone asks, how is global warming going to impact us?” says Cook. “But the first question to answer is, are we warmer than normal? There’s no doubt that we’re warmer — frighteningly warmer.”
Temperature data for Elgin demonstrate that all but one of the last ten years had less chill accumulation than the long-term average. The past three years were the warmest in the past thirty years, and last year was the warmest year on record.
“This used to be a 700–800 chilling-hours region,” states Cook. “Now it’s a 400-hour region. The big question is, what chill model do we use, or do we create a new model?”
Bonus: Applying chill models to real-life data
The solid lines in the chart represent cumulative chill units calculated with the Utah model. In year 1, the model has been correctly calibrated, so units only begin to accumulate from the inflection point, which happens to be 1 May. However, in year 2, chill units are also accumulated from 1 May, but the inflection point only occurs at the end of May.
The result is an accumulation of negative chill units during May of year 2, which incorrectly reduces the reported total.
The dotted lines in the chart represent cumulative chill units calculated with the daily-positive model. It can be seen that this model gives a higher number — this is because a negative total number of units for any specific day is recorded as zero.
For year 1, the difference between the results from the Utah and the daily-positive models is 47 units. For year 2, the difference is 120 units. The main reason for the larger disparity in year 2 is the incorrect inclusion of negative chill units in the total for the Utah model, because chill accumulation did not begin at the inflection point.
Thanks to Dr Nigel Cook for contributing the data set for this chart.