THE CLIMATE KELPIE BLOG: Using past records to better understand frost risk

Posted by BCG on 21st September 2021

Some years are frostier than others and there seem to be frostier and less frosty decades.

Why is historical minimum temperature data important?

In the last decade there has been an explosion of NRM and on-farm weather stations and many of these are now networked. The lower cost and availability of temperature loggers such as Tiny Tags and iButtons provide further local information. This is valuable information, but unfortunately it is all recent. Even 10 years of measurement is a small sample to try and understand year to year variability. In the future there may be clever ways to connect the on-farm weather data with the long-term data. At the moment we rely on BoM sites for long term records

These long-term records are the only way to:

  1. rank the frostiness of a season like 2021,
  2. make long-term comparisons of the frostiness of regions,
  3. check the impact of climate drivers like negative IOD,
  4. check on how frost risk is changing over decades and
  5. analyse ‘optimal’ or ‘safe’ flowering windows. 

Measuring and ranking the frostiness of a season or a region is somewhat similar to what we do with deciles for rainfall. Dryland farmers would agree that a statement like “the growing season was a decile 3 season” tells us more than “it was a dry season, not the worst season, but dry”. Of course, deciles for the growing season don’t cover everything. The statement of a decile 3 season might be followed with “a late break but the rain came just as the crops were flowering”. Deciles as a way to rank the rainfall of the season are only possible because we have reliable long-term records of rainfall. We have fewer temperature records and even when we do have good records, it is more difficult to provide a frostiness decile. Candidates for ranking years include the number of frosts, date of last frost, coldest night, accumulated nights or ‘frost sum’ under a threshold of 2 degrees or 0 degrees and simulated damage from frost using a crop model like APSIM (YieldProphet). Over the coming months different ways to measure and rank frostiness will be explored so any suggestions will be gratefully received.

What do you mean by Optimal Flowering Windows?

In addition to ranking a given season, historical information on frost risk is important to identify the optimum flowering window. This is a “window” of days in spring that balances the risks of frost, heat and moisture stress.  At recent frost workshops organised by MSF and GRDC we asked farmers and agronomists: “About what date in spring would you drive past a crop in head, think it was too early and pushing against frost risk? What date would you think a crop was a bit late to be in head and prone to heat and moisture stress?” Farmers and agronomists in the room were able to give us a consensus based on their experience, local trials and because they had participated in discussions at recent GRDC updates, for example Stabilising the flowering time of wheat. Farmers and agronomists at the workshops were quick to point out that hitting the optimum flowering window with different varieties and sowing times isn’t always easy, especially if a season is warmer or cooler than usual. Further discussion covered how the targeted flowering window might shift earlier or later depending on the risk appetite of the farmer and the frostiness of the paddock. It is obviously a mistake to determine the Optimal Flowering Window after a single year, or even a three-year project; we need to use long term climate records.

How can farmers access local information on frost and heat?

The APSIM model behind Yield Prophet captures the dynamics of crop development and how critical stages interact with the risks of heat and frost but also rainfall, crop water use, nitrogen supply and the other driver of modelling biomass, radiation.   

A great tool for farmers and agronomists to freely access climate data is the CliMate App. A recent GroundCover article ‘Dodging frost a numbers game‘ gives a good example of using CliMate as one input into managing frost risk. A user of CliMate app can set the percent chance of reaching a cold or hot threshold. For example the 10% chance of being warmer than zero degrees and the 30% chance of being cooler than 30 degrees. Because this is only based on frost and heat it is better understood as a user defined ‘safe flowering window’.  Unless farming in a highly frost prone area, the Optimal Flowering Window is usually earlier than the Safe Flowering Window.  This is because earlier flowering minimises the risk of moisture stress as well as heat stress. What this means is that an over-emphasis on flowering after the last frost or a very low chance of the last frost can be costly insurance because yields are reduced by moisture and heat stress.

What is the risk of frost and heat at Birchip?

Long term temperature data is not available for Birchip but data is calculated from the nearest recording stations with correction for altitude etc. This is the same data that is used in CliMate and YieldProphet for Birchip.

Figures 1 and 2 show the risk of being colder than zero degrees and warmer than 30 degrees for Birchip. The x axis of both graphs is the day of year and the Y axis is frost risk presented as a percentage. The difference is that Figure 1 is the risk for a single day whereas Figure 2 is the risk at least one day in seven.  The very sensitive stage of wheat to frost is at least 7 days. The chance of one frosty night over 7 days or one hot day over the week is much higher than the chance of a frost or heat on a single day.   

Figure 1. Chance of a single day measured at Birchip being zero degrees or colder (blue line) and 30 degrees or warmer (red line).
Figure 2. The same thresholds as Fig 1 but chance of at least one day in seven meeting the threshold.

One way to think about this is that you are crossing a six-lane highway with no safety island in the middle and you want to know the best hour of the day to do this. One analyst tells me the risk of being hit by a vehicle in one lane, another tells me the risk of being hit on at least one lane. They will both tell me how the risk changes by hour over the day, but the risk of being hit in at least one lane is a better measure of the risk for my problem.  The chance of one event over the period gives a more realistic measure of risk. This should not be interpreted as a criticism of CliMate, it is a great tool and the how cold/how hot component was not designed specifically for wheat, it can be applied to many agricultural applications.

The analysis in YieldProphet assumes that frost has impacts depending on the severity of the minimum temperature and the stage of the crop. Taking the analogy of crossing the six-lane highway even further, YieldProphet is taking into account how quickly you cross the road. Late in spring in a warm location wheat will move through the sensitive stages much more quickly than earlier in spring.  

Is the risk of frost and heat changing?

Farmers and agronomists with long experience in Victorian cropping regions have observed more frost damage and in many cases, earlier heat events.  There is a paradox whereby in a warming world, we seem to be getting more frost, or at least noticing more frost damage. Figure 3 presents the heat and frost risk in the same way as Figure 2 (at least one event in seven days).  The data for Figure 2 comes from 61 years (1960 to 2020). Figure 3 compares the risk for the last 20 years (2001 to 2020) with the risk in the previous 20 years (1981-2000).  It is clear that days over 30 degrees C are coming earlier; the red line showing risk in recent decades is shifted above and to the left of the brown line. The risk of frost is also much greater in the last 20 years. It is hard to say whether this is a permanent shift but it supports local experience that recent decades have had damaging frosts at a critical period. This increase in August frosts may be contributing to stem frosts. 

Figure 3. The chance of at least one day in seven measured at Birchip being zero degrees or colder and 30 degrees or warmer in the last 20 years from 2001 to 2020 (blue line and red line) or the previous 20 years from 1981 to 2000 (grey line and brown line). Because weather stations can be moved, we need to be careful interpreting a single station, however there are many sites across southern Australia that show an increase in frosts in recent decades).

What to expect in 2021 

The Bureau of Meteorology has declared a negative Indian Ocean Dipole (IOD). A negative IOD is associated with warmer waters to the northwest of Australia and wetter conditions. The increased cloud cover and wetter conditions are generally less conducive to frost. However, analysis conducted by SARDI Climate Applications for key sites in South Australia shows a minor decrease in the number of frosts over the season (not consistent at all sites) and even less impact on late spring frosts. Unfortunately, late spring frosts remain a random low frequency but high consequence event. The negative IOD suggests an increased chance (but no guarantee) of wetter conditions, but only a small shift in the odds towards slightly lower frost risk.

The last negative IOD was 2016 and this was a wet spring with relatively low frost incidence. There were some damaging frosts however especially in the Clare region. The experienced agronomist Mick Faulkner made the important point that frost in a drought adds insult to injury, but the main damage to the farm enterprise is the low yield due to the lack of rain.  The cost of a frost in a good season is much greater. Another problem of a frost in a wet spring is cutting hay is more challenging and finding a profitable market can also be difficult.   

For more information contact:

The SARDI Climate Applications Team

Peter Hayman,

Leave a Reply