Understanding frost risk in a variable and changing climate

Author: Steven Crimp, David Gobbet, Dane Thomas, Shuvo Bakar, Uday Nidumolu, Peter Hayman and Garry Hopwood | Date: 26 Feb 2013

Steven Crimp, David Gobbet, Dane Thomas, Shuvo Bakar, Uday Nidumolu, Peter Hayman and Garry Hopwood, 

CSIRO Climate Adaptation Flagship

GRDC project code: MCV 00010

Keywords: climate, frost, phenology, trends

Take home messages

  • In southern regions of Australia, despite global warming, the number of spring frosts has increased and the period of frost occurrence has changed (i.e. broadened over the southern NSW, Victoria and part of South Australia and become later over WA and western parts of South Australia)
  • Frost occurrence is linked to a long-term southerly shift in position and intensification of the band of high pressure typically located over central Australia in spring.
  • Over many parts of NSW the frost season length has broadened by as much as 40 days and the mean number of consecutive frost days has increase to 5 days.
  • Field and simulation studies of wheat have shown flowering and maturity occurring at a rate of 7 days earlier per 1oC of warming. This quicker phenology combined with changes to the underlying likelihood of frost will contribute to a significant increase in the risk of frost damage.

Introduction

Current research is targeted at understanding what is driving changes in both frost frequency and occurrence across Australia’s southern regions. This is to understand how frost risk has changed in the past and may change into the future and the implications for cereal production. 

 

Using high quality climate data from the Australian Bureau of Meteorology (BOM) for the period 1960 to 2010 inclusive we have analysed changes and trends for a range of climate variables including:

  • Frost season length
  • Consecutive frost days
  • Cold wave duration
  • 10th percentile cold spells
  • 10th percentile cold days
  • 10th percentile cold nights
  • The proportion of cold nights/ days below the 10th percentile.

 

The results from these analyses highlight strong decadal and regional variation for the different measures of frost mentioned above.  Consistent increases in the number of frosts are found in both southern NSW and northern Victoria since 1970.  The spatial analysis shows much of the increase in frost frequency has occurred in August, although the date at which the last frost occurs has increased by as much as three weeks in some parts of eastern Australia (Figure 1).  Our analysis to date shows a broadening of the frost window in the east and a shift in the window to later in the year in the west (Figure 1).

Figure 1: Trends in the frost season duration (number of days) for the period August to November (1961 to 2010) based on the BOM high quality gridded minimum temperature dataset.

Figure 1: Trends in the frost season duration (number of days) for the period August to November (1961 to 2010) based on the BOM high quality gridded minimum temperature dataset.  Regions coloured in blue represent areas of increasing frost season length areas depicted in red indicate declining frost season length.  The legend values represent the change in mean length of the frost season in days per year.

 

Much of Victoria, NSW, southern WA and SA also show a mean increase of up to 6 consecutive days with minimum temperatures at or below 2oC (Figure 2).  In the eastern parts of Victoria and southern coastal regions of NSW the number of consecutive days below 2oC has declined by 2.5 to 4 days for the period 1960 to 2010.

 

Figure 2: Trends in the number of consecutive frost events (temperatures below 2oC) for the period August to November (1961 to 2010) based on the BoM high quality gridded minimum temperature dataset.

Figure 2: Trends in the number of consecutive frost events (temperatures below 2oC) for the period August to November (1961 to 2010) based on the BoM high quality gridded minimum temperature dataset.  Regions coloured in blue represent areas of increasing frost number and areas depicted in red indicate declining frost number.  The legend values represent the change in length of consecutive events per month, per year

Our research suggests that increases in number and change in the period of frost occurrence is driven by a long-term southerly shift in position and intensification of the band of high pressure typically located over central Australia in spring.  The east to west variation in effect on frost indices suggesting that the strongest driver of change i.e. the southward displacement and strengthening of subtropical high pressures is most acutely felt in the eastern states.

 

We have also examined individual frost events at six sites in Victoria, six sites in NSW and eight sites in WA using the National Centre for Environmental Prediction (NCEP) and National Centre for Atmospheric Research (NCAR) climate reanalysis data set known as Reanalysis 1. The dominant synoptic systems were identified on each day that the minimum temperature fell to 2°C or less at any of the stations.  The analysis shows that 92% of frost events in Victoria, in September to November, occurred in association with high pressures systems situated between 35S and 37.5S and between 138E to 139.5E (Figure 3).  More extreme minimum temperature events are associated with highs centred further south (i.e. 37S) and further west (i.e. 139.5E) than for less severe events (Figure 3).  Similar results were found for NSW and WA with 86% and 83% of frosts were associated with high pressures south of 32.5S and 35S respectively and highs at or around 142.5 to 147.5E and 108 to 110E, respectively.

Figure 3: Mean (a) latitudinal and b) longitudinal position of MSLP high centres for minimum temperatures below thresholds of 2°C to 0°C; 0°C to -2°C and less than -2°C for six Victorian sites for the period August to October 1960 to 2010.

Figure 3: Mean (a) latitudinal and b) longitudinal position of MSLP high centres for minimum temperatures below thresholds of 2°C to 0°C; 0°C to -2°C and less than -2°C for six Victorian sites for the period August to October 1960 to 2010.

Given that changes in frost risk have been identified by almost all of the participating farmers as a significant management issue we attempted to understand how change in frost risk are interacting with production risk across Victoria.  With the widespread adoption of no‐till seeding across northern Victoria, during which weeds are controlled before and at sowing with contact and residual herbicides, many crops are now sown by the calendar.  An increasing number of farmers begin sowing in mid‐April.  Early sown crops are at risk from frost damage during flowering, and late sown crops are at risk of dry conditions and heat shock events during the grain filling period. Farmers are finding that targeting the least risk flowering window using different crop types and varieties is not easy, particularly as the current sowing program can now extend across multiple weeks.

 

To provide contextually relevant information about frost risk to Victorian farmers modelling was undertaken to determine when frost risk corresponds with production risk.  We have examined the interaction between sowing and flowering dates for three widely used wheat varieties in northern Victoria: the early flowering AxeA, mid variety YitpiA and the late-maturing Rosella. Using APSIM (the crop model behind Yield Prophet) we planted each variety each week from April 1 (day 90) to July 15 for the period 1960 to 2011.  We then calculated a mean flowering date for each sowing date for the three varieties across the 17 locations.  The time of flowering was then graphed against the risk of frost (Figure 4).

 

For the Birchip site 90% of the frosts occurred at, or before, day 240 (August 27), 10% of frosts occurred after day 290 (October 16). If you wanted to avoid all but 10% of the frosts that may occur in any one year then planting on day 170 (June 18) for early maturing, day 160 (June 8) for medium and day 150 (May 29) for late, would achieve that end.

Figure 4: Left: Simulated flowering dates as a function of planting date (1 April = day 90; 15 July = day 197) for an early (e.g. cv ‘AxeA’), medium (e.g. cv ‘YitpiA’) and a late (e.g. cv ‘Rosella’) maturing wheat variety at a Birchip (DS) site. Right: Corresponding frost probabilities, based on the entire temperature record for the period 1950 to 2011.

Figure 4: Left: Simulated flowering dates as a function of planting date (1 April = day 90; 15 July = day 197) for an early (e.g. cv ‘AxeA’), medium (e.g. cv ‘YitpiA’) and a late (e.g. cv ‘Rosella’) maturing wheat variety at a Birchip (DS) site. Right: Corresponding frost probabilities, based on the entire temperature record for the period 1950 to 2011.

The results from the analyses show a consistent broadening of the frost window and therefore the occurrence of the 10th percentile frost risk occurs much later in the growing season.  In some instance the 10th percentile frost risk window ends 3 weeks later. As a result, later planting of the medium growing season length wheat variety appears to be an effective method to avoid frost damage in the northern Victorian region given the current trend. 

Contact details
Steven Crimp

CSIRO Ecosystem Sciences, GPO BOX 1700, Canberra ACT 2601

02 6246 4095

Steven.Crimp@csiro.au