Frosts, food and satellites

Developing a national frost monitoring system

Frost on grass

What is the problem with frosts and growing food?

Frosts commonly occur across much of southern Australia and don’t pose much of a problem for cereal farmers during the coldest months. It is late frosts – ones that occur in Spring when frost-sensitive flowers and fruit are forming – that can cause widespread damage to crops overnight. Over a typical year, frosts cost the Australian agricultural industry over a billion dollars.

Once a crop is planted, there is little that can be done to avoid late frost damage. There is still value though in knowing when, where and how severely a frost has happened as it can help growers decide whether cutting now for hay or continuing through to grain harvest will maximise their profits.

Despite the value of such frost information, there is currently no means of remotely detecting frost events both at fine scales and across large areas.

A national frost monitoring system would also allow the grains industry to gain regional and state-wide insights into the impact of frost on grain supplies as the growing season progresses. This would be equally true for other agricultural commodities that are disrupted by frost. As the frost event record builds over time, this information may also assist grain growers in building frost risk maps of their properties and will provide invaluable training data for building future frost prediction systems.

What did we do to help solve the problem?

A map of south eastern Australia

Imagery showing minimum overnight land surface temperature of southern Australia during the winter of 2016. White squares are where clouds have been detected and masked-out.

We tested the idea that land surface temperature data derived from a weather satellite can provide a useful indicator of frost occurrence. The Japan Meteorological Agency’s Himawari satellite provides thermal imagery at 2 km resolution, every 10 minutes. From this we worked on developing a near-real-time, national frost event detection system for continuous information on frost occurrence, duration and severity within 12 hours of it happening.

As the system relies on underlying temperature data, the aim was it would  simultaneously function as a national land surface temperature monitoring system.

Why was this a challenge?

We addressed two specific challenges. The first was how to detect night-time cloud cover. At night clouds can be the same temperature as frosted ground and may falsely indicate a frost event. During the day, the data can be filtered based on cloud brightness. This can’t be done at night of course so we have had to develop an alternative means of cloud filtering.

The second challenge was finding a way to convert the relatively ‘coarse’ 2 km land surface temperature data to a resolution that is useful for individual paddocks to something less than 100 m. To do this, we used field observations, topography and wind speed to figure out which parts of the landscape cool faster than others.

Where did the work get up to?

We developed a prototype system for detecting frost occurrence and applied it across south-eastern Australia for two past seasons (2016 and 2019). We tested the results and further research focussed on maximising their accuracy. Once established, we plan to build these models into an automated, national monitoring system.

How can we help industries – such as horticulture – right now?

We can now monitor land surface temperatures from satellite every 10 minutes over night. This provides unprecedented intelligence on sub-paddock-scale cooling processes including the extent, severity and duration of frost events. Industry specific information can be generated that describes current-season progression of growing season temperatures such as chilling, and that notify of the extent and severity of shock frost events. This can help frost-sensitive industries such as cherries, vegetables, tree nuts, grapes, apples and berries.

How can I find out more?

Watch Randall Donohue, frosts project leader, talk about the science behind the work.

Frosts, food and satellites: the science – Randall Donohue (9:54 minutes)

If you’d like to find out more still or would like to view some of our prototype frost occurrence data for 2019, you can get in touch with Randall.

  • Randall researches vegetation and remote sensing such as crop yield prediction, nation-wide and global estimation of pasture biomass, and remotely monitoring vegetation condition.