AgScore™: testing climate forecasting models for agriculture

AgScore™: the first diagnostic tool that tests seasonal climate models for their relevance to agriculture

What climate model ‘problem’ does AgScore™ solve?

The demand for accurate and reliable seasonal climate forecasts is an unmet need in Australian agriculture. Having a clearer picture of the upcoming season can help bolster farm preparedness and resilience, allowing farmers to better manage their business and reduce climate-related risks.

The most widely used seasonal forecasts are produced by the Bureau of Meteorology and several international agencies such as the UK Met Office that use complex numerical models of the earth’s atmospheric and oceanic processes.

However, increasingly, government and industry climate services are providing seasonal outlooks via multiple seasonal climate models. While using multiple forecasts can improve confidence and accuracy in seasonal outlooks, there has been limited oversight into their comparative performance for a given farmer’s decision or location.

The ‘skill’ of seasonal climate forecasts can be difficult to assess for predicting agricultural metrics such crop yield or pasture growth. Therefore improving these forecasts is often done with little consideration of an industry for which they probably have the most potential value.

If climate modellers had a way of evaluating their seasonal climate forecasts that were more streamlined and agriculture-focussed, they could better target future research towards key areas that would directly improve forecasts for agriculture.

The AgScore™ interactive report card

How can AgScore™ help agricultural industries?

Our AgScore™ tool provides a robust approach for benchmarking the performance of seasonal climate models using agriculturally relevant metrics. It is in the form of a service to climate and agricultural scientists and forecasting agencies to help them understand how useful particular forecast systems might be for predicting productivity for a range of sectors.

For instance, rather than just answering ‘how well does this climate model predict rainfall?’, AgScore™ can answer ‘how well does this climate model predict wheat yield?’

A sample AgScore™ assessment for a user-submitted forecast

The tool draws on the computing power of CSIRO’s unique scientific workflow platform, Senaps, to combine climate forecasts and weather observations with crop and pasture simulation models such as APSIM. It then produces succinct and interactive report cards for a given seasonal climate forecast dataset.

Essentially, AgScore™ does the heavy lifting by using pre-defined test locations and simulation configurations so users can get critical feedback quickly without prior expertise in agricultural modelling and high-performance computing.

With AgScore™ we hope to change how investment and research effort is directed towards model development and services derived from seasonal climate models by providing forecast assessments that are timely and comparable moving forward. It will also aid communicators and service providers by identifying which models they may consider useful and of sufficient quality to include in their seasonal outlooks.

How can I find out more about AgScore™?

Watch team member AgScore™ Javi Navarro present on where the science is up to.

‘AgScore: an online tool to assess climate models’ yield prediction skill’ by Javi Navarro (9:44 minutes)

Contact Dr Pat Mitchell to discuss using AgScore™ to address issues around seasonal forecast performance and value.

Dr Pat Mitchell
  • Pat uses his expertise in plant biology, climate science and systems modelling to improve climate decisions in agricultural and forest systems. He leads CSIRO's AgScore, Farming Forecaster and Pasture API work, among other projects.
  • Javi was originally a Telecommunications Engineer and is now an agricultural systems analyst.