The challenge
The agricultural sector is one of the largest users of seasonal climate forecasts (SCFs). A skilful SCF can offer confidence and benefits to grain growers, providing valuable information to support on-farm decision-making.
However, seasonal climate models (and the forecasts derived from them) often lack information on the value of the forecast to agricultural users. As a result, the adoption of forecast information is hampered by perceptions of poor forecast quality and low relevance.
As a result, growers are left asking questions around the quality and usefulness of seasonal forecasts. This is especially the case when the climate model used in preparing the forecast has not been translated into agriculturally relevant information.
A stark recent example is climate forecasts that predicted a super El Nino for the summer of 2023-24 and, therefore, exceptionally hot and dry conditions. Persistent rain and storm systems over that summer subsequently caused financial and production losses, especially for sectors that maintain livestock.
The response
In 2019, GRDC invested in assessments of SCF performance and their value to the grain sector. The project was headed by CSIRO and had two main components:
- a side-by-side evaluation of different forecast systems from a suite of international forecasting agencies for the major agricultural regions across Australia; and
- further examination of a subset of these forecast systems in terms of their ability to predict wheat yield.
In all, 12 different forecast systems were assessed, including 10 dynamical global climate models and two statistical models. About 25 years of past forecasts were analysed, which allowed forecast performance to be assessed against observed climate data. This resulted in measures of accuracy, reliability and skill for the different forecast systems.
Included in the analysis were the ACCESS-S1 and S2 models used by the Bureau of Meteorology (BoM).
In the second part of the project a new software service – AgScore™ – was used to assess forecasts in terms of farm productivity, particularly wheat yield. The AgScore™ service was used to test five different forecast datasets, including BoM’s ACCESS-S1 and the more recently released ACCESS-S2 models.
The AgScore™ service ingests forecast datasets for a select group of locations and automatically runs simulations of wheat crop growth (using the Agricultural Production Systems sIMulator model) and performs verification analyses. The results for a particular forecast dataset are then provided to the user as a report card, providing a summary of the performance of the data from an agricultural perspective.
The impact
The study did not identify a single model with superior skill in all locations and seasons.
For grain growing regions, there are several models that provide skill for southern and eastern regions during winter and spring. While the western region has limited skill across the winter growing season, BoM’s model (the most widely used seasonal outlook in Australia) ranked highly among the top-performing models.
Forecasts that translated into yield-based predictions were found to provide skill during mid to late stages of the winter wheat growing season (July onwards). This is likely to offer benefit to in-season management decisions around fertilising, marketing and logistics.
In terms of seasonal rainfall forecasts, gross margins varied little between scenarios that did and did not incorporate seasonal rainfall considerations. Overall, the models had limited skill to forecast rainfall to inform winter crop sowing decisions.
As such, an economic analysis of this project found little gross margin benefit to growers from changing decisions based on SCFs. Rather, the project recommended taking a consensus approach to presenting seasonal climate information from multiple models in one information service.
The economic analysis was performed by Dr Marit Kragt, Professor David Pannell and Dr Hue Vuong at the Centre for Agricultural Economics and Development.
More information: view the Delivering impact case studies.