Gro Intelligence has added an Australian Wheat Yield Model to its suite of high-protein wheat yield forecasting applications for monitoring global wheat crops.
Using Gro’s Australian Wheat Yield Model model, standard users can get an early read on Australia’s high-protein wheat yields at the sub-state level. For example, confirmation of a bumper crop in Australia can buffer the crop losses elsewhere and allow users to make better procurement decisions.
Australia is forecasted to harvest a bumper high-protein wheat crop starting in October 2021. That could help offset the disappointing high protein spring wheat harvests recorded in Canada, the US, and Russia.
Recently, the Australian Bureau of Agricultural and Resource Economics and Sciences’ increased its domestic wheat production estimate to 32.6 million tonnes, 17% higher than its June estimate but below last year’s bumper crop. Typically, Australia produces 17-33 million tons of wheat annually; yields often vary considerably in drought years.
As Australian white and amber wheat represents 11.5% of global exports and an even bigger portion of high-protein wheat exports, its wheat plays an important role in global wheat supply.
This means that having insight into Australian wheat crop conditions is essential for building a global wheat supply and demand balance sheet and for understanding how trade flows will react. With this in mind, we are incorporating our new Australian Wheat Yield Model into Gro’s automated balance sheets, which are used by procurement, sovereign food security, shipping, and asset management market participants to manage supply risk, forecast price, and predict trade flows.
Today, Australia is the Asian market’s closest producer of high-protein white wheat, a type of wheat commonly used for noodle making. Australian wheat is also gaining a larger following among food and beverage procurement buyers in China.
Our new Australian Wheat Yield Model uses the Australian Bureau of Statistics (ABS) as its ground truthed data source. And like Gro’s other wheat yield forecasting applications, our Australian Wheat Yield Model:
Factors in weather and climate variables
Uses a satellite imagery-based crop mask
Drills down to the sub-state level
With a mean absolute percentage error (MAPE) of 4.5% (on November 1), Gro’s Australia wheat yield model yield estimate was significantly more accurate than the USDA’s estimate, which had a MAPE of 7.0% (November WASDE report).
Gro’s Australian Wheat Yield Model’s accuracy aligns with Gro’s other wheat yield models, despite a limited history of ABS’ sub-state data, which reaches back 10 years. To address this mismatch, the Australian Wheat Yield Model’s machine-learning algorithm, XGBoost, tests more data points and improves its accuracy as it goes. The inputs to our Australian Wheat Yield Model include:
Normalized Differential Vegetation Index (NDVI)
Gro Drought Index (GDI)
Rainfall (NASA GPM)
Evapotranspiration Anomaly (ETA)
Land Surface Temperature (LST)
Soil Moisture (SMOS)
Gro’s wheat yield models cover over 50% of the world in wheat production. Our other wheat models include:
To learn more about the Australian Wheat Yield Model or any of our other Forecasting models, contact our sales team email@example.com for a demo of the Gro Platform.