Enhancing Rice Provenance and Quality Prediction

In Partnership With:

SunRice

Charles Sturt University

Agrifutures (Rural Industries Research & Development Corporation)

The Challenge

Rice grains can crack during processing, significantly reducing the amount and quality of product that can be sold commercially. For Australia's leading rice processor and global retailer, SunRice, this is a problem.

By collecting data on the cracking issue and sharing the findings with its network of Australian growers, SunRice hopes it can help growers select varieties less prone to breakage. This small change has the potential to be significant, representing millions of dollars to local growers.

The Solution

The project looked at how head rice yield (HRY) could be improved through the use of data and machine-learning to create sets of rules or ‘models’ that explain the hidden relationship between various environmental, crop-management, phenology factors and HRY.

During the 2021/2022 rice crop, the developed models and knowledge discovery methods have had multiple industry applications. In season ,a variation of the model was used to provide in-season forecasts of HRY based on the weather to date and long-term daily climate data for the remainder of the season. These forecasts were used by the SunRice finance, sales and operations teams to help inform budgets, storage site configurations and potential sales and purchases of whole and broken rice.

Before harvest, the models were integrated into SunRice grain elevator sample stands for real-time predictions on each incoming load. The 2022 rice harvest ran from mid-March into June, where the trained models made predictions on over 20,000 deliveries of rice.

Accurate prediction of future rice yield and quality will ultimately allow SunRice to improve industry profitability by maximising yields and segregating quality types for high-value markets. Growers will further benefit through near-real-time feedback.

The project also created software systems that enable users with no previous programming experience to build machine-learning models from that data and perform comparative analysis on historical and recent weather conditions at any location in Australia.

It also saw the development of a web-based comparative weather analysis framework that enables fast analysis of up to 18 different weather features for any location anywhere in Australia at the click of a map. This will give growers more targeted information regarding the climatic conditions experienced by their crops and help in delivering better long-term outcomes for growers.

Outcomes

This project developed a suite of digital tools to enhance product quality and provenance for whole grain rice.
A centralised data repository, allowing the development of predictive models for rice milling quality
A model-building tool for future researchers to work with the data using a graphical user interface
A location-specific weather data analysis tool called ‘CLOWD'