The 2020 rice harvest in the Murrumbidgee Irrigation Area (MIA) marks an important milestone for PhD student Allister Clarke, providing an opportunity to put his PhD research to the test.
Mr Clarke’s research through the Australian Research Council (ARC) Industrial Transformation Training Centre for Functional Grains aims to develop models to allow for rice to be graded on the basis of quality at the point of delivery.
Since late March rice crops from across the Riverina have been harvested and delivered to receival stands across the region.
In contrast to other cereals like wheat, where growers can find out quality parameters like protein and therefore the pay grade at delivery, the rice sample taken from the truck needs to be stored, dried and milled before the quality can be appraised.
“After delivery growers have to wait a few more months before they know the quality appraisals of their crops,” Mr Clarke said.
“The quality appraisal score represents whole grain yield (WGY) which is the percentage of rice grains that remain un-broken during the milling process.
“Without an ability to physically test WGY at the delivery stand, pre-milling segregation is also limited and growers are left in the dark regarding the quality of their crop and the final payment they’ll receive.”
Mr Clarke’s research aims to generate variety specific predictive models that will allow the grading of a load, based on WGY, at the delivery point.
Mr Clarke’s developing the models through application of sophisticated machine learning algorithms to a dataset of historical rice production records which incorporate the key factors that influence rice WGY during crop development.
The 2020 rice harvest was an opportunity to validate preliminary models on the new post-harvest data being captured.
A pilot study was designed to focus on production of the most widely grown rice variety, Reiziq, in the MIA.
Mr Clarke said the dataset construction was made up of three key stages.
“The first step involved collation of historical industry data, where grower recorded crop management surveys were joined to the delivery stand data and the resulting quality appraisals of that crop,” he said.
“Next, rice phenology data, as reported by the NSW Department of Primary Industries (NSW DPI), was used to approximate the timings of critical development stages of each crop.
“Finally, the rice crop phenology and location data was used to extract the raw weather data from the closest weather station, allowing the climatic conditions experienced during each crop development stage to be added to the dataset.”
Initial model development from the pilot study dataset involves a cycle of ‘Feature Selection’, ‘Algorithm Selection and Parameter Optimisation’ and ‘Validation of Model Results’.
“This process is critical in the identification of the most impactful variables, selection of the algorithm that best fits the dataset, and validation of model accuracy when predicting the new results,” said Mr Clarke.
“Repetitions of the cycle allow for further optimisation and improvement of prediction accuracy and in turn the development of the best possible model.”
Mr Clarke said preliminary validation results are showing strong prediction capabilities, particularly after application of feature selection.
“Testing is ongoing to further enhance prediction accuracy and identify best practice methods, particularly for data pre-processing, feature selection and algorithm optimisation,” he said.
“It’s anticipated that this will allow streamlining of future model development for each active variety on enriched datasets which include records from all growing regions and increased resolution of weather station data.”
The final results and findings of the pilot study are expected to be released to coincide with the end of the 2020 harvest and grower meetings to be held by the Rice Extension and the SunRice Grower Services teams across the Riverina.
Mr Clarke hopes his research will deliver benefits to growers through better understanding of the factors affecting WGY but also through improved storage and milling.
“One of the key aims of the project is to develop the ability to segregate rice at the delivery point on the basis of milling potential, which would allow processing operations to be planned to maximise WGY.
“The price for whole grain is double that for cracked and broken grains, so that means if we can reduce the percentage of broken grains, through better management of storage and milling, it could have a big impact on the value of the crop, and hopefully lead to better prices for growers,” Mr Clarke said.