A/Prof. James Brinkhoff1, Brian Dunn2, Mark Groat3, Sharon McGavin1, Josh Hart2, Tina Dunn2, Alex Schultz2, Peter McDonnell3
1University of New England, 2NSW Department of Primary Industries, 3Ricegrowers Ltd. (SunRice)
Biography:
James Brinkhoff received a PhD in electronic device modelling from Macquarie University in 2005. Among other subsequent positions, he worked with Broadcom for 5 years, designing radio frequency circuits for WiFi chips for smartphones. In 2015, James moved into agriculture research, focusing on wireless sensor networks and remote sensing for irrigation optimisation at Deakin University. He joined the Applied Agricultural Remote Sensing Centre at UNE in 2019, where he leads projects including yield forecasting for macadamia growers, and developing real-time analytics for rice grower and industry applications.
Abstract:
Rice crop yield and input use efficiency optimisation relies on timely and informed decisions, particularly regarding fertiliser and water management. We deliver information including near-real time growth curves benchmarked against historical crop trajectories, aiming to support such decisions.
Growth curves were based on the biomass-sensitive NDVI (NIR-R / NIR+R) and a chlorophyll-sensitive ratio index (NIR/RE), derived from Sentinel-2 satellite imagery with revisit of 2-5 days. We created a database of curves from > 7,500 Riverina rice fields in 6 seasons (2018-2023). Poor quality observations due to cloud were removed from the time-series data, and a Savitzky-Golay smoothing filter was applied. The data were grouped into variety, region and sowing-method subsets, from which the quantiles (10, 50, 90%) were derived on 2 bases: per day-of-season, and per day before/after flowering. These growth curves were sensitive to growth stage, applied nitrogen, mid-season drain events, grain moisture dry-down, and are predictive of crop yield.
In the 2023-2024 rice season, imagery for each of >2000 fields was processed daily. Through online dashboards, growers could access the interactive growth curves of their crops, compared to the historical quantiles of similar crops. The dashboards also included weather, predictions of growth stages, nitrogen status and grain moisture dry-down. Growers and their advisors used the provided information to benchmark their crops, investigate causes of abnormal trajectories, and to guide in-crop nitrogen application and irrigation.
Challenges remain however, including imperfect cloud masks, data gaps and satellite view angle effects. We suggest possible avenues to ameliorate these issues.