Dr Tim Weaver1, Dr Michael Bange2, Mr Chris Teague2, Mr Brint Gardner3, Ms Lauren Stevens3, Dr Stuart Gordon4
1CSIRO, Myall Vale, Australia, 2CSD Ltd, Wee Waa, Australia, 3CSIRO, Clayton, Australia, 4CSIRO, Werribee, Australia
Biography:
Dr Tim Weaver is a senior research scientist at Myall Vale, NSW with CSIRO and involved in research exploiting the genetic, environment and management interactions in cotton systems. Tim leads a number of projects with industry partners and start-up companies investigating machine learning models for cotton yield and micronaire prediction as well as near infrared spectroscopy for real-time in-situ analysis of cotton leaf and petiole for nitrogen management.
Abstract:
Environment and crop management can play an important role in determining upland cotton fiber quality. One of the important quality parameters is fiber micronaire, which is an indirect measure of fiber linear density (fineness) and maturity. Predicting micronaire in-season would allow growers to make an informed and strategic decision in their management to maintain/improve fiber quality, for example, at harvest time (picking). Prior to picking, defoliants are applied at maturity, and the timing is one trigger where micronaire could be optimised through optimal harvest aid applications. It also has the potential to assist growers with marketing their cotton. A cotton yield prediction model, BARRY (Biometric Agronomy for Realising Representative Yield), was developed by CSD Ltd and CSIRO using crop data and XGBoost (extreme gradient boosting package in R) collected from CSD Ltd. ambassador grower sites. BARRY was developed to equip growers with a tool to make informed in-crop decisions to maintain/optimise their yield (bales/ha). The same process was utilised to develop a predictive micronaire tool (i.e. an RStudio Shiny app. like BARRY) using the XGBoost package. The model was shown to produce a strong relationship with climate, location, sowing date and micronaire with an r2 of 0.845 and an RMSE of 0.18. The main factors that influenced micronaire were seed imbibition date, latitude, number of hot days and average temperature from seed imbibition to defoliation. These variables support previous findings of temperature impacting micronaire, and how latitude reflected local practices and cultivar choices.