Improving canola harvest management decisions with remote sensing

Abstract

Key findings • Using advanced predictive modelling approaches, we have successfully used both satellite and drone-based multispectral imagery to predict canola maturity parameters to a high degree of accuracy (seed colour change, root mean squared error – RMSE of <10%). • Simple normalised difference vegetation index (NDVI) based regression modelling was unable to account for location- and variety-induced variation resulting in significantly higher prediction errors than when using more advanced predictive modelling approaches. • Significant potential exists for using this technology in a canola windrow-timingdecision support tool that would overcome the many challenges of current industry practice. However, additional investigation is required to validate the performance of this technology application across multiple seasons and further progress modelling approaches.