Скачать или смотреть Optimising Yield Forecasts with Red-Edge Bands and Deep Learning Models

Abstract:
Accurate crop yield prediction and forecast is critical for precision agriculture, especially at field scale. In this seminar, approaches for improving crop yield prediction/forecast will be presented.

First, we will explore the potential of red-edge (RE) based vegetation indices, especially the Triple Red-Edge Index (TREI), for improving crop yield prediction. Using the three RE bands provided by Sentinel-2, the TREI outperformed traditional vegetation indices by integrating all three RE bands. The discussion will focus on why this index was developed and how it helped to improve the prediction of canola and wheat yield at the field scale across large regions in Australia.

Secondly, the performance of three deep learning models for wheat yield forecasting will be compared using structured and unstructured data. The advantages of spatial and temporal data representation will also be discussed.

Dr Dhahi Al-Shammari:
Dhahi is a postdoctoral researcher in the School of Life & Environmental Sciences, the University of Sydney. Dhahi’s research interests are in modelling (e.g. crop yield and crop type models) in space and time. Specifically, he is interested in developing machine learning algorithms in agriculture for crop yield prediction, crop type mapping, and soil carbon prediction. Dhahi has completed his master’s degree in science in agriculture at the University of New England (Armidale, NSW, Australia), and a PhD at the University of Sydney.