PS Webinar Series: Machine Learning, Satellites, and Crops – The (very near and exciting) future of space-based plant biology
Abstract - This seminar will discuss the terabytes of unused satellite data that observe the natural world, yet have not been widely used for field biology, in the context of agriculture. Using a basic, low-resolution analysis of experiments containing 850,000 plant populations, and fundamentally simple machine learning algorithms, I develop models that predict substantial variation in the yield and fitness traits of eleven major crop species. This analysis is extended to develop testable hypotheses: not the usual ‘black box’ models of high accuracy and low utility. This analysis highlights how much biology, and plant biology in particular, could achieve using satellite data and unified environment models.
Biography - Saul conducts interdisciplinary research across evolutionary biology, bioinformatics, and evolutionary demography. His current projects include the prediction of crop yield and fitness, the testing of inclusive fitness and life history models of ageing in humans and model organisms. Saul formerly held a position at the CSIRO, and currently works at both the ANU Research School of Biology and the newly-formed Biological Data Science Institute (http://bdsi.anu.edu.au/), applying machine learning models to wheat genome and large-scale experimental data to predict flowering and yield. Saul has several sideline projects including meerkat geriatrics, resolving the cause of late-life mortality plateaus, and convincing the world’s oldest people of their own nullibiety.