At the 10th International Summer School on AI and Big Data, Lilly-belle Sweet will talk about Advancing agricultural modelling with machine learning through interdisciplinary coordination.

Talk: Advancing agricultural modelling with machine learning through interdisciplinary coordination

Agricultural models are vital tools for the study of the complex effects of climate change on global and regional agricultural systems. Over the last years, the increasing power of machine learning (ML) methods and the availability of data and computation power has opened up new opportunities in this domain. ML methods have been used for applications such as yield forecasting, data gap-filling for process-based model input, downscaling crop model simulations and for the analysis of complex biophysical processes via interpretable or explainable ML. However, using ML for scientific research requires awareness of the underlying assumptions of these methods and their potential pitfalls.

Interdisciplinary, community-wide efforts are needed to identify and establish domain-specific best practices for model development, evaluation and usage. These collaborations can also help to drive research progress in ML by (for example) curating and maintaining benchmark datasets that correspond to active areas of ML research and reflect the needs of a wide range of stakeholders. By facilitating dialogue and knowledge exchange across fields, communities and countries, more robust and trustworthy models can be developed that can then be used to tackle the many challenges facing our global food system.

Bio

Lily-belle Sweet is a PhD student at the Helmholtz Centre for Environmental Research – UFZ, where she uses interpretable and explainable machine learning to identify compound meteorological drivers of agricultural yield failure. Prior to her studies, she worked as a data scientist in industry for several years in multiple sectors, after obtaining a masters’ degree in theoretical physics. In late 2022, she co-founded the AgMIP Machine Learning team (AgML), which brings together crop modellers, machine learning experts and other relevant scientists and stakeholders from around the world to conduct model intercomparison studies and create essential resources such as benchmark datasets.