Markus Keller

Jointly modeling distinct phenological tasks in a shared multi-task RNN overcomes severe temporal label scarcity, decisively outperforming established domain-specific scientific process models.

Cold hardiness predictions can now be accurately made in new regions with minimal data, thanks to a framework that learns transferable latent representations.

Achieve 60% better phenology and 40% better cold hardiness prediction by using a neural network to dynamically parameterize biophysical models.