Pathways to resolving the issues in land models seen in PLUMBER2
Gab Abramowitz
1. Land models appear to be underperforming, as shown in the PLUMBER2 land model comparison experiment, and we are trying to develop tools to help understand what the problem is.
2. In this analysis of PLUMBER2 model output, we try to identify the meteorological conditions in which land models do poorly relative to out-of-sample machine learning estimates.
3. A relatively small subset of meteorological conditions are responsible poor performance for many of the models, suggesting parametrisation pathways in these models to target for improvement.
4. Land modellers, climate modellers, ecologists
5. Land model, machine learning, benchmarking
Please use this thread for discussion about this talk.