Impact of more rigorous observation quality control on GraphDOP performance

Ascertain the effect of applying more rigorous quality control to the training dataset—such as removing degraded AMSU-A channels—on GraphDOP’s forecast performance across variables and lead times.

Background

The authors describe that despite attempts to remove low-quality observations, many remained in the dataset, including degraded AMSU-A channels. They explicitly state that the impact of more rigorous quality control is not yet determined, highlighting an open issue regarding data curation standards and their influence on model skill.