Replicating computational atmospheric science

Scientific claims can be verified by reproducing or replicating the experiment.  Replication involves new authors writing new code following the original method.  Reproduction involves new authors running the original code.  Doing nothing means that scientific claims remain unverified.

This talk is based on the work of Lorena Barba and was also inspired by Philip Stark, Carl Boettiger and Jonathan Peelle.  I present my own experiences trying to replicate the advection test of Schär et al. 2002, and the computational reproducibility study of the aerodynamics of flying snakes conducted by Mesnard and Barba 2016.