Latent risks and dangers in the state of climate model software development
Shawn M. Freeman, Thomas L. Clune, Robert W. Burns · 2010
As the importance of climate modeling dramatically increases due to concerns about global climate change, the quality of model software will come under ever more intense scrutiny. Software defects that alter model predictions could negatively impact important climate policy decisions, and even if detected in time could reduce policy makers' confidence in the science. Effective protection against software defense against errors requires the anticipation of vulnerabilities for every facet of this important work. Unfortunately, existing climate model implementations and associated software engineering practices are inadequate to defend properly against some concerns over defects and untested parameter regimes. Organizations that wish to avoid scrutiny, whether deserved or not, should ensure that appropriate software engineering practices are established with all due haste. In this paper we examine some of the inadequacies of existing software development methodologies and suggest strategies for fundamentally changing the established culture.