A Scientist’s Perspective on Sustainable Scientific Software

Abstract

Software underpins most of our daily activities, from banking and finance to interactions with the internet, to weather forecasts and reports. Software also impacts individuals, groups, and societies through policy implementation, since information for decision and policy making is frequently derived from software ranging from climate and weather models to financial forecasting systems. One way to gauge the extent to which specific software needs to be sustainable, accessible, and transparent essentially hinges on the degree to which scientific analysis software, models, and model output are used to help inform and guide policy. Climate models and related scientific results are perhaps the most obvious example of the need for sustainable and transparent software, due in part to the public forum in which the results are scrutinized and the implications on environmental management policy. Without almost ubiquitous adoption of best practices for scientific software development, maintenance, and use, the credibility of scientific results and of ourselves as scientists is substantially at risk; sustainable and transparent research processes are thus at the heart of maintaining and increasing our collective reputations. [The authors want to make clear that, by using climate models as an example of software with policy impacts, we are not claiming that these models, are written with little “best practices” in mind, nor that they are inherently unsustainable as software.]