Articles | Volume 20, issue 9
Hydrol. Earth Syst. Sci., 20, 3739–3743, 2016

Special issue: HESS Opinions 2016

Hydrol. Earth Syst. Sci., 20, 3739–3743, 2016

Opinion article 12 Sep 2016

Opinion article | 12 Sep 2016

HESS Opinions: Repeatable research: what hydrologists can learn from the Duke cancer research scandal

Michael N. Fienen1 and Mark Bakker2 Michael N. Fienen and Mark Bakker
  • 1US Geological Survey Wisconsin Water Science Center, Middleton, Wisconsin, USA
  • 2Water Resources Section, Faculty of Civil Engineering and Geosciences, Delft University of Technology, Delft, the Netherlands

Abstract. In the past decade, difficulties encountered in reproducing the results of a cancer study at Duke University resulted in a scandal and an investigation which concluded that tools used for data management, analysis, and modeling were inappropriate for the documentation of the study, let alone the reproduction of the results. New protocols were developed which require that data analysis and modeling be carried out with scripts that can be used to reproduce the results and are a record of all decisions and interpretations made during an analysis or a modeling effort. In the hydrological sciences, we face similar challenges and need to develop similar standards for transparency and repeatability of results. A promising route is to start making use of open-source languages (such as R and Python) to write scripts and to use collaborative coding environments (such as Git) to share our codes for inspection and use by the hydrological community. An important side-benefit to adopting such protocols is consistency and efficiency among collaborators.

Special issue
Short summary
In the field of cancer research, a scandal occurred at Duke University in the USA in which independent researchers were unable to repeat analysis performed by another group. This led to recommendations by the university and governing organizations to motivate the use of scripting languages to enhance repeatability of research. The hydrology community can easily adopt similar protocols to enhance the integrity of our data analysis and modeling.