Workshop: Replicability and Reproducibility in Geospatial Research

About the workshop:

Replicability and reproducibility (R&R) have always been core requirements of scientific research. Recently, cases of failure to replicate previously published findings have received widespread public attention. As research grows more complex and increasingly reliant on data and software, concerns about replicability will grow rather than diminish. For example, different software packages may produce different results even when the same technique of spatial analysis is applied to the same data or analysis results cannot be reproduced by the same software due to the lack of proper metadata or provenance documenting the spatial processing and parameters used. Geospatial researchers may need to be especially concerned about replicability, such as when results from one geographic area fail to be replicated in other geographic areas.

This small and focused workshop addressed the following questions and related issues:

What forms of failure to replicate exist in the geospatial sciences? Can a formal framework be devised?
In what areas of geospatial research is the danger of non-replicability most severe?
What mechanisms can be used to avoid or minimize the danger of such failures?
Do we expect the results of model calibrations to be constant over space and if not, what are the implications for spatial analysis?
How should R&R be incorporated into the design and implementation of future spatial software?
How should students be made aware of these issues?
What follow-on activities might draw greater attention to these issues?
 

Read the full summary of the workshop

Update (11/10/2020): Forum Articles on Reproducibility and Replicability, a product of the SPARC R&R Workshop, has now been published in the Annals of the American Association of Geographers. Read more.