Reproducible Geocomputation: an open or shut case?

2019-09-18T02:40:55Z (GMT) by Mark Gahegan
This submission tackles the issue of reusability and replication of experiments and other forms of spatial analysis from the perspective of eScience or eResearch—a community that has been deliberately grappling with this issue for many years—and applies it to our geocomputational and GIScience methods and models.
Most of the effort that routinely goes into our analysis and modelling efforts is not documented in any consistent or accessible way. It might be in part referable from scripts and code we use, and from the ‘methods’ section of papers that we write. But by and large much of the background, context and process of research remain unrecorded. This makes reproducibility hard (or intractable) and reusability nigh on impossible. Funding agencies worldwide are beginning to insist that we do better, and this pressure is now beginning to be felt in geography and the spatial sciences as well as in
psychology and biology (where the science ‘reproducibility crisis’ began).
We describe several approaches to improve reusability and reproducibility for geocomputational work, and point to best practice from other disciplines as appropriate. (The presentation will elaborate on each of these headings and give examples relevant to geocomputation and spatial analysis.