Do-a-thon: best practice of sensitivity analyses

do-a-thon
aarhus-2019
validation

#1

Many effort was and is being spent on getting the code and data of energy system models right and open. Complementary to these ongoing activities, this do-a-thon will focus on a subsequent step of modeling exercises: sensitivity analyses.
Most of us will already have performed a bunch of sensitivity analyses in order to test the robustness of our obtained results. This is of crucial importance as we don’t know for sure how sensitive our obtained model results are against different data and model assumptions. In literature you often find variations of single parameters seemingly selected on the basis of the modeler´s experience. As our models are growing continuously in complexity, the question remains if this approach is still sound enough. Are there more sophisticated tools to perfom sensitivity analyses? How can we evaluate the appropriateness of our chosen approach? Centred around those questions, we will:

  • Exchange about and collect our experiences with and ideas about sensitivity analyses (read in papers or done by ourselves)
  • Identify possible options for improvement (and maybe scan literature for this in addition)
  • We may also manage to test some of those identified improvements to evaluate the differences in results and conclusions (People can bring their own test cases for this or we will make use of some test case already available for several open source model frmaeworks.)

Based on these steps our main output will be:
A guideline on best practice for sensitivity analyses on the openmod wiki and potentially writing a review paper