1 Reproducible pipeline

…This places an obligation on all creators of software to program in such a way that the computations can be understood and trusted. This obligation I label the Prime Directive.

  • John Chambers (Software for Data Analysis: Programming with R)

2 Adaptive management

  • Ideally, we can rerun all decision support tools over and over again
  • Must be MUCH easier the second, third, etc. time
  • If there is an incremental change in input data, then we shouldn’t need to redo the same effort as the first time around.

3 Why a Discrete Event Simulator is necessary?

  • There is much fanfare about using Rmarkdown as a tool for reproducibility (i.e, Rmarkdown = “code, output, narrative”)

  • But this allows exact reproducibility

    • Everything will be run exactly as before
  • What happens if the goal is to run it again, but with slight changes to the input data?

  • The Rmarkdown approach can deal with this to a certain extent, but it will be clunky

4 Adding novelty

  • What happens if you want to add something new, but keep it mostly the same
  • You have a good vegetation model and a good caribou model, but you want to add a wolf model
  • With Rmarkdown, you would have to edit the original file
  • With SpaDES, you can use the other modules by calling them… no need to know how they work
  • You add your own wolf model, looking at outputs of the vegetation model and caribou model
  • Using the model metadata from those other modules

5 Deploying for policy and management

  • Need the reusable work flow, and reproducible science …
  • But with a GUI interface
  • shine function is a start
  • We created a demo version of a generic web interface for models
  • It is currently broken because we decided to make a complete overhaul, but it is coming :)
  • A non-scientist can ask questions of a system of modules without knowing which modules are necessary

6 Caching

  • For the above to work for, say, a national question, it can’t run the model on the fly
  • Could have massive data requirements, long simulation times, even on supercomputers
  • Caching allows for near instant results from very complex systems
?cache

wolf model shows a caching use case

7 Data-to-decisions

  • This is the current vision for SpaDES
  • We are looking for people who are interested in being part of this adventure

Bringing the best science, data, models into the hands of policy makers in real time, on their phones