tomalrussell
commented
3 years ago Just to add few variations on the user story:
- I would like to run a transport model over ~5 timesteps with the same parameter values and input datasets so that I can simulate the transport system into the future under a consistent set of assumptions
- I would like to run a transport model over the same set of ~5 timesteps with around 4-6 variations on parameters and input datasets to conduct a scenario analysis
- I would like to run a water supply model over ~100 ensemble members x ~25 years x ~3 climate scenarios x ~2 demand scenarios x ~2 population scenarios, so that I can simulate the system under uncertainty and accounting for variability within a climate/hydrology ensemble.
BethanPerkins
roseDickinson
I've got a model to run over multiple steps, with mostly the same configuration, varying one or two inputs.
It would be handy to be able to duplicate a workflow step, then change the relevant values in the web user interface. This would reduce the amount of manual entry when working with a non-default set of inputs that should mostly stay the same between steps.
Another approach to the same problem would be a (possibly sequential, possibly parallel) iterator, which could take a series of values and apply them to model datasets // inputs. In the sequential version, e.g. running a model over multiple timesteps, it may be necessary to make outputs from one step in the iteration available in the next. In the parallel version, e.g. running a model over multiple scenarios or with parameters varying to conduct a sensitivity analysis, there may be no data dependency between each of the iteration steps, but there would likely be an aggregation step after the iterator (a "reduce" step in map-reduce terms).
Another approach would be an API or config-file interface to creating workflows, where repetitive confiugration could be scripted or copy-pasted and quickly edited by hand.