Is there a good way to document the process of tuning and validation? There are certain large-scale global patterns that Charlie had in mind when originally developing COBALT. As we start applying it in regional contexts, there will be other specific features we will be trying to capture that may require some fiddling with parameters, but we want to make sure that in doing so we don't accidentally "break" things that we don't happen to be focusing on at the moment but may still be very important.
Some input parameters are constrained pretty tightly by measurements or theory, but others are less mechanistic and have a lot of leeway as tuning knobs. For the latter category in particular, it would be nice to document the "story" behind their implementation, and what things future developers should keep in mind when tampering with them.
(I'm picturing an annotated generic_COBALT document, with lots of asides discussing the sorts of things we've been covering in our weekly discussions, maybe?)
Is there a good way to document the process of tuning and validation? There are certain large-scale global patterns that Charlie had in mind when originally developing COBALT. As we start applying it in regional contexts, there will be other specific features we will be trying to capture that may require some fiddling with parameters, but we want to make sure that in doing so we don't accidentally "break" things that we don't happen to be focusing on at the moment but may still be very important.
Some input parameters are constrained pretty tightly by measurements or theory, but others are less mechanistic and have a lot of leeway as tuning knobs. For the latter category in particular, it would be nice to document the "story" behind their implementation, and what things future developers should keep in mind when tampering with them.
(I'm picturing an annotated generic_COBALT document, with lots of asides discussing the sorts of things we've been covering in our weekly discussions, maybe?)