dingaaling
commented
7 months ago - clim-recal project timeline?
- ibicus packaging and documentation took a full 3 month focus
- ibicus designed not to download dataset but to make it easy to apply/compare; their engagement also shows there is demand for the BA datasets
- happy if you use it (open license) and happy to open up to R users
- there's many components to the evaluation framework that you have to manually walk through as a user
- in convos with UKCH, conferences, Met Office users primarily; coding & climate science abilities high
- "status quo" question: people are used to using their org's method
- depends on level of operational constraints
- interest in developing a scorecard in the long term (point user to evaluations that may be important to them)
- health researchers and public agencies have high demand but its misunderstood how much work is needed before it can be integrated or applied with health models
- climate variables of particular interest are heat and humidity (and rainfall)
- clim-recal apply ibicus to expand the BC methods clim-recal currently applies via python
- are methods all inheriting from the debiaser class? yes
- methods implemented currently: cmethods package, QDM, SDM (pycat)
- RB: start with methods in ibicus as starting point and stretch goal is to use the methods again in R to compare (reticulate); wrapper would be preferrable as ibicus does not plan to implement in R
- FS: more than happy for you to use the methods, and give feedback if things aren't working correctly
- FS: also happy to bring your other methods into debiaser class
- can clim-recal incorporate ibicus visualisations into our assessment process
- FS: happy to for you to integrate
- are ibicus team members open to helping with assessment review/decisions
- not much time for ibicus software maintenance
- web app visualisation
- interactive more focused on population side, but as part of that we need to visualise the predictions anyway
- scorecard and front end app don't need to be together but they have the same goal of making the outputs more accessible
- ways of working:
- 3 month software sprints; uncertainty quantification and cross validation (starting in march)
- really interested in stakeholder insights