-
Notifications
You must be signed in to change notification settings - Fork 28
Thetis meeting agenda 2020 07 24
Date/time of meeting 2020-07-24 13:00 UTC (14:00 BST)
Stephan Kramer, Matthew Piggot, Tuomas Kärnä, Mariana Clare, Joe Wallwork, Simon Warder
See call for papers. Could be an opportunity to publish a review paper on Thetis adjiont.
Sounds like a good idea, we'll revisit in September. Could also submit to another journal which has more ocean modeler audience. Matt: Think about the novelty, what's different compared to conventional ocean modelling. Add a roadmap for Thetis adjoint; what are the missing bits for operation modeling applications.
HDiv spaces no longer accept variant=equispaced. Stephan has a fix in a PR. Soon the default variant will be integral; this variant causes 3 Thetis test to fail (2 error tolerance and one solve failure). Tuomas will take a look at the failing tests.
Mariana and Stephan have been working on updated version, ready for a review soon.
In 3D, horizontal velocity lives in spaces like HDiv(RT2 x DG1). This defines a horizontal vector space iff bathymetry is constant. On a deformed mesh, the vector space follows the layers, i.e. is not strictly horizontal. To get a strictly horizontal space, we need to augment U with suitable vertical vector space. This adds more DOFs to the problem (up to 1.5x) which makes 3D HDiv solver (even more) slow. Is it possible to define a strictly horizontal HDiv velocity space on deformed mesh? Unclear, Tuomas will talk with Colin and Lawrence.
Tuomas is interested in doing a 2D model for the Baltic and North Sea, for which the model should be on the sphere. Previous tests suggest that rt-dg version just works on a sphere mesh. One may need to increase the quad rule degree (currently hard coded to 2p+1). Should be tested (Williamson tests?). Also, it is unclear when one has to switch to spherical domain, could be tested.
Matt shared an ESE jupyter notebook as an example of nice documentation including python and math, hosted on github.io. We could develop similar for Thetis; maybe we could organize a sprint in the fall.