JACDec82020

Progress achieved Updates:

Ian:

Claire: I met with Danny to go over the handover information. We decided to try out Jhan's branch of the code JaC code ported to 5.9: JaC-5.9_draft1. We found nci_intel_loobos_gl4_cable test wasn't working for us. It was a problem of formatting in the rose-app-cable.conf file. This morning I successfully reformatted that file and got a successful test. But other tests are failing. There are still a few formatting issues to fix in the file (easy for me to do).

More important the metachecker test fails because normally we can't have cable_pftparm and jules_pftparm defined in the same job. Jhan is using jules_pftparm to define the LAI and canopy height. It will have to be moved to cable_pftparm. I'm pretty certain the trigger about using only one of the pftparm namelist was requested by UKMO. I'm happy to look into this.

Jhan: MPI version of CABLE itself is fine at previously presented CABLE-3.0 vn. However, globally (GSWP2), we ran into a problem. This is no doubt due to the relatively infinite heterogeneity compared to running at single sites. Suspicion is that it is due to treatment on sites where it snows. (Although single site Hyytiala included snow precipitation - it was the most sensitive site tested and still worked, however GSWP2 intros many more sites of this variety that have varying configurations).

Backing up to the trunk version and proceeding to incrementally to merge CABLE-3.0, whilst maintaining a "working" MPI we got back to a CABLE-3.0 vn that included modifications along the way which may indeed, either individually or in combination, led to the global MPI model working. Generally these were inclusion of initialisations of snow related variables, for offline, in the absence of a restart. Upon re-merging it was these sections of code which were examined more closely as it is possible that that they were never invoked in previous tests of the model, hence the reason the model was successful in the past.

Progress planned Updates:

Ian:

Danny/Claire:

Jhan:

Following above: With CABLE-3.0, I will show results for

• JAC5.7 versus CABLE-3.0 (offline@Loobos)

• CABLE-3.0 versus trunk @ PLUMBER sites

• CABLE-3.0 versus trunk (MPI) @ GSWP2

Following "Comparing CABLE-CABLE TO CABLE-JULES" have to update JAC side and CABLE "trunks".

Need to get this into NCI trunk.

Need to update to JULES-5.9.

Issues for discussion

• Will there be other cases where things don't add up? No doubt. An immediate candidate is JAC5.7 versus CABLE-3.-0 (offline@GSWP). However, it is also true of the trunk version that we periodically encounter incidences of things the don't work.

The community was challenged to break CABLE-3.0 back in July. This was long before these improvements and yet there still hasn't been a single report.

It is somewhat critical to implement a CABLE-3.0 trunk. Is it prohibitive to JAC that it isn't? Not entirely, however the realistic outcome of trying to maintain an ACCESS version of CABLE and a local version of CABLE would be shifting the problem we have now of updating to each new version of ACCESS

Other business

• New time for meetings