Skip to content

JACNov2020Minutes

Jump to bottom
JhanSrbinovsky edited this page Nov 23, 2020 · 1 revision

Progress achieved Updates:

Ian: First steps towards scafolding* for "radiation CABLE" and "explicit CABLE" section under JULES5.9 drafted - involves changes to control.F90 and below.

Currently sitting on NCI (mirrored from MOSRS branches)

By scafolding I mean intermediate layers that do the land-sea-sea ice split and the sequence of science-based subroutines below that. In particular this is code that sits between the surf_couple routines and cable_cbm.F90

Danny:

Jhan:

Comparing CABLE-CABLE TO CABLE-JULES

Progress planned Updates:

Ian: Compile equivalent scafolding for "implicit" and "extras" branches under 5.9.

Danny:

Jhan:

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

  • There is always going to be a difference between CABLE-CABLE and CABLE-JULES due to the "initialization". i.e. The parameters, prognostics at dt=0, and the conversions and so on. Ultimately we would like to observe the difference between CABLE andJULES due to the different processes assumed in each LSM. To achieve this we have to remove the differences due to initialization of each model. This can be tweaked on a site by site basis. I imagine that in the longer term we make this the default behaviour of the model?

  • Overall strategy: section by section (inputs+scafolding+science) - or inputs+scafolding then science? JULES5.9 is quite different in terms of scafolding code. We maybe able to port the 5.7HAC structure into 5.9 as a means of testing but I doubt we will get that onto the trunk even as an interim measure.

  • Where are we allocating CABLE's typed variables? If at control.F90 level or higher then we have to make changes to the UM (we'll have to do this anyway but dummy variables of zero can be used) and get CABLE typed variables into the UM to get even scafolding code onto the trunk. If within surf_couple* then we have to do more passing+filling into/out of CABLE-typed variables in the surf_couple* layer \

  • How can we test bits of the code without doing everything - even just to check compilation? We may be able to implement some scafolding code without CABLE-typed variables etc \

  • It looks like there's been some changes to tiled-soil capabilities of JULES at 5.9 - with implications for the ancillary reading/passing.

  • explicit vs implicit calls don't have much (if any) impact Standalone

Clone this wiki locally