-
Notifications
You must be signed in to change notification settings - Fork 0
jhHaC
HaC version 1.0 is the first "working" implementation of CABLE @ Point Loobos. It is based on JULES 5.4.
We do not expect this version to be physically correct. The intent is mainly to have a version to progress JaC submissions by making incremental changes in HaC. A direct discussion of a specific example will help to clarify this.
In the working HaC1.0, CABLE is being called in entireity as the LSM, excluding JULES, and only using JULES i/o. There are parts of JULES (especially online) that are required to be called, and consistently so, with the CABLE implementation. An example of this in the online case is the treatment of fluxes over sea-ice. CABLE has no mechanism for computing fluxes over sea-ice and so in the online implementation we are compelled to keep these mechanisms in JULES active. Some care was taken to implement these exclusions and inclusions correctly in HaC1.0. However, there were complications, at which point it was decided to proceed with working version of HaC1.0 to focus on the radiation call. The complications arose on the implicit call and could be revisited later when we get to that stage.
Our idea for now is that although the "output" may not be 100% physically correct, HaC1.0 still allows us progress JaC submissions. In our current implementations of CABLE in all ACCESS (JULES)variants, we skip the radiiation call on the first timestep. Ultimately we want to be able to call cable in the radiation pathway on the first timestep. The obstacle to this, in short, is that some properties of the surface which are needed, are not defined until the explicit pathway is activated.
To test a "quasi" bitwise reproducibility, every timestep we sum the fluxes, soil temps etc and dump this to a file dummy000.txt*.
We can then cmp this file to a baseline version of the file.
We note that many of the changes result in changes to the SW forcing seen by CABLE and so the fluxes etc.
However, there are no changes for the first 18 timesteps where the SW forcing = 0 anyway.
We make use of a variable
Modification to SW in (and so far a lot of the changes) we get diff in output bitwise reproducibility variable