I’m interested in setting up semi-idealised experiments with ACCESS-rAM3, that start with a real case and modify some aspects of the surface properties. I’ve tried a few things, including modifying the ancillaries using ANTS. I’m not at all sure that this is doing what I’m expecting it to though, and wondering whether I’ve got something a bit wrong in the work flow.
For example, I’ve tried modifying a 1X1 degree patch of surface in the centre of the domain in an attempt to control some aspect of the atmosphere near the surface - i.e. changing the albedo, soil surface roughness, or soil thermal capacity so that the ancillary looks like this, with the modified area matching in the outer and inner domains:
I’m a little bit surprised that the changes I’m making don’t have more impact, since I’ve changed the surface conditions over the patch to quite extreme values (i.e. the maximum that is found anywhere in the outer domain). I know that the effects might be non-local and advected away, but I think I would have expected a little more difference. Here are the results of five different experiments (each experiement only changes one thing), showing temperature at the bottom model level, the difference with the control, and a version zoomed in over the modified box. I’ve looked at a few different fields, including surface temperature, surface sensible heat flux and wind speed, and see similarly small differences.
Is this a reasonable way to proceed, or would these changes to the ancils be overwritten and modified somewhere in the code? Are there other suggested ways of doing this? Has anyone else done something like this?
Hi Claire, can you give a bit more info on the workflow you have used to modify the ancils and run the model to read them?
Tagging @reyhan.respati as I know he has experience with modifying rAM3 ICs and BCs. Not the ancils to my knowledge, but he may still be able to help here.
As @bethanwhite mentioned, I have some experience modifying the atmospheric boundary conditions, so I might be able to help if that’s what you intend to do. However, I never touched and played with the ancillary variables (e.g., land cover, physical constants, etc.).
Try outputting the field you’ve modified to make sure that the model is actually using it - it may be that some fields are not being sourced from ancillaries and are instead regridded from the parent domain
Hi Claire, you’re using a different filename from the filename the suite expects (qrparm.soil_roughness).
If you haven’t overwritten the original ancil with the modified one, using the same filename as the original, the suite will still read the original file.
Yes, definitely, I then overwrote the original soil_roughness file with the modified one. Sorry, I forgot to include that step when I explained what I did
@Claire_Vincent@cbengel I was misremembering just now in ModSci, the person who has recently told me (by email) that they paused a suite mid-run and modified files before letting the suite continue was @reyhan.respati
So perhaps this thread is the place to continue the discussion?
OK thanks. I had a few questions following on from the modsci meeting just now:
When you mentioned pausing the suite and modifying the start-dump - can you tell it in advance where to pause, or did you mean watching the scheduler and stopping it manually?
I see your point about modifying the startdump rather than the ancils, but I was just curious what is the issue with modifing the ancils first?
@cbengel you mentioned using MULE, which I’m fine to look into, but is it OK to use ANTS? This seemed to work well for modifying and ancils (see code earlier up in this thread).
Thanks all! I still need to follow up on Scott’s suggestion of outputting the actual fields that I thought I’d modified, so will try that first.
Hi @Claire_Vincent, I suggest that you continue with what @Scott suggested and confirm whether or not the fields you have modified have got through to the ‘start dump’ initial conditions file. To do this you need to look at your “start dump” or initial conditions file.
You can find the “Start dump” in the “ics” (initial conditions) folder. To find the “ics” folder please go through this description of the outputs from the RNS is given in the run-a-model. Inside the “ics” folder is a folder named with the scientific choice (currently GAL9 or RAL3P3 depending on what nest you are looking at). The start dump is the file inside that directory called {scientific choice}_astart, i.e. GAL9_astart.
If the field you had modified is not modified there then you need to come back to me and I think we need to put together a tutorial for users to modify their initial conditions using mule. A tutorial of that nature is way beyond a forum post and may be something to discuss in terms of a tutorial and something on the Hive.
