I’m planning to do some model experiments to test the importance of meridional moisture gradient to some synoptic weather systems in the tropics. My plan for now is to use ACCESS-rAM3 to simulate a case of non-tropical cyclone (non-TC) low pressure system over the northern part of Australia. The scientific hypothesis is that this kind of system owes its existence and propagation to the moisture advection from the equator side of the vortex. I want to test this by experimenting on different north-south gradients of moisture using the regional model.
I have tried to get my hands dirty on running the ACCESS-rAM3. I have successfully modified the released Lismore suite by moving the centre of the domain to the Gulf of Carpentaria, as depicted below, and changing the simulation dates, but leaving everything else untouched (e.g., cycling frequency, STASH output, etc.).
What I’m planning to do next is to enlarge the d1000 domain and steepen the north-south gradient of moisture (probably start with surface humidity) and see how this increased moisture gradient affects the strength of the non-TC low pressure system over the Gulf of Carpentaria.
Is this something possible to do with ACCESS-rAM3 and to get help with from the ACCESS-NRI Team? I really need some help as I’ve never done any modelling experiments before.
Most of us at ACCESS-NRI are not scientists so I want to make sure we understand your request correctly.
If I understand you want to force the surface humidity to a higher value as in a free running experiment, is that correct? Do you want to force this higher humidity during the whole simulation or just as an initial condition? If you want to force the humidity during the whole experiment, do you want to provide your own field at each time step (or a given time frequency) or do you want to provide a multiplying factor for the model to apply to the values it calculates? Do you know if you force the surface humidity, do you need to apply any compensating forcing on other values to ensure conservation is respected (or you don’t mind about conservation in that experiment)?
I want to steepen the north-south gradient of surface humidity in the initial conditions and let the model free-run for several days. For now, I think I’m going to ignore any conservation. Is this possible to do?
To steepen the gradient, I think I’m going to start by multiplying the initial surface humidity by a linear scale function where it is 1.0 at the equator and 0.5 at 30deg S.
Please let me know what you think.
Thank you!
clairecarouge
(Claire Carouge, ACCESS-NRI Land Modelling Team Lead)
5
Thanks for clarifying. I have asked our staff for someone to help you, I’ll let you know when they can help.
I am also wondering if you have a strong reason to keep this request private. I think that people in the community would be interested to learn about the process to modify the initial conditions. So I wondered if you would be ok to change this help request to be public. You are more than welcome to refuse obviously without giving me any reason.
Hi @reyhan.respati, it is definitely possible. The issue is reproducibility.
If you are happy with a manual-approach, you would need to run ACCESS-rAM3 until all the reconfigurations are finished (just before the forecast step), then modify the start dump sitting in the directory. Modifying a start dump uses a package called “mule”.
We have a few experts in mule in ACCESS-NRI who could help you with that step. Does sound like something you would be willing to do?
I’m very unfamiliar with the terms here as I’m really new to using ACCESS-rAM3. Is there any sort of documentation/tutorial that I can learn from to do the three steps you outline above?
This post might be better placed on Cumulus rather than the Hive (for scientific rather than technical help), although lots of NRI people are across both platforms. However this is definitely the place to be asking about the use of things like mule.
To pause a suite that is running, you can hit the “pause” button in the rose gui (if you’re using the gui).
You mention you’ve left the cycling frequency at its default setting. For this kind of scientific study I would suggest you would be better using a free-running simulation (i.e. set the cycling frequency to the length of the simulation you want to do. You can edit the conf file if the option you want isn’t in the gui).
I’ll be at Monash tomorrow (Weds) morning if you want to have a quick chat about your science plan for this. I’m happy to give you suggestions that will hopefully put you in a better position to ask for technical help from NRI.
Replying here for anyone else who may read it in the future (Reyhan and I met in person at Monash today) - this will work, it will throw a warning in the GUI if you don’t also add the option in the suite files to add 96H to the drop-down GUI options, but that can be ignored.
I wonder if clicking the following Hold option at the start of the run would “pause” it just before the forecast tasks, so that I don’t have to monitor it constantly to decide when to “pause”.
You can right-click on (or press both fingers down on a Mac) on the “fcst_000” job at the beginning of the run. A drop-down menu will pop-up, and you can select Hold.
Apologies for not getting back to you, I thought you would talk to @bethanwhite.
I do think it would be worthwhile you discussing your idea with the scientists first because there are lots of different considerations.
Mule is a library module that can be loaded in Python, but it is not that easy to use.
Maybe get back to me after talking to the scientists and we can talk options going forward.
