@cbengel, @reyhan.respati’s model setup is only going to use one grid.
I just thought, would it be easier to modify the moisture gradient for the required region in (local copies of) the ERA5 grib files that are used to produce the start dump, rather than the start dump itself?
Yes, that is definitely another option.
@reyhan.respati this is a science comment but the existing discussion is already here so I’ll post it here:
You haven’t mentioned what you plan to do with your boundary conditions, but I think the best approach would be to apply the moisture gradient to your LBCs as well as your initial conditions.
If you don’t, you will introduce a large discontinuity in the moisture field inside the model’s domain and along its boundaries. The model will be out of balance along this region and at the very least will take time to re-equilibrate. It might run ok, or it might complain, or it might crash, or the boundary effects might be very big and propagate deep into the domain.
If your domain is sufficiently large, applying the moisture gradient to the boundaries won’t significantly affect the evolution of systems deep within the domain.
@bethanwhite I’m also planning to apply the scale function to the LBCs, if possible. Or, do you have other suggestions?
No, that sounds ideal. In that case I suspect it would be easiest to make a local copy of the ERA5 grib files you want to use, apply the scale function to the grib files, then use those in the ancil suite to generate your start dump and LBCs. That would avoid having to use the mule package to edit your start dump.
(I haven’t used mule but I have heard that it can be tricky, and I’ve seen several posts where the model complains about start dumps that have been edited with mule. I don’t want to put you off using it but the path of least resistance if you are only using one nest would probably be to edit your ERA5 input files).
Let’s see if @cbengel agrees or is in favour of using mule?
Hi @bethanwhite and @reyhan.respati, the answer is both technical and scientific.
ERA5 data is at 0.25 degrees. If you are happy for that data to be modified then yes, modifying the grib would easier. In fact the grid files can be converted to netcdf and back-again. So the work would be in netcdf.
For modifications to input files at the nesting resolutions, most likely the tool would have to be mule – although there are other options in terms of field replacements that are also complicated.
It gets tricky and complicated, depends of the timing and scientific choices. Hence it is better that you sort out the science choices alongside the technical difficulty of the solution provided.
I think @reyhan.respati is planning to use a 10km grid.
Reyhan, can you tell us what sort of gradient you are intending to use? The question would be whether 0.25 degrees is sufficient to resolve the gradient you want, or whether you need represent it more finely.
We can take this offline if you want to discuss the scientific details in private.
Oh, good point, thanks for that reminder about public / private posts, Chermelle!
Hi @cbengel and @bethanwhite, could you please let me know how to start a private discussion on this? Is it just the DM feature?
@reyhan.respati I suggest you start a Cumulus message thread. Go to Cumulus, click ‘My Messages’ in the left hand column, click the blue ‘new message’ button in the top right, then add me and Chermelle to the message and start the thread 
(You can also add anyone else to the message that you want to bring into the discussion, e.g. supervisors - or not if you don’t want to)
Hi Reyhan,
I have been working on modifying ERA5 initial and boundary conditions and bringing them into ACCESS-rAM3 simulations for Tiwi Island thunderstorm simulations. It sound like you want to modify the 3D moisture field in the initial conditions to increase the moisture gradient. I think this could be done using a similar method where you tell ACCESS-rAM3 to use your modified ERA5 instead of the standard ERA5 data but bear in mind that this is quite experimental at this stage. Happy to chat about this if you want to try this method. I am aiming to document the process here but this github is only barebones at the moment: GitHub - youstormorg/Tiwi: Tiwi Islands ACCESS