So: you have been told to do a JULES run for your project?

Here are some standardised steps to follow that should see you right in most situations:

First, a couple of disclaimers:

• There are many ways to run JULES and this webpage is not intended to be a comprehensive list of all of them: this is just my (Toby's) recommended way to do it, which may very well not be the way that is most appropriate for your project (!).
• This webpage is also about fine-resolution runs (sub-10km resolution, usually 1km). Because of this, I am discarding any advice about global runs: all examples are about smaller areas (ranging from provincial/state-sized areas to subcontinental). 
• We are also exclusively doing gridded runs, not single-site runs.
• This webpage is about JULES-Standalone only (i.e., the version of JULES that is run on its own outside any coupling arrangements with an atmosphere model or an ocean model). For coupled runs, see (Fig. 1 in my report here and):

1.    the Regional Coupled Suite (RCS) developed by RMED (e.g. Castillo et al. 2022), or
2.    the UKESM project, or
3.    the GC5 configuration, or
4.    the huge current efforts focused around NGMS in Exeter, UK.

• First step is to identify the MODEL DOMAIN or BOUNDING BOX over which you want the model to run. I use this online lat/long finder tool to identify the coordinates of a box around my area of interest.
• This can be a box in rotated coordinates (e.g. Australia right, where the edges of the box are not straight east-west and north-south lines) or a straight latitude/longitude box (=GRATICULAR or RECTILINEAR) where they are straight east-west and north-south lines (e.g. the Africa domains right).
• Note that the graticular CORDEX domains are popular choices for model runs, but often larger than required by particular projects.
• Note also that for JULES runs with rivers activated, remember that you need to include whole river catchments, i.e., to simulate an area around Kinshasa, you will most likely need to include the entire catchment of the Congo, not just the area around Kinshasa.
• Also, I recommend to add a 'halo' around your catchments of at least 20 km or so to avoid having edge effects.
• Finally, remember that JULES ignores all ocean points, so having a domain like my SE Asia one below right is not as costly in terms of runtime as you might think.

• Next decide the resolution you want to run JULES at. Usually this is a balance between accuracy (higher resolution is better) and available resources (high resolution costs runtime). At the moment, I am trying to develop a standard workflow based on 1km resolution runs.
• Once you have decided domain and resolution, I recommend to create a shapefile describing the domain and put it on a QGIS map straight away (e.g. as I did, right). Recommend to start with this shapefile of a box around Perak in Malaysia, then use the Vertex Editor in QGIS to edit the corners to be correct for your domain. Having a shapefile of your model domain is useful because it allows you e.g. to clip other layers to it in QGIS.
• For general QGIS tips, see here.

Next, you need to start the process of getting yourself set up to use JULES.  You will definitely need an MOSRS login, and I recommend environmental researchers (whether in the UK or not) to apply for a JASMIN login too (this allows you to run JULES on the JASMIN super data cluster). Follow these links to get these:

MOSRS: https://jules.jchmr.org/getting-started
JASMIN: https://www.tobymarthews.com/jules-on-jasmin.html

When I was putting together the JULES website, I stated here that you need three things to run a land surface model - driving data, ancillary data and control files:

• ANCILLARY DATA: Run my script GENHSA.r with menu option to generate ancillary files (n.b. these are what I call HSANCILS rather than normal ancil files: see here). Note that my script is (a) slow (takes at least a few days) and (b) quite memory-hungry (which means you probably need to run it on an HPC rather than a laptop)
• DRIVING DATA: Then run GENHSA.r with menu option to generate driving data. My script pulls in ISIMIP3b data, which is already bias-corrected.
• Next, if you are going to have rivers simulated, download river directions from HYDRO1k, correct them and append them to the ancillary file
• CONTROL FILES: Now choose an appropriate suite. I usually use a 15 PFT configuration these days.

Now you're ready to try running the model. See here if you have never done that before, but here I'm going to assume that you know how to do that at least in outline.

- Adjust the timestep_len (for RFM) until you don't get the warning about thetas being set to zero (e.g. for rivers_dx=1500km I needed timestep_len<=600s).
​
I am going to need a webpage about comparing to station data (inc. examples about mismatched coordinates).



Notes for me:
- I need to generate an ancil file for at least a small subgrid (if I do it for a subgrid, choose NOT to crop the output file).
- Then upload that to ~/ancils on JASMIN (if you generated this only for a subgrid, rename it to lose the "_part" in the filename, but note somewhere that this file is only for a subgrid)
- Change choice9 to 2 and make sure subgrid=FALSE
module load jasr
R
source("GENHSA.r")

- While it's generating the ddata on JASMIN, upload the scripts to my VM as well and generate the subgrid=FALSE ancil file there too.