Downscaling from GCMs

CCAM is often used to dynamically downscale GCMs for regional climate projections.  This downscaling can be performed using the script run_ccam.  Below are the basic steps for configuring a dynamically downscaled simulation

Step 1: Choose the type of downscaling experiment

The first step is to decide what type of downscaling experiment will be undertaken using CCAM.  The two main options are:

  • Downscale reanalysis or Global Climate Model (GCM) data.  These files contain the atmospheric data from the GCM, converted to the conformal cubic grid using CDFVIDAR.  Previously prepared datasets are available if the user has registered to use the dataset (e.g., ERA-Interim).
  • Simulate a global atmosphere using bias corrected Sea Surface Temperatures (SSTs) from a GCM.  This global simulation has reduced some of the GCM errors and then can be downscaled for a region.

Step 2: Number of processes to use for the experiment

The number of processes to use for the experiment can be determined with nproc.  Note that CCAM will reduce the number of processes being used if it cannot identify a valid grid decomposition.  Usually nproc is set to the number of processes available (e.g., $SLURM_NTASKS for the SLURM scheduler)

Step 3: Specify the simulation name, directory and spatial domain

Using the run_ccam script, specify the directory name (hdir), center latitude (midlat), center longitude (midlon), spatial resolution (gridres) and grid size (gridsize).

For simulations using bias corrected SSTs must use a global grid (gridres=-999.) and a gridsize=96 or gridsize=192, depending on the choice of input data (see below).  This global simulation can be later downscaled to finer resolution.

For simulations downscaling from atmospheric data (e.g., a previous CCAM simulation or from reanalysis data that has been processed by CDFVIDAR, we recommend a change in grid resolution between 1:20 (e.g., 100 km to 5 km resolution).  This ratio is considerably higher than that used for limited area models (e.g., 1:4), since CCAM has no lateral boundaries.  Rather the limit is determined by the scale-selective filter used to assimilate/nudge atmospheric data into CCAM.

Step 4: Specify a simulation start and end date

The simulation start year (iys), start month (ims), start day of month (ids), end year (iye), end month (ime) and end day of month (ide) can be specified.  In some cases it is useful to also useful to specify the start time (ihs) as some GCMs provide data starting at 3Z, but for most GCMs using a ihs=0 is appropriate.

Step 5: Specify the experiment mode

The dmode option must be selected, depending on the experiment chosen in step 1.  There are three options

  • Downscaling non-native CCAM data (e.g., reanalysis data processed with CDFVIDAR such as ERA-Interim)
  • Use bias corrected Sea Surface Temperatures (SSTs), requiring a global simulation
  • Downscaling from a previous native CCAM output (e.g., downscaling a previous CCAM simulation to finer resolution)

Step 6: Choose the RCP or SSP pathway

Using  rcp can specify which emission scenario to employ for a future climate.  Historic emission scenario is automatically specified if the simulation year is prior to 2005 for CMIP5 and 2015 for CMIP6.  Two emission scenarios have been prepared for rcp45 and rcp85 under CMIP5, although other scenarios can be constructed based on the CMIP5 datasets.  It is also possible to select a CMIP6 scenario using cmip, with rcp then specifying the required SSP.

Step 7: Location of input datasets

The final step is to specify the location of the input datasets.  The options then depend on the choice of dmode in step 5.

  • When downscaling from reanalysis or previous CCAM output (e.g., dmode=0 or dmode=2), then the host dataset directory is specified with bcdom, and the prefix for the input dataset is set with bcdom.
  • When running a bias corrected SST global simulation, then the input SSTs and sea-ice from the GCM is specified using sstfile, the directory containing the sstfile is specified with sstdir, the initial atmospheric conditions are specified using sstinit, and the directory containing the atmospheric initial conditions is selected with bcdir.

Once the run_ccam.sh script is configured, it can be submitted to a scheduler or run from the command prompt.  The script will simulate a limited number of simulation months (ncountmax) and then resubmit itself to the scheduler.  This allows users to optimise the run_ccam script for their scheduler.