The codes in this folder are made to run tsunami propagation models for every unit-source on the source-zone. The unit-sources themselves must already have been defined using code in ../EQ_SOURCE/.
Currently the folder is set-up to run the SWALS model on raijin.nci.org.au. The codes would require significant modification to be run on another machine, or using another model.
To run the models, the process is
Check the variables defined in config.R and see that they point to the correct unit sources, and that the model output files will go to a location that you are happy with.
create_files_for_all_simulations.R is used to make directory structures for all model runs (using some definitions in config.R).
We make directories for all model runs by modifying template files in the template folder. This allows us to define separate output directories and initial conditions (i.e. tsunami unit sources) for every model run. Note that files in the template folder define other model configuration options (e.g. resolution), which may need to be modified depending on your applications.
Once the template files are ok, to create a separate SWALS model that runs for each unit-source, do:
Rscript create_files_for_all_simulations.R
This script is quite light-weight (i.e. you might not need to use the job queue).
run_16.sh is used to submit 16 of the SWALS models (which are not already slated for running) to the job queue, to be run on a single node. The number 16 was chosen because that is efficient for the hardware on raijin.nci.org.au. Note that significant performance benefits were gained by using 'numactl' when running multiple jobs on a single node. It is run with:
qsub run_16.sh
The above needs to be done repeatedly, until all model runs are completed.
NOTE: If you are running multiple source-zones, it may be preferable to use the alternative script ../../run_16.sh. The latter is run from it's own directory, and has the advantage of being able to run models from more than one source-zone (if it doesn't find 16 models to run on a single source-zone).
Perform some basic checks that the models have run correctly.
check_runs_complete.R is used to check that the SWALS model runs have finished, and to do some basic checks. It is run from within R using:
# Read the functions into R
source('check_runs_complete.R')
# This is used to check for models that seem to have not completed
check_models_have_been_run()
# If the above is ok, then try this to check that gauge outputs exist and
# are correctly ordered
check_model_gauge_integrity()If there is a problem it should print out an error message. Otherwise it will be silent (and return NULL, which is what R does for functions that don't return anything).
Note that this process can be run over multiple source-zones at once, by using the alternative script ../../checkruns.R.
check_logfile.R is used to plot the evolution of peak-stage and mass balance in every tsunami model, using the log-files. To run it, do
Rscript check_logfile.R
and then visually investigate the resulting pdf file 'tsunami_log_check.pdf', which will show the time-evolution of peak-stage and mass-balance, for every model.
run_permute_netcdf_gauges.sh is used permutes the dimension of the tide-gauge netcdf output files, to ensure fast access to single-station data [which is extremely convenient for later analysis]. It is run with:
qsub run_permute_netcdf_gauges.sh