GENIE:GENIE Tuning rel-2-6

=Tunings of release 2.6.n=

This page provides comprehensive details of the parameter estimation exercises employed for releases 2.6.n.

Potential tuning exercise strategy:


 * Standard physical tuning of the 16-level  model
 * Tuning of the biogeochemistry parameters for the  model holding the tuned physical parameters constant
 * Co-tuning of
 * Addition of ENTS

Summary
The source code repository has been tagged at release  to indicate a stable release of the code base following the completion of major developments in a number of the modules (see release page for details). The branch  has been created at the same point in the repository in order to manage the tuning process for this version of the model. It is intended that the branch will be merged back into the trunk when tuning is complete if this proves appropriate.

Base configuration
The initial physical tuning of the model is applied to the 32x32x16 resolution of the model and is based upon the configuration file.

The source code is exported from http://source.ggy.bris.ac.uk/subversion/genie/branches/rel-2-6-tuning.

The model is built in the usual way using genie_example.job:

./genie_example.job -f configs/eb_go_gs_itfclsd_16l_tuning.xml

An archive of the model binary and supporting data files and libraries is generated:

zip rel-2-6-tuning.zip $(cat ArchiveContents.txt)

Tuning specification
The config file  is a copy of   and provides the specific changes required for the tuning exercise. In particular the duration of each simulation is set to 4000 years to ensure equilibration in the model for each parameter set that is evaluated. The frequency of output is adjusted in order to facilitate restarts of the model:

Tuning configuration
The parameter estimation problem is:

Find:

Boundary conditions
For the record the boundary conditions are (ie. these files are read by the model during initialisation):

Target data
Detail the target data files and the code that is used to compare the model output with the data to produce the RMS error scores.

The root mean square error function code is found in. This code compares model output with observational data across the four fields ocean temperature, ocean salinity, atmospheric temperature and atmospheric humidity. Four root mean square error values are returned, one for each field, plus a composite error function value. The multi-objective optimisation algorithm works with the four individual field errors and works to generate a pareto-optimal set of solutions to the problem.

The model configuration includes entries that define which data files are to be used to make the model-data comparison:

Validation
An independent expert will provide details of the error score for the vanilla build of  and ideally will provide NetCDF output from their run. The tuning software must be able to reproduce the vanilla results for the default parameter set using the infrastructure supporting the tuning exercise (ie. the DAGMan execution of the Win32 model build on the UoS Condor Pool must reproduce known results before the tuning exercise can be performed).