4. Standard AnalysisΒΆ

The gcmfaces standard analysis consists of an extensive set of physical diagnostics that are routinely computed to monitor and compare MITgcm simulations and ECCO state estimates [FCH+15], [FCH+16]. The computational loop is operated by diags_driver.m which stores intermediate results in a dedicated directory (mat/ in Demo Directories Organization). Afterwards, the display phase is operated via diags_display.m or diags_driver_tex.m as explained below.

In order to proceed, user should have completed the installation procedure in Section 1 and organized directories as shown in Demo Directories Organization. They can then, for example, generate and display variance maps from the ECCO v4 monthly mean climatology (12 monthly fields) by opening Matlab and executing diags_set_B.m as follows:

%add paths:
p = genpath('gcmfaces/'); addpath(p);
p = genpath('MITprof/'); addpath(p);
p = genpath('m_map/'); addpath(p);

%set parameters:
dirMat=[dirModel 'mat/'];

%compute diagnostics:

%display results:

which should take \(\approx5\) minutes. Each generated plot has a caption that indicates the quantity being displayed. Results of diags_driver.m can, alternatively, be displayed via diags_driver_tex.m to save plots and create a compilable tex file. This process should take \(\approx\)10 minutes:

dirTex=[dirModel 'tex/']; nameTex='standardAnalysis';

Other diagnostic sets can be computed and displayed accordingly by modifying the setDiags specification: oceanic transports (A), mean and variance maps (B), sections and time series (C), and mixed layer depths (MLD). Each set of diagnostics (computation and display) is encoded in one routine named as diags_set_XX.m where XX stands for e.g., A, B, C, or MLD.

These routines can be found in the gcmfaces_diags/ subdirectory. Computing all four diagnostic sets from ECCO v4 r2 climatology takes \(\approx\)1/2 hour. Computing them from the 1992-2011 monthly time series (nctiles_monthly/ in Demo Directories Organization) by typing

dirModel='release2/'; dirMat=[dirModel 'mat/'];

takes \(\approx20\) times longer and typically runs overnight. However, to speed up the process, computation can be distributed over multiple processors by splitting [1992:2011] into subsets.


The above diags_driver calls rely on default parameters that are adequate for the Section 1 solution, but yours may differ. Using the doInteractive option (see help diags_driver) is therefore the generally recommended method, since it gives you the opportunity to review and, if needed, edit the relevant parameters.