This is the directory tree of G-larmS/testcases: here LOG_DIR is './testcases', and <EVENT_ID> is the testcase name.

./testcases
    ├── 2010_El_Mayor_Grapenthin_etal_JGR_2014
    ├── 2014_South_Napa_Grapenthin_etal_GRL_2014
    └── Hayward_Grapenthin_etal_JGR_2014
        ├── bin
        ├── config
        ├── events
        ├── gps_timeseries
        │   └── 2013
        │       └── 20130905
        │           └── 0000
        ├── log
        ├── model_logs
        ├── movie
        ├── offset_logs
        └── plot
        ├── log_validate
        │   ├── model_logs
        │   │   ├── BestFit
        │   │   ├── MtDiablo
        │   │   ├── SAFconjugate
        │   │   └── SAFparallel
        │   ├── movie
        │   ├── offset_logs
        │   └── plot
        │       ├── BestFit
        │       ├── MtDiablo
        │       ├── SAFconjugate
        │       └── SAFparallel
        └── simulations

One testcase has been set up so far (minus the actual comparisons): 

        Hayward_Grapenthin_etal_JGR_2014
        
The names of the testcases indicate:
        - Year of the event and event name (if it is based on an actual earthquake)
        OR Name of the fault a simulation is running on
        
        - The rest is publication information, if the results/data were used in a 
          paper (here, Grapenthin et al., JGR/GRL 2014).

Each testcase replicates the G-larmS directory structure to the most extent. That 
brings some overhead in terms of files, but keeps everything separated and easy
to parse through for the human tester, too.
 
Directory structure explained:

    ./bin
        Each testcase is mostly self-contained. Meaning it brings its own scripts 
        in this directory to call G-larmS. The folder contains the following files:
        
            - clean_testcase.sh - removes all the stuff created by calling run.sh, 
                                  resets testcase directory
            - make_plots.sh     - collects fault regimes that were processed and 
                                  calls plotting scripts from G-larmS' plot_scripts
                                  directory
            - offset_est.inc    - called by run.sh, takes care of the offset estimation,
                                  (sequentially), uses global values set in run.sh
            - param_est.inc     - same as offset_est.inc, but for parameter estimation.
        
            - run.sh            - calls G-larmS in its root directory to run this 
                                  testcase with the current version of G-larmS

    ./config
        very similar to G-larmS' config directory

            - G-larmS.spec      - specification of cfg file syntax, variable types, etc.
            - NMT_bard_research.cfg
                                - is mainly used to get the baselines that were specified 
                                  for this testcase
            - NMT_FaultPolygons - definition of fault provinces / fault geometries
            - NMT_template.cfg  - header information of config file - is being rebuilt  
                                  repeatedly during calls by run.sh/offset_est.inc/param_est.inc
            - sites.list        - Bay Area GPS site database, contains coordinates and 
                                  names of stations

    ./events
            - HaywardTemplate.xml
                                - Template XML message file that is similar to messages sent by 
                                  ShakeAlert Decision Module as alarm. EVENT_ID, ORIG_TIME will
                                  be edited by shell scripts to create event that works for the
                                  specific G-larmS call

    ./gps_timeseries
            This directory emulates the 'archive' structure for real-time GPS data implemented
            at the BSL: YYYY/YYYYMMDD/HHMM/ The files contained in the last subdirectory
            are for 2013-05-09, a day that I picked to use for the Hayward examples in the JGR
            paper. These data files were created in real-time by trackRT rather than in 
            rewind mode using trackRTr! The files contain the time period used in the 
            JGR paper, which is specified in run.sh in the GLARMS_START/END parameters

    ./log
            This is the output directory for the testcase. It replicates the directory structure 
            that G-larmS expects. All results for this one testcase can be found here. The
            plotting scripts operate on this directory.
            
    ./log_validate (read-only!)
            This contains the pre-computed results for a full run of this example, including log 
            file output from g-larmsOE/PE runs, movies etc. The text validation runs on files 
            stored in here. These should not be modified!
            
            Results in BestFit, SAFparallel should be similar to the Grapenthin et al. 2014, JGR
            paper results for the simple Hayward simulation (Figure 4, HaywardSyntheticM7.0_0km_buried.mpg).
            
    ./simulations
            Contains everything that's necessary to run a simulation, i.e. add pre-computed 
            offsets to the archived GPS position time series. 

                - HaywardTemplate.xml
                                    - Contains the definition for this simulation, including when
                                      after simulation start the pre-computed offsets should be 
                                      added, which files these are stored in (PATHS are set by run.sh
                                      upon start to remain system independent), location of the
                                      event etc. 
                - Hayward_baseline_disp_horizontal.gmtvec, Hayward_baseline_disp_vertical.gmtvec
                                    - contain simulated baseline-offsets for a M7.0 event on the 
                                      Hayward fault with ~1m of slip on it.