NOTES ON EEPAS SOFTWARE (EEPSOF) INSTALLATION IN GNS TESTING CENTRE
====================================================================

1. The following models are submitted for testing as three month models.

PPE model with parameters specified on p. 22 of EQC report "Testing and 
development of eq forecasting models" (GNS2008/70) 

Four versions of EEPAS: EEPAS_0r,EEPAS_0f,EEPAS_1r,EEPAS_1f with parameters 
specified on p25 of GNS2008/70.

The necessary example files are found in 
/home-s/david/extra/eq/janus/eepsof/relmtestsnz/files.

2. Forecasts for all models are computed by running the EEPSOF package 
with different parameter and command files. Note that the PPE model forecast 
is produced from any of the four EEPAS runs.

3. The approprite version of the EEPAS software (EEPSOF) package is 
eepasof19z.f, which
compiles with f77 or g77. The f77 executable is called eepasof19z. 
The g77 executable is called eepasof19zg77. This is a later version than the
one  being used by the SCEC testing center (eepasof19p.f). The manual 
for this package is 'ManualForEEPAS_Software19z.doc' (in Microsoft Word 
format.  

[The only practical difference between the 19p and 19z versions for these
models is that 19p writes "0" and "30" for the depth limits in the grid cells
even though it computes the expected numbers corresponding to the depth range
(depmin, depmax) entered using the CATALOG directive. The 19z version writes
the actual values of depmin and depmax, which are 0 and 40 for the NZ tests.] 

4. Parameter files for each model are supplied (eg nzeepas_0r.par). These
will not be changed throughout the tests. Demo command files are also 
supplied (eg newgridnzeepas_0r.eep). Only the dates defining the end of
the input catalogue and the beginning and end of the test period, and the 
name of input catalogue, and (if you wish) that of the output files need to 
be changed to produce the forecast for a different time period.

5. Although the models will all accept data up to just before the three 
month test period, I am interested in testing them with a 50 day lag,
i.e., with data up to 50 days before the start of the test period. Hence
I have specified "dely 50" in the parameter files. This can be changed 
if you want to test the models with a different lag, or no lag at all
("dely 0").

6. For efficiency of computations, I compute the forecasts on a rectangular 
grid larger than the test region. Post-processing is necessary to select
the grid cells that are inside the test region. The values are meaningless 
for grid cells outside the test region.

7. Example: To compute the grid for PPE and EEPAS_0r models, enter

eepasof19zg77 newgridnzeepas_0r.eep 

The grid expectations for PPE are then written
to the file 'ppemodeloutput.grd' and those for EEPAS_0r to the file 
'eepasmodeloutput.grd'.

8. Note that it is necessary either for the files "licencekey.dat",
"nztestpoly.dat" and "normalintegral.dat" to be sitting in the directory from 
which the program is run, or for the pathname to these files to be changed in
the command files. 

9. I don't know whether you need this, but the program reformatscec.f changes
the catalog from the format they were previously using at SCEC to the format 
required by EEPSOF. Compile with g77.

10. The PPE model is also submitted as a five year model for the whole 
catalog and the declustered catalog. The files of grid expectations are 
"ppe5yrnzwhole.grd" and "ppe5yrnzdec.grd", respectively. The analogous five 
year SUP model forecasts are given in "sup5yrnzwhole.grd" and 
"sup5yrnzdec.grd". 

11. My version of the ETAS model is submitted for 1 day testing. The input
command file is newgridnzetas.eep. This is similar to the EEPAS command files
except that it has few less directives (WEGHTS, COMP, and LIKE are not 
needed). It runs in under 20 minutes on totara with a full catalogue of 
M>2.95 from 1952. Most of the time is consumed in computing the PPE component 
of the model. It runs with eepasof20k.f, i.e. 

eepasof20kg77 newgridnzetas.eep

The etas output file is "etasmodeloutput.grd".

