how_to_build_the_GEAR1_XML_file (and also, how to display and test it): This is an expanded version of the forecast-creation instructions in: Bird et al. [2015, Bull. Seismol. Soc. Am., "GEAR1: ..."]. Notice that it is written for users of Windows; others may be able to get the same results on other computer O/Ss, but there will be more steps. -------------------------------------------------------------------------- 1) You will need a 64-bit Windows computer to build this forecast. Near the end of the job, the memory requirement is about 3.8 GB. Allowing some memory for the Windows O/S, you will need a total of 5~6 GB of RAM. 2) You have a choice of two 64-bit versions of the executable: GEAR1_for_CSEP-Win64seq.exe is the sequential-execution version (needs ~2 days); GEAR1_for_CSEP-Win64par.exe is the parallel-execution version (needs ~3-12 hours). 3) If you choose to run the parallel version, then be sure that you read and follow the directions in my file "each_WinNNpar.EXE_requires_one_DLL.txt", which will point you to my folder REQUIRED_libiomp5md.dll_file. You must have a copy of the appropriate version of Intel's libiomp5md.dll located in the same folder where you store the input data files for GEAR1_for_CSEP-Win64par.exe. 4) Download 2 other necessary files from my folder GEAR1_forecast: *GEAR1_parameters.dat.txt *GL_HAZTBLT_M5_B2_2013.zip 5) File GEAR1_parameters.dat.txt lists the 6 required data files in its first 6 lines. There are 5 other files that you will need to download from my other folder(s): From my folder SHIFT_GSRM2_forecast: *PB2002_tectonic_zones.grd.txt *underwater.grd.zip *GSRM_average_strain_v2.1.txt.Z *GCMT_shallow_m5p767_1977-2013.eqc.txt *PB2002_steps.dat.txt (which is also in my folder PB2002) 6) Uncompress the compressed data files: *GL_HAZTBLT_M5_B2_2013.zip --> GL_HAZTBLT_M5_B2_2013.TMP *underwater.grd.zip --> underwater.grd *GSRM_average_strain_v2.1.txt.Z --> GSRM_average_strain_v2.1.txt (Note that the first 2 can probably be uncompressed by right-clicking and selecting "Extract all...". To uncompress the .Z file, you might need a third-party app such as WinZip, which is on the Windows Store, and free to use for the first 15 days.) 7) Look at the contents of GEAR1_parameters.dat.txt, in a text editor: *You may NOT change anything in the first 6 lines. *You MAY change the forecast start-date, end-date, and duration. (However, it would NOT be scientifically valid to choose any start date before 2014-01-01.) *You may NOT change minTargetMagnitude. *You MAY change the name of the output file (GEAR1.global.forecast.xml), but this is probably not a great idea. 8) Rename any files that have 2 file-name extensions, where the second is ".txt", by removing the second ".txt" file-name extension: *GEAR1_parameters.dat.txt --> GEAR1_parameters.dat *PB2002_tectonic_zones.grd.txt --> PB2002_tectonic_zones.grd *GCMT_shallow_m5p767_1977-2013.eqc.txt --> GCMT_shallow_m5p767_1977-2013.eqc *PB2002_steps.dat.txt --> PB2002_steps.dat Check that your file-names now agree EXACTLY with those in GEAR1_parameters.dat. [HINT: DO NOT OMIT THE FOLLOWING STEP!] 9) Move ALL of these files off your Windows Desktop, to another folder (e.g., C:\T). (If you fail to do this, you will create a 4 GB file on your Desktop, and it will be included in every Windows System Protection restore-point file, and therefore your hard drive will be full in a few weeks!) 10) Adjust your computer's power-management Settings to Sleep: Never. (Otherwise, your job will not make progress at night. Also, based on experience with previous versions of Windows, one may not have complete confidence in the Sleep/wake-up process applied to a complex, intensive parallel-processing job.) 11) Start a Windows Command Prompt (text-mode user interface w/ black screen). Navigate to C:\T using commands such as "CD C:\T". Confirm that all your files are there by typing "DIR". Start the job by typing either "GEAR1_for_CSEP-Win64seq.exe GEAR1_parameters.dat", OR "GEAR1_for_CSEP-Win64par.exe GEAR1_parameters.dat" (depending on your choice). 12) Wait about 3 hours~2 days [see above] for the job to finish. (Notes: The parallel version will use parallel processing in the most time-consuming steps, but not in all steps. Everything you may try to read on the black screen is also being copied to a log file, so do not be concerned if some text goes by too fast to follow.) 13) Assuming the job finishes normally, it will produce a 3.8-GB GEAR1.global.forecast.xml file in the format required by CSEP = Collaboratory for the Study of Earthquake Predictability. This XML file has 3 independent variables: *longitude (in 0.1-degree steps) *latitude (in 0.1-degree steps) *magnitude (in 0.1 steps) and states the number of earthquakes expected, during the forecast time window, in each longitude/latitude/magnitude cell. (Floating-point numbers are used, and most will be quite small.) 14) To reduce the forecast to a 2-D .GRD file format, suitable for plotting as a colored map, you can download and run my utility program XML_2_GRD. When it runs, you choose a threshold magnitude, and it condenses all forecast seismicity at-or-above that magnitude into one number (epicentroid rate density, in earthquakes/m^2/s) for each 0.1 x 0.1-degree geographic grid cell. The result is a global .GRD file of about 64 MB. My GRD format is documented at http://peterbird.name/guide/grd_format.htm 15) If you are only interested in a particular geographic region, you can further cut down this global .GRD file by downloading and running my second utility program Extract_Regional_GRD. 16) Either a global or a regional .GRD versions of the forecast can be plotted with my program NeoKineMap. This will produce an Adobe Illustrator .AI file. (You will have to buy a version-- possibly a cheap academic version-- of Adobe Illustrator to view, edit, & print this. Also, notice that AI menu item "File / Save for Web & Devices" will allow you to export graphics in bitmap formats like .JPG or .GIF or .TIFF, etc.) 17) Forecasts in .GRD format can be compared to earthquake catalogs in my .EQC format [see documentation attached to Bird & Kagan, 2004, BSSA: http://peterbird.name/oldFTP/2003107-esupp/eqc_format.pdf ] by either of my forecast-scoring programs: *Kagan_2009_GJI_I_scores, AND/OR *PseudoCSEP which are also available on this web site. Please read Bird et al. [2015, BSSA, "GEAR1: ..."] for retrospective score results which will provide some context for your new results Also, check the CSEP web site for "official" prospective-test results that may begin to appear around(?) October 2016... Best wishes, and Good Luck! Peter Bird UCLA pbird@epss.ucla.edu 2015.08.21