Advanced Real-Time Gempak Tutorial

(if all goes well)

Rather than providing an exhaustive description, I shall provide just enough information to point you in the right direction, and allow the Gempak online help (or a few well-placed questions) to fill in the gaps.

Actually, everything here should be considered in a state of flux. If something doesn't work, try one of the alternatives. Also, email me at n-g@tamu.edu whenever you encounter a systematic problem, or mail rmwhite@tamu.edu whenever you encounter a temporary (data-flow-related) one.

Real-time data display

Most files have the time (in Greenwich) encoded in the file name. For example, hourly data files have the two-digit year, month, day, and hour in the file name. The file for 0000 UTC 24 March 1996 would be 96032400_sfc.gem, and the file for the next hour would be 96032401_sfc.gem. In this document, such filenames are abbreviated as yymmddhh_sfc.gem. So don't type the y's and d's, substitute the appropriate dates and times.

1. Surface data

   • Useful surface data programs include sfmap and, less commonly, sfgram. Explaining the intricacies of sfgram is beyond the scope of this document, but you can use sfmap to plot surface weather maps.

   • Surface airways observations (a.k.a. "hourlies") are stored in two places. Files for each individual hour are kept in /data/gempak/surface/yymmddhh_sf.gem, while files for a full day are kept in /data/gempak/surface/sao/yymmdd_sfc.gem. Ideally, both types of files will have identical data, but in practice you can sometimes find data in one file but not the other.

     We also receive "synoptic" observations from all of North America. They are available every six hours in /data/gempak/surface/syn/yymmdd_sfcsyn.gem. Finally, hourly buoy observations are found in /data/gempak/surface/buoy/yymmdd_buoy.gem, and six-hourly ship observations are found in /data/gempak/surface/buoy/yymmdd_shp.gem. You can use sfmap with clear=no to overlay several different types of data on a single map.

   • WARNING: Always specify the dattim (dd/hh). Don't just use dattim=last. Often, one or two observations come in with the wrong time attached to them, making the "last" time one hour into the future. While you may like to look at data from the future, there's not much available, and the future can change without warning.

   • Key parameters:

     AREA

     Usually, you should use area=dset. This tells Gempak to plot all data from that data set that fits within garea.

     SFPARM

     This parameter controls not only what gets plotted at the station, but how big the symbols and wind barbs are. You should customize this to some preferred set of parameters. A good start is my own preference: sfparm=skyc:.7;tmpf;wsym;smsl;brbk:1:2;dwpf. Type phelp sfparm from within sfmap to find out more (lots more), including what all of that means. Note that not all possible parameters listed under the phelp are plottable; just the ones that are either directly decoded or can be calculated from directly decoded data.

     COLORS

     Use the list of colors to plot different parameters in different colors. If you have a color terminal, it makes surface maps a lot easier to read.

     FILTER

     You can avoid plotting data if it would overwrite other data. Filter determines whether to do this, and how much overlap is allowed.

     TEXT

     Since what you're plotting are numbers, the text font size, thickness, etc. is particularly critical for readability.

     MARKER

     Usually, if you're plotting sky condition, you won't also want a separate station marker. So use marker=0.

     WIND

     As with any program which plots winds, you should use wind=bk1 until you know what you're doing.

2. Upper air data

   • Useful upper-air data programs include snmap, snprof, and sncross. Snmap basically works just like sfmap (see above), so it won't be described in much detail here. Snprof plots a sounding diagram, while sncross plots a cross section.

   • Upper-air data can be found in one place: /data/gempak/upperair/yymmdd_upa.gem.

   • WARNING: The most common "mystifying problem" is with the parameter ptype. For snprof, the best ptype is skewt, but if you try this in sncross or some other program, you'll get an error. Most other programs like ptype=log.

   • Some parameters should just be set, because that's the way it is. For snprof, I suggest the following:
     • snparm=tmpc;dwpc
     • line=2;3/1/4
     • ptype=skewt
     • vcoord=pres
     • wind=bk1
     • winpos=1
     • yaxis=///;1
     • xaxis=///;1
     • filter=.5
     • thtaln=5
     • thteln=6
     • mixrln=3/2
     To specify the station, use the area variable. Specify the station id after an @ symbol, such as area=@lch. Nearby stations are LCH, CRP, and GGG. For other stations, see the next paragraph.

   • The other parameter to worry about, in sncross is cxstns. This specifies the stations which make up the cross section. To find out the available stations, run snmap with STID as one of the snparms. Then, based on the map and the desired location of the cross section, you might specify something like cxstns=bro;crp;fwd;oun;top;oax;mpx;inl;wpl;yyq;ybk;yrb;weu, especially if you like cross sections that span fifty degrees of latitude.

     Don't forget to specify something plottable, like thta, for snparm. A cross section of station ID's is not very interesting. Also, if you've been following along in order and haven't changed line yet, you'll get a Christmassy looking plot. Come on, take control of your Gempak! Specify the line colors, contour intervals, etc.! Do something useful!

3. Analyses

   Sorry, folks, analyzing surface hourly data and plotting contours is beyond the scope of this document. For the time being, you'll have to content yourself with plotting the analyses from the NMC (recently renamed NCEP) forecast models, described below.

4. Forecasts

   • There are lots of programs for gridded data. All of them start with "gd". I'll describe gdcntr and gdinfo here, and discuss the rest of them in a later section.

   • We receive data from lots of models. Here is the table of filenames. The interval between runs is how frequently the model is run by NCEP. The output frequency is the interval between output grids. For example, the NGM is run every 12 hours by NCEP, and we receive output at six-hourly intervals out to 48 hours for each model run.

   NWP FORECAST GRID FILES
   ...........Model............	File name	Interval between runs (hours)	Period of run (hours)	Output frequency (hours)	Grid levels
   Nested Grid Model	/data/gempak/hds/yymmddhh_ngm211.gem	12	0-48	6	every 50 mb
   Early Eta Model	/data/gempak/hds/yymmddhh_eta.gem	12	0-48	6	every 50 mb
   Rapid Update Cycle	/data/gempak/hds/yymmddhh_ruc.gem	3	0-12	3	every 50 mb
   Global Spectral Model (high resolution, N. Amer.)	/data/gempak/hds/yymmddhh_avn211.gem	12	0-48	6	mandatory levels
   Global Spectral Model (low resolution, global)	/data/gempak/hds/yymmddhh_thin.gem	12	0-72	6-12	mandatory levels
   Global Spectral Model (medium range forecasts)	/data/gempak/hds/yymmddhh_mrf.gem	00Z only	84-240	12	surf, 500mb only
   European Center Model	/data/gempak/nps/yymmddhh.ecmwf_b.gem	12Z, late	84-240	12	surf, 500mb only

   • These data trickle in throughout the day. You can either confidently try to use these files, or use the Unix ls command to check to see if the file size of the most recent file is similar to the file size of older, complete files.

   • In gdcntr, or any other grid program, you must specify the grid function, or grid vector function, to plot. To do this, you have to know what raw grids are in the file. Gempak is very particular about spelling, so if you make any mistake in gfunc, gvect, glevel, gvcord, or gdattim, you'll get an error saying the grid was not found. And then, if Gempak is being particularly sinister, it will plot some other grid instead.

   • To find out what's in a grid file, use gdinfo. This program takes only three parameters, the file name, the list flag, and the output device. Specify the grid file as one of the entries in the table (substituting dates and times for the yy's and dd's, etc.), and use lstall=yes and output=t. Then run the program. You'll get a list of all grids in the file. Either memorize some of them, or write them down, including the grid level, vertical coordinate and everything else. Use this information to set the appropriate parameters in gdcntr, or any other grid program.

   • Once you've successfully plotted a simple grid, read the help for gfunc. There's an amazing morass of possibilities. Make sure you keep scalars (gfunc) and vectors (gvect) separate. Gdcntr will plot scalars or scalar functions of vectors; gdwind and gdstream will plot vectors or vector functions, and gdplot will plot both. Other potentially useful grid programs are gdprof and gdcross. Try them all.

5. Radar and satellite imagery

   • WARNING: To successfully display radar and satellite imagery, it may be necessary to remove other graphics processes and images from your screen. The symptoms will be (1) no radar or satellite image ever actually gets drawn; and (2) you get the message [GEMPLT -46] NCLRAL - Can not allocate read/write colors. If this happens to you, delete other graphics windows, particularly those which might be using lots of colors. Some unfortunate people never seem to be able to get this program working.

   • ADVICE: If you're giving a map discussion and want lots of graphics, begin your session by plotting a radar or satellite image with Gempak. Once Gempak allocates the colors successfully, it won't let them go.

   • For all programs that do radar and satellite overlays, you will need to worry about two parameters: the radar or satellite file name and the projection. To get you started, try plotting the current radar composite. The program that just plots the radar or satellite is gpmap. Go into that program, set radfil=/data/nexrad/MASTER (the capital letters do matter in the file name), set proj=rad, and set clear=yes. Pick a good garea, or just set garea to the two-letter state abbreviation of your choice, and run it. (Good luck...)

   • All other map-drawing programs, if you look closely, also have radfil and satfil parameters. So you can overlay observations or gridded data, by just specifying a radar file and setting the projection to rad. Conversely, if you've figured out a satellite file name and want to overlay satellite data, set the projection to sat.

   • Besides the composite file /data/nexrad/MASTER, we receive radar products from several radars in the area. The naming convention is: 1. the directory (/data/nexrad/), the radar site (EWX, FWS, GRK, HGX, JAN, LCH, MLB, or TLX), and the product type (such as BREF1, VEL1, PRET, or VIL). For example, to plot the elevation angle 1 base reflectivity from Houston, the radfil would be /data/nexrad/HGXBREF1. Of course, your garea must include the radar site; something like garea=hgx* would be appropriate.

   • The satellite data is in the directory /data/ssec/data. The file naming convention is AREA followed by the four-digit area number. The area numbers are the same as in McIDAS: 1200's for West Coast visible, 1300's for West Coast infrared, 1400's for East Coast visible, 1500's for East Coast Infrared, 1700's for West Coast water vapor, and 2100's for East Coast water vapor.

   • To find out which area file to use for the SATFIL parameter, you need to do a listing of the data directory. For example, if I want to plot the current East Coast visible image, I would issue the command (to the shell, not to GEMPAK)
     ls -l /data/ssec/data/AREA14* and see what happens. Here's some sample output:

   -rw-rw-rw-    1 ldm      data      607776 Feb  3 04:37 /data/ssec/data/AREA1400
   -rw-rw-rw-    1 ldm      data      607776 Feb  3 05:39 /data/ssec/data/AREA1401
   -rw-rw-rw-    1 ldm      data      607776 Feb  3 06:39 /data/ssec/data/AREA1402
   -rw-rw-rw-    1 ldm      data      607776 Feb  3 07:38 /data/ssec/data/AREA1403
   -rw-rw-rw-    1 ldm      data      607776 Feb  3 08:40 /data/ssec/data/AREA1404
   -rw-rw-rw-    1 ldm      data      607776 Feb  3 09:39 /data/ssec/data/AREA1405
   -rw-rw-rw-    1 ldm      data      607776 Feb  3 10:39 /data/ssec/data/AREA1406
   -rw-rw-rw-    1 ldm      data      607776 Feb  3 11:39 /data/ssec/data/AREA1407
   -rw-rw-rw-    1 ldm      data      607776 Feb  3 12:39 /data/ssec/data/AREA1408
   -rw-rw-rw-    1 ldm      data      607776 Feb  3 13:40 /data/ssec/data/AREA1409
   -rw-rw-rw-    1 ldm      data      607776 Feb  3 14:40 /data/ssec/data/AREA1410
   -rw-rw-rw-    1 ldm      data      607776 Feb  3 15:52 /data/ssec/data/AREA1411
   -rw-rw-rw-    1 ldm      data      607776 Feb  3 16:42 /data/ssec/data/AREA1412
   -rw-rw-rw-    1 ldm      data      607776 Feb  3 17:40 /data/ssec/data/AREA1413
   -rw-rw-rw-    1 ldm      data      607776 Feb  3 18:39 /data/ssec/data/AREA1414
   -rw-rw-rw-    1 ldm      data      607776 Feb  4 04:40 /data/ssec/data/AREA1415
   -rw-rw-rw-    1 ldm      data      607776 Feb  4 05:38 /data/ssec/data/AREA1416
   -rw-rw-rw-    1 ldm      data      607776 Feb  4 06:39 /data/ssec/data/AREA1417
   -rw-rw-rw-    1 ldm      data      607776 Feb  4 07:40 /data/ssec/data/AREA1418
   -rw-rw-rw-    1 ldm      data      607776 Feb  4 08:41 /data/ssec/data/AREA1419
   -rw-rw-rw-    1 ldm      data      607776 Feb  4 09:42 /data/ssec/data/AREA1420
   -rw-rw-rw-    1 ldm      data      607776 Feb  4 10:41 /data/ssec/data/AREA1421
   -rw-rw-rw-    1 ldm      data      607776 Feb  4 11:42 /data/ssec/data/AREA1422
   -rw-rw-rw-    1 ldm      data      607776 Feb  4 12:40 /data/ssec/data/AREA1423
   -rw-rw-rw-    1 ldm      data      607776 Feb  4 13:44 /data/ssec/data/AREA1424
   -rw-rw-rw-    1 ldm      data      607776 Feb  4 14:53 /data/ssec/data/AREA1425
   -rw-rw-rw-    1 ldm      data      607776 Feb  4 15:48 /data/ssec/data/AREA1426
   -rw-rw-rw-    1 ldm      data      607776 Feb  4 16:42 /data/ssec/data/AREA1427
   -rw-rw-rw-    1 ldm      data      607776 Feb  4 17:41 /data/ssec/data/AREA1428
   -rw-rw-rw-    1 ldm      data      607776 Feb  4 18:39 /data/ssec/data/AREA1429
   -rw-rw-rw-    1 ldm      data      607776 Feb  5 04:42 /data/ssec/data/AREA1430
   -rw-rw-rw-    1 ldm      data      607776 Feb  5 05:38 /data/ssec/data/AREA1431
   -rw-rw-rw-    1 ldm      data      607776 Feb  5 06:38 /data/ssec/data/AREA1432
   -rw-rw-rw-    1 ldm      data      607776 Feb  5 07:39 /data/ssec/data/AREA1433
   -rw-rw-rw-    1 ldm      data      607776 Feb  5 08:42 /data/ssec/data/AREA1434
   -rw-rw-rw-    1 ldm      data      607776 Feb  5 09:39 /data/ssec/data/AREA1435
   -rw-rw-rw-    1 ldm      data      607776 Feb  5 10:41 /data/ssec/data/AREA1436
   -rw-rw-rw-    1 ldm      data      607776 Feb  5 11:41 /data/ssec/data/AREA1437
   -rw-rw-rw-    1 ldm      data      607776 Feb  5 12:48 /data/ssec/data/AREA1438
   -rw-rw-rw-    1 ldm      data      607776 Feb  2 10:38 /data/ssec/data/AREA1439
   -rw-rw-rw-    1 ldm      data      607776 Feb  2 11:39 /data/ssec/data/AREA1440
   -rw-rw-rw-    1 ldm      data      607776 Feb  2 12:38 /data/ssec/data/AREA1441
   -rw-rw-rw-    1 ldm      data      607776 Feb  2 13:38 /data/ssec/data/AREA1442
   -rw-rw-rw-    1 ldm      data      607776 Feb  2 14:39 /data/ssec/data/AREA1443
   -rw-rw-rw-    1 ldm      data      607776 Feb  2 15:39 /data/ssec/data/AREA1444
   -rw-rw-rw-    1 ldm      data      607776 Feb  2 16:39 /data/ssec/data/AREA1445
   -rw-rw-rw-    1 ldm      data      607776 Feb  2 17:38 /data/ssec/data/AREA1446
   -rw-rw-rw-    1 ldm      data      607776 Feb  2 18:38 /data/ssec/data/AREA1447
   

   • The most recent image will be the one created most recently; in this case, AREA1438 is the most recent image. (You may need to scroll to the right in your Web browser to see the file names.) So, in my GEMPAK plotting program, I would specify SATFIL=/data/ssec/data/AREA1438 and PROJ=sat to display this satellite image.