IDL Routine: CYCLONE_TRACKS

Platform(s) procedure has been tested on (print of !VERSION):

{ sparc sunos unix 5.3 Nov 11 1999}
{ sparc sunos unix 5.4 Sep 25 2000 64 64}

Note date: 9 Jul 2001

• for users running IDL on solaris: thanks to RSI and the folks at the CIRES computing facility, it looks like there's a more robust fix for the hanging-up problem described in the 10 May 2001 note...it has to do with a bug in the FlexLM license manager. if you just do a search in google groups in the comp.lang.idl-pvwave newsgroup with the following keywords, you should see the pertinent thread: genver, flexlm, asleep, johnny.

Note date: 18 May 2001:

• if you have a very long timeseries, use the Assoc_Slp keyword option to enable reading the SLP data directly from file.

Note date: 10 May 2001:

• for users running IDL on solaris: for some reason during the test simulation month, IDL kept hanging up. i thought it was a license manager progrma, but the machine i'm running IDL on uses a FLEX manager, and is its own license server, so we're tentatively ruling that out as a problem. for some reason, if i have IDL dump out (using SAVE) all the system variables at the end of each timestep, this sometimes prevents IDL from hanging up. so, i've addded a keyword Save_Each_Delt to CYCLONE_TRACKS for solaris users to activate that SAVE command.

Note date: 8 May 2001

i've tested this on a 1 month control run of ARCSyM, and it seems to compare favorably with a manual check. a few caveats/notes:

• the thing is SLOW :( w/ a 45x27 2-D grid, with 31 days of 6 hrly values, it took a little less than 1 hr of wall-clock time using c. 40% of the cpu on a sun enterprise 250 running solaris (sun os 5.8).

• the timeseries Serreze used was 12 hrly. since ARCSyM output is 6 hrly, i've set the default search parameters in the function for a timestep of 6 hrs. if you have data of different time resolution, you'll have to set the keywords to different values. see the commenting for both CYCLONE_LOC and CYCLONE_TRACKS to see what keywords can be changed/called at the CYCLONE_TRACKS call line. note that the _Extra keyword is used on the CYCLONE_LOC call within CYCLONE_TRACKS, so keywords that CYCLONE_LOC needs are seen if defined at the CYCLONE_TRACKS call line.

• in my test month, i found it worked better using Search_Rad_Min=600e3 as a keyword (you can set this in the CYCLONE_TRACKS call line) instead of the Serreze (1995) and Serreze et al. (1997) derived default value. note that even though i change Search_Rad_Min, i didn't change my value Edge_Distance (which is used to filter out edge effects) from the default value.