HYSPLIT Forum: hysplitbbs.arl.noaa.gov

I downloaded registered version "hysplit.v5.3.0_RHEL8.8.tar.gz" and installed it on our Centos 8 Linux. Then I converted the WRF output files into ARL format for the time periods that I am interested in, and plotted the trajectories. When I examine the output gif plot, the forecast initialization time there is confusing me while trajectory time period is correct, so I am trying to understand how the title is created on the HYSPLIT plot.

The attached sample gif file shows that backward trajectories were plotted for 24h period ending at December 24, 2023 12Z using the ARL files (I could not attach one as a sample as it was too big, even after compression). For this particular time period, the input files for WRF runs were created by WPS starting from September 15, 2023 12Z for several months but as hourly input files for WRF runs and the model is run for 5 day period with a 1 day spin-up using the appropriate input WPS files. Then I expect that the title of the graph should/might show September 15, 2023 12Z as the forecast initialization date, but instead it shows the forecast initialization time as November 12, 2023 20Z.

Here is the information from WRF output:

// global attributes:
:TITLE = " OUTPUT FROM WRF V4.4.2 MODEL" ;
:START_DATE = "2023-09-15_12:00:00" ;
:SIMULATION_START_DATE = "2023-09-15_12:00:00" ;

and the output date/times are given as:

data:

Times =
"2023-12-24_12:00:00",
"2023-12-24_13:00:00",
"2023-12-24_14:00:00",
"2023-12-24_15:00:00",
"2023-12-24_16:00:00",
"2023-12-24_17:00:00",
"2023-12-24_18:00:00",
"2023-12-24_19:00:00",
"2023-12-24_20:00:00",
"2023-12-24_21:00:00",
"2023-12-24_22:00:00",
"2023-12-24_23:00:00",
"2023-12-25_00:00:00",
"2023-12-25_01:00:00",
"2023-12-25_02:00:00",
"2023-12-25_03:00:00",
"2023-12-25_04:00:00",
"2023-12-25_05:00:00",
"2023-12-25_06:00:00",
"2023-12-25_07:00:00",
"2023-12-25_08:00:00",
"2023-12-25_09:00:00",
"2023-12-25_10:00:00",
"2023-12-25_11:00:00" ;

I tried to find the code that creates the title of the graph within hysplit.v5.3.0_RHEL8.8 directory, but couldn't. I attached the zip file that includes CONTROL, TRAJ.CFG, MESSAGE, WARNING, tdump.Calexico.bwd.24hr and trajplot.Calexico.bwd.gif files for your information. Please let me know if I am doing something wrong.

I also have a second question. When I am calculating forward or backward trajectories for a 24 hour period, how many 24 hour input files do I need to provide within the CONTROL file? My trial and error showed me that I needed to provide 3-4 days before and after time period of interest for a successful plot run.

Thank you for your help.

We do not see a trajectory dump file in your zip file. Please have it available for diagnosis.

Examination of your MESSAGE file shows that meteorology data from Dec 24, 2023 at 12Z to Dec 23 at 12Z were used for your 24 hour backward trajectory run. Supposing your daily meteorology files start at 12Z and end at 11Z the following day, you would need two files starting at Dec 23 at 12Z and Dec 24 at 12Z.

Thanks for the feedback. I provided the tdump file (I thought I included that earlier. My mistake. Sorry about that.).

Thanks for making the trajectory file available. We see all of the meteorology forecast hours are 999. Subtraction of 999 hours from 1200Z, Dec 24, 2023 yields 2100Z, Nov 12, 2023 which is just one hour off from the date and time of meteorology initialization on your plot. We will need to look into why there is one hour offset. However, this explains why the meteorology initialization on your plot is shown as Nov 12, 2023 not as Sep 15, 2023 as you expected.

Thank you for your explanation. In general, it is important to know when the forecast start time was, for example, 48 hour forecast into the future. But for our own multi-month simulations, the forecast start time of Sep 15, 2023 12Z that is dumped into the WRF output files is not actually forecast start time and is not relevant to us since we generate the WRF input files for several months via the WPS system and the forecast start time dumped into the WPS outputs is not really the start time of the forecast. This is because we run WRF in 5-day chunks with 1-day spin up added onto the 5 day for every 5 day over multiple months. In this setup, actual forecast start time is when we start the model run for that particular episode, which is at the beginning of 6th day for that episode. But WRF reads the forecast start time embedded into WPS output files and that information is pushed down into the WRF outputs.

Just for generation of visually correct title purposes, I tweeked the tdump file to reflect information for the header with Sep 15, 2023 12Z as the forecast start time. Original 999 hours given in the tdump file needs to be 2399 hours. I attached the modified tdump file and generated plot output here for your information. Whether or not to modify the graph code to use the forecast start time given in WRF header, if any, to show the posted forecast start time is certainly in your discretion. Thank you very much again for providing the explanation of the data in tdump file for the title generation.