Accumulated precipitation in ARL files

Topics about the HYSPLIT trajectory model.
Post Reply
samhyon
Posts: 3
Joined: September 11th, 2019, 8:29 am
Registered HYSPLIT User: Yes

Accumulated precipitation in ARL files

Post by samhyon »

I am comparing precipitation from 3-hourly, 1-degree GDAS ARL data downloaded from the NOAA archive with precipitation from 6-hourly, 1-degree ERA-Interim ARL generated using grib2arl. I extracted the first couple days worth of timesteps from each and made a timeseries of area weighted global mean precipitation (see figure).

Looking at GDAS, there is an oscillation in global mean as the timesteps progress whereas there is not for ERA-Interim. ERA-Interim was originally processed using grib2arl from a sequence of 12-hour forecasts with a 6hr step in their native GRIB format. The resulting ARL files showed exactly the same oscillation as GDAS due to the way the precipitation accumulation is treated. It’s only because we explicitly corrected for this by “de-accumulating” (i.e. subtracting the preceding step=6 accumulated precipitation from each step=12) that the oscillation went away for ERA-Interim.

Question 1: how were the GDAS ARL files generated using grib2arl? Can you please confirm what causes the observed oscillation in the GDAS ARL files?

My understanding of NOAA generated GDAS ARL files is that the high values in the oscillation must be labelled as TPP6 and the low values TPP3 for HYSPLIT to interpret these files correctly. Is this the case?

Question 2: is HYSPLIT able to correctly handle the “de-accumulated” format used in the ERA-Interim ARL file?

Specifically, does HYSPLIT always assume a certain pettern of precipitation accumulation in the ARL files? Or is HYSPLIT able to handle different accumulation times to the GDAS format?

Question 3: does HYSPLIT expect the accumulation of precipitation to be based on the time before the current timestep? For example, accumulated precipitation at hour 00:03 is the total precipitation that occurred between 00:00 and 00:03. For 00:06, the time window would be 00:03-00:06 etc.

Hopefully my questions are clear! Any help is very much appreciated!
Attachments
gdas_erai_comparison.png
alicec
Posts: 416
Joined: February 8th, 2016, 12:56 pm
Registered HYSPLIT User: Yes

Re: Accumulated precipitation in ARL files

Post by alicec »

Different NWP models output precipitation differently. Many output precip accumulated over a certain time period.
See an explanation for 6 hour accumulations (TPP6) here.
viewtopic.php?t=433

HYSPLIT can handle precip in different formats.
https://www.ready.noaa.gov/hysplitusersguide/S141.htm

If you are doing your own conversion from grib2 to ARL format, then you need to make sure
that the configuration file uses the appropriate code (TPP3 for 3 hour accumulations, TPP6 for 6 hour accumulations etc.)

HYSPLIT calculates a precip rate (https://www.ready.noaa.gov/hysplitusersguide/S614.htm) which is valid for the time period over which the precip was accumulated.

An example:
NWP model outputs TPP6 (6 hour accumulated precipitation). Outputs are every 3 hours.

At 3z precip=2 (2 units of rain fell from 0 to 3z)
At 6z precip=5 (5 units of rain fell from 0 to 6z)

For times > 0 and <= 3z HYSPLIT uses precip rate of 2/3h
For times >3z and <=6z HYSPLIT uses precip rate of (5-2)/3h

Hope that helps.
Post Reply

Return to “Trajectory Model”