Why are some "good" trajectories not clustered?

Topics about the trajectory clustering program for HYSPLIT.
Post Reply
wuy
Posts: 5
Joined: April 8th, 2014, 12:50 pm
Registered HYSPLIT User: Yes

Why are some "good" trajectories not clustered?

Post by wuy » April 16th, 2014, 2:57 pm

Some trajectories seem good, but not are clustered or excluded by the current HYSPLIT cluster algorithm.
In fact, these trajectories are close to the clustered mean trajectory.
Why are these 'good' trajectories not clustered?
Is there a way to include them?

Thanks for your assistance.

Wu

barbara.stunder
Posts: 260
Joined: November 9th, 2012, 4:23 pm
Registered HYSPLIT User: Yes

Re: Why are some "good" trajectories not clustered?

Post by barbara.stunder » April 22nd, 2014, 8:02 am

Are you using the cluster "Trajectory skip" feature that skips (does not include) some of the trajectories? A value of 2 means the cluster program will use every other trajectory, though the same INFILE is created no matter what the skip value is. The output CLUSLIST file does not itemize the files skipped as cluster 0.

Look at the INFILE and the file CLUSLIST. A skip of "1" means there is no skip and all files are used. They should all be in CLUSLIST as either cluster 0 (excluded because the trajectory terminates before the specified "hours to cluster") or another cluster.

wuy
Posts: 5
Joined: April 8th, 2014, 12:50 pm
Registered HYSPLIT User: Yes

Re: Why are some "good" trajectories not clustered?

Post by wuy » April 23rd, 2014, 10:29 am

Thank you.

A skip of "1 is used.
In the total 62 trajectories, 4 of them looks "good" but not used or missed in the final cluster results.

Wu

barbara.stunder
Posts: 260
Joined: November 9th, 2012, 4:23 pm
Registered HYSPLIT User: Yes

Re: Why are some "good" trajectories not clustered?

Post by barbara.stunder » June 30th, 2014, 9:19 am

In the trajectory cluster program, you select the time duration of the trajectories over which the clustering is done. For example if you have 36-h trajectories, you might choose 36-h. The trajectories that terminated early (before 36-h) would not be used in the clustering. Trajectories would terminate early if they reach the edge of the horizontal domain of the meteorology file, if they vertically reach the top of the meteorology file (related to the "top of model" value specified in the CONTROL file), if they reach the end time of the meteorology file, if there are missings in the meteorology file. Information in the MESSAGE file may describe why the trajectory terminated early.

The trajectories not used are given as cluster number zero (0) in the CLUSLIST file.

Post Reply