SEED not affecting particle positions in dispersion mode

[kstudens]

I'm running HYSPLIT v5.4.2 in particle dispersion mode from the Windows shell. I want to set a seed for my dispersion runs to allow for replicability, and to generate multiple runs from the same starting locations with different particle trajectories. However, I'm finding that different values of SEED in the SETUP file doe not result in different particle positions across runs. I'm using the default KRAND = 0 with only 3 starting locations and NUMPAR = 1 (as a test). I found that changing a single starting coordinate's latitude by 0.001 (out of 3 total coordinates) effectively changed the seed, as the particle positions originating from the unaffected starting points were different from the previous run.

I realize I could increase NUMPAR to effectively achieve the same goal but I'm trying to keep my individual runs small since I need to repeat them in my modelling framework, and I have several thousand starting points.

Any advice on this would be greatly appreciated, thanks!

[sonny.zinn]

If you are using HYSPLIT 5.4.2 or an earlier version, set the SEED value in SETUP.CFG to a negative number to reach your goal.