Weird Vertical Velocities Post-PyDDA-Retrieval


I’ve run PyDDA over two PNW coastal radars, KLGX and KRTX. The initial horizontal wind field looks good (Fig. ADS1), but when I try to plot vertical velocity, I get very strange values, especially around the cone of silence of a given radar. Fig. ADS2 shows a vertical cross-section through the KLGX radar, the northernmost intersection point at (0,0) in Fig. ADS1. I’m not sure how to diagnose this issue. Has anyone seen this before?

Fig. ADS1

(see Part 2 for Fig. ADS2)

Thank you,

1 Like

Fig. ADS2

This to me looks like the horizontal winds inside the background field (cone of silence) are probably zero. This is making PyDDA think there is horizontal convergence/divergence near the cone when there isn’t and therefore is producing this artificial updraft/downdraft you see. I would impose a background constraint from a sounding to ensure that the horizontal winds inside the cone of silence are nonzero.

Hi Robert,

Thanks for your help. Using a sounding did produce better results for vertical velocity, but I’m still concerned that the results are not entirely physical and are somehow responding to the cone of silence. I’ve attached Figures ADS3 and ADS4 for reference (ADS4 is in a follow-up post - I am still somehow a “new user”, and am warned I cannot embed a second media). Here also is the code I execute to get the retrieval:

grids = pydda.retrieval.get_dd_wind_field(allg, u_init, v_init, w_init, Co=1.0, Cm=1.0, Cx=0.0, Cy=0.0, Cz=0.0, Cb=0.0, frz=3000.0, wind_tol=0.5, mask_outside_opt=True, upper_bc=1, engine=‘tensorflow’, vel_name=newvel, refl_field=newref)

Any suggestions? Perhaps I’m missing a parameter? One thing I can try is to ensure values in the cone of silence are set to NaN and not 0. I wonder if anyone who uses NEXRAD data has had to do this (or something like this) in their pre-processing to account for the cone of silence.

  • Anthony

Figure ADS3

Figure ADS4