I’m interested in understanding the filter_window and filter_order parameters in the get_dd_wind_field function. Specifically, I’d like to know how they affect the output and why I don’t see much difference between filter_order=2 and filter_order=3, but notice a difference between filter_window=15 and filter_window=10. Can anyone provide more information on the working process behind these parameters?"
Hello. After the retrieval finishes, PyDDA will use a low pass filter on the wind field to smooth out low level noise in the retrieval. The filter is done on a rolling window of filter_window pixels. The filtering is done by interpolating the data in the filter window to a filter_order order polynomial.
Thank you for the clarification! I’ve tested several configurations and found that using a filter_window of 10–15 with filter_order=3 provides smoother and more stable results, particularly for the vertical wind component. Do you have any additional suggestions that could help further improve the quality of the retrieval?