Filtering Results from RI/QI Smoothing

Hello all,

The script is (finally) working relatively quickly so I have some results from the smoothing algorithm.  To refresh the memory, I’m smoothing the fields of reliability index (RI) and quality index (QI) because the refractivity algorithm as it now stands has to smooth the received phase and refractivity output.  The relative resolutions are ~4 km for refractivity but only 250 m for RI and QI (calculated for each gate).  In order to do an accurate comparison we must smooth RI & QI on the order of ~4 km.

Below is an example of the response of the Gaussian curve to an input value of 1 at the center of the map.  As you can see, the effects are limited to a radius of ~5 km.  In fact the response drops below 1% (3 N-units) beyond ~6 km and below 0.1% (0.3 N-units) beyond ~7 km.  (For those who are curious: I set the variance equal to that of refractivity N, which is ~4 km.)

Gaussian Response to Single Observation of 300 N-units

The initial pre-smoothing data, as can be seen below, has high resolution, especially compared to a map of refractivity. The smoothing alluded to above is needed to bring the effective resolutions closer together for comparison purposes.

Unsmoothed Quality Index

After smoothing the resolution reduces accordingly, more in line with that expected from refractivity data.

Gaussian-smoothed Quality Index

Now we can make the comparison between the variance of refractivity/scan-to-scan and fields of RI and QI.  Time series of all quantities at one location, scatter plots of various data fields, and comparing maps of the fields together will help to qualitatively determine any correlation between them.  Again, the goal is to find a link between RI and/or QI and fields of refractivity/scan-to-scan variance, which can be used as a an additional clutter filter when refractivity is used for assimilation into the ARPS computer model.

Results of this comparison will follow…

Advertisement