NSF Meeting (cont’d)

Continuing the thoughts from yesterday’s post…

Since the comparison is not the greatest (point vs. smoothed measurement), it’s unfair at this point to make any solid conclusions about the correlation of variance statistics and the reliability index (RI).  However, the preliminary data did show little to no correlation between the variance of N, dN/dt and RI.  This, again, may be purely an artifact of the area that each pixel represents.  I will be performing a 2-D Gaussian filter on the RI field and seeing what kind of comparison we get from that.  One thing to note is that even visually, the two fields look completely different as far as the area each pixel represents.

As you can see from this image, variance of the refractivity field is smoothed to approx. 4 km.  This is obvious in the pixels near the edges of the domain, where outlying points have a (roughly) spherical, 4-km radius shape.  The variance, as has been shown in the spring, is much higher SE of the radar (KTLX) and near the edges of the domain.  The large region SE of KTLX with high variance is an area with complex terrain, filled with hills and river beds and trees.  This kind of terrain is largely absent from the rest of the domain.  The high variance near the edges of the domain is likely due to clutter quality which is just good enough to pass though our filtering, but the phase varies so much that steady refractivity measurements in time are not possible.

This image shows RI for the same time period.  It is fairly obvious now why the variance is likely so high SE of the radar (KTLX).  RI in this region is very low, meaning the phase of the returned radar signal is very unstable as a function of time.  Since refractivity is derived from the signal phase, unsteady phase directly leads to unsteady refractivity measurements, leading to high variance of N over this area (see Fig. 1 above).  The suburban and urban landscape near I-35 (N-S highway west of KTLX), I-240 (W-E highway just south of Oklahoma City) and Norman (just east of I-35 and west of KTLX) has very high RI and very low variance statistics, which is due to the large number of power poles, buildings, radio towers, etc. present.

Also easily seen are the high-voltage power lines.  The clearest one runs SE from KTLX, shown by a relative spike in RI.  This leads to reduced variance even though the terrain is still complex here, likely because the returned power from the large supporting towers is much larger than power return off of trees or other clutter objects, and since the towers are stationary their returned phase will only be a function of N changes and not due to motion of the target.  Other power lines are seen N and NE of KTLX.

One of the reasons we had wanted to do a comparison of RI with variance is that RI is a direct measure of phase.  If phase was stable, then our derived N values should also be stable.  During the NSF meeting yesterday we thought of using our quality index (QI) to do further correlation statistics.  QI is a function of the observed power, radial velocity and spectrum width from a target.  Correlating QI and RI with variance may shed more light on what exactly is causing the high variance, whether it can be explained straight-forward through phase stability, or whether we must use the derived radar moments (power, velocity and spectrum width).  More on this later…

Advertisement