Part 2: Set verification region with varying observation times

Does the ETKF simply pick up on the areas where there is a large ensemble spread or is it able to locate important upstream dynamic features that are correlated with the verification region. To attempt to answer this question I produced summary maps for the easternmost verification region, centered at 40N, 0, for observation times from one to five days in advance of the verification time. I then looked at the available analysis data for the target regions indicated by the ETKF. I had access to wind pressure and temperature data for various levels in the atmosphere. The ETKF uses the variances in wind speed and temperature at the 250, 500 and 850 mb. levels to calculate the predicted signal variance that indicates the optimal target regions. I compared the target regions with the available data to see what variables might most influence the signal. Although surface pressure is not one of the variables used by the ETKF at this time it does give an indication of the dynamic features that would be expected to impact a forecast.

The target regions were all in regions that contained a high ensemble spread as

indicated by the ensemble perturbation energy but the strength and structure of the target region was significantly different from the ensemble spread. The optimal target regions were not as a rule the regions greatest ensemble spread. The ETKF did pick out areas of high ensemble spread which would appear to be dynamically connected to the verification region at the verification time. There were however many minor targets that coincides with high ensemble spread which would not be expected to propagate into the verification region at the verification time because they were located too far downstream. These long distance correlation would appear to be spurious and are most likely an artifact of the model. The target regions moved in the downstream direction between 20 and 30 degrees of longitude per day as the forecast time was decreased. This speed is in agreement with my initial assumption of the propagation speed used in part one.

The target regions for the most part coincided with the general location of the jet stream but not with the regions of maximum wind speed. There were also significant target regions that could not be associated with the jet stream, mainly over the Baja peninsula and along the U.S. eastern coastline. The target regions generally seemed to encompass regions where there is a sharp change in the direction of the upper-level winds with longitude across the target area. This is apparent from both the 200mb wind analysis and the 850-200mb averaged winds. I was unable to locate an reanalysis of vorticity but it is obvious that the target regions are generally regions of large relative vorticity. The target regions are located in ares of both large positive and large negative relative vorticity. It is not so obvious how the target regions are situated relative to the position of maximum and minimum regions of relative vorticity. In relation to the surface pressure the target regions are generally located in areas where the pressure gradient changes sign (high or low pressure centers) or in the region between a high and a low pressure center. When the target region is between a high and a low it is on the upstream side of the low and on the downstream to northern side of the high. The vorticity would be expected to change sign across this region as well since the flow around high and low pressure centers are of opposite vorticity.

A comparison of the target regions to the 850mb and 500mb temperature fields shows that the targets are approximately parallel to the isotherms rather than cross isotherm. The target regions with maximum predicted signal variance tend to be in regions with rather strong temperature gradients but not necessarily where the sharpest gradients exist. The 250mb temperature field seems to have a greater influence on the signal than the 500 and 850 mb. temperatures. There is an isolated warm pocket at 30N, 110W over the Baja on the 21^st and22nd that is co-located with the strongest target region. This targeting of this area and the ensemble spread here are likely due to this temperature anomaly. The target region that appears near Maine on the same days the also coincides with a 250mb warm spot. Again the temperature anomaly would seem to be the cause for the high ensemble spread and the targeting of this area. As noted earlier these were the two main regions that were not associated with the location of the jet stream. By the 23^rd the region off Maine is associated with the surface cyclone that was initially targeted. At this time the region is a targeted for both dynamic and thermodynamic reasons. Except for these two instances where there was an isolated heat anomaly it would seem that the target regions are more closely related to the dynamic features.