NCEP Service Center Evaluations SPC Steve Weiss/David Bright

• The max-min products showed that the addition of 6 WRF members tended to produce larger (or smaller) extreme values compared to the operational SREF, indicative of an increased spread.  This suggests that the number of observed outliers should be reduced in the 21 member SREF.

• The parallel SREF intersection (or grids points where all members exceed a threshold value) showed a tendency to contract in area, whereas the union (or grid points where at least one member exceeds a threshold value) showed little change from the operational SREF.  This is consistent with an increase in variance but not an excessive increase in spread.

• The combined probability products in the parallel SREF identified similar areas as the operational SREF, but probability values were typically of lower value.  This is also indicative of an increase in parallel system spread.  This may require a small “human recalibration” by SPC forecasters when interpreting the combined probabilities in the new SREF, especially near the northern extent of severe potential in return flow situations where the probability of moisture/CAPE exceedance can exhibit lower values compared to the operational SREF. 

• The calibrated severe probabilities from the parallel SREF are generally similar in location and orientation to products from the operational SREF.  Again, there is a tendency for the probability maxima from the parallel SREF to have lower values relative to the product from the operational SREF.  We expect this result to diminish with time as the “moving” calibration tables gradually become better tuned to the 21 member SREF performance after implementation.  Overall, we expect a smooth transition once the parallel SREF is implemented.