much they differ. Since the polygon method is more punitive on scoring than the county method when events are not forecast, at least a portion of the discrepancy in scores is likely due to a lack of warning experience by the NWS interns. Although no data on more experienced forecasters are currently available, future data will allow further study on this issue.
Figure 1. Scatter plot diagrams of POD as a function of warning frequency for (a) the current NWS verification method and (b) the proposed NWS verification method. The least-squares best-fit line and equation, as well as the correlation coefficient, are provided for each diagram.
4.2 Analysis of Verification Methods
To help with the analysis, the standard verification variables, along with some additional variables, were plotted on scatter charts to determine if any clear linear relationships exist. If such a relationship existed between the variables then a line was fit to the plot using a least squares approximation. A correlation coefficient was also computed for these relationships. Much of this analysis is ongoing, so only some preliminary results are shown.
Figure 1 shows such a chart for the variables warning frequency (WF) and POD. For both methods, a direct linear relationship exists between these two variables. There is a slightly greater correlation for the county verification method, but a more sloping increase in POD for the polygon verification method. The statistics from the DLOC workshop suggest that, regardless of the verification method used, issuing more warnings generally leads to higher POD scores. Since WDM I participants were working on the same scenarios, Fig. 1 is very straightforward. With a set number of severe weather events, it makes sense that increasing the number of warnings issued will result in a higher POD regardless of the verification method.
Figure. 2. Scatter plot diagrams of leadtime as a function of warning frequency for (a) the current NWS verification method and (b) the proposed NWS verification method. The least-squares best-fit line and equation, as well as the correlation coefficient, are provided for each diagram.