It has long been known that the atmospheric conditions preceding significant weather events bear remarkable resemblance to previous occurrences of the same type of phenomena. This COMET sponsored partnership project (UCAR S03-44672) hypothesized that significant weather events had a repeatable signature that should be objectively identifiable from the anomalies of specific atmospheric analysis fields from the NCEP Reanalysis data set (Kalnay, 1996, BAMS). It was further proposed that the quantification (strength and location of the anomalies) of these patterns could be used in conjunction with real-time prediction of these fields (and their anomalies) to alert the user to the risk of such events far in advance.

Significant weather events of heavy snow, ice storms, record heat and cold, damaging winds, severe thunderstorms, hail, tornadoes, flash floods, widespread fires and forest fires were selected for part or all of the domain of the Middle Atlantic River Forecast Center (approximate area from southern New York to central Virginia) for the period from 1948-2002. These events were documented from several sources (mainly NCDC’s storm event database). From the date and time of the events, the locations of the interior maxima and minima of the climatological anomaly of the fields and their associated values were acquired, from NCEP Reanalysis data, and placed into a relational database (MYSQL). Along with those values, the distance and direction from the point of occurrence of the significant weather event occurred to location of the anomaly maxima and minima were also entered into the database.

A maximum is defined herein as a location in gridded data where its value is greater than or equal to the eight surrounding grid points. A minimum is defined as a location in gridded data where its value is lower than or equal to the eight surrounding grid points. Each field contains information for the two highest maxima and the two lowest minima.

The collection of these anomalies produced clusters of departures, i.e. recurring locations of the maxima and minima of anomalies in the various fields. An objective method of determining these clusters was developed in order to automate the entire process. The method, called “Strong Point Analysis”, determines for each field the gridded spatial distribution of maxima and minima locations. Scaled thresholding of these counts is then used to locate grid points that are unambiguously part of a cluster (i.e. - “Strong Points”). Any group of Strong Points forming a closed network on the grid is defined as a cluster. Gaps left in the cluster by finite sample size are filled in using a combination of density and structural criterion on the surrounding network.

A relationship for the predictive weight of each important anomaly was also developed and will be implemented in the algorithms used to scan real-time prediction fields and assess the risk of any of the significant events. This relationship quantifies the degree to which the maxima and minima of each anomaly field is indicative of the occurrence of the event. In operational use, the greater the number of predicted maxima and minima that matches up with highly weighted clusters, the higher the risk will be for that event. An array of the risks of these events in the MARFC domain will be presented using the 6 hour forecast fields from the operational runs of the ETA model to 84 hours. Further refinement is expected as the NCEP Regional Reanalysis data set becomes available and the significant events are ranked and regionalized.