duration series and, therefore, did not have the same likelihood of occurrence 

 as the other 26 members in the ensemble. Therefore, these two extra storms 

 were combined with a fewer number of tides. To determine the number of events 

 which should be formed using these two storms, the following simplified analy- 

 sis was used. The 1945 and 1978 storms had the first and second largest 

 surges in a 58-year annual series (the length of available data). Assuming a 

 Weibull plotting position formula, p = m/N+1, where m is the rank and N = 58, 

 the 1945 and the 1978 surges would have frequencies of 1/59 and 2/59, respec- 

 tively. Assuming the other 26 members of the storm ensemble to have fre- 

 quencies of 1/20, and using the ratios of these frequencies, the 1945 storm 

 was combined with 170 tides and the 1978 storm with 340 tides. For example, 

 (1/59) / (1/20) x 500 = 170 . This analysis is not rigorous from a statis- 

 tical standpoint and was done primarily to prevent the two storms with the 

 strongest winds and largest waves from being overrepresented at low and medium 

 water levels. Approximately 13,500 possible surge plus tide time-histories 

 resulted from this process (26 x 500 + 340 + 170). Events to be simulated 

 were selected from this file of possible surge plus tide time-histories. 



Selecting Events to Model 



17. A flood-causing event is multidimensional. The severity of the 

 damage caused by the event is determined by several factors, among which are 

 the magnitude and duration of winds, waves, and water levels. Because of the 

 difficulty of ranking multidimensional entities, as well as the lack of avail- 

 able data for doing so, it is necessary to reduce the dimensionality. There- 

 fore, only one dimension, maximum still-water level, was used to measure the 

 severity of an event. This criterion was chosen for two main reasons. First, 

 it was deemed the most important; and, second, there was a large volume of 

 available data. NED has established a stage-frequency curve (Figure 4) at 

 (NED 1983) relating maximum still-water level with its frequency of occurrence 

 the Boston N0S tide gage. This stage-frequency curve was used as the basis 

 for both event selection and the assignment of probability to simulated 

 events. 



18. Based upon previous experience (Prater, Hardy, and Butler, in prep- 

 aration) it was estimated that by simulating 50 events the frequency of still- 

 water level would be accurately represented throughout the study area. 



15 



