Because of the frequency of tsunamis occurring in the Pacific Ocean, 

 a tsunami warning system has been developed for the inhabitants of 

 Pacific coastal areas. A similar warning system has not been developed 

 for other areas . 



2 . Probability of Occurrence . 



Where sufficient historical data are available on tsunami flood 

 levels, the probability of tsunami flooding at any elevation can be 

 determined by the same methods used for determining the probability of 

 floods on rivers. For a known period of record, the recorded flood 

 levels can be ranked from the largest to the smallest; i.e., the highest 

 flood level is ranked 1, the next highest is ranked 2, and so on. Linsley, 

 Kohler, and Paulhus (1958) show that the probability of each flood level 

 is then given by 



P(Z) = -^— (1) 



n + 1 



where 



P(Z) = the probability of flooding to the elevation Z in any year 



Z = the elevation above a defined datum 



m = the rank of the flood level 



n = the period of record in years 



Houston, Carver, and Markle (1977) have determined the probability of 

 tsunami flood levels for the Hawaiian Islands. For recurrence intervals 

 greater than 10 years, i.e., P(Z) < 0.1, they give 



h 2Q0 = -B -A log 10 P(h 2Q0 ) (2) 



where h 200 is the elevation of the maximum tsunami wave crest above 

 mean sea level (MSL) 200 feet (61 meters) shoreward of the coastline, 

 P(h ) the probability of a flood level occurring at elevation h 2Q0 

 in any given year, and A and B the empirical coefficients which are 

 determined for each point on the coastline. Where sufficient historical 

 data were not available, they generated additional data using a mathe- 

 matical model. The model data were multiplied by correction factors 

 and compared to historical data. This produced additional data at points 

 along the coastline where historical data were not available, and allowed 

 a determination of the empirical coefficients A and B at all coastal 

 points. 



It should be noted that there is a probability of some error in the 

 predicted flood elevations based on available historical data. For 

 example, there is a 37-percent probability that a 100-year flood level 



21 



