57 



from the record. The method is likely to perform better in shallow water, or with 

 records that are measured closer to the water surface. Despite this limitation, the 

 method was able to capture much of the detail of a irregular laboratory record, and 

 provide fairly accurate estimations of the local kinematics. 



The analysis of regular waves provides guidelines for the parameters to be used 

 in the analysis of irregular waves. In shallow water, higher order solutions and wider 

 windows must be used than in deep water. Window widths of one fifth of the zero 

 crossing period and a sixth order potential function are adequate for the shallowest 

 waves, and window widths as small as one tenth of the zero crossing period and a 

 third order potential function are adequate for deep water. 



The laboratory results indicate that the method requires good precision and care in 

 the measurements. Any high order method demands very accurate data, but this need 

 is exacerbated by the ill posed problem of determining the near surface kinematics 

 from a subsurface record. While not particularly sensitive to random noise in the 

 record, the decay with depth of the high frequency information makes it very difficult 

 to capture the high frequency modes. Fundamentally, the LFI method is designed to 

 capture the dominant free mode in each given window. As the higher frequency modes 

 decay faster with depth, and if the measurements are taken far below the surface, the 

 dominant mode will always be one of lower frequency modes. This difficulty would 

 be exacerbated by any limitations in the frequency response of the gauges. 



The computer resources required for the method are substantial, but not pro- 

 hibitive. As computers continue to get faster, computation time is not the considera- 

 tion that it once was. The method was developed and all computation done using the 

 MatlaB computational package. MaTLAB is an interpreted language that provides 

 a very easy to use interface to a robust and complete library of computational and 

 visualization routines, allowing for rapid development of new methods. Being an in- 

 terpreted language, the resulting code is not as fast as it might be if the routines were 

 programmed in a compiled language, such as Fortran or C. The computational speed 

 is also affected by the degree to which the program pauses to provide visual output. 

 Perhaps a better measure of the computational intensity of the method is a count of 

 the total number floating point operations (flops) used to compute the solution. 



