The consistent and exact recovery of assigned directions achieved 

 for the simulated 16-record wave trains is in part due to the use of 

 the high computational resolution available in a large computer. Use 

 of less exact data as available from recording instruments is expected 

 to result in a less consistent and accurate recovery of the true direc- 

 tion. Simulated observations 1 to 4 were rerun truncating the computed 

 profile values to three digits as is commonly available from recording 

 systems. The effect of this truncation is estimated to have introduced 

 an error of the order of ±0.127 centimeter (0.05 inch) in the instanta- 

 neous values of the profiles. No appreciable differences in computed 

 directions resulted by this truncation. 



3. Identification of Wave Trains from the High-Resolution Spectrum . 



A wave train in the real ocean is conceivably made up of several wave 

 components with nearby periods propagating in approximately the same direc- 

 tion. Simplistic idealizations of such wave trains are exemplified by 

 simulated observations 1, 3, 5, and 7. For wave trains 1 and 3, the 

 long-crested wave model gave the correct direction of v/ave propagation 

 for all 10 combinations (within 5°) , not only at those spectral periods 

 closest to the periods of the sinusoids combined, but at several adjacent 

 ones on either side of these periods (see App. D, Figs. D-1 and D-3) . At 

 some of these adjacent spectral periods, the contribution to the energy 

 was several times the background level and nearly the same at the five 

 gage locations. Thus, it can be assumed that a wave train in the ocean 

 will give rise to a number of adjacent spectral periods in the high- 

 resolution energy spectrum with energy content several times the back- 

 ground level. This background level can be estimated by inspection of 

 the minima in the energy spectrum, as discussed previously. A criterion 

 for what energy level will be considered "high-energy content" can be 

 set, and groups of adjacent spectral periods in the spectrum with high- 

 energy content identified. These groups may each be assumed to arise 

 from the presence of a wave train in the field with a mean wave period 

 within the range of spectral periods in the group (a wave packet) . The 

 number of adjacent spectral periods in each group will be used as a 

 measure of the width of the energy peak in the spectrum and indicates 

 the spread in periods of the wave train. The spread in computed direc- 

 tions at adjacent spectral periods in a group is an indication of the 

 degree of directional organization in the wave train. Large spread in 

 directional results may indicate the possibility that crossing wave 

 trains with nearly the same period are present. As results for simulated 

 wave train 4 indicate, the long-crested model based on the assumption of 

 a single wave train at each frequency is not suitable for a determination 

 of wave direction in such cases. Multiple wave trains at a single fre- 

 quency may result from refraction around a shoal or island or from reflec- 

 tion by a vertical wall. 



4. Spectra and Direction of Wave Propagation for Field Data . 



The energy and direction of wave propagation at each spectral period 

 were computed for 44 field observations where the average uncompensated 



28 



