147 



except that the window width used was much narrower, tq = 0.171). There are large 

 discontinuities in the horizontal velocity in the x direction (third plot), near both 

 crests. Figure 6.10 shows the parameters of the potential function from this solution. 

 ^, where 9 = tan~^ {ky j k^) ^ is the local direction of propagation of the wave. There is 

 a sudden shift in direction of 80° at the first crest, and over 100° at the second crest. 

 This would result in unrealistic predictions of huge accelerations. Figure 6.11 shows 

 the parameters of the solution computed with tq = O.STj. With the wider window, 

 the propagation direction varies smoothly around the peak direction of the record, 

 e = 83°. 



Array Size The need for a fairly wide window for the field solution is likely due 

 to the small size of the pressure gauge array. Compact size is one of the major 

 advantages of the DWG-1 array, as it is a single unit that is easily deployed, but 

 it makes it susceptible to difficulties with local analysis. The propagation direction 

 within each window is determined by the spatial gradients of the dynamic pressure, 

 as estimated by the measurements. In segments of the record where the gradients 

 are small, the predicted direction of propagation can fluctuate a great deal with only 

 small change in any of the measured values. These changes could be the result of 

 measurement error, or be small fluctuations in the actual dynamic pressure. In either 

 case, the large resulting change in predicted propagation direction can be a source of 

 error. 



The linear dispersion relation predicts that the wavelength corresponding to the 

 peak period of this record in this depth of water would be over 42 m. The DWG-1 

 array is only 1.6 m on a side, which is about 0.04 of the approximate wavelength of the 

 peak period waves. As a result, near the crest of the wave, where the water surface 

 is close to horizontal, all three gauges are near this point, and the local estimate 

 of the gradient of the pressure is very sensitive to small fluctuations. This eff'ect 

 is mitigated by using a larger window in time, so that a part of the wave with a 

 higher gradient is part of the window, allowing the propagation direction to be better 

 defined, and smoothing out any measurement errors. As minimum window size has 

 been determined to be about 1/10 of the zero crossing frequency, an array size of 



