32 



0.25 cm) of water surface change. To reduce these noise levels, a 3.33-Hz. 

 low-pass filter was applied to the wave records. 



Identification of sand bed. The sand bed was exposed at some gauges in 

 between the wave down-rush and the subsequent wave uprush. In the wave 

 recoxds, this exposure appears as a flat or dipped wave trough. For gauges 

 located seaward of the still-water shoreline, these bed exposures were infre- 

 quent, unless the wave period was long so that there was little interference 

 between the wave run-down and the next wave uprush. Landward of the 

 shoreline, however, the bed was frequently exposed. Near the runup limit, 

 the bed was often exposed more than it was submerged, as only a few waves 

 reached the wave gauri 



For gauges with frequent bed exposures (gauges landward of the initial 

 shoreline), the bed generally eroded or accreted during a test so that the eleva- 

 tions of the bed decreased or increased, respectively, over time. This pro- 

 duced large trends in the measured wave record. To remove these trends, an 

 algorithm was developed to identify the sand bed during the exposures, to 

 interpolate sand beds in between exposures (during the wave uprush), and then 

 to remove the sand bed record from the original signal. 



The sand bed surface was defined numerically as a minimum in the wave 

 record for which the signal remained essentially constant (flat) for longer than 

 1 sec 16 tiara points). This was based on trial calculations in which it was 

 found that fiat strings of less than 1 sec were often actual water surface rather 

 than the sand bed. Because of small variations and remaining noise in the 

 signals, the bed was considered flat if elevations were within ±0.05 in. 

 '. 15 cm). This criterion was also verified by trial calculation. 



Once sand bed exposures were identified, an approximate time series for 

 the sand bed elevation was generated. For the duration of the bed exposure, 

 from the end of one back-rush to the beginning of the next uprush, the aver- 

 age bed elevation was determined. Between bed exposures, as the next wave 

 passed, bed elevations were linearly interpolated to connect one exposed bed 

 level to the next. As a result, a bed elevation time series was generated at the 

 same 16-Hz sampling rate used in the gauge measurements. This bed eleva- 

 tion record was then subtracted from the original signal (with offset also sub- 

 tracted) to produce a time series of the water surface elevations with any trend 

 due to erosion or accretion removed. It is noted, however, that the resulting 

 signal may not be stationary, as, for example, the average wave height may 

 increase over time if the sand bed eroded significantly. 



Filtering. Once the sand bed signal was removed, the remaining water 

 surface signal was filtered again to separate the signal into low- and high- 

 frequency components. This procedure was identical to that performed on the 

 other hydrodynamic data in which the filter cut-off frequency was selected as 

 one-half of the peak wave frequency based on the target wave spectrum input 

 into the wavemaker. Following this filtering, three time series were available 

 for analysis: laj the original (de-trended but unfiltered) total wave record, 



