Technical Note No. 99/1 - Measuring Waves in Lakes and Estuaries is Quite Tricky, data can be Improved by Adjusting Sensor Noise Level
It sounds paradoxical, but the most difficult task when measuring waves (Task 4/0 or Task 4/1) is knowing when there aren't any! The danger is that sensor noise - not real signal - gets transformed into estimates of wave height, which will be wrong. The "reality checks" applied by DOBIE have been described in a previous Technical Note, and these are meant to catch those bursts where sensor noise has been (incorrectly) transformed. An even better trick, which is done when you are running PEDP after the data have been collected, is to adjust the sensor noise level. Run PEDP and set the sensor noise level very low (say 0.001). Now plot the standard deviation (Column G in the output) and the mean spectral period (Column H) and have a close look. You will see times when the standard deviation rises above a base level, and you should also notice that the period rises at those same times. That base level is just the intrinsic sensor noise. Choose a value of the standard deviation that just crosses above that base level (not through the base level), and run PEDP again using that value for the sensor noise. You will see that PEDP now ignores all data below that level and produces a nice clean output with the true wave events standing out clearly from the clutter (as shown above). This works well on data from lakes and estuaries, where waves only appear intermittently.
MOG May, 1999