Zeiss operator to follow the vertical movement of the CRAB, but some sig- 

 nificant features were probably missed. 



30. A third type of operator error is caused by the instrument 

 being triggered to begin a reading prior to properly aiming at the prisms. 

 This short cut of triggering, then sighting the point, results in sampling the 

 azimuth and zenith angles at the time of triggering and later use of these 

 incorrect angles with the distance to the prisms to calculate the CRAB loca- 

 tion. Errors affecting single points are usually easy to detect and remove. 



31. An additional but unquantified source of error results from the 

 thermal expansion and contraction of the CRAB frame and its liquid-filled 

 tires. There is also some "error" introduced into the data by averaging the 

 elevation over the 8.2-m base of the CRAB. The prism and the stadia board 

 were mounted in the vertical plane of the back wheels equidistant between 

 them. Some small error also occurs when sediment is compressed by the weight 

 of the CRAB. Divers have observed tread marks on the bottom following the 

 passage of the CRAB. No adjustment to the data has been made to account for 

 these effects. It is suspected that they may be on the order of 3 to 6 cm. 

 This amount of variation combined with the slight uncertainty of the over 100 

 different instrument setups results in a survey noise level that obscures 

 small bed-level changes along the offshore reaches of the profiles. 



Error minimization 



32. Two steps were taken to minimize the errors associated with the 

 operation of the Zeiss. First, only known locations were chosen as sites for 

 the setup of the instrument, random locations were not used, and each survey 

 always began with a shot to a prism in a known location. If the measured 

 location of the prism was outside certain tolerances, +6 cm horizontal or +1.5 

 cm vertical, the instrument was restationed. Second, periodic checks to the 

 reference prism during the course of the survey helped ensure that the instru- 

 ment had neither gone out of level nor was in need of restationing. 



Error identification 



33. Errors in the data are most easily recognized through comparison 

 with the past data collected for that profile. All data were compared with 

 the immediately prior survey of the profile. More often than not, this com- 

 parison shows where possible errors occur. The suspect points can then be 

 inspected more carefully and a decision made as to whether the point or points 

 were in error or represented real changes on the profile. Figure 7 shows the 



16 



