Measurements were made using an engineer's scale with 24 divisions per 

 centimeter (60 divisions per inch). This allows an accuracy of 1/2 division 

 which at the photo scale equals ±0.76 meter. Therefore, the effect of measure- 

 ment error, e™, can be determined by estimating the standard deviation o™. 



err, = ±0.76 meter 

 1.52 _ 



a^ = ±J.d±. = 0.25 



m 



0.064 



The errors in identifying and marking the bluff line, e^ , and in trans- 

 ferring it to the 1974 photo set, e^, are unique to the analysis used in this 

 study. The width of the ink line is 1.5 meters and the line is drawn so that 

 one edge of the line traces the bluff line. To minimize interpretation errors, 

 photos were selected at times when trees were without leaves and when the bluff 

 line was well defined. 



If a conservative error of ±1.5 meters is assumed for each process, then 



eji = e^ = ±1.5 meters 



4= 'I- 0-25 

 Finally, collecting all the error terms 



9 9 9 9 9 



t m b z 



0^ = 0.27 + 0.064 + 0.25 + 0.25 = 0.834 

 a 7 = 0,91 meter 



Therefore, individual distances from the reference line to the bluff line 

 can be measured to an accuracy on the order of ±1 meter. This is a reasonable 

 amount of error which has been kept small by (a) a large photo scale (1:3,600), 

 (b) short measuring distances, (c) optical matching of photos, and (d) with all 

 photos at the same nominal scale. 



The accuracy in determining bluff-line changes can also be determined by 



where 



6 = change in bluff line 



Dj^ = measured distance at time t = t-^ 



D2 = measured distance at time t = t2 



V(6) = a?, +0?, = variance of 6 



72 



