exceeded 26 kilometers per hour. Clearly, only W has the right phase and 

 amplitude variations as the bluff recession rate, B, Simple regressions be- 

 tween B and the other variables yielded the correlation coefficients given 

 in Table 12. Interestingly, although the rate of bluff recession correlated 

 well with W (correlation coefficient = 0.87), explaining 76 percent of the 

 variation, it negatively correlated with all of the lake level variables. 



Attempts were made to obtain better correlation by combining variables 

 and by multivariate analysis, but no significant increase was found in the 

 correlation coefficient above 0.87. More meaningful conclusions might be 

 possible if the data were further refined and the data set expanded. 



One of the weakest variables is W, which estimates storm wave activity 

 during a period. Muskegon, Michigan, data were used because of the uniform 

 quality, but hourly data taken at the powerplant indicate different and gen- 

 erally higher values. The powerplant wind data were not used because of gaps 

 in the data and because of problems in resolving which of two anemometers (at 

 different elevations) were used. 



Since the ultimate interest is in the wave action and energy reaching the 

 beach, wave data (either actual or hindcasted) should be included. Unfortu- 

 nately, the Warren Dunes State Park LEO data did not cover a long enough period 

 to be useful. Quigley (1976) examined the relationship between the bluff re- 

 cession rate and wave power on Lake Erie and found a strong linear correlation 

 (r = 0.79). However, he proposed a more realistic, nonlinear, relationship in- 

 volving the combined effects of low and high waves and varying lake levels. 



Other problems with the data in Table 12 which may affect the correlations 

 include the different period lengths, the incomplete knowledge of ice periods, 

 and the imperfect split of the storm periods since the effect of September and 

 October storms fall into the longer summer periods. Another study of the aerial 

 photos (or a similar set) should make more frequent measurements (every month or 

 every other month) and should include a monitoring program of waves, ice, and 

 lake levels. 



Though the lake level variables did not correlate well with the bluff reces- 

 sion rates in Table 12, its importance on bluff recession is well known. Berg 

 and Collinson (1976) showed that there is a phase lag between bluff recession 

 and lake levels. Part of the reason for the lag is the time required to denude 

 bluffs of protective vegetation as the levels rise and the time needed to re- 

 vegetate the bluffs after the levels start falling. In this study with its 

 unique point in the lake level cycle (a rising peak), the bluffs within reach 

 A were already actively eroding at the beginning of the study and the lag 

 effect may not be significant. The average lake level steadily increased to 

 its peak and then stabilized at a high level while the recession generally in- 

 creased until 1974 when it dramatically decreased. 



The con^iutations of the lake level variables in Table 12 were based on 

 average values for each ice-free period. If the average lake level, AL, is 

 shifted one period forward to improve the phase relationship with B, the corre- 

 lation coefficient between B and AL changes from -0.41 to 0.01. No increase 

 occurred in the correlation coefficient between B and W when the shifted AL 

 was included as a third variable. 



56 



