8 



The difficulties of extracting the sea level rise (SLR) "signal" from 

 a record containing substantial noise has been studied carefully by Sturges 

 (1987). The coherency of spatially separated tide gage records was 

 investigated with the hypothesis that coherent signals with no lag could be 

 interpreted as global sea level rise whereas lags with a certain character 

 could be interpreted as due to atmospheric forcing or long water wave 

 (Rossby wave) motions. As an example, the records at San Francisco and 

 Honolulu were found to be coherent at periods of 5 to 10 years and longer, 

 although with a phase lag. A comparison of the energy spectra obtained 

 from these two stations is presented as Fig. 2.1a and other spectral 

 information is presented in Figs. 2.1b,c,d. The amplitudes of these 

 coherent components are 5-15 cm. Similar coherence results were found for 

 tide gage records located on both sides of the Atlantic. Sturges concluded 

 that the available records are contaminated by substantial energy with 

 periods up to 40 to 50 years , thus exacerbating the problem of identifying 

 any change in the rate of SLR. The ability to extract the SLR signal may 

 possibly be enhanced through an analysis which recognizes the probable 

 cause of the noise components, thereby guiding their removal from the 

 record. 



Aubrey and Emery (1983) applied the method of eigenanalysis to United 

 States tide gage data in an attempt to identify fluctuations that were 

 spatially and temporally coherent. This method, among the most 

 sophisticated applied to date, has the potential advantage of retaining in 

 the first few temporal eigenfunctions , those fluctuations that have the 

 same form and that are either exactly in or exactly out of phase. The 

 principal disadvantage is that the method is purely statistical and does 

 not recognize the physics of the phenomenon, although it may isolate 

 features that will assist in identifying physical components. A particular 

 drawback is that the method only recognizes correlations which are either 

 in phase or exactly out of phase as "signal." Thus a very long and slowly 

 propagating wave would be rejected as noise whereas a pure standing wave 

 would be recognized as "signal." Aubrey and Emery first applied the 

 technique to 12 U.S. gages each of which encompassed 61 years of data and 

 secondly to 41 tide gages with a common time base of 40 years of data. 

 Different rates of rise were found for the East and West coasts. From the 



