2. EXPLANATION OF TRENDS AND VARIABILITY 



Yearly mean sea level is the arithmetic mean of hourly sea 

 level heights obtained from an analog tide gage over a period 

 of one calendar year. The tide gage, often located on a pier, 

 continuously measures sea-level heights relative to the land 

 adjacent to the station location. The gage is connected to 

 bench marks on the adjacent land by precise first-order lev- 

 eling. If possible, the bench marks are located in bedrock. 



One table and nine illustrations show the trends and 

 variability of yearly mean sea level at permanent tide stations 

 operated by the National Ocean Survey (NOS). Column 1 of the 

 table lists all of the NOS-operated stations that were in opera- 

 tion by 1939 and that had very few and short breaks in measure- 

 ment. In addition, all permanent stations in the greater New 

 York Bight area are included. The inclusive dates of each 

 station series are given in column 2. Where the length of a 

 break in the series is sufficient to invalidate a yearly mean, 

 the missing year is shown in column 3. 



If a series of yearly mean sea level values is plotted on 

 a graph of height against date, an apparent secular trend and 

 yearly variability become evident. "Secular" means nonperi- 

 odic; "apparent" means it is not known whether the trend is 

 nonperiodic or is merely a segment of a very long oscillation. 

 Apparent secular trends in sea level result from glacial- 

 eustatic, tectonic, and climatological and oceanographic appar- 

 ent secular trend effects. Columns 4 and 7 show the apparent 

 secular trend as the slope of a straight line mathematically 

 fitted through the yearly mean sea level values (see note a on 

 table). About two-thirds of repeated calculations of the ap- 

 parent secular trend will differ from the true apparent secular 

 trend by less than the standard error of slope listed in col- 

 umns 5 and 8 (see note b on table). About 957o of repeated 

 calculations of the apparent secular trend will differ by less 

 than two times the standard error of slope, and practically 

 all repeated calculations will differ by less than three times 

 the standard error of slope. 



