U.S. EAST COAST SEA LEVEL ACCELERATION: RECENT EVIDENCE FROM TIDE GAUGES
BOON, John D., Department of Physical Sciences, Virginia Institute of Marine Science, P.O. Box 1042, Gloucester Point, VA 23062, boon@vims.edu
The present study of relative sea level (RSL) rise, based on a linear-quadratic regression model applied to seasonally-adjusted monthly mean sea level (mmsl) observations, finds statistically significant acceleration at 16 tide stations along the mid-Atlantic and U.S.-Canadian north Atlantic coastal region based on regression over the 43-year period 1969-2011. Linear-quadratic models assume acceleration (or deceleration) to be constant during the period regressed and thus the starting and ending years are critically important. An empirical technique, serial trend analysis, was used to evaluate U.S. east coast mmsl records 76 years or more in length with the aim of finding periods when this assumption might be valid. Applied to data at eight tide stations from Norfolk, VA, to Halifax, Nova Scotia, this technique identified a pronounced linear rate increase (constant acceleration) common to all eight stations beginning in 1987 and continuing through 1994 using serial trends of fixed length (36 years). A similar analysis at four tide stations from Charleston, SC, to Key West, FL, showed neither acceleration nor deceleration.
A key point in serial trend analysis is that trend rate estimates apply at the central or median year of a fitted mmsl series whose starting (ending) year is half the series length earlier (later). A 36-year series centered on 1987 marking the beginning of a constant acceleration period would have 1969 as its starting year and 2011 as its ending year. Guided by this information, linear-quadratic regression was applied at 24 U.S. and Canadian Atlantic coast tide stations with complete or near-complete mmsl records over 1969-2011. Analysis of variance F-tests subsequently found the quadratic term contribution to be significant at one station (Kiptopeke, VA) at the 95 percent level of confidence and 15 additional stations from Sewells Point, VA to Halifax, Nova Scotia at the 99 percent level of confidence. Quadratic contribution to regression was non-significant at St. John’s, NL as well as at five tide stations from Wilmington, NC to Key West, FL.
