
Response of the Adriatic sea level to the air pressure and wind forcing at low frequencies (0.01–0.1 cpd)
Author(s) -
Pasarić Miroslava,
Pasarić Zoran,
Orlić Mirko
Publication year - 2000
Publication title -
journal of geophysical research: oceans
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.67
H-Index - 298
eISSN - 2156-2202
pISSN - 0148-0227
DOI - 10.1029/2000jc900023
Subject(s) - forcing (mathematics) , atmospheric pressure , sea level , mediterranean sea , climatology , environmental science , wind stress , wind speed , linear regression , pressure gradient , geology , meteorology , atmospheric sciences , mediterranean climate , oceanography , mathematics , physics , statistics , geography , archaeology
Low‐frequency (0.01–0.1 cpd) variability of air pressure, wind, and sea level is examined through 6‐ to 8‐year records of data collected at three locations along the east coast of the Adriatic and one on the west coast. Seasonal energy spectra show that processes at these timescales are more energetic in winter than in summer. There is substantial wind energy at timescales corresponding to planetary atmospheric waves. In order to explain the stronger‐than‐isostatic adjustment of sea level at low frequencies to the air pressure forcing, recorded in different parts of the Mediterranean, the present empirical analysis is based on a physically more tractable model, relating sea level slope to the air pressure gradient and wind stress integral. The multiple input regression and the cross‐spectral analysis yield a spatially variable response: over the deeper sea region sea level slope is fully explained by isostatic adjustment to the air pressure gradient alone; over the shelf a much stronger‐than‐isostatic response (−1.7 cm/mbar) is greatly reduced (−1.3 cm/mbar), but not fully accounted for, by the action of wind. Next the multiple linear regression method is carefully reexamined; a simple statistical model is developed to show that in multiple‐input linear models with mutually correlated inputs, small errors in one of the inputs produce biased estimates of all the response parameters. The apparent discrepancy between the theoretically predicted and the estimated response is attributed to the bias.