GNGTS 2017 - 36° Convegno Nazionale

106 GNGTS 2017 S essione 1.2 geophysical sources. (i.e. slow earthquakes) and therefore they must be filtered out from tilt signals. �� We show in Fig. 1 the time behaviour of the hourly values of the atmospheric pressure and temperature and of �� the E-W and N-S components of the ground tilt �� at the station. The analysis of the tilt records points out a clear influence of the fluctuations of the air temperature on ground deformation, mainly at annual and diurnal periods. We observe a significant correlation between the seasonal trends of the temperature T and N-S component (r = 0.8) while a weak inverse correlation affects the E-W component.�� The spectral analysis of the cross-correlation function between the deviations of T and tilt components from the seasonal trends, highlights the presence, in both E - W and N - S components, of significant energy in the band of the diurnal frequencies. �� Being T ( t ) and p ( t ) mutually correlated (Fig, 2), it is hard, and practically impossible, to build the functions f p [ p ( t )] and f T [ T ( t )] accounting for the influence of p ( t ) and T ( t ) on the tilt signal s ( t ). �� The Ocean Tide Load contribution to the tilt at the station is here neglected as it results of the order of magnitude of 10 -3 µrad for the main tidal waves. We propose here a statistical numerical procedure aimed at removing �� from the ground tilt measurements �� the joint effects of the atmospheric temperature and pressure changes. Let s ( t ) be the signal and c ( t ) the joint contribution to s ( t ) of both the atmospheric pressure p ( t ) and temperature T ( t ); then s ( t ) could be expressed by the general relationship: s ( t ) = c ( t ) + r ( t ) = f p [ p ( t )] + f T [ T ( t )] + r ( t ) (1) where r ( t ) is the residual part of the signal not affected by p ( t ) and T ( t ). The most general relationship correlating the signal s(t) with p(t) and T(t) is a time convolution relation: s ( t ) = � ∞ –∞ a ( t' ) p ( t–t' ) dt' + � ∞ –∞ b ( t' ) T ( t–t' ) dt' + r ( t ) (2) The admittance functions a ( t ) and b ( t ) in eq. (2) are the solutions of the multiple regression simultaneous equations (Hannan, 1970): Φ pp (ω) A (ω) + Φ pT (ω) B (ω) = Φ sp (ω) Φ Tp (ω) A (ω) + Φ TT (ω) B (ω) = Φ sT (ω) (3) where A (ω) and B (ω) are the Fourier Transform (FT) of a ( t ) and b ( t ) and Φ ij denotes the cross power (FT of the cross correlation function) between any couple of the variables p ( t ), T ( t ) and s ( t ); ω is the angular frequency. Once A (ω) and B (ω) have been computed they can be back Fig. 2 - Contribution to the tilt of the temperature (left) and the pertinent spectral content (starting time of the plot is 2011, May 9 at 4 h 00 m 00 s GMT).