Model design for detection of gravity changes in Marmara region

This study gives a processing strategy with variance component estimation aiming to perform a proper stochastic model in a gravity monitoring network. First, a program which is able to make some necessary reductions and to adjust gravity observations was written in Matlab. The basic properties of the program called GrAd as follows: 1) Gravimetic tide reductions are made according to some tidal potential models, 2) robust estimation with Denmark weight function and median methods are used for outlier detection, 3) the observations belonging to the lines of a network are combined properly in the adjustment, 4) gravity values of the points are estimated with a trace-minimum adjustment for deformation analysis, 5) deformation analysis is performed for different epochs with the global test and localization procedures and the temporal gravity changes are obtained according to the statistically-verified reference points in the network. In order to estimate the variance components, the method of BIQUE (Best Invariant Quadratic Unbiased Estimation) was used in the study. An algorithm in which the components of different meters are estimated iteratively (iterative-BIQUE) was adapted to GrAd. For numerical applications, Marmara Region Gravity Network (MAGRANET) with 25 gravity points was used. The network was measured with 2 LaCoste&Romberg relative gravity meters between 2003-2005 semi-annually as 5 periods. For free adjustments of different periods, the weigths of the meters? readings, a) were selected equally (=P) as usual, b) were obtained according to the estimated variance components. As a result of the estimation of the components, it is shown that the meters? weigths are different (approximately P and P/2, respectively). Comparing the estimated gravity values for the cases of a and b, the differences between ?7 and 11 ?Gal were observed. Moreover, using the estimated gravity values and their cofactor matrices from the two cases separately, the periods of October 2004-October 2003, October 2005-October 2004 and June 2005-June 2004 were analysed with localization procedure. According to the deformation analyses? results, the temporal gravity changes computed for the two cases vary each other significantly. Especially, for the June 2005-June 2004 period, the variations reach up to 17 ?Gal. As a result of the statistically significant differences, it is concluded that the improper stochastic part of a Gauss-Markoff model for gravity networks affects adversely the interpretation of the temporal gravity changes. With the temporal gravity changes obtained for the case b, it is seen clearly that the region between Istanbul and Izmit as well as the southern part of the Izmit Bay have different anomaly patterns related to the changes on the other points of the MAGRANET. The magnitudes of temporal gravity changes in the corresponding parts of the network reach up to 50 ?Gal per year. Key Words: Temporal gravity changes?Variance component estimation?BIQUE method? Gravity networks?Deformation analysis?Marmara Region