GNGTS 2013 - Atti del 32° Convegno Nazionale

References Bertoni C. and Cartwright J.A.; 2006: Controls on the basinwide architecture of late Miocene (Messinian) evaporites on the Levant margin (Eastern Mediterranean ). Sedimentary Geology, 188-189 , 93-114. CIESM; 2008: The Messinian Salinity Crisis from mega-deposits to microbiology - A consensus report N.33 in CIESM Workshop Monographs (F. Briand, Ed.), 168 pp, Monaco. Fabbri A. and Curzi P.; 1980: The Messinian of the Tyrrhenian Sea: seismic evidence and dynamic implications . Giorn. Geol., ser.2, 43, 1, 215-248. Finetti I.; 1982: Structure, stratigraphy and evolution of Central Mediterranean. Boll. Geof. Teor. Appl., 24 , 75-155. Finetti I.R. (ed.); 2005a: CROP Project: Deep seismic exploration of the Central Mediterranean and Italy . Atlases in Geoscience 1, Elsevier, 794 pp. Hsü, K. J., Cita, M. B., and Ryan, W.B.F; 1973: The Origin of the Mediterranean Evaporites . In Ryan, W.B.F, Hsu, K. J., et al. , Init. Repts. DSDP, 13, Pt. 2: Washington (U.S. Govt. Printing Office), 1203-1231. Lofi J., Déverchère J., Gaullier V., Gorini C., Guennoc P., Loncke L., Maillard A., Sage F., Thinon I.; with contr. J. Benkhelil et al. .; 2011: Seismic atlas of the “Messinian salinity crisis” markers in the Mediterranean and Black Seas Paris . Memoire de la Societe Geologique de France, 179 , 71 pp. ISBNs:2853630978, 9782853630979, 9782853630979 . Rehault J. P., Boillot G., Mauffret A.; 1984: The Western Mediterranean Basin geological evolution . Mar. Geol., 55 , 447- 477. Scrocca D., Doglioni C., Innocenti F., Manetti P., Mazzotti A., Bertelli L., Burbi L. and D’Offizi S. Eds; 2003: CROPAtlas – Seismic Reflection Profiles of the Italian Crust. Memorie descrittive della Carta Geologica d’Italia, 52. Stampfli G.M.; 2005: Plate Tectonics of the Apulia-Adria microplate. In: Finetti I.R. (ed), CROP Project: Deep seismic exploration of the Central Mediterranean and Italy, Atlases in Geoscience 1, Elsevier, pp. 747-766. Comparison between Neighborhood and Genetic Algorithms on two Analytical Objective Functions and on a 2.5D Synthetic Seismic Inverse Problem A. Sajeva 1 , M. Aleardi 1 , A. Mazzotti 1 , E. Stucchi 2 1 Earth Sciences Department, University of Pisa, Italy 2 Earth Sciences Department, University of Milan, Italy Introduction. A geophysical inverse problem consists in obtaining the earth model for which the predicted data best fit the observed one (Tarantola, 1986). The problem is often non- linear and can be solved using a local linearization method (such as Gauss-Newton, steepest descent or conjugate gradient) or using a global optimization method (such as Grid Search, Simulated Annealing, Genetic Algorithms, Particle Swarm and Neighborhood Algorithm). In this work we compared and evaluated the efficiency and the limits of a Genetic Algorithm (GA) and of the Neighbourhood Algorithm (NA) varying the dimensions of the model space. We first tested these methods on two analytical objective functions: a multidimensional convex parabola and a more complex egg-box functional. Lastly we performed an acoustic full waveform inversion considering a small and smoothed portion of the Marmousi model. The Neighborhood Algorithm. The Neighborhood Algorithm is a direct search method which is based on the concept of proximity. The method is composed by two parts which deal with two different stages of the optimization problem: the sampling of the model space (Sambridge, 1999a), and the appraising of the ensemble (Sambridge, 1999b). The sampling part deals with finding models of acceptable data fit in a multidimensional parameter space. One of its key ideas is to be best guided by all previous models for which the forward problem has already been solved. Therefore, it makes use of the previous models to approximate the misfit function everywhere in model space. The misfit functional is interpolated to the value of its nearest neighborhood in model space. The method uses Voronoi cells (Voronoi, 1908) to determine the nearest model space sample. The algorithm is conceptually simple, in fact, it requires just two ‘tuning parameters’: the number of new models to be generated per iteration ns ; and the number of best fitting models to be selected among these new models per iteration nr . The user can also set ns to decrease as the number of iterations increases. Furthermore, the method makes use of only the rank of a data fit error rather than its numerical value. 60 GNGTS 2013 S essione 3.1