Broadband inversion and source localization of vertical array data from the L-antenna experiment in 1999

Matched field processing techniques have been demonstrated on vertical array data from the 1999 L-antenna experiment. Both single-frequency and broadband inversion of geoacoustic and geometric parameters were carried out. The baseline model consisted of three water layers and a sediment layer overlying a half-space substrate. The environment was assumed range-independent with no shear waves. Broadband inversion showed to be more robust than single-frequency inversion. For broadband inversion the estimate of the water depth was usually between the depths at receiver and source positions. The p-wave sound velocity in the sediment was estimated to be about 1600 m/s, which was 200 m/s lower than in the baseline model. The sediment thickness was estimated to be thicker than in the baseline model, which conflicted with the measured two-way travel time. Broadband source localization (range and depth) showed to be most successful when using a modified baseline model and at the same time inverting the water depth. The error in the range estimate was only 1 – 6 %, while the error in the depth estimate was 4 – 16 %. The depth was underestimated for the 91 m and 244 m shots, and usually overestimated for the 18 m shots. The range was usually underestimated. The best matches in both inversion and localization were found for the 18 m shots. The OASES module OAST and SAGA with genetic algorithms was used in forward modelling and inversion.