Accurate modelling of sonobuoy refraction data to determine velocity variations in oceanic crust

Modern disposable sonobuoys can provide a simple and cost-effective alternative to ocean bottom seismometers for marine refraction experiments over oceanic crust. Unfortunately, the fact that they are free to drift with the prevailing ocean currents can introduce significant travel-time errors into the modelling process if the seafloor topography is large. For sonobuoys recorded during and after turns the drift rate and direction can be uniquely determined by inversion of the shot-receiver ranges derived from the water-wave arrival. The same method can be used to determine a best fitting average drift vector for the whole dataset. A modification to conventional two-dimensional travel-time modelling techniques has been developed to account for this drift. Each sonobuoy profile is divided into several subsets, typically of 100 shots each, and each subset is then modelled as a separate common receiver gather, significantly reducing the errors in the calculated travel-times. For realistic bathymetry, the magnitude of these travel-time errors is up to 200 ms, significantly larger than the estimated picking uncertainty. Real data from a typical sonobuoy refraction experiment on the Mid-Atlantic Ridge were modelled with and without the drift correction applied. Much of the lateral variation in the velocity structure was removed when the drift correction was applied, indicating that this structure was due to variations in the travel-times caused by sonobuoy drift.