Application of Gaussian and Percentile filters in Particle Swarm Optimisation for 3D gravity modelling and its implementation on Sinanpaşa graben gravity data in SW Turkey

3D Modelling of gravity data is generally performed as a multi-objective optimisation, trying to minimise observed-calculated data misfit while providing models with certain properties, such as smooth or sharp boundaries. We demonstrated that controlling the model properties in global optimisation schemes is possible through basic image processing filters, and developed a Particle Swarm Optimisation (PSO) algorithm that benefits from Gaussian and Percentile filters to avoid ambiguous boundaries that are generally seen in 3D smooth inversions of gravity data. The effectiveness of the algorithm is shown on a synthetic model consisting of two dipping structures with anomalous density contrasts. Thereafter, the algorithm is implemented to recover subsurface density distribution from a field data set. The field data is collected at the south-western part of the Sinanpaşa graben, Turkey. Due to the lack of previous geophysical studies in the area, 3D Euler decomposition, tilt angle, and 3D smooth inversion methods are also implemented to help interpretation and to compare to the model recovered using the PSO algorithm. The developed approach is observed to be resulted with a model, which is more compatible with the known geology of the region.