A Parallel Scheme for Multi-scale Data Integration

Abstract

In this paper, we propose a parallel modeling approach for solving large and complex inverse problems involving multiple data sources each with different scale of observation. This parallel approach relies on building property models on multiple grids with different resolution at the same time, rather than selecting a single modeling grid. By keeping a high resolution model and its upscaled, coarsened model in constant consistency with each other during the inversion process, a fully consistent integration of all data sources is achieved.