Towards a Fast, Practical Alternative to Joint Inversion of Multiple Datasets: Model Fusion

In many areas of science and engineering, we have different sources of data. For example, in geophysics, there are many sources of data for Earth models: first-arrival passive seismic data (from the actual earthquakes), first-arrival controlled-source seismic data (from the seismic experiments), gravity data, etc. Datasets coming from different sources can provide complimentary information. In general, some of the datasets provide better accuracy and/or spatial resolution in some spatial areas and in some depths, while other datasets provide a better accuracy and/or spatial resolution in other areas or depths. For example: each gravity data points describes the result of measuring the gravity field at some spatial location; this field is generated by the joint effects of many locations; as a result, gravity generally measures the average density over a reasonably large spatial region. Thus, estimates based on gravity measurements have (relatively) low spatial resolution. In contrast, seismic waves generally travel a narrow trajectory from a seismic source (earthquake or explosion) to a recording senor. Thus, the spatial resolution corresponding to this data is much higher than gravity. At present, each of these datasets is often processed separately, resulting in several different models reflecting different aspects of the studied phenomena. It is therefore desirable to combine data from different datasets. An ideal approach would be to use all the datasets to produce a single model. However, in many research areas – including geophysics – there are no efficient algorithms for simultaneously processing all the different datasets. While such joint inversion methods are being developed, as a first step, we propose a practical solution: to fuse the models coming from different datasets. c