Hydrogeophysical applications for full-waveform inversion (FWI) of ground-based ground-penetrating radar (GPR) data
Abstract
Flow in porous media can be dominated by preferential flow mechanisms making it difficult to predict the dynamics and impacts of hydrologic processes when data from in-situ sensors are solely considered. However, accurate measurements of hydrogeologic features and associated hydrologic dynamics are crucial to advancing the conceptual and mechanistic understanding of hydrologic processes. e.g. the complex multi-scale nature of the critical zone. Ground penetrating radar (GPR) is a valuable tool in this regard, given the ability to image subsurface structures, the sensitivity of electromagnetic waves to water content, and the mobility of the instrument which allows for relatively larger-scale subsurface imaging when compared to in-situ sensors. Full-waveform inversion (FWI) of surface-based GPR data, coupled with petrophysical relationships, shows promise for high-resolution imaging of complex near-surface targets and for mapping variability in hydrologic state.
FWI has been used extensively in the oil and gas industry over the past decade and has seen more recent attention in the near-surface community. To advance the FWI algorithm for application to surface-based GPR and hydrogeophysical applications, we present results from numerical simulations where varying degrees of complexity are represented and discuss application to field data. For the inversion algorithm, we use the adjoint wavefield method where the misfit between the model and the data are back-propagated from receiver locations. The adjoint field is cross-correlated at each time step with the forward modeled source function to calculate the solution gradient and iteratively update the permittivity field. We obtain the starting model for inversion by smoothing the velocities obtained from normal moveout analysis. Filtering of the gradient is performed to smooth the gradient and mute energy near the source locations. We also present results where we begin the inversion with a low-passed version of the source function and iteratively step up the frequency content of the source function for added solution stability. Results of our synthetic tests show that we can significantly improve model resolution over other available velocity analysis tools, which holds promise for improved geophysical characterization of hydrologic systems.- Publication:
-
AGU Fall Meeting Abstracts
- Pub Date:
- December 2018
- Bibcode:
- 2018AGUFMNS31B0753M
- Keywords:
-
- 0545 Modeling;
- COMPUTATIONAL GEOPHYSICSDE: 3260 Inverse theory;
- MATHEMATICAL GEOPHYSICSDE: 3270 Time series analysis;
- MATHEMATICAL GEOPHYSICSDE: 7270 Tomography;
- SEISMOLOGY