Time-lapse joint inversion of geophysical data 1 with automatic joint constraints and dynamic attributes

Authors: RITTGERS, JUSTIN - Colorado School Of Mines; REVIL, ANDRE - Centre National De La Recherche Scientifique; MOONEY, MICHAEL - Colorado School Of Mines; KARAOULIS, MARIOS - Colorado School Of Mines; WODAJO, LETI - University Of Mississippi; HICKEY, CRAIG - University Of Mississippi

Submitted to: Geophysical Journal International
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 9/14/2016
Publication Date: 9/14/2016
Citation: Rittgers, J.B., Revil, A., Mooney, M.A., Karaoulis, M., Wodajo, L., Hickey, C. 2016. Time-lapse joint inversion of geophysical data 1 with automatic joint constraints and dynamic attributes. Geophysical Journal International. 207(3):1401-1419. https://academic.oup.com/gji/article/207/3/1401/2194601?searchresult=1.

Interpretive Summary: This paper shows the effectiveness of both time-lapse constraints and joint constraints used in concert in improving inverse models by reducing random noise that contaminates geophysical data both spatially and temporally. Furthermore, we demonstrate the marked improvements in recovered models when both spatial and temporal constraints are combined and specific attributes are used. This is successfully demonstrated for both synthetic and real data examples. Finally, we present a new Automatic Joint Constraints technique for further improving recovered time-lapse joint inversion models. The AJC technique is demonstrated to be effective at focusing models recovered for both synthetic and real time-lapse geophysical data sets. When combined with the definition of specific attributes describing the target and its expected temporal behavior, it can be effectively used to track a dynamic target such as the development of internal erosion and piping over space and time. The detection of soil piping and internal erosion are important for assessment of aging earthen dams. The methods described here can be used to improve the detection of faults that could lead to catastrophic dam failures.

Technical Abstract: Joint inversion and time-lapse inversion techniques of geophysical data are often implemented in an attempt to improve imaging of complex subsurface structures and dynamic processes by minimizing negative effects of random and uncorrelated spatial and temporal noise in the data. We focus on the structural cross-gradient (SCG) approach (enforcing recovered models to exhibit similar spatial structures) in combination with time-lapse inversion constraints applied to surface-based electrical resistivity and seismic traveltime refraction data. The combination of both techniques is justified by the underlying petrophysical models. We investigate the benefits and trade-offs of SCG and time-lapse constraints. Using a synthetic case study, we show that a combined joint time-lapse inversion approach provides an overall improvement in final recovered models. Additionally, we introduce a new approach to reweighting SCG constraints based on an iteratively updated normalized ratio of model sensitivity distributions at each time-step. We refer to the new technique as the Automatic Joint Constraints (AJC) approach. The relevance of the new joint time-lapse inversion process is demonstrated on the synthetic example. Then, these approaches are applied to real time-lapse monitoring field data collected during a quarter-scale earthen embankment induced-piping failure test. The use of time-lapse joint inversion is justified by the fact that a change of porosity drives concomitant changes in seismic velocities (through its effect on the bulk and shear moduli) and resistivities (through its influence upon the formation factor). Combined with the definition of attributes (i.e. specific characteristics) of the evolving target associated with piping, our approach allows localizing the position of the preferential flow path associated with internal erosion. This is not the case using other approaches.