Constraining CO2 simulations by coupled modeling and inversion of electrical resistance and gas composition data

https://doi.org/10.1016/j.ijggc.2013.04.011Get rights and content

Highlights

  • We invert time-lapse geophysical (ERT) and hydrological data from a CO2 injection.

  • Simplified geophysical and hydrological models reveal complex flow phenomena.

  • Model structure (reservoir width) and permeability values are estimated.

  • Inversion results are improved with ERT data despite high measurement noise.

  • Using ERT for monitoring CO2 has great potential but also significant limitations.

Abstract

This study investigates how model predictions of subsurface CO2 migration can be constrained and improved with time-lapse electrical resistance tomography (ERT) data for a pilot experiment located at Cranfield, Mississippi. To this end, we first invert the time-lapse ERT dataset using structurally constrained and unconstrained inversions. With the ERT time-lapse inversions, we image the increasing supercritical CO2 saturation in the reservoir and find that including the reservoir boundaries as structural constraints significantly improves the images. We then use ERT-derived changes in subsurface electrical resistivity along with gas composition data to constrain and calibrate hydrological models. We use the inversion framework iTOUGH2 and test several simplified conceptual models for the reservoir. Our analysis shows that the reservoir response cannot be adequately reproduced with a radial model; rather, the system exhibits 1D behavior. A model with three 1D layers, whose permeability values and width were estimated by inversion, is able to explain the ERT and gas composition data. Derived permeabilities agree with those from core measurements and a well test. Despite high noise levels, the ERT data provided crucial information in the inversion thanks to its high sensitivity at the inter-well scale, its stabilizing effect on the inversion, and the direct link it provides between electrical resistivity and CO2 saturation.

Introduction

Geologic sequestration of carbon dioxide (CO2) is a promising approach to offset anthropogenic carbon emissions. Deep saline aquifers have been identified as key target formations for sequestration due to their large spatial extent and high potential storage volumes. For the safe and efficient operation of underground CO2 storage, it is important to monitor the migration of the CO2 (Benson et al., 2005). Monitoring subsurface CO2 distribution can help to identify leakage through faults and abandoned wells, so that remedial measures can be taken before contamination of potable aquifers occurs. Knowledge of the spatial extent of injected CO2 is also useful for assessing and updating pre-injection estimates of available storage volumes and refining models of subsurface flow properties.

Crosshole geophysical methods could play a key role in integrated monitoring programs for deep CO2 injections and have recently been tested at a variety of pilot storage sites (e.g., Hovorka et al., 2006, Giese et al., 2009, Hovorka et al., 2011). Their advantage is the comparably high spatial resolution at the inter-well scale of tens of meters and the sensitivity to changes in bulk properties, such as supercritical-phase or gas saturation. Examples of successful geophysical monitoring of injected CO2 include seismic tomography (e.g., Ajo-Franklin et al., 2013, Daley et al., 2008), continuous active-source seismic monitoring (CASSM) (Daley et al., 2011) and electrical resistance tomography (ERT) (Bergmann et al., 2012, Carrigan et al., 2013).

Crosswell electrical resistance monitoring for environmental applications in the shallow subsurface is a well-established technique (e.g., Daily and Ramirez, 1995, Ramirez et al., 1993, Slater et al., 2000) and, recently, fully automated systems for long-term monitoring have become available (e.g., Coscia et al., 2012). ERT monitoring of CO2 migration can build upon this knowledgebase while developing field systems better suited for the hostile downhole environment present in deep storage wells. ERT is well suited for monitoring of CO2 due to the strong electrical resistivity contrast between highly conductive reservoir brine and practically non-conducting supercritical or gas-phase CO2. Modeling studies have demonstrated the potential for ERT monitoring of CO2 sequestration projects (e.g., al Hagrey, 2012, Christensen et al., 2006, Ramirez et al., 2003). Successful CO2 field experiments with ERT monitoring have been reported from experiments in Nagaoka, Japan (Nakatsuka et al., 2010), Ketzin, Germany (Bergmann et al., 2012), and Cranfield, Mississippi, USA (Carrigan et al., 2013).

While ERT is sensitive to the bulk properties of interest, it suffers from inherent resolution limitations (Ellis and Oldenburg, 1994). Under the difficult environmental conditions (e.g., high temperature and pressure) found in deep brine reservoirs, these resolution limitations can be amplified by high noise levels. Resolution limitations can be partly overcome by constraining ERT inversions with structural information from borehole logs or other geophysical data (e.g., Doetsch et al., 2012). Including interfaces separating zones with strong resistivity contrasts is one piece of a priori information that dramatically improves the resistivity estimate of each zone or its evolution over time.

Geophysical data or inversion results can improve the understanding of hydrological processes and improve hydrogeological models (Scheibe and Chien, 2003), but links between the geophysical data and hydrological properties are often weak or indirect. If they are known, the petrophysical relationships that link hydrological state variables or properties of interest and geophysical parameters are often poorly understood or site specific. The limited sensitivity of the geophysical methods to hydrological properties, the imprint of the geophysical inversion (e.g., regularization) and resolution limitations further complicate the problem.

Some of these issues can be addressed by using a coupled hydrogeological–geophysical inversion framework that integrates the geophysical information directly into hydrological parameter estimation. Hydrological parameters are estimated jointly from geophysical and hydrological data, where simulated geophysical measurements become a function of the hydrological processes. Examples of coupled hydrogeophysical inversions include the use of ground penetrating radar data to estimate spatial distributions of permeability (Finsterle and Kowalsky, 2008, Kowalsky et al., 2005), and ERT data to estimate hydraulic properties of a dam (Huisman et al., 2010) and those controlling recharge dynamics (Kowalsky et al., 2011) and water infiltration and redistribution (Hinnell et al., 2010). One of the challenges of the fully coupled inversion is to identify and model all physical processes and parameters that control the different data sets. While different data sources can improve the estimated model and reduce its uncertainty, it is often difficult to adequately fit all available data.

In this study we evaluate how time-lapse ERT data can help to constrain simulations of CO2 migration for the SECARB pilot site located in Cranfield, Mississippi. After describing the field site and experiment (Section 2) and the inversion methodology (Section 3), we analyze time-lapse ERT data with structurally constrained and unconstrained ERT inversions in an effort to identify a reliable form of the data that can help calibrate our flow and transport models (Section 4). We then use ERT-derived changes in subsurface electrical resistivity along with gas composition data for model calibration. We test different conceptual flow models and invert the data for permeabilities and reservoir structure (Section 5). Finally, we test our inversion results by using various parameters that were estimated using simplified 1D models to build a 3D model (Section 5.3) and comparing the predicted system response with data that were not used in the inversions (Section 6).

Section snippets

SECARB Cranfield field site

The SECARB Cranfield pilot site (∼20 km east of Natchez, Mississippi) was chosen for a large scale (1 M tonne) CO2 injection into a brine-filled reservoir at ∼3200 m depth. The injection interval is a segment of the Lower Tuscaloosa Formation referred to as the Tuscaloosa D/E sand. This unit consists of relatively permeable fluvial sandstones and conglomerates. The system is described as fluvial point-bar and channel deposits by Lu et al. (2012b), who found that mineral reactions during CO2

Methodology

For combining the hydrological and geophysical data, we first invert the ERT data to get space–time distributions of electrical resistivity. These resistivity distributions are linked to gas saturation using Archie's law. The resistivity distributions along with gas composition measurements from U-tube sampling are used as observations to calibrate the hydrogeological model by running flow and transport simulations in iTOUGH2. This section gives an overview of these different components used in

Analysis of time-lapse ERT data

As mentioned in Section 3.2, we invert the ERT data for changes in subsurface resistivity prior to the coupled inversion. This section describes inversion of the time-lapse ERT data using different meshes and parameterizations. The meshes used for the ERT inversion are not related to those used in the flow and transport simulations, except that some of them include the same three layers.

Hydrological model development and fully coupled inversion

The hydrological modeling with different meshes and geometrical assumptions aims at identifying the core characteristics of the CO2 migration in the reservoir. We first consider a radial model, and then use the coupled inversion scheme to optimize the parameters of a model consisting of three 1D layers and use the inversion results to construct a 3D model. Sketches of the three meshes are shown in Fig. 7.

The complex interplay between CO2 and CH4 and the two flow paths that were inferred from

Discussion

The results of the ERT inversion (Section 4) show both the high potential and the limitations of electrical resistance monitoring of CO2 in deep reservoirs. The main advantage is clearly the sensitivity to changes in gas saturation at the inter-borehole scale that are otherwise very difficult to monitor. The challenges in the specific case of the Cranfield field experiment relate to the complex and involved field effort and the comparably poor data quality. Both factors are due to the harsh

Conclusions

In this work we investigate how electrical resistance monitoring data can be used for constraining models for simulation of the flow and transport of CO2 at the SECARB pilot site in Cranfield, Mississippi. We use the coupled modeling and inversion scheme of iTOUGH2, where simulation results are directly compared to available geophysical and hydrological data. In our approach, we first invert the ERT monitoring data for changes of subsurface resistivity. We can image the increasing resistivity

Acknowledgments

The authors would like to acknowledge the Cranfield and SECARB team for their scientific contributions and collaboration. We thank Barry Freifeld and Paul Cook (LBNL) for the U-tube design and many colleagues at the Texas Bureau of Economic Geology and Gulf Coast Carbon Center for U-tube data collection and Douglas LaBrecque (MPT) for assistance in acquiring the ERT data. We also would like to acknowledge David Freeman of Sandia Technologies for well design, instrumentation deployment and field

References (52)

  • J.A. Huisman et al.

    Hydraulic properties of a model dike from coupled Bayesian and multi-criteria hydrogeophysical inversion

    Journal of Hydrology

    (2010)
  • G. Landrot et al.

    Measurement of accessible reactive surface area in a sandstone, with application to CO2 mineralization

    Chemical Geology

    (2012)
  • J. Lu et al.

    CO2–rock–brine interactions in Lower Tuscaloosa Formation at Cranfield CO2 sequestration site, Mississippi, U.S.A

    Chemical Geology

    (2012)
  • J. Lu et al.

    Reservoir characterization and complications for trapping mechanisms at Cranfield CO2 injection site

    International Journal of Greenhouse Gas Control

    (2013)
  • Y. Nakatsuka et al.

    Experimental study on CO2 monitoring and quantification of stored CO2 in saline formations using resistivity measurements

    International Journal of Greenhouse Gas Control

    (2010)
  • L. Slater et al.

    Cross-hole electrical imaging of a controlled saline tracer injection

    Pure and Applied Geophysics

    (2000)
  • S.A. al Hagrey

    2D optimized electrode arrays for borehole resistivity tomography and CO2 sequestration modelling

    Pure and Applied Geophysics

    (2012)
  • G.E. Archie

    The electrical resistivity log as an aid in determining some reservoir characteristics

    Transactions of the American Institute of Mining, Metallurgical and Petroleum Engineers

    (1942)
  • S.M. Benson et al.

    Underground geological storage of carbon dioxide

  • P. Bergmann et al.

    Surface-downhole electrical resistivity tomography applied to monitoring of CO2 storage at Ketzin, Germany

    Geophysics

    (2012)
  • C.R. Carrigan et al.

    Application of ERT for tracking CO2 plume growth and movement at the SECARB Cranfield site

  • N.B. Christensen et al.

    Monitoring CO2 injection with cross-hole electrical resistivity tomography

    Exploration Geophysics

    (2006)
  • M. Commer et al.

    Advances in hydrogeophysical joint inversion

  • I. Coscia et al.

    Estimating traveltimes and groundwater flow patterns using 3D time-lapse crosshole ERT imaging of electrical resistivity fluctuations induced by infiltrating river water

    Geophysics

    (2012)
  • B. Dafflon et al.

    Monitoring CO2 intrusion and associated geochemical transformations in a shallow groundwater system using complex electrical methods

    Environmental Science and Technology

    (2013)
  • W. Daily et al.

    Cross-borehole resistivity tomography

    Geophysics

    (1991)
  • Cited by (33)

    View all citing articles on Scopus
    View full text