Work package 2: monitoring to understand detailed flow processes and improve in situ quantification, testing hypotheses 3 and 4 (Edinburgh, BGS, NOC)

Summary

A key UK regulatory requirement is to demonstrate understanding of in-reservoir processes. This forms the basis of risk management and medium- to long-term performance prediction. Evidence so far suggests that, because of its very low viscosity, injected CO₂ tends to accumulate and migrate in the subsurface as thin, very mobile layers, typically a few metres thick (Williams and Chadwick, 2012). Characterising the static and dynamic properties of these layers is essential to understand CO₂ propagation processes, and for verifying and calibrating predictive flow models. Seismic techniques have the necessary resolution to accomplish this, but a range of significant uncertainties need to be overcome. The relationship between seismic velocity and CO₂ saturation depends on the nature of the mixing between the CO₂ and water phases. With uniform mixing a rather small quantity of CO₂ can radically reduce the low frequency seismic velocities of otherwise water-saturated rock, but further increasing CO₂ saturation has only a small additional effect. With more 'patchy' mixing, velocity change with saturation is more linear. This leads to a fundamental difficulty in obtaining reliable saturation estimates from seismic velocity. In addition, it is difficult to disentangle the competing effects of pressure and saturation changes and the effects of fractures, which can influence both the flow properties and the resultant seismic signatures.

Task 2.1: integrated very high-resolution flow modelling and seismic analysis

Very high-resolution numerical flow simulations of spreading layers (with decimetre-scale cell thicknesses) will be used together with synthetic seismic forward modelling to calibrate and control analysis of real datasets from Sleipner. Temporal layer thicknesses and layer velocities will be addressed by a number of methods, each with attendant uncertainty: structural analysis, reflection amplitudes, high accuracy time-shift measurement, spectral decomposition and spectral inversion. Different methods will be used in tandem to reduce non-uniqueness of solutions and results will complement the novel analysis in Task 2.2.

Task 2.2: novel methods for layer characterisation

Seismic attenuation is much more sensitive to the degree of gas saturation than seismic velocity. Moreover, while the relationship between velocity and saturation is approximately monotonic, the relationship between attenuation and saturation is peaked, with small values of attenuation for full-water and full-CO₂ saturations, and a maximum for intermediate CO₂ saturations. This fundamental difference can be exploited to greatly enhance our ability to estimate CO₂ saturation where both velocity and attenuation data are available. Application to seismic data from the Vienna basin (Wu, Chapman and Angerer, 2011) showed great potential for enhanced discrimination of gas saturation using a frequency-dependent amplitude versus offset technique.

We will test hypothesis 4 by attempting rock physics model-based inversion of CO₂ saturation from seismic data through measurements of both velocity and attenuation attributes. The key issue will be to quantify the various uncertainties in the method, which are of three forms. First is intrinsic uncertainty about how to model anelastic properties of fluid saturated rocks. Many models have been proposed and, while there is agreement over trends, differences exist in the details. We hope that the work package 3 experiments can greatly reduce these uncertainties. Secondly, fluid information can only be derived if we assume that background rock properties are known, but in reality these properties vary spatially. A further complication is the presence of fractures and seismic anisotropy, so that azimuthally varying seismic attributes will have to be considered. Finally, deriving attenuation estimates from seismic data is error prone, whether we use classical techniques such as the spectral ratio method or more modern techniques based around frequency-dependence of reflectivity. In particular, multiple scattering effects, whether related to layer thickness or lateral variation in saturation, can be expected to corrupt the measurements. We will employ careful processing and modelling to mitigate these issues.

Although some obstacles to successful application are evident, we believe that our approach, combining state of the art rock physics theory and experiments, extensive seismic modelling and data analysis, is the correct way forward and offers the possibility of a step-change in our ability to monitor CO₂ saturation. In addition to testing the method on the Sleipner data, it is hoped that the uncertainty analysis will be able to provide some guidelines on the conditions under which the techniques will be applicable for other areas.