CO2 capture and storage is considered as one of the key measures for mitigation of climate changes due to emissions from fossil fuel combustion. For safe geological storage of CO2, very good knowledge about the subsurface storage and sealing formations, as well as about the behaviour of CO2 in the subsurface is required. Specifically, accurate and reliable monitoring methods have to be used both before and during injection, but also for subsequent surveillance of the storage site.
Commonly used methods provide an image of the subsurface which may be used for instance for early detection of leakage. This makes monitoring methods crucial for reduction of risks associated with CO2 injection and storage. World-wide efforts have been made to study and improve the accuracy of these methods, but so far very little has been done to quantify how certain the information provided by the monitoring images is. Sound knowledge of the uncertainty in an image is a vital component in quantification of risks during injection. In addition, this increases the confidence in the assessment of a storage site prior to injection.
In this project, the focus is on developing a methodology for quantification of uncertainties in geophysical monitoring. For given synthetic cases, the resulting uncertainties are kept at a minimum by tuning the monitoring methods. The methods will also be tailored such as to provide monitoring at Sleipner (off-shore Stavanger) with the smallest possible amount of uncertainties.
During the first half of the project, the work was mainly focused on the following tasks: development of a "work bench" of monitoring (imaging) methods, development of an efficient forward modelling scheme, development of a first strategy for uncertainty quantification, incorporation of this quantification in the monitoring methods, and preparation of realistic synthetic models for testing of the methods.
The work bench is meant to facilitate all further development of the monitoring methods and the incorporation of functionalities for uncertainty quantification. Based on existing SINTEF codes (FWI and CSEM), a new modular software has been implemented for which a much larger degree of flexibility is obtained.
For the uncertainty quantification, a few different potential strategies have been considered, some of which have been previously described in literature for other research topics. The first strategy implemented is one which is accurate, but with a high computational cost. This strategy will serve as a baseline case for further studies. First tests on the synthetic models were carried out during the Summer 2015 and results were presented at SIAM Conference on Mathematical and Computational Issues in the Geosciences.
During the first half of 2016, a Postdoc studied alternative, more efficient and accurate methods for uncertainty quantification. The focus was for a while on studying to what extent it is possible to find the model that optimally (in a global rather than local sense) explains the gathered geophysical data. After a while it was concluded that this was not a realistic alternative since the required computational power was far too large. Instead, studies have been performed to find out how to get most information from the matrix (the Hessian) that describes, locally, how much an alternative model can deviate from the best one, without explaining the gathered data worse. The Postdoc has also worked on some necessary updates of SINTEF's CSEM-code. SINTEF har under andre halvdel av 2016 utført en omfattende studie av en ny metode for usikkerhetskvantifisering. During the second half of 2016, SINTEF has carried out a study of the use of a new uncertainty quantification method with very good results both for FWI and CSEM at Sleipner. The work was presented at GHGT-13 and has been accepted for publication in Energy Procedia in 2017. The most recent work on the uncertainty quantification will be presented at TCCS-9, EAGE, and the EAGE/SEG workshop on CO2 monitoring.
In addition, there are on-going studies of the use of AVO analysis on Sleipner data for improved a priori information for the above mentioned FWI code. This was not part of the original project plan, but was deemed to be such an interesting alternative to using a priori information from thin layer analysis (work at BGS), that an additional effort was made on this topic. The first parts of this AVO work was presented at GHGT-13 and has also been published in Energy Procedia in 2017.
A major publication of quantitative monitoring at Sleipner was also published in a special edition of Interpretation in November 2017.
In this project, we propose the development of a method for the quantification and the reduction of uncertainties in geophysical models during CO2 monitoring.
While the development in this project is implemented for seismic Full Waveform Inversion (FWI), and Controlled Source Electro-Magnetics (CSEM) the approach should be valid for other geophysical methods as well.
We will use a benchmark environment to facilitate the implementation and the testing, as well as the collaboration and the communication b etween the different research partners. We aim to develop and implement the uncertainty quantification techniques for CO2 monitoring methods, addressing both the uncertainties of the model parameter amplitudes, and the spatial uncertainties of structure ( e.g., the extent of the CO2 plume). The analysis of the uncertainties will be based on the analysis of a posteriori covariance matrix. Various ways of quantifying uncertainties in monitoring methods and how to reduce these uncertainties for a particular s ite will be investigated.
We will utilize the uncertainty information to optimize the CO2 monitoring for different scenarios. This includes the investigation of various misfit functions, inversion schemes, and regularization strategies for realistic synth etic models and data. Accurate a priori information based on interpretative data will also be included in the investigation. The resulting methodology will be applied to Sleipner and Snøhvit
The knowledge of uncertainties of geophysical models will incre ase the value of these models significantly, and may improve the interpretation of geophysical monitoring, and the quantification of CO2. We also anticipate that the results of this project may be used in planning CO2 injection projects, and the design of efficient and cost effective monitoring programs, both for CO2 storage and EOR.