Carbon capture and storage (CCS) is a potentially important
technology for the mitigation of industrial

Geological storage of carbon dioxide (

Previous work on the subject includes the influence of estimated storage
capacity due to uncertainty in thermophysical properties (pressure and
temperature of the reservoir; Calvo et al., 2019). A rigorous Monte Carlo
approach has been demonstrated using the

Many published regional studies of

Furthermore, capacity assessments will largely depend on expert
interpretation of geological data, and are therefore dependent on the prior
knowledge and experience of individual experts (see Curtis, 2012, for
summary). Studies have shown that geological experts are subject to a range
of cognitive biases, as are all individuals (Kahneman et al., 1982), that
combined with differences in prior experience can influence their
interpretation of data leading to subjective results (e.g. Phillips, 1999;
Polson and Curtis, 2010; Bond et al., 2012). As a result, an estimate of the
uncertainty in single-value storage capacities is of practical use, not
least with assessments already published but lacking an assessment of
uncertainty. This is of particular practical importance where a storage
estimate falls close to a cut-off value, below which, for example, a
potential storage unit may be rejected as having a storage capacity that is too low
to be economically viable. For example, a regional screening study
(Wilkinson et al., 2010) rejected all units below an arbitrary 50 Mt of
estimated

Here, we test the hypothesis that the uncertainty in storage estimates is a significant proportion of the estimated storage capacity, and should hence be evaluated as a part of any assessment. For this study, an assessment of the precision of storage capacity estimates was conducted as part of a study of an area of the UK territorial waters, in the Inner Moray Firth area of the North Sea (Fig. 1). Subsurface geological data were available from boreholes drilled by the petroleum industry, both as individual well records released by the UK Government and summarised in scientific publications. The subsea strata are largely siliciclastics of Devonian to Jurassic age. They rest are unconformably on strata that were affected by the lower Palaeozoic Caledonian orogeny (Andrews et al., 1990), which are here considered to be basement (i.e. to have no storage potential). To the east of the area there is a variable-thickness cover of Cretaceous chalk, a fine-grained pelagic limestone, here not considered a potential store as it lacks an obvious seal. Questions concerning the presence of a suitable seal, trapping structures and potential leakage pathways were addressed in the wider study but are not reported here.

Location map of study area.

A group of 13 graduate students, who had been trained in the methodology of
storage capacity estimation and in at least basic geology relevant to

For surface area, the experts were directed to maps within Cameron (1993) and Richards et al. (1993); each expert independently estimated the area. Uncertainty in this parameter is therefore due to the variable interpretation of the same data from expert to expert. For the other parameters, the experts were expected to locate suitable data, primarily using web-based search tools. The uncertainty in these parameters is therefore determined by the total number and range of published values, the ease with which experts could find relevant information, and the interpretation by the experts of the applicability and reliability of the data that they located.

For the purposes of this paper, the values for each variable provided by the
experts were combined with constant values of

In order to determine the full range of possible estimates from the
expert-derived values, storage estimates were calculated for all possible
combinations of the variables. The resulting distribution of the storage
estimates,

There are seven geological units (which are either formations or members in
formal nomenclature; Cameron et al., 1993; Richards et al., 1993) that are
potential storage reservoirs in the area, henceforth called storage units.
Figure 2 shows

Range of individual expert and distribution (

Range of storage capacity estimates using the different values for variables found by group of experts for seven saline aquifers. Range is shown as a cumulative density function but does not represent the true probability density function for each aquifer.

The median values of the expert estimates tend to be similar to the median
of the distribution (within 10 %, except the Hopeman Sandstone Formation which is
within 20 %). The individual expert estimates tend to cover the range
from the 5th to 95th percentiles of

The 5th to 95th percentiles expressed as a percentage of the
median value of

The range of

The storage capacity estimates of seven saline aquifers by a group of experts shows that any single estimate by one expert might be a gross under- or overestimation of the median storage capacity. Even using a cohort of experts to provide independent estimates of the storage capacity does not cover the full range of possible values using just the data that those same experts collected. In particular, the range of expert estimates underestimated the highest values of the storage capacity by at least 40 % (and up to 120 %). As there are no reasons to assume that any one combination of variables is more or less likely than any other, all possible combinations must be assumed to have the same probability. Hence the storage capacity calculated using all minimum or maximum values for all variables are equally likely as any other individual combination, though there are more combinations of variables that will produce storage capacities around the median value than the extremes, making an estimate around the median more likely overall.

The number of experts in the study was necessarily limited; however, using more experts would not alter the outcome of the study. More experts may increase the range of estimates produced but would certainly not decrease it. Having more experts might be predicted to decrease the standard deviation of the mean estimate; however, as above, there is no reason to consider that the mean estimate is a better estimate of the true (unknown) value of the storage capacity than any other value.

It is therefore evident that the uncertainty associated with a single
estimate of

A further potential source of variability in the storage estimates is the influence of the individual assessors. Both personal judgement and previous experience have been shown to influence geological interpretation (Polson and Curtis, 2010). In this case, personal judgement is exercised when faced with parameters for which several data values are available, with no indication of which are more representative of the regional mean, and with no objective method of ranking the precision or importance of the values. One approach under these circumstances is simply to average the available values; the resulting mean clearly depends on which data have been located by the individual expert.

Personal judgement is required when estimating net : gross ratio as the most common source of data are borehole logs with a summary lithology column showing whether the sediments within the reservoir interval are interpreted as sandstone, silty sandstone, siltstone or mudstone (there are no significant limestones in the study area). Clearly mudstone is a non-reservoir, and sandstone is a potential reservoir, but a more or less arbitrary boundary between the two must be drawn. A more experienced wireline log interpreter might choose to ignore the summary lithology column of the composite log, and choose a value of, for example, the gamma ray log as an arbitrary cut-off between reservoir and non-reservoir, or estimate porosity (see below) and use an arbitrary minimum value of ca. 10 % for reservoir.

The most important control on the quality of the estimate of reservoir thickness is probably the number of borehole logs used to estimate the mean value. The most commonly used sources of data in this study (Cameron, 1993; Richards et al., 1993) typically present three summary borehole logs of each storage unit. However, the experts had access to 28 other composite (summary) borehole logs from the region, released by the UK Government. Some experts chose to use the entire suite of logs provided, while others used only a subset. Even if all logs are used, it is possible to use a range of methods to calculate mean regional thickness. For example, one can simply calculate the mean of the storage unit thickness data; alternatively, one could construct a map and interpolate contours, then estimate mean thickness by some simple graphical method involving dividing the storage unit into zones of constant thickness interval and calculating an average thickness weighted to the areas of the zones. It is also possible to use commercial software to perform both the contouring and the reservoir volume calculation, in which case calculating the mean thickness is unnecessary. Each of these approaches will result in different estimates of the thickness of the reservoir (or final gross reservoir volume).

For porosity, literature values can be utilised if they exist but if a range is given then the mean must be estimated. Sometimes porosity data are only provided graphically (as a cross-plot of porosity versus log of permeability) and the mean value can only be estimated visually as the points are frequently too dense to be read individually from the graphs. Alternatively, porosity can be calculated from borehole logs using standard methods – using formation density compensated (FDC) and compensated neutron logs (CNLs) for example – either manually or by using petrophysical computer software if the wireline logs are available in digital form. Again, the choice of method will influence the result. Measured porosity data are most commonly from within hydrocarbon fields, where the spatial density of boreholes is greatest. Whether the porosity of oilfield reservoirs is representative of the associated aquifer, or is systematically higher and thus introduces a systematic error in the estimate of aquifer porosity, is a controversial issue (e.g. Wilkinson and Haszeldine, 2011) for which a judgement is necessary. In a commercial study, it is possible to purchase porosity data measured from borehole core; unsurprisingly, none of the experts chose this option in this study.

The study reported here could be considered to be typical of regional studies conducted with the aim of ascertaining which geological units in a region are worthy of further study, i.e. a scoping study. The data available to the experts will be only a fraction of the total data collected from the area, and the data must obviously be located before being utilised. In any hydrocarbon province, it is unlikely that all possible data can be used in a regional scoping study, due to the large (often very large) volumes of data that have been collected historically and due to the non-availability of some (or much) of the data due to commercial confidentiality. Unless there are previously published syntheses of data with calculated averages of parameters such as the thickness of storage units, some proportion of the total data will be selected and utilised, inherently introducing uncertainty into the result.

Furthermore, the experts in this study could not spend unlimited periods of time searching for data, or in processing it once obtained. Again, this restriction is likely to be encountered in a regional scoping study, where many potential stores must be assessed within a fixed budget. The North Sea is also typical of hydrocarbon provinces in that there are a large number of boreholes drilled into relatively small areas (i.e. producing hydrocarbon fields) and relatively small numbers of boreholes in the much larger intervening areas. The spacing of the boreholes (data density) is probably not atypical of other offshore hydrocarbon provinces, though onshore hydrocarbon provinces may have much higher borehole densities (i.e. boreholes per square kilometre). Borehole records in the UK are released by the UK Government, so the density of available data may be comparable to other areas of the world where borehole density is greater but where drilling results are not so readily available due to commercial confidentiality.

While the uncertainty in estimated storage capacities will vary from study
to study, and can be reduced by costly data collection (or possibly
purchase) for any given geological unit, the results here suggests that
there is significant uncertainty in any storage capacity estimate that does
not include a site-specific estimate of uncertainty. Note that this analysis
does not take account of uncertainty in

It is not possible to estimate the likely uncertainty of any single storage
capacity estimate as there is no way to know whether it is at the lower, middle
or upper range of

Data for this study were limited to that in the public domain, which is probably realistic for a regional study where a potentially large number of candidate aquifers are assessed for first-order suitability for storage (e.g. Scottish Centre for Carbon Storage, 2009). It is probably not applicable to a detailed study of a single aquifer, where every effort is made to reduce key uncertainties and where confidential data may be available. For example, in the estimation of aquifer thickness, every borehole log that penetrates the storage unit could be utilised, removing the subjective element of choice associated with taking a subset of the available data. It is also likely that a more rigorous approach to uncertainty would be used in a single aquifer study, generating a reliable estimate of the likely range of capacity (e.g. Keating et al., 2011; Pawar et al., 2016). For this reason, the range of uncertainty for a detailed single-aquifer study should be substantially less than that derived here, and more comparable to the 4 % relative uncertainty at the 95 % confidence interval found in a detailed study by Deng et al. (2012).

The average standard deviation in

Uncertainty documented in this study is due to a mixture of spatial variability in the parameters combined with only limited availability of data, the number of independent (prior) estimates that are located for each parameter and the variation in interpretation of the same data by different experts. The range and standard deviation values in this study should be considered to be minimum values. The overall uncertainty is likely to be significantly larger as several sources of uncertainty are not accounted for in this study, in particular uncertainty due to storage efficiency could be larger than the geological uncertainty assessed here. Therefore a single assessment of a storage capacity of a geological unit, with no associated assessment of uncertainty, should be considered to have at least this degree of uncertainty in the absence of other information.

All data are available in the Supplement.

The supplement related to this article is available online at:

MW designed the initial concept and supervised the storage assessment exercise. DP performed the majority of the data analysis and interpretation.

The authors declare that they have no conflict of interest.

This article is part of the special issue “Understanding the unknowns: the impact of uncertainty in the geosciences”. It is a result of the EGU General Assembly 2018 , Vienna, Austria, 8–13 April 2018.

Thank you to all of the students of the Carbon Capture and Storage Masters of Science degree (2009–2010) at The University of Edinburgh, who gave permission for their results to be used in this paper. Borehole logs were from the Common Data Access database, which was kindly made available by Schlumberger.

This paper was edited by Juan Alcalde and reviewed by Ran Calvo, Philip Stauffer, and one anonymous referee.