OR/21/006 Manual picking
|Newell, A J, Woods, M A, Graham, R L, and Christodoulou, V. 2021. Derivation of lithofacies from geophysical logs: a review of methods from manual picking to machine learning. British Geological Survey Open Report, OR/21/006.
Contributor/editor: Kingdon, A
Basic principal and rationale
The manual approach to lithofacies classification of geophysical logs is a digitising process. This is generally (and most efficiently) undertaken within log interpretation software such as WellCAD where lithofacies boundaries can be inserted, dragged and edited using a mouse within a lithofacies track (Figure 2). The lithofacies track is placed adjacent (or overlain) on other log tracks such as gamma-ray and sonic transit time. These input logs form the basis for picking the lithofacies boundaries. The input logs are not (necessarily) required for any quantitative analysis so it is possible to use scaled raster scans of the logs in place of the digital versions if required. This can be advantageous in negating the need for log digitisation where digital versions are unavailable.
The primary advantage of this manual approach is that the geologist can place the lithofacies boundaries where they wish. This brings into play all of the experience and tacit knowledge that the geologist possesses on the formation under investigation. This might include the typical range of lithofacies that are found, the usual thickness of beds and any known recurring patterns of lithofacies that define cycles and rhythms within the formation. Moreover because the logs are not being used quantitatively the quality (and quantity) of the log data is less critical than other methods. The geologist can use their experience to interactively recognise, correct and account for poor log quality throughout the classification process. The disadvantages of this technique is that the geologist’s preconceived ideas of what they expect to see can influence their decision making and can lead to over interpretion that identifies subtle variations that are not effectively demonstrated at the resolution of logging tools or conversely missing subtle facies variations. Such issues can be effectively remediated by robust peer-review.
The technique can be particularly applicable to some superficially ‘homogeneous’ carbonate- dominated systems such as the Chalk Group where log responses are often subtle compared to other types of formation (e.g. siliciclastic). Moreover units such as the Chalk Group of the UK are penetrated by boreholes drilled for many different purposes (hydrocarbon, water, geotechnical, research) resulting in extremely heterogeneous geophysical log suites of vastly varying quality. This can hamper blanket automated recognition of lithofacies within the Chalk. To illustrate some of these issues and demonstrate the value of manual approaches in these types of scenario the Chalk is considered in more detail below.
Manual classification of lithofacies example: the chalk group
Optimal geophysical log combinations
The most widely available geophysical logs that are useful for interpreting stratigraphy and facies in the Chalk are gamma, resistivity and sonic. Induction logs are also valuable for deriving resistivity profiles for air-filled borehole intervals that cannot be recorded by resistivity tools, and digital image logs (where they exist) allow direct feature observation and refine understanding of corresponding geophysical responses. Ideally, a combination of gamma-ray and resistivity/sonic logs is desirable for maximum confidence of interpretation, but this rarely exists for the total depth range of Chalk in boreholes. In deep boreholes drilled for hydrocarbons exploration, the only logs typically run in the Chalk Group are caliper and low-resolution (often through casing) gamma-ray. For these boreholes, the availability of digital log data and ability to adjust the scale range is key to optimising interpretation value. For shallower boreholes (typical those less than 150 m deep), which in the Chalk Group have usually been drilled in connection with groundwater or major civil engineering investigations, there is usually complete coverage of gamma-ray logs and partial coverage of resistivity/sonic logs, the latter typically being restricted by the location of the water table and/or borehole construction (particularly borehole lining). However these often drilled by civil engineering focussed contractor recording in non-standardised units with quality standards below those adopted for oil industry work making quantitative assessment very difficult. In some instances, caliper logs may be a useful guide to stratigraphical boundaries (particularly where these correspond to changes in cementation), and for identifying particular types of feature (e.g. fracturing).
Overarching approach to classifying a thick carbonate formation.
For the purpose of geophysical log interpretation, the stratigraphy of the Chalk (Mortimore, 1986) can be thought of as comprising broad-shifts in mud content and/or cementation, within which is embedded a detailed framework of marker-beds. The marker-beds include the following:
- Very thin calcareous mudstones (marls) including bentonites (typically 50–100 mm thick)
- Hardgrounds, locally enriched with pyrite, glauconite and phosphate
- Sponge beds (cemented, typically with sponges preserved as iron-rich mesh-works)
- Flint bands (named flints typically ca.+30 cm & closely spaced/semi-continuous)
The marker-beds and facies units each have particular geophysical log features that allow their recognition, and these combine into packages of markers and facies defining lithostratigraphical units with distinctive associations of inflection patterns.
Recognition of key marker-beds
The methodology for interpreting marker-beds in the Chalk is described below:
Marls: Typically seen as localised sharply developed peaks in gamma-ray log values, corresponding with localised sharp drops in the resistivity log profile, and localised sharp increases in interval transit time on sonic logs. Geophysical log response varies with marl-type, from relatively stronger, higher amplitude signatures in solid closed marl seams and bentonitic marls, to weaker responses in anastomosing plexus marls that intercalate significant chalk sediment, and thin marl wisps and coatings on stylolitic surfaces. In Northern England, the development of high concentrations of thin marl seams in the Flamborough Chalk is signalled by an increase in “spikiness” of gamma-ray logs, and corresponding response of sonic logs (Woods, 2018; Figure 3).
Hardgrounds: Hardgrounds correspond to localised, very sharp increases in resistivity log values and localised sharp reductions in interval transit time. Corresponding gamma-ray logs typically show sharp increases because of the tendency of these features to concentrate iron and phosphate minerals (e.g. glauconite & apatite). Where both gamma-ray and resistivity/sonic logs are available, then the coupled response described above is strong evidence for hardground identification (Figure 4). When only a single log type is available, significantly more caution is required, and interpretations should take account of the stratigraphical context provided by interpretations of adjacent intervals and any laterally related boreholes. The shape of the peak in gamma-ray log values is typically somewhat different from marl seams, usually broader and blunter (reflecting the fact that hardgrounds are typically thicker (dm) than marls (cm), with much higher amplitude. Sharp peaks on resistivity and sonic logs can also be produced by very thick flint seams (typically with low gamma-ray values), or where flint seams are highly concentrated, and might be difficult to distinguish from hardgrounds if no gamma-ray log is available.
Hardgrounds imply significant stratigraphical omission, but other thin, hard cemented units in the Chalk that are not hardgrounds, and likely represent reductions in the rate of sedimentation (e.g. sponge beds), may have a similar geophysical log response. Typically both the gamma-ray and resistivity/sonic responses are lower amplitude compared to hardgrounds (because cementation is less extreme & mineral enrichment is limited to thin films of iron-oxide/hydroxide associated with sponge preservation), but understanding this relative difference can be difficult if undoubted hardground signatures are not available for comparison. Again, the stratigraphical context provided by interpretation of adjacent intervals and laterally related boreholes will be an important guide to feature discrimination.
Flints: Although flint is widely distributed in the Chalk, and there are a number of important named flint marker-beds that can be traced laterally for 100s km. However, it is often difficult to recognise their individual signatures on geophysical logs. This may partly reflect the brittle character of flint, and its tendency to fragment during coring rather than providing a clean, solid surface for instrument detection. There are a few examples of where recognition has proved possible, particularly in the Chalk of northern England, where thick (+30 cm) bands of laterally continuous flint form a series of named markers at the base of the Late Turonian Burnham Chalk Formation (Figure 3). These flints can be identified by sharply defined high resistivity/‘fast’ sonic (low interval transit time) peaks and low gamma-ray log values.
Manual log interpretation of facies units
The creation of a regional geological model of the Chalk across southern England (Woods et al., 2016; Newell et al., 2018) included information about chalk facies units derived from borehole geophysical logs. In addition to facies types represented by marker-beds (see above), these included:
- Hard chalk
- Marly chalk
- Phosphatic chalk
In the above list, ‘Limestone’ in this context refers to cemented carbonate units that are not dominated by nannofossils and lack typical chalk fabrics. There is a close association of the limestone and mudstone facies, and in this report (below) they are discussed as ‘Interbedded limestone/mudstone facies’. Although not forming part of the previous facies modelling work, it is also possible to identify very flint-rich chalk, also described below.
Hard Chalk: Strong lateral shifts in resistivity and sonic logs to higher values (resistivity) and fast interval transit times (sonic), that are sustained over 10s of metres of strata, can generally be used to infer the presence of hard chalk. In combination with marker-bed information, and knowledge of coeval successions at outcrop and in boreholes, inference of ’nodular’ fabrics is possible. Interpretational ambiguity can be caused by high concentrations of flint, which collectively may cause similar but usually less marked shifts in resistivity and sonic profiles. If the shift itself occurs gradually, over 10s of m, then in the context of the known mechanisms that typically influence broad patterns of hardness in the Chalk (relative sea level change and basin architecture), a response to flint content should be suspected. Additionally, regional knowledge and stratigraphical context can provide valuable supporting evidence about the potential influence of flint on geophysical logs.
Flint-rich chalk: Some intervals in the Chalk contain high concentrations of flint over narrow stratigraphical intervals that are laterally persistent, particularly in the Late Turonian (e.g. Brandon Flint Series; Mortimore & Wood, 1986). Individual flints and flint-rich intervals typically correspond with very low background gamma-ray log values, serving to distinguish the corresponding fast sonic and elevated resistivity log responses from highly cemented chalk intervals.
Marly Chalk: This facies represents chalk with a greater background mud content, but is distinct from a marl seam where there is greater separation of mud and chalk phases. On geophysical logs, marly chalk is represented by an increase in average gamma-ray log values with correspondingly lower resistivity/slower sonic responses. In the Chalk, the admixture of mud and chalk usually results in a relatively poorly cemented lithology. In contrast, carbonate systems that are not dominated by nannofossils, the equivalent facies (wackestone) is typically highly cemented. The Late Cenomanian Grey Chalk Subgroup provides a good example of marly chalk facies (Figure 6), in which massive-bedded units of creamy-grey chalk have significantly higher gamma-ray log values than the purer chalk deposited in the Early Turonian.
Intercalated mudstone/limestone facies: This facies is typical of mixed carbonate systems, its development in the lower part of the Chalk Group (Grey Chalk Subgroup) reflecting the influence of Milankovitch cycles immediately prior to deep flooding of continental shelves (and corresponding restriction of clastic input) in the earliest Turonian (Gale et al., 1999). Gamma, resistivity and sonic logs show a pronounced, high frequency, out-of-phase oscillations over wavelengths of 1–5 m, with high gamma-ray log intervals corresponding with low resistivity/fast sonic response (Figure 7).
Sandy (including glauconitic) chalk: This facies is exemplified by intervals in the basal and lower parts of the Chalk Group (Grey Chalk Subgroup), and corresponds with significantly elevated gamma-ray log values that may extend for 5–10 m. The strong gamma-ray log response is driven by glauconite and dispersed mud content, usually boosted by significant quantities of phosphatic clasts. Patterns of log serration are a response to the piping-down of more chalky sediment infilling burrow systems that are usually pervasive. Moderately elevated resistivity and intermediate interval transit times on sonic logs reflect the significant sand content (Figure 8), but the responses of these logs are generally less amplified than for cemented chalk and hardground intervals.
Phosphatic Chalk: In the Chalk, this facies may be developed as a conglomerate of phosphatic nodules and phosphate-encrusted limestone cobbles, usually less than a metre thick, or as intervals of fine granular phosphate, sometimes more than 10 m thick. Where developed as a phosphatic conglomerate, the geophysical log response can be hard to separate from that of a hardground, since in both gamma-ray log values tend to be sharply elevated and coincide with high resistivity/fast sonic log responses (Figure 9). In many ways these intervals resemble hardgrounds, being associated with condensed sedimentation with nodules/cobbles bored and encrusted with marine serplulids and molluscs. A distinction is that erosion and winnowing appear to dominate over in-situ sediment lithification that characterises hardgrounds.
Understanding the stratigraphical context of these inflection patterns, and knowledge about the likely distribution of erosion surfaces, provides a valuable guide to their correct interpretation.
Relatively thick intervals (+15 m) of granular phosphatic chalk have recently been reported and geophysically logged along the alignment of a proposed road tunnel adjacent to Stonehenge (Mortimore, 2014, figure 3.26c). The gamma-ray log signature rapidly builds out over a few metres, from background values of 40 API in the host chalk, to 140 API in the phosphatic interval. The signature shows high frequency oscillations and longer wavelength variability, the latter probably reflecting detail of the internal sedimentary geometry. In Chalk, there is usually a close association between the development of these features and patterns of channelling developed in response to local structure (Mortimore et al., 2017). These may be evident on seismic profiles (Evans and Hopson, 2000).
Manual log interpretation of facies packages that define formational units
Onshore UK Chalk Group stratigraphy was rationalised by Rawson et al. (2001), and comprises a highly distinctive arrangement of marker-beds and facies units (Mortimore, 1986; Wood & Smith, 1978; Whitham, 1991, 1993) into broader packages with particular geophysical log patterns. These facies packages correspond with Chalk formational units, and the methodology for their recognition is described below.
- Look for any broad shifts in log profile that act to anchor the interpretation at one or more levels. A good example is the Cenomanian/Turonian boundary, at the base of the Holywell Nodular Chalk, where the mud-rich Plenus Marls are overlain by the highly cemented Melbourn Rock. This stark contrast in physical properties is marked by a major inflection in both resistivity/sonic and gamma-ray logs, and gamma-ray values below the Plenus Marls are consistently higher than above because of higher overall mud content (Figure 10). Another distinctive ‘event bundle’ occurs in Late Turonian Chalk successions, where a closely spaced successions of marls and hardgrounds in the lower part of the Lewes Nodular Chalk produce a series of sharp oscillations in resistivity/sonic and gamma-ray log signatures (Figure 11a). Confirm that this is consistent with any local or regionally available control data (e.g. published log interpretations, SOBI borehole data, and regionally relevant thickness data).
- Look for finer-scale marker-bed event bundles. These second-order features are usually slightly less obvious because they are either changes to a single log type, or trends in the pattern or amplitude of log values through a broader depth range. Typically these events only become more meaningful for interpretation in the context of the parameters set by the first order log features. Examples of second-order events are: 1) increase in gamma-ray log serrations 15–30 m above the base of the Plenus Marls (indicative of the marl-rich New Pit Chalk); 2) significant low-resistivity punctuations coincident with high gamma-ray peaks in the middle and higher part of the New Pit Chalk, coinciding with increased frequency and thickness of marl seams; 3) discrete series of typically 6 gamma-ray log peaks just above the base of the Seaford Chalk.
- Where interpreting a series of spatially associated boreholes, explore the detailed pattern of inflections between the anchor points formed by marker-beds in adjacent boreholes. This has three important functions: 1) patterns that closely match enhance confidence of interpretation; 2) trends in the spatial organisation of these inflections provides refined understanding of where sedimentary packages are expanded or condensed locally within successions; 3) where there is variability in the quality or availability of log data for different boreholes, but high confidence in likely correlation, then flattening adjacent boreholes on a common stratigraphical datum can provide increased information about the likely position of stratigraphical boundaries.
- Confirm that stratigraphical picks are consistent with regional structural trends and up-to-date geological map data from the borehole vicinity. Consider the availability of outcrop biostratigraphical data to constrain near-surface log interpretations. If inconsistencies are apparent, determine if there are other data to support a stratigraphical/structural explanation for this. If not, re-evaluate correlation and decide if the error is likely to be with first or second-order features. If first-order features are suspected to be incorrectly assigned, consider the wider implications of this for other borehole correlations and if necessary re-evaluate the correlation of all first-order features in all boreholes.
- Refine interpretation of stratigraphical boundaries to ensure consistency. In the Chalk Group, consistency of interpretation is affected by two issues: 1) the transitional nature of facies changes at stratigraphical boundaries, and 2) log resolution. In stratotype sections the boundaries of Chalk units are named marker-beds, but these are not always easily recognisable as discrete entities on borehole logs, and the associated shifts in facies patterns that they signal are rarely precisely coincident. In such cases, trends in log patterns, for example increasing/declining resistivity/gamma-ray profiles become important for boundary interpretation, and a consistent decision needs to be made about where along the trend line a boundary is likely to be located.
- Inability to identify stratigraphy-defining marker-beds can be associated with geophysical data of particular vintage. A significant number of single point resistivity logs in BGS archives record data at widely spaced depth increments. Although the trends are still meaningful for stratigraphical interpretation, multiple marker-beds that are usually separately resolved on continuous logs may be covered by a single inflection on vintage logs. This problem can be mitigated if gamma-ray logs are also available. A good example of facies transition at a stratigraphical boundary is the junction of the Holywell Nodular Chalk and New Pit Chalk (Figure 11b), and Figure 12 compares the difference in marker-bed resolution revealed by vintage single point and modern continuous resistivity logging. Therefore log age and quality is a proxy of data uncertainty; borehole with less uncertain data should be prioritised and the most uncertain data considered for exclusion.
Manual facies interpretation in offshore Chalk successions (North Sea)
Manually picked geophysical log responses underpin the stratigraphical nomenclature for the Chalk in the North Sea (here including strata of earliest Palaeocene age) (Johnson & Lott, 1993; Lott & Knox, 1994; Gradstein & Waters, 2018; Figure 13). Formations are recognised by broad increases or decreases in gamma-ray and sonic log values, and by recognition of particular patterns of inflections in a series of reference boreholes, with microfossils (foraminifera, dinoflagellate cysts & nannofossils) where available providing a correlatable framework of age-related marker-horizons. Patterns of intraformational log variability (that are not stratigraphically formalised) have been described in the context of facies variation, for example, significantly high gamma-ray log responses in the basal part of the Ekofisk Formation, corresponding with a mud-rich chalk interval (Johnson & Lott, 1993). The logs illustrated by Johnson and Lott (1993) and Lott and Knox (1994) suggest significant potential for identifying facies subdivisions, and Mortimore (2014, figs 5.5–5.9) highlighted some of these on logs that he illustrated. In the context of the history of North Sea Chalk sedimentation (Mortimore, 2014), these facies are likely to be significantly more variable and diachronous than for onshore Chalk successions.
Pros and cons of manual facies interpretation on geophysical logs in carbonate-dominated systems
For carbonate systems like the Chalk, the main advantage of manually interpreting carbonate facies on geophysical logs is that it overcomes the problem of subtle log-responses to facies contrasts, and the difficulty this creates for establishing appropriate cut-off values for automated systems of facies interpretation. For the Chalk Group, recognising changes in the frequency and overall pattern of low amplitude geophysical log responses is the most important aspect of facies and stratigraphy interpretation. This conclusion likely reflects the fact that the Chalk was deposited on a deeply flooded shelf, with more subtle facies responses to sea level oscillation than might be predicted for a shallow carbonate platform or ramp setting. The main disadvantages of the manual picking methodology are that it is usually time-consuming; requires detailed knowledge of stratigraphical trends and regional facies variation, and may be guided by a conceptual model that is flawed and over-looks the significance of log responses that are inconsistent with this model. Where large-scale log responses are developed in carbonate facies systems, there is a potential role for using threshold cut-off values to highlight first-order anchor points for log interpretation and correlation.
- MORTIMORE, R N. 1986. Stratigraphy of the Upper Cretaceous White Chalk of Sussex. Proceedings of the Geologists’ Association, Vol. 97, 97–139.
- WOODS, M A. 2018. Geophysical log characterization of the Flamborough Chalk Formation (Late Cretaceous, Santonian—Early Campanian), East Yorkshire, UK: implications for understanding the onshore and offshore occurrence of the Late Campanian Rowe Chalk Formation. Proceedings of the Yorkshire Geological Society, Vol.62, 201–209.
- WOODS, M A, NEWELL, A, HASLAM, R, FARRANT, A, and SMITH, H. 2016. Predicting the Subsurface. Geoscientist, Vol.26, 10 13.
- NEWELL, A J, WOODS, M A, FARRANT, A R, SMITH, H, and HASLAM, R B. 2018. Chalk thickness trends and the role of tectonic processes in the Upper Cretaceous of southern England. Proceedings of the Geologists’ Association, Vol.129, 610–628.
- MORTIMORE, R N, and WOOD, C J. 1986. The distribution of flint in the English Chalk, with particular reference to the ‘Brandon Flint Series’ and the high Turonian flint maximum. In: The Scientific Study of Flint and Chert, Sieveking, G de G (ed), Cambridge University Press: Cambridge, 7–20.
- GALE, A S, YOUNG, J R, SHACKLETON, N J, CROWHURST, S J, and WRAY, D S. 1999. Orbital Tuning of Cenomanian Marly Chalk Successions: Towards a Milankovitch Time-Scale for the Late Cretaceous. Philosophical Transactions: Mathematical, Physical and Engineering Sciences, Vol.357, No.1757, 1815–1829.
- MORTIMORE, R N. 2014. Logging the Chalk. Whittles Publishing: Caithness, pp.357 pp.
- MORTIMORE, R N, GALLAGHER, L T, GELDER, J T, MOORE, I R, BROOKS, R, and FARRANT, A R. 2017. Stonehenge—a unique Late Cretaceous phosphatic Chalk geology: implications for sea-level, climate and tectonics and impact on engineering and archaeology. Proceedings of the Geologists’ Association, Vol.128, 564–598.
- EVANS, D J, HOPSON, P. 2000. The seismic expression of synsedimentary channel features within the Chalk of southern England. Proceedings of the Geologists’ Association, Vol.111, 219–230.
- RAWSON, P F, ALLEN, P , and GALE, A.S. 2001. The Chalk Group—a revised lithostratigraphy. Geoscientist, Vol.11, 21.
- WOOD, C J, and SMITH, E G. 1978. Lithostratigraphical classification of the Chalk in North Yorkshire, Humberside and Lincolnshire. Proceedings of the Yorkshire Geological Society, Vol.42, 263–287.
- WHITHAM, F. 1991. The stratigraphy of the Upper Cretaceous Ferriby, Welton and Burnham formations north of the Humber, north-east England. Proceedings of the Yorkshire Geological Society, Vol.48, 227–254.
- WHITHAM, F. 1993. The stratigraphy of the Upper Cretaceous Flamborough Chalk Formation north of the Humber, north-east England. Proceedings of the Yorkshire Geological Society, Vol.49, 235–258
- JOHNSON, H, and LOTT, G K. 1993. 2. Cretaceous of the Central and Northern North Sea. In: Knox, R W O'B, and Cordey, W G. (eds.) Lithostratigraphic nomenclature of the UK North Sea. British Geological Survey, Nottingham.
- LOTT, G K, and KNOX, R W O’B. 1994. Post Triassic of the Southern North Sea. Volume 7. In: Knox, R W O’B. & Cordey, W G. (eds), Lithostratigraphic nomenclature of the UK North Sea. British Geological Survey, Nottingham.