Biostratigraphic correlation based on the geological succession of fossil species remains one of the fundamental workhorse approaches to geological correlation in the exploration industry. The method is only as good as its sampling, and even in the most thoroughly sampled sequence a plethora of problems conspire to confound the interpreter. For example, the lowest or highest observation of a particular fossil species in a well will never represent the true first or last occurrence in time – a problem that reflects sample spacing, stochastic effects related to limited sample sizes and examination effort, and the incomplete nature of the stratigraphic and fossil records. Furthermore, other potential sources of confusion include cavings, fossil reworking, post-depositional mixing and biogeographic and facies effects. For these reasons, the observed orders of lowest or highest occurrences of particular species vary from well to well.
Despite these problems, global geology has been extremely well served by biostratigraphy because scientists have tended to focus on a small subset of ‘well behaved’ fossil events that tend to show minimal variation in order. These key events are well behaved because they are relatively widely spaced in time and thus have less likelihood of showing event-order contradiction from well to well. They have also been selected because they are apparently relatively insensitive to facies effects and are geographically widely distributed.
Constrained Optimization
Before and after: correlations between two well-sampled, onshore sections based on marine dinoflagellate microfossils. These two sections are in the East Coast Basin and span all of the Campanian (Late Cretaceous, ~83.5–70.6 Ma). On the left, blue lines link the observed lowest or highest occurrences of species in each section. On the right, red lines are the inferred correlations based on the CONOP best-fit composite sequence of biostratigraphic events; red and yellow blocks indicate the limits of traditional biostratigraphic resolution using biozones. (Data from Crampton et al., 2006, Geological Society America bulletin 118: 975-990.) Source: GNS ScienceAlthough still an important part of the biostratigraphic toolbox, the use of key fossils ignores a very large amount of biostratigraphic data and inevitably limits the resolution of correlation that can be achieved. This issue is becoming increasingly important as explorationists and basin modelers seek to squeeze ever greater time and depth resolution from available data, with the ultimate aim of correlating to absolute (geochronological) age. Age is a key property in generation modeling and is commonly assigned to structural surfaces as a value selected arbitrarily from an age range or from age maps that better capture the diachronous nature of many geological surfaces. Increased geochronological resolution and improved lithological detail are critical to better identification of potential hydrocarbon carrier beds and seals.
In response to these issues, biostratigraphers have developed a suite of fully or semi-quantitative techniques that use many or most available events and determine ‘best-fit’ solutions to event-order contradictions. The task of identifying the best-fit event ordering is known as the ‘correlation problem’. The oldest and simplest approach to this is graphic correlation, with events integrated graphically, well-by-well, into a best-fit composite sequence. Although graphic correlation is used widely, it is time-consuming and retains a subjective, operator-dependent element.
Newer techniques include Constrained Optimization (CONOP1), which is fully automated, handles large datasets, and has been shown to be a versatile, efficient and powerful tool in hydrocarbon exploration2. Importantly, recent developments have allowed for the incorporation of non-biostratigraphic data, such as seismic picks, chemostratigraphic and chronostratigraphic datums, with great flexibility in the assumptions attached to each particular data type. The CONOP method has been applied successfully in two of New Zealand’s geologically complex basins containing thick, dominantly terrigenous clastic successions, which have undergone intense polyphase deformation resulting in highly complex stratigraphies.
The aim of CONOP is to produce a best-fit composite sequence of events, to age-calibrate this composite, and to achieve the most refined correlation possible by locating the likely ‘true’ position of each event in each well. Importantly, the approach also yields various measures of well- and event-reliability that can be used to guide future sampling and interpretation. Typically, CONOP analysis is iterative, with run-time settings and data being refined progressively in order to converge on the best solution.
Untangling the Correlation Knot
The utility of CONOP in solving the correlation problem is illustrated on page 31, based on 15 onshore sections from the East Coast and northern Canterbury basins. The East Coast Basin is located on the eastern North Island and occupies the accretionary prism of the present-day Pacific-Australian plate boundary, although it initiated on a mid-Cretaceous passive margin. The Canterbury Basin, on the east coast of the South Island, initiated in response to mid-Cretaceous rifting. Late Cretaceous formations in both basins are characterized by thick and commonly monotonous mudstone or very fine sandstone successions.
Even in the absence of caving and associated sampling issues, there is a dramatic contradiction in biostratigraphic correlations based on marine microfossils. Following CONOP analysis, these contradictions are eliminated and the most parsimonious, internally consistent correlation is identified. In this study, the composite sequence was age-calibrated using a subset of dated biostratigraphic events, to yield an average composite age resolution of 130,000 years for the interval between ~90 and ~65 Ma. In addition to the highly refined correlation and dating of sections, the CONOP analysis also helped to identify both basin-wide and local unconformities – a topic that is explored in more detail in the following section.
Comparison of CONOP and seismic correlations for three wells in the Taranaki Basin. Both panels are flattened on seismic pick P60. The CONOP correlation also shows the time-calibrated composite on the left. Some seismic picks are duplicated in the CONOP panel – this reflects the limits of uncertainty intervals, as explained in the text. A few discrepancies are visible between the panels – such as apparent stratigraphic condensation in the lower part of Tane-1 in the CONOP correlation, which relates mainly to presently unresolved questions regarding treatment in CONOP of a problematic seismic pick, P20, a regional unconformity that is probably strongly diachronous. Source: GNS Science
Quantitative Stratigraphy in the Taranaki Basin
Composite log of Maui-4 well, Taranaki Basin. Source: GNS ScienceThe Taranaki Basin, New Zealand’s only producing basin, is located along the west coast of the North Island. Taranaki initially formed as an Early Cretaceous to Paleogene intraplate rift basin during fragmentation of the Gondwana continental margin. The basin was structurally overprinted in the Neogene as a result of the development of the modern convergent Pacific-Australian plate boundary. Exploration targets include structural and stratigraphic traps formed by rift-related processes and compressional inversion structures.
Data derive from 20 offshore wells with the highest available quality biostratigraphic and seismic information. The initial dataset contained depths to biostratigraphic events for 1,376 species of foraminifera, dinoflagellates, and pollen, derived from 2,829 cuttings, sidewall core, and conventional core samples. Biostratigraphic events are mainly lowest- and highest-occurrences of species, but include some foraminiferal coiling-event boundaries and pollen assemblage-zone boundaries. The study also included 77 estimated depths to 18 seismic reflectors in 12 of the wells, and 106 depths to 34 lithostratigraphic unit tops in 18 of the wells. Depths to seismic reflectors were encoded conservatively as uncertainty intervals to take account of various sources of error: vertical resolution of seismic data, mis-picks, including those due to bulk shifts applied during phase-matching, and time-versus-depth relationship errors; the summed errors are typically on the order of ± 100–200m.
Importantly, these various event types were treated quite differently during analysis, and event treatments were tested repeatedly and modified during interative CONOP runs. Thus, different biostratigraphic event types were given varying weights in the analysis to reflect assumptions regarding their reliability and likelihood of misplacement. For example, observed lowest-occurrences were down-weighted and allowed to move freely in the solution, to account for the likelihood of caving and low confidence in these observations. In contrast and following testing, some observations of foraminiferal coiling events were treated as high-value time-planes, albeit with uncertainty in stratigraphic placement. Seismic reflectors were initially treated as independent events in different wells and allowed to ‘float’ in the composite. Those that converged in the resulting solution were subsequently encoded as time-planes that were constrained to lie within their uncertainty intervals, and these intervals were ‘shrunk’ according to biostratigraphic constraints. Lithostratigraphic tops, on the other hand, were regarded as precisely located in each well but were never treated as time-planes.
New Information Revealed
CONOP absolute age to depth data points can be plotted on a well composite log relative to the well depth. Plotting event ages along a proportional depth scale means that the age-calibrated time scale of the composite becomes non-linear, expanding in intervals with high sedimentation rate and shrinking in intervals of low or negative (erosion) sedimentation rate. Viewed in this way, the data allow the interpreter to identify unconformities or condensed sections, relate them to petrophysical log and sedimentological patterns, and – via seismic time to depth conversion and synthetics – to the seismic wiggle, as seen above. Interpretation of unconformities is subject to some caveats, of course: clustering of events at one horizon may indicate the position of a local unconformity, a regional (sample-wide) unconformity, times of elevated evolution or extinction rate, or simply wide sample spacing. Once an unconformity is identified, however, the amount of missing geological time can be determined from the composite age model.
In the high resolution dataset used in Taranaki, CONOP-inferred unconformities often correlate with changes in wireline log pattern at the resolution of sample spacing. Not all CONOP-inferred unconformities are expressed on the wireline logs for any particular well, however. This may indicate that these unconformities are missing locally or, alternatively, that they have no obvious lithological or petrophysical expression. For example, in the well Maui-4 (left) there is good correlation between the pre-CONOP unconformities P20, P50 and P60, and log breaks – the occasional slight offset between the two reflects the fact that CONOP sample horizons are discrete. The identification of these three unconformities was primarily based on seismic interpretation and classic biostratigraphy and they correlate well with lithology and petrophysical property changes in the well. Based on CONOP results, we can now quantify the absolute age and the time missing in each event (e.g., P60, age 35.36-21.26 Ma, 14.1 m.y. missing). Also, based on CONOP, several new unconformities have been revealed (i.e., at 2,713m, 2,465m and 2,158m) and their durations and log expressions identified. The lack of log-expression of a CONOP-inferred unconformity at 2,713m may indicate that it is associated with a fault or, alternatively, that this unconformity is locally missing in Maui-4. A comparison between ages based on traditional biostratigraphic analysis and the CONOP ages shows generally good agreement and greatly increased age resolution of the CONOP results. For example, in Maui-4 the Cretaceous-Tertiary boundary (65.5 Ma) is bracketed by a CONOP unconformity with age-span of 58.29–66.62 million years.
It is important to note that CONOP does not set out to duplicate existing traditional age assignments but, instead, aims to create an alternative, reproducible correlation model that integrates seismic and lithological observations with observations from traditional biostratigraphy. Only future stratigraphic work, crucially incorporating CONOP, will be able to resolve apparent contradictions and pave the way for a unified age model.
The authors acknowledge the work of Hugh Morgans, Ian Raine, and Poul Schiøler of GNS Science.
GNS Science is New Zealand’s premier provider of natural resources research and consultancy services. Open-file data and reports are available from the Petroleum Basin Explorer website: http://data.gns.cri.nz/pbe.
References:
1. Developed by Professor Peter Sadler at the University of California.
2. Cooper, R.A., Crampton, J.A., Raine, J.I., Gradstein, F.M., Morgans, H.E.G., Sadler, P.M., Strong, C.P., Waghorn, D. and Wilson, G.J. (2000). Quantitative biostratigraphy of the Taranaki Basin, New Zealand: A deterministic and probabilistic approach. AAPG Bulletin V. 85 No 8. 1469-1498.
