Firstly, this study re-considers past approaches to inform correlation panels by guiding the lateral tracing of each individual channel element, in the light of information derived from the large architectural knowledge base (FAKTS) that is now available. It is not within the scope of this work to provide a full account of the drawbacks of analogue-based or palaeohydrology-based approaches, and neither is this necessary since these pitfalls have already been discussed in detail by Bridge & Mackey (1993), Bridge & Tye (2000) and Miall (2006). Instead, this work further highlights the inadequacy of approaches based on the correlation of each single channel body by focusing once more on the wide architectural variability that might stem from adopting such methods without checking for the geological realism of the modelled succession. This problem is emphasized by the considerable
scatter observed in the architectural data presented here, which highlights the difficulty of reliably inferring channel-body width from the formative-channel bankfull depth, of inferring formative-channel bankfull depth from the thickness of a channel sandstone body, or of inferring channel-body width directly from its thickness. For example, considering bankfull depths observed in the 7 to 23 m range, FAKTS channel-complex widths cover as much as four orders of magnitude (figure 5.1);
overall the two variables yield a Pearson correlation coefficient of 0.341. The architectural database stores both the inferred/measured bankfull depth of channels and the geometry of lower-scale units (architectural elements) contained within the channel complexes; since architectural-element thickness, in some cases, may relate to formative channel bankfull depth, some architectural elements whose thickness was interpretable as the entirely preserved thickness of the associated in-channel geomorphic element (barform) were therefore considered to estimate bankfull depth (cf. Bhattacharya & Tye 2004). With regard to the relationship between measured or inferred bankfull depths and channel-complex thickness (figure 5.2), FAKTS data do not fit well with the relationship given by Fielding &
Crane (1987) in the form of channel depth = 0.55 sandstone thickness, or with a linear relationship altogether (application of a linear best fit to the FAKTS dataset returns R2 = 0.0656). The FAKTS channel-complex width-to-thickness scatterplot (figure 5.3) displays substantial scatter, even if only real widths are considered, with three to four order of magnitudes in width are possibly associated with any given value of thickness; importantly, the power-regression best fit of all FAKTS channel-complex real-width data shows a significant discrepancy with the most-likely case predicted by Fielding & Crane (1987), especially for channel complexes that are thicker than 8 m.
Consequently, the strict application of quantitative relationships clearly entails significant risk; even a flexible application of analogue information, based on the range of natural variability in architectural characteristics would still potentially lead to many correlation panels that are architecturally very different but equally acceptable, given that they would honour geometrical constraints. Although it is important to recognize that there is value in basing such models on geometrical information, and that it is useful to synthesize such information into empirical relationships, in this study a different set of constraints are used – again derived by outcrop or modern analogues – to better inform or rank well-to-well correlation frameworks of subsurface fluvial architecture.
120
Figure 5.1: scatterplot of channel-complex width against formative-channel bankfull depth based on all suitable data contained in the FAKTS database, including data published by Fielding & Crane (1987), Jordan & Pryor (1992), Fielding et al. (1993), Friend &
Sinha (1993), and Tye (2004). The power-regression curve is plotted as a continuous line, whereas the equation given by Collinson (1978) – included for comparison – is represented as a dashed line.
Figure 5.2: scatterplot of channel-complex thickness against formative-channel bankfull depth or architectural-element thickness based on all suitable data contained in the database. Architectural elements represent lower-scale units contained within channel complexes and that are interpretable as the preserved product of geomorphic units, such as barforms; geomorphic elements whose thickness appears to be completely preserved and which are considered reasonable and useful indicators of channel bankfull depth are depicted as filled data-point markers.
Figure 5.3: scatterplot of channel-complex width against channel-complex thickness;
apparent widths refer to measurements made from exposures that are oblique with respect to the channel-belt-scale flow axis or from situations where palaeoflow was uncertain; real widths refer to the entire body lateral extent along a direction normal to the flow axis; following the terminology by Geehan & Underwood (1993), partial widths refer to measurements of channel complexes for which one lateral termination is not exposed, whereas unlimited widths refer to bodies for which both lateral terminations are not exposed. The curve expressing the “most-likely scenario” of Fielding & Crane (1987) is also plotted, for comparison with a power-regression curve obtained from all FAKTS channel-complexes for which real-width data are available.
Specifically, in the approach taken in this study we do not consider relationships that refer to individual elements that need to be correlated over several wells;
instead we consider relationships that refer to either the sedimentary succession as a whole, or to specific portions thereof. In particular, we provide probabilistic tools that can be employed to check the realism of a given fluvial reservoir/aquifer model, so that the interpretation can be iteratively adjusted to match with a target quantity describing the correlability of channel bodies over a given inter-well distance for an ideal synthetic analogue made of architectural data obtained from several real-world case studies (cf. Colombera et al. in press, Chapter 3). Analogue data on which estimates of target system correlability are based can be customized to fit interpreted palaeo-environmental or system-descriptive parameters (e.g. bankfull discharge, channel pattern), but the use of this approach does not require palaeo-environmental or palaeo-hydrological interpretation, as it potentially only involves the use of relationships describing associated architectural properties of the preserved record (e.g. geometry and proportions as shown in a specific model later). Clearly, the method can be used in conjunction with expressions for estimating the lateral extent of individual bodies; for example, relationships linking
122
channel body thickness with range in width can be flexibly used to inform the lateral extent of any given sandstone body, provided that the width distributions are such that they match the target correlability given by the model presented below. The approach can be used either to guide or validate/evaluate a model in cases where well spacing is fixed; later in this work a set of previously-interpreted correlation panels are used to perform an example quality check.