Quantifying the sensitivity of seismic facies classification to seismic attribute selection: An explainable machine learning study

During the past two decades, geoscientists have used machine learning to produce a more quantitative reservoir characterization and to discover hidden patterns in their data. However, as the complexity of these models increase, the sensitivity of their results to the choice of the input data becomes more challenging. Measuring how the model uses the input data to perform either a classification or regression task provides an understanding of the data-to-geology relationships which indicates how confident we are in the prediction. To provide such insight, the ML community has developed Local Interpretable Model-agnostic Explanations (LIME), and SHapley Additive exPlanations (SHAP) tools. In this study, we train a random forest architecture using a suite of seismic attributes as input to differentiate between mass transport deposits (MTDs), salt, and conformal siliciclastic sediments in a Gulf of Mexico dataset. We apply SHAP to understand how the model uses the input seismic attributes to identify target seismic facies and examine in what manner variations in the input such as adding band-limited random noise or applying a Kuwahara filter impact the models’ predictions. During our global analysis, we find that the attribute importance is dynamic, and changes based on the quality of the seismic attributes and the seismic facies analyzed. For our data volume and target facies, attributes measuring changes in dip and energy show the largest importance for all cases in our sensitivity analysis. We note that to discriminate between the seismic facies, the ML architecture learns a “set of rules” in multi-attribute space and that overlap between MTDs, salt, and conformal sediments might exist based on the seismic attribute analyzed. Finally, using SHAP at a voxel-scale, we understand why certain areas of interest were misclassified by the algorithm and perform an in-context interpretation to analyze how changes in the geology impact the model’s predictions.

Seismic Attribute-Aided Characterization of Margin Backstepping and Advance along Isolated Carbonate Sequences, Sirt Basin, Libya

Isolated carbonate platforms are common and contain significant hydrocarbon accumulations, particularly in the tectonically complex Sirt Basin in Libya. This study investigates the margin cyclicity of two carbonate stratigraphic sequences developed on an isolated carbonate platform in the NW Sirt Basin using 3-D post-stack seismic volume and wireline log data. The two sequences (sequences 4 and 5) are bounded by unconformity surfaces from the base and top. Seismic attributes show that each sequence displays a cycle of margin backstepping followed by margin advance for several hundred meters. This study concludes that the margin backstepping and advance are mainly influenced by sea-level changes. A rapid sea-level rise caused the backstepping, whereas slow sea-level rise caused the margin advance.

Stratigraphic Study for the Cretaceous-Tertiary period in Qasab-Jawan area in north weastern Iraq by using 2D seismic survey

A seismic study was conducted to re-interpret the Qasab and Jawan Oil fields in northwestern Iraq, south of the city of Mosul, by reprocessing many seismic sections of a number of field surveys by using the Petrel software. Two reflectors, represented by the Hartha formation, deposited during the Cretan age, and the Euphrates formation, formed during the Tertiary age, were delineated to stabilize the structural picture of these fields. The stratigraphic study showed that the Qasab and Jawan fields represent areas of hydrocarbon accumulation. Seismic attribute analysis showed low values of instantaneous frequency in the areas of hydrocarbon accumulation. Instantaneous phase was used to determine the limits of the sequence, the nature of sedimentation, and the type of vanishing, i.e. onlap vs. toplap. Low instantaneous amplitude values were recorded, indicating hydrocarbon reservoirs in the studied area. Various other seismic stratigraphic features were studied , including the distribution mound, flat spot, and channels in the two formations, but they were discontinuous because of the tectonic effects. These activities explain reasonably the distribution of hydrocarbons in the studied area.

Seismic attribute analysis of a fault zone in the Thebe field, NW Shelf, Australia

Seismic data is one of the main ways to characterize faults in the subsurface. Faults are 3D entities and their internal structure play a key role in controlling fluid flow in the subsurface. We aim to characterize a geologically sound fault volume that could be used for subsurface model conditioning. We present an attribute analysis of a normal fault from a high resolution seismic dataset of the Thebe Field, offshore NW Australia. We merge together a series of common attributes for fault characterization: dip, semblance and tensor (DST), and we also introduce a new Total Horizontal Derivative (THD) attribute to define the edges of the fault zone. We apply a robust statistical analysis of the attributes and fault damage definition through the analysis of 2D profiles along interpreted horizons. Using the THD attribute, we interpret a smaller width of the fault zone and a more straightforward definition of the boundaries than from the DST cube. Following the extraction of this fault volume, we define two seismic facies that are correlated to lithologies extracted from our conceptual model. We observe a wider fault zone at larger throws, which corresponds also to syn-rift sequence, hence more complex internal fault damage. Our method provides volumes at adequate scale for reservoir modeling and could therefore be used as a proxy for property conditioning.

Geostatistics assisted by machine learning for reservoir property modeling: A case study in presalt carbonates of Buzios Field, Brazil

An embedded model estimator (EMBER) petrophysical modeling algorithm has been applied to obtain effective porosity and permeability within the presalt carbonate reservoirs of the Barra Velha Formation in Buzios Field, Santos Basin. This advanced methodology was used due to the heterogeneity and complexity of the reservoirs, which makes modeling by conventional geostatistical methodologies difficult. For effective porosity modeling, we chose one facies model, one stratigraphic seismic attribute (acoustic impedance), and one structural seismic attribute (local flatness) as secondary variables. Permeability was modeled by using the best effective porosity simulation result as a secondary variable. Our results demonstrate that average effective porosity and permeability were 0.10 v/v and 440 md, respectively, indicating good reservoir quality throughout the studied area. A vertical trend of high effective porosities and permeabilities for the basal and uppermost reservoir sections was identified in our results, as well as a trend with lower values for these reservoir properties for the intermediate reservoir section. The lower section of the formation presented more continuity, and we infer it to be the best reservoir interval. We observed two horizontal trends for these reservoir properties at the formation top: one of higher values aligned to the north–south direction at the structural highs and another of lower reservoir properties related to isolated structural lows within structural highs. Correlation between modeled results and the blind test ANP-1 well upscaled properties was high, and upscaled well-log property distributions were preserved in the EMBER simulations, proving the predictive capacity of the algorithm. Finally, conditional distributions analysis indicated that the basal section of the Barra Velha Formation presents higher uncertainty for the estimation of effective porosity. Even though this interval is considered to have the best reservoir characteristics, decision making should be done with caution for this section.

The Importance of Seismic Attributes in Complex Area, Case Study from Vodianivske Field, Dnieper-Donets Basin, Ukraine

The Dnieper-Donets Basin (DDB) is the principal producer of hydrocarbons in Ukraine and reserves are found in lower Permian and in Visean-Serpukhovian from Lower Carboniferous. The Vodianivske field is located halfway between Poltava and Kharkiv in east Ukraine with proven reserves at depth of 5-6km. Previous studies based on legacy seismic data show thickness changes of the upper Visean towards the main structure and dim small-scale structures on the block boundary. A recent 3D data reprocessing using 5D interpolation and advanced prestack time migration provides a broad frequency content image and imparts detailed high-resolution geological events. While traditional exploration is focused on gas traps in the Visean and below, current study aims to scan for potential traps in the Serpukhovian and above. In order to reveal thin section features, multiple seismic attributes were tested, and spectral decomposition was found to be a powerful tool that delineated thin sand bodies in river valleys and allowed interpretation of high-resolution small-scale faults and pinch-outs not seen before. Frequency tuning analysis on mapped horizons associated with upper Serpukhovian supported the presence of a large deltaic structure revealing SE-NW thin ∼1km wide sand body and developed set of crossing meanders. Similar approach was applied on legacy data expanding to the east and while seismic quality was limited, it was possible to identify a narrow ∼25km length meander and highlight a fault set. Upon seismic attribute study we were able to identify and map thin units associated with sands that can be considered as future targets in hydrocarbon exploration in the area.