Bivariate Analysis in Petroleum Geoscience: Crossplots for Formation Evaluation and Completion Design
Bivariate analysis in petroleum geoscience and reservoir engineering is the quantitative examination of the statistical relationship between exactly two variables, using graphical tools (crossplots, scatter plots, probability plots) and numerical metrics (correlation coefficients, regression equations, rank-order tests) to determine whether and how systematically one variable changes in response to changes in the other, and to build predictive relationships between measurable properties and target reservoir parameters that cannot be measured directly. Bivariate analysis occupies the methodological space between univariate statistics (single variable description: mean, standard deviation, histogram) and multivariate analysis (three or more variables simultaneously: principal component analysis, cluster analysis, neural networks), and remains the most widely used quantitative data interpretation tool in everyday WCSB petroleum geoscience practice because it directly addresses the most common problem in formation evaluation: predicting one reservoir property from another that is easier or cheaper to measure. The Pearson correlation coefficient (r, ranging from -1.0 for perfect negative correlation to +1.0 for perfect positive correlation, with 0 indicating no linear relationship) is the standard summary statistic for bivariate relationships when both variables are continuous and approximately normally distributed; for non-normal petrophysical data (which is common, since porosity, permeability, and water saturation are often log-normally distributed in heterogeneous formations), the Spearman rank-order correlation coefficient (rho) provides a more robust correlation measure that does not assume a specific distribution shape. In WCSB petrophysical analysis, the most frequently used bivariate crossplots are: the density-neutron crossplot, which distinguishes lithology and gas-bearing zones by the vertical and horizontal separation of data points from the mineral lines (limestone, dolomite, sandstone) plotted on the Schlumberger chart Z-1 or Halliburton equivalent, identifying gas effect as a characteristic upper-left displacement (low density, high apparent neutron porosity reduction from gas evacuating the pore space); the porosity-permeability crossplot (phi-k crossplot), which quantifies the empirical reservoir quality relationship from core plug measurements and allows estimation of permeability from wireline log porosity using the form k = a times 10^(b times phi) or k = a times phi^b (different formation-specific constants for each WCSB reservoir type); and the acoustic impedance-Vp/Vs crossplot, used in seismic reservoir characterization to distinguish gas sands (low impedance, low Vp/Vs) from brine sands (low impedance, higher Vp/Vs) and tight carbonates (high impedance, variable Vp/Vs) in amplitude-versus-offset (AVO) analysis of Montney and Cardium seismic data. In completion engineering, bivariate analysis of treatment parameters versus production outcomes — fracturing fluid volume (m3) versus estimated ultimate recovery (e3m3), or proppant mass (tonnes) versus first-year production — identifies the completion variables most strongly correlated with production performance across a population of Montney horizontal wells, informing the completion design optimization that drives pad-level capital allocation decisions for multi-well programs where the per-well completion cost can reach CAD 4-8M.
Key Takeaways
- Density-neutron crossplot for WCSB lithology identification: The density-neutron crossplot is the most commonly used bivariate display in WCSB wireline log interpretation, plotting bulk density (rho_b, g/cm3 on the y-axis, decreasing upward to match the log convention) versus apparent neutron porosity (phi_N, %, on the x-axis) for each depth sample in the interval of interest. The three primary WCSB lithologies plot at characteristic locations: sandstone data clusters along the sandstone mineral line (rho_b approximately 2.65 g/cm3, phi_N approximately 0% for tight sand); limestone along the limestone line (rho_b 2.71 g/cm3, phi_N 0%); and dolomite along the dolomite line (rho_b 2.87 g/cm3, phi_N 0%). Gas-bearing sands show a characteristic "gas crossover" where density-derived porosity exceeds neutron-derived porosity, displacing data points to the upper-left of the sandstone mineral line. In Montney Formation interpretation at 2,800-3,200 m depth, the density-neutron crossplot is used to distinguish the productive gas-siltstone facies (data cluster above the dolomite line due to gas effect) from the tight carbonate interbeds (tight dolomite cluster near phi_N 0%, rho_b 2.85-2.90 g/cm3) and confirms the gas-bearing intervals for perforating recommendations in horizontal completions.
- Porosity-permeability crossplot and Montney reservoir quality prediction: The phi-k crossplot from Montney core plug analysis typically shows a power-law relationship: k = 0.0002 times phi^3.8 (for Upper Montney siltstone from Groundbirch core, phi in fraction, k in mD) with R2 = 0.72. The correlation coefficient of 0.85 (square root of R2 for a linear regression on log k versus phi) indicates a moderately strong relationship sufficient to predict order-of-magnitude permeability from wireline log porosity, but with significant scatter that reflects the heterogeneity of the Montney silty facies (natural fractures, dolomite cements, and lamination orientation relative to core plug axis all contribute to scatter above and below the regression line). Wells where core plug permeability is systematically above the phi-k regression line are associated with natural fracture enhancement — a bivariate diagnostic that triggers fracture characterization studies (FMI image log, shear-wave birefringence analysis) before the completion design is finalized for the horizontal well penetrating that section.
- Vp/Vs versus acoustic impedance for seismic fluid discrimination: The bivariate crossplot of Vp/Vs ratio versus acoustic impedance (AI = Vp times rho_b) is the standard seismic reservoir characterization tool for fluid identification in WCSB Montney and Viking amplitude anomalies. Gas sands plot at low AI (1,800-4,500 (m/s)(g/cm3)) and low Vp/Vs (1.5-1.8) in the gas sand quadrant; brine sands at similar AI but higher Vp/Vs (1.8-2.2); tight carbonates at high AI (greater than 7,000 (m/s)(g/cm3)) with variable Vp/Vs. Discrimination between gas and brine sands using this bivariate crossplot requires well control from wells with measured Vp and Vs (full waveform sonic) to calibrate the boundaries between fluid quadrants for the specific formation and depth of interest, which then allows the geophysicist to color-code the seismic amplitude anomaly map using the calibrated Vp/Vs — AI crossplot quadrant assignment — a powerful pre-drill fluid prediction tool for Cardium and Viking light oil play exploration in the WCSB that reduces false-positive amplitude anomaly drilling decisions by approximately 25-40% compared to amplitude-alone analysis.
- Completion parameter bivariate analysis for Montney optimization: After 40 horizontal wells have been completed in a Montney pad program at Sunrise, BC, the completion engineer performs bivariate analysis of all completion variables versus first-year production (IP365). The strongest bivariate correlations (Pearson r greater than 0.70) are found between total proppant mass (tonnes/m of lateral, r = 0.74) and IP365, and between completion fluid volume per stage (m3, r = 0.68) and IP365. Weaker correlations (r = 0.35-0.55) are found between number of perforation clusters per stage and IP365. The bivariate regression for proppant mass versus IP365: IP365 = 2.8 times proppant_per_m + 12, with the equation suggesting an additional 1 tonne/m of proppant adds approximately 2.8 e3m3/year to IP365, worth approximately CAD 10,600/year at AECO CAD 2.80/GJ — providing a clear economic optimization framework for proppant loading decisions across the next pad program in the same formation and area.
- Bivariate analysis limitations: correlation versus causation in WCSB data: A bivariate correlation between two production or geological variables does not establish causation: a high correlation between measured depth (MD) and production rate in a Montney well database may reflect confounding by formation quality improvement with depth (the deeper wells are in a better part of the fairway, not necessarily because depth itself causes higher production). The standard statistical tests for bivariate significance (t-test for Pearson r significance, Spearman rho significance test) assess the probability that the observed correlation arose by chance given the sample size, but do not rule out confounding. WCSB reservoir engineers and completion engineers address confounding by building multivariate regression models that control for the other variables (depth, lateral length, formation thickness, TOC) when testing the bivariate completion parameter-production relationship — using the bivariate crossplot as the starting point for hypothesis generation and the multivariate model as the verification tool before committing to a new completion design philosophy for a pad program.
Density-Neutron Crossplot: Montney Gas Identification
A petrophysicist evaluates the gas saturation and lithology of a Montney horizontal well at Groundbirch using the density-neutron crossplot of LWD data from the 2,800-2,900 m landing zone selection interval. Plotting all depth samples at 0.1 m resolution (1,000 data points): Upper Montney siltstone (2,810-2,850 m) plots with bulk density 2.45-2.60 g/cm3 and apparent neutron porosity 12-18%, clustering above the dolomite mineral line with a clear gas crossover signature (density porosity 14-20% greater than neutron porosity at the same depth). Lower Montney carbonate (2,850-2,870 m) plots tightly at rho_b 2.82-2.88 g/cm3 and phi_N 2-5%, near the dolomite mineral line with no crossover. The bivariate crossplot confirms the landing zone for the horizontal well: the Upper Montney siltstone gas-bearing interval (2,810-2,850 m, approximately 40 m gross thickness) is the target, with the tight Lower Montney carbonate confirming the lower boundary of the reservoir window. The directional well engineer sets the bit walk target at 2,830 m TVD, center of the 40 m gas-bearing siltstone interval, before the horizontal section begins at 2,820 m TVD after the build section. The density-neutron crossplot is included in the completion design package for the perforation interval selection across the 2,400 m lateral section, with the gas crossover pattern used to exclude tight carbonate intervals from the perforation plan.