Reservoir characterization

Production Response in the Denver-Julesburg Basin

The authors used a high-quality digital-log data set to characterize reservoir quality accurately in the Niobrara and Codell Formations in the Denver-Julesberg (DJ) Basin.

Carrizo Oil and Gas

The authors used a high-quality digital-log data set to characterize reservoir quality accurately in the Niobrara and Codell Formations in the Denver-Julesberg (DJ) Basin. A petrophysical work flow was developed, and detailed mapping of the reservoir attributes was completed. The log-derived parameters, along with an aeromagnetic and vitrinite-reflectance data set, provided excellent insight into which geologic parameters could be tied best to well-production response.


In 2013, the authors began to evaluate production response in an area where nearly 50 Niobrara wells were completed by a single operator with a similar completion design for all wells. There was a wide variation in production results after 180 days of production, ranging from 4 to 16 BOE/lateral ft. The amount of proppant pumped per lateral foot changed very little and ranged between 800 and approximately 1,000 lbm/ft. The dramatic change in production response in light of the absence of major completion changes led to the early conclusion that geology is of great importance in the the Niobrara and Codell.

In the early days of DJ production, ­horizontal-well-development operators did not generally make radical changes to completion designs, making it harder to evaluate the effect of these changes. Only since 2014 has a significant change from previous approaches been seen, with a new focus on a reduction in cost per BOE.

Starting in 2009, stage count for mostly short (approximately 4,300-ft) laterals varied between 10 and 20 stages, with average stage intensity of 300 ft/stage. One of the first horizontal wells in the basin started with five stages, after which stage count quickly jumped to 16 to 20 stages in a short lateral. In recent years, stage count has increased significantly, partly because of longer extended-reach laterals and partly because of higher stage intensity. Stage intensity has dropped below 200 ft/stage, with some operators now experimenting with 125 ft/stage. 

Fluid and proppant volumes on a per-lateral-foot basis have not changed as dramatically in the DJ Basin as they have in other major US shale plays. Rate and rate per lateral foot show a similar lack of change over the first few years of DJ ­horizontal-well development; average rates are relatively low, most likely driven by the early-stage limitations of sliding-sleeve completions. Only recently have higher-rate jobs been seen. 

In response to the initial lack of DJ completion changes and the associated apparent lack of effect on production (resulting from production impact being hidden by larger geological changes), the authors developed a petrophysical work flow in an attempt to capture some of these geological parameters and assign them to every horizontal well. This led to calculation of the hydrocarbon pore volume (HPV), a proxy for bulk rock quality, for each of the wells. The conclusion was reached that any statistical model built only on completion parameters will be insufficient and will have to rely on a combination of completion and petrophysical/geological parameters. 

Petrophysical Work Flow

Production and completion parameters are easily gathered and widely available for analysis for all horizontal wells across the DJ Basin. Less widely available, especially on a basin wide scale, are high-quality reservoir-quality parameters. To identify variability in reservoir quality, the authors undertook a project aimed at gathering high-quality information pertaining to the overall bulk rock quality of horizontal targets across the basin.

The bulk of the data for this exercise came from openhole wireline well logs across the DJ Basin. The data set contained 562 wells with high-quality digital-log curves across the basin. This spread of data allowed excellent coverage across the extent of current horizontal-well activity. With all pertinent formation tops picked, a work flow was developed to calculate and map all possible bulk-rock-quality parameters. The parameters include, but are not limited to, thickness, total vertical depth, HPV, water saturation, shale volume, original oil in place, and total organic carbon. Fig. 1 shows four examples of such properties for a select area of the DJ Basin in northeast Weld County. These parameters were chosen as examples to demonstrate how the geologic/reservoir/petrophysical parameters change over a given area and how spatially coincident they may or may not be with production results. The production bubbles on each map represent 365‑day cumulative production in BOE/lateral ft, from the Niobrara Formation only. The authors understand that a limitation in this approach is the exclusion of other critical geological and reservoir information from the data set that may or may not have a significant effect on production.

Fig. 1—Four examples of petrophysical/geological parameters projected on a grid along with 365-day BOE/ft bubbles: (a) water-cut grid inferred from actual well production; (b) HPV grid; (c) average resistivity×net pay (Ω•m-ft); (d) aeromagnetic anomaly. All data are for Niobrara only, except aeromagnetic data, which is not formation-specific.


The end goal of the petrophysical work flow was to assign to each horizontal well a value for each mapped parameter. By incorporating such data, statistical models would be stronger in terms of describing the variance in production, which variables are most important, and the relative influence of those parameters on the production response. To this end, each mapped variable was sampled at a resolution of 1 sq mile (section level) and assigned to any producing horizontal well with a surface location within that section.

Database Maintenance and Limitations

The current DJ database contains completion, petrophysical, and production data for nearly 4,000 horizontal wells. For every well, this database contains 18 production parameters, 37 geological and reservoir-engineering parameters, and 59 completion parameters. The authors recognize that this production data­base is lacking in a few important respects (missing completion parameters, incomplete geological parameters, and production interference), but it also has significant strengths. For one, it is a very quick tool to conduct comparisons and a scoping analysis of trends to obtain an initial understanding of what appears to work in a specific area. It is very easy to filter and compare wells by location, operator, formation, vintage, and a range in completion types.

Brief Multivariate-Analysis Background

The purpose of multiple regression, or more-general multivariate analysis, is to determine the effect of completion and petrophysical (geological) parameters on production. In general, the following steps are taken as part of an analysis:

  • Define objective of study and quality-control data.
  • Perform exploratory data analysis and check for outliers. Apply transformation if necessary. Identify the dependent and independent variables.
  • Run statistical diagnostics.
  • Check residual plots to ensure that regression assumptions are not violated. Also examine the plot of predicted vs. observed response.
  • Assess the candidate-model-prediction performance by running sensitivities.

The goal of a typical multivariate analysis is to understand as much about the reservoir and stimulation as possible and how each variable contributes to the overall success of a well. 
Finally, what the reader needs to understand is that this is a “dumb” analysis. A multivariate-analysis model could produce potential correlations between parameters, but it follows no physical logic to determine these correlations. A potential problem associated with the lack of experimentation for DJ Basin fracturing designs is that the multivariate analysis does not excel at predicting beyond the range of the data set; in other words, it cannot predict what has not been done. The authors recommend that, at the end of the scoping analysis, a cross check be conducted with a “smart” calibrated fracture-growth model tied to a reservoir model.

Simple Well-Cost Model

To determine cost sensitivity to all completion parameters that might come out of the multivariate analysis, the authors broke down the cost of a fracture treatment by the parameters used in the multi­variate analysis. This can be straightforward for some parameters. However, if the multivariate analysis uses lbm/ft, the cost needs to be calculated in USD/lbm/ft by dividing the total proppant cost for a well by the proppant loading per lateral foot. A similar logic applies to fracturing cost as a function of barrels of fluid pumped. This information can simply be stripped from a fracture-treatment-cost proposal.

Sensitivity to a few parameters that sometimes are produced by the multivariate analysis is harder to obtain and requires multiple cost proposals using a fixed proppant volume. 

Ultimately, these fracturing-cost changes are compared relative to the change in well cost, or the numerator in the USD/BOE ratio. The approximate well cost for different operators in the DJ Basin is generally available through investor presentations.

The authors are interested in lowering USD/BOE and in determining the effect of a completion change in terms of how it changes the total well cost and production history.

A main assumption in the cost analysis is that the uplifts observed in production [within the time frame of the multivariate analysis (i.e., 365 producing days in this case)] are also the uplifts in the estimated ultimate recovery (EUR) of the well. Therefore, if a 10% uplift in 365-day cumulative production is observed, it can be assumed that 10% uplift also applies to the EUR of the well. 

When incremental cost is compared with incremental revenue after 365 producing days for Niobrara and Codell wells, it is seen that suggested ­fracture-design changes aiming to increase production by 10% will reduce USD/BOE even on a 365-day production metric and, in most areas, at a net oil price as low as USD 20/bbl.

Incremental revenue gains will far outstrip the incremental costs—spending a few percent of well cost on the completion improvements generally reduces USD/BOE by more than double the well-cost increase.

This article, written by JPT Technology Editor Chris Carpenter, contains highlights of paper SPE 180217, “The Impact of Petrophysical and Completion Parameters on Production in the Denver-Julesberg Basin,” by Fred Miller, Carrizo Oil and Gas; Jon Payne, Eureka Geological Consulting; and Howard Melcher, Jim Reagan, and Leen Weijers, Liberty Oilfield Services, prepared for the 2016 SPE Low Permeability Symposium, Denver, 5–6 May. The paper has not been peer reviewed.