John H. Schuenemeyer

Uncertainty Estimation for Resource Assessment—An Application to Coal

The U.S. Geological Survey is conducting a national assessment of coal resources. As part of that assessment, a geostatistical procedure has been developed to estimate the uncertainty of coal resources for the historical categories of geological assurance: measured, indicated, inferred, and hypothetical coal. Data consist of spatially clustered coal thickness measurements from coal beds and/or zones that cover, in some cases, several thousand square kilometers. Our procedure involved trend removal, an examination of spatial correlation, computation of a sample semivariogram, and fitting a semivariogram model. This model provided standard deviations for the uncertainty estimates. The number of sample points (drill holes) in each historical category also was estimated. Measurement error in the thickness of the coal bed/zone was obtained from the fitted model or supplied exogenously. From this information approximate estimates of uncertainty on the historical categories were computed. We illustrate the methodology using drill hole data from the Harmon coal bed located in southwestern North Dakota. The methodology will be applied to approximately 50 coal data sets.

A Framework for Quantitative Assessment of Impacts Related to Energy and Mineral Resource Development

Natural resource planning at all scales demands methods for assessing the impacts of resource development and use, and in particular it requires standardized methods that yield robust and unbiased results. Building from existing probabilistic methods for assessing the volumes of energy and mineral resources, we provide an algorithm for consistent, reproducible, quantitative assessment of resource development impacts. The approach combines probabilistic input data with Monte Carlo statistical methods to determine probabilistic outputs that convey the uncertainties inherent in the data. For example, one can utilize our algorithm to combine data from a natural gas resource assessment with maps of sage grouse leks and piñon-juniper woodlands in the same area to estimate possible future habitat impacts due to possible future gas development. As another example: one could combine geochemical data and maps of lynx habitat with data from a mineral deposit assessment in the same area to determine possible future mining impacts on water resources and lynx habitat. The approach can be applied to a broad range of positive and negative resource development impacts, such as water quantity or quality, economic benefits, or air quality, limited only by the availability of necessary input data and quantified relationships among geologic resources, development alternatives, and impacts. The framework enables quantitative evaluation of the trade-offs inherent in resource management decision-making, including cumulative impacts, to address societal concerns and policy aspects of resource development.

A framework for expert judgment to assess oil and gas resources

In frontier areas, where well data are sparse, many organizations have used expert judgment to estimate undiscovered resources. In this process, several important issues arise. How should the knowledge be elicited? At what level of aggregation (geologic process model, play, petroleum system, country, etc.) should the assessment be performed? How and at what stage of the assessment process should feedback be given to assessors? Is independent replication of estimates possible? How are issues of dependency treated? When and how should uncertainty be specified? The context for this presentation will be the methodology used in the US Geological Surveys 1998 1002-Arctic National Wildlife Refuge assessment of oil and gas resources.

Oil- and Gas-Resource Assessment in Certain South American Basins An Application of ARDS (Ver. 5.0) to Complex Exploration and Discovery Histories

The modified Arps-Roberts Discovery Process Modeling System [ARDS (Ver. 4.01)] has recently been upgraded [ARDS (Ver. 5.0)] and applied to a wide variety of field discovery and wildcat drilling data with differing characteristics. ARDS is designed to forecast the number and sizes of undiscovered fields in an exploration play or basin by using historical drilling and discovery data. Fields used as input may be grown or ungrown. Two models for field growth—one offshore and the other onshore—have been implemented (Schuenemeyer and Drew, 1996). Uncertainty attributable to field growth is estimated via simulation. This upgrade of ARDS has been designed to handle situations when the data cannot be partitioned into homogeneous regions, but where estimation of the number of remaining oil and gas fields is still meaningful. In this upgrade of ARDS, many restrictions, which include those on the number of fields and wildcat wells required to forecast the size distribution of the oil and gas fields that remain to be discovered in an exploration play, a basin, or other target area, have been removed. In addition, flexibility has been gained by reforming the criteria for convergence of the model. In all, 32 basins and subbasins in South America were examined, 18 of which had sufficient data to be amenable to forecasting the field-size distribution of undiscovered oil and gas resources directly by using the Petroconsultants Inc. (1993) field discovery and wildcat drilling data. Overall, ARDS (Ver. 5.0) performed well in estimating the field-size distribution of undiscovered oil and gas resources in the 18 basins and subbasins. The aggregate volume of undiscovered petroleum resources was characterized by using histograms of the distribution of resources and the following five statistics: the mean, the 80% trimmed mean, and the 10,50 (median), and 90 quantiles. More than 38 billion barrels of oil equivalent (BOE) in fields that contain more than one million BOE individually were forecast as remaining to be discovered. The largest basin, the Campos (Brazil), is forecast to contain nearly 10 billion BOE undiscovered resources. The East Venezuela Basin (excluding the Furrial Trend) is forecast to contain about 8 billion BOE; the Austral-Magallanes Basin (Argentina and Chile), about 7 billion BOE; and the Napo (Colombia and Ecuador) and the Neuquen (Argentina) Basins, between 3 billion and 4 billion BOE. A subset of these basins that illustrate the increased flexibility of ARDS are discussed.

Uncovering Influences on the Form of Oil and Gas Field Size Distributions

A detailed understanding of the processes that led to empirical oil and gas field size distributions, especially the dynamic character of the discovery process, is needed to improve the quality of forecasts of oil and gas resources. An empirical distribution results from a complex interaction of economic, technical, and social factors with geology in the form of a distribution of deposits. These factors may cause an empirical distribution to mutate nonrandomly through time. Changes in the price of oil, the cost of exploration and development, technology, and access to prospects influence the discovery process. Failure to recognize and account for them in the modeling process can result in serious bias in estimates of the number and volume of future discoveries. In addition, the broad range of some forecasts for a given region may be explained by differences in perspective of those involved in the process. Geologists who understand the basic processes and collect the data may be scientific determinists. Statisticians who model and analyze the data are trained to think in terms of random variables and stochastic processes.

Application of the modified Arps-Roberts discovery proces model to the 1995 U.S. National Oil and Gas Assessment

ARDS (version 4.01), a modified version of the Arps-Roberts discovery process model, was used to forecast the remaining oil and gas resources in more than 50 provinces, super-exploration plays, and individual plays in the onshore and offshore United States for the 1995 National Oil and Gas Assessment. The size distribution of oil and gas fields was estimated for the underlying distribution of fields; the size distribution for the remaining fields was calculated to be the difference between this distribution and that of discovered fields. The guidelines that govern the 1995 National Assessment require the underlying size distribution of fields to be estimated by using only data from two standard commercial data files (the NRG Associates field file and the Petroleum Information Inc. well file). However, a variety of situations required further modification of the discovery process modeling system; for example, multiple exploration plays that occurred nearly simultaneously and also displaced each other in time, and the phenomenon of field growth introduced a large bias in the forecasts produced by the discovery process models for some provinces.

The space-time structure of oil and gas field growth in a complex depositional system

Shortly after the discovery of an oil and gas field, an initial estimate is usually made of the ultimate recovery of the field. With the passage of time, this initial estimate is almost always revised upward. The phenomenon of the growth of the expected ultimate recovery of a field, which is known as “field growth,” is important to resource assessment analysts for several reasons. First, field growth is the source of a large part of future additions to the inventory of proved reserves of crude oil and natural gas in most petroliferous areas of the world. Second, field growth introduces a large negative bias in the forecast of the future rates of discovery of oil and gas fields made by discovery process models. In this study, the growth in estimated ultimate recovery of oil and gas in fields made up of sandstone reservoirs formed in a complex depositional environment (Frio strand plain exploration play) is examined. The results presented here show how the growth of oil and gas fields is tied directly to the architectural element of the shoreline processes and tectonics that caused the deposition of the individual sand bodies hosting the producible hydrocarbon.

Characteristics of water-well yields in part of the blue ridge geologic Province in Loudoun County, Virginia

Loudoun County, Virginia, which is located about 50 km to the west of Washington, DC, was the site of intensive suburban development during the 1980s and 1990s. In the western half of the county, the source of water for domestic use has been from wells drilled into the fractured crystalline bedrock of the Blue Ridge Geologic Province. A comprehensive digital database that contains information on initial yield, location, depth, elevation, and other data for 3651 wells drilled in this 825.5-km2 area was combined with a digital geologic map to form the basis for a study of geologic and temporal controls on water-well yields. Statistical modeling procedures were used to determine that mean yields for the wells were significantly different as a function of structural setting, genetic rock type, and geologic map unit. The Bonferroni procedure then was used to determine which paired comparisons contributed to these significant differences. The data were divided into 15 temporal drilling increments to determine if the time-dependent trends that exist for the Loudoun County data are similar to those discovered in a previous study of water-well yields in the Pinardville 7.5-min quadrangle, New Hampshire. In both regions, trends, which include increasing proportions of very low yield wells and increasing well depths through time, and the counterintuitive result of increasing mean well yields through time, were similar. In addition, a yield-to-depth curve similar tothat discovered in the Pinardville quadrangle was recognized in this study. Thus, the temporal model with a feed-forward-loop mechanism to explain the temporal trends in well characteristics proposed for the New Hampshire study appears to apply to western Loudoun County.

Predicting Molybdenum Deposit Growth

In the study of molybdenum deposits and most other minerals deposits, including copper, lead and zinc, there is speculation that most undiscovered ore results from an increase (or “growth”) in the estimated size of a known deposit due to factors such as exploitation and advances in mining and exploration technology, rather than in discovering wholly new deposits. The purpose of this study is to construct a nonlinear model to estimate deposit “growth” for known deposits as a function of cutoff grade. The model selected for this data set was a truncated normal cumulative distribution function. Because the cutoff grade is commonly unknown, a model to estimate cutoff grade conditioned upon the deposit grade was constructed using data from 34 deposits with reported data on molybdenum grade, cutoff grade, and tonnage. Finally, an example is presented.