Estimation of Uncertainty In Coal Resources

Official estimates of United States coal resources published during the past 15 years vary from less than 1.5 to 3.5 trillion metric tons (1.7 to 3.9 trillion short tons). These differences imply that a high degree of uncertainty exists in resource assessment. This report identifies sources of uncertainty in coal resource estimation. Our report focuses on the comparison of variability in coal resource estimates in areas of different ancient depositional environments. The Texas Gulf Coast Basin was chosen for this study because it exhibits a full range of ancient depositional environments: (1) upper alluvial plain, (2) lower alluvial/upper delta plain, (3) delta plain, and (4) strandplain/lagoonal. Four lignite deposits, each representing one of these depositional environments, were evaluated. Important sources of uncertainty in resource estimation include variability of seam thickness, areal distribution, and the number of seams. To test the degree of uncertainty caused by variations in seam thickness, the numbers of boreholes considered in each lignite deposit are reduced and resources calculated for each reduction in data. Various techniques of resource calculation (manual, computer, and geostatistical) are used to investigate the uncertainties associated with each method. Classical statistics is the method used to determine the number of boreholes required to obtain resource estimates of individual seams within a given confidence interval under specified conditions; geostatistical methods (variograms and kriging) are used to measure variability in resource estimates. Classical statistical methods show that the minimum number of evenly distributed boreholes required to characterize resources of a lignite seam to within a precision of 20 percent is substantially less than might be expected intuitively, and depends on the coefficient of variation of seam thickness. Geostatistical methods indicate that a substantial further reduction in the minimum number of boreholes is possible when a spatial dependency structure can be established by means of a variogram. Resource figures for seams calculated by manual, computer, and geostatistical methods at various levels of data density are well within those predicted by classical statistical theory. These studies demonstrate that the maximum seam thickness variation occurs at the margins of lignite seams and that variations in thickness of individual lignite seams are not a major source of uncertainty in resource estimation, given the level of data usually available. However, determination of the areal extent and seam boundaries of coal beds is a major source of uncertainty. Data availability for regional-scale resource analysis nullifies seam-by-seam (deposit) methodologies. Our regional test area was the Wilcox Group outcrop in east-central Texas. Geostatistics did not yield a dependency structure for the entire area, therefore alternate methods were used: (1) equal weighting over the entire area: (2) equal weighting within grid cells; and (3) equal weighting within internally homogeneous blocks chosen using statistical or geologic parameters. Our methodology was successfully transferred to the Tongue River Member, Wyoming, and the Allegheny Formation, Ohio. Tonnages calculated for Wyoming and Ohio exceeded official estimates because we included deep-basin, thick continuous coals. Depositional models were used to calculate resources for the entire Gulf Coast. Calculated resources indicate the magnitude of total resources but do not quantitatively measure the associated uncertainty.