Uncertainty, Subjectivity and Bias
Andrew Curtis and Matthew Walker
University of Edinburgh, U.K.
Lack of detailed knowledge about heterogeneities in subsurface carbonates renders the injection of expert-derived information important, yet uncertainties are seldom analyzed or quantified. We discuss the sources of expert uncertainty, and show how expert knowledge may be elicited so as to encapsulate uncertainties and reveal natural human bias.
Compared to many other areas of industrial endeavour, geologically-related industrial problems are often characterised by large uncertainties. Success rates of exploration wells intersecting hydrocarbon reserves are generally lower than 60% in the USA (data: PetroStrategies, Inc.) showing that uncertainties in prospect evaluation are significant. Likewise, uncertainties in reservoir characterisation for industrial storage of CO2 have been shown to be an order of magnitude higher than are acceptable in conventional business practice for the power companies who wish to store the CO2 (Polson et al. 2011, 2012), and because these uncertainties stem from the lack of suitable physics to interrogate the Earth’s subsurface at resolutions comparable to the reservoir heterogeneity, all methods applied to geological problems must account for intrinsic uncertainty if they are to provide robust results.
Assessing the prospectivity of carbonate reservoirs is particular challenging due in large part to the large range of scales of variability (<1 mm to ~1 km) in pore space characteristics that contribute to flow properties of single or multi-phase fluids. Each of the individual geophysical, geological, well testing, well log and core analysis methods used to assess such heterogeneity is only sensitive to a relatively small range of scales, and homogenising or even calibrating information provided across different scale ranges is extremely difficult. Between wells the range of methods available to interrogate reservoir properties directly reduces to geophysical remote sensing, well testing, and (post-reservoir development) production history matching techniques, all of which provide information only (significantly) above length scales of ~1 m. This results in an information scale gap and huge uncertainties in properties across the majority of any subsurface carbonate reservoir. The only existing way to inject information into this gap is to use Geological Prior Information and interpretation.
AAPG Search and Discovery Article #120034©2012 AAPG Hedberg Conference Fundamental Controls on Flow in Carbonates, Saint-Cyr Sur Mer, Provence, France, July 8-13, 2012