Verification and calibration of volume forecasts in petroleum exploration: insights from the Norwegian Continental Shelf

Abstract

Exploration decisions rely on pre-drill volume estimates that shape drilling choices, portfolio ranking, and economic appraisal. This study evaluates the reliability and accuracy of such forecasts and tests whether simple calibration can improve them. Using prospect-level forecasts from 1998–2022 on the Norwegian Continental Shelf, we compare pre- and post-drill volume distributions and assess forecast quality using complementary diagnostics of reliability, accuracy, and uncertainty. Results reveal systematic optimism and miscalibration, as realized volumes fall too often in the lower tail of predicted distributions and uncertainty bands are generally too narrow, with little improvement over time. We then develop an operational calibration framework consisting of (i) a reference class benchmark based on historical analogs and (ii) simple distribution-level adjustments to central tendency and quantiles. Reference class adjustment provides the most stable improvements while maintaining forecast sharpness. In contrast, methods that enforce probabilistic reliability mainly widen uncertainty, highlighting a trade-off between reliability and focus. Overall, the outside view, particularly the reference class benchmark, offers a simple, low-cost improvement to exploration workflows and supports routine updating.