Debiasing Probabilistic Oil Production Forecasts on the Norwegian Continental Shelf

Future production forecasts are the final output of the uncertainty quantification workflows in the oil and gas industry and are essential for decision making. Exploration and production companies have every interest in being unbiased in their forecasts and getting uncertainty estimates right, since it is an intrinsic part of making oil field development profitable. Here, a set of forecasts submitted to the Norwegian Petroleum Directorate at the time of the investment decision for fields with production start between 1995 and 2019 (72 fields), is analyzed. To focus on subsurface uncertainty quantification, the forecast time series is aligned with actual production start in the analysis. The presence of persistent optimism and overprecision bias is shown. A simple and robust calibration model is proposed, and the value of using the forecast triplet is shown to be marginal at best relative to only using the mean forecast. Our interpretation is that these findings lend support to results in behavioral economics - access to more insider details does not necessarily translate into better forecasts and decisions. It is thus necessary to calibrate such forecasts for systematic bias