Abstract

Quantifying uncertainty is essential in forward geological modeling, especially in settings with limited data during early exploration. This study presents two proxy modeling approaches: kriging-based response surface modeling (RSM) and neural network modeling (NNM) to accelerate sensitivity and risk analysis while maintaining reliability. Applied to stratigraphic and basin modeling scenarios, these methods reduce computational time by one to two orders of magnitude compared to full Monte Carlo workflows. Case studies include a carbonate platform in the Middle East and a frontier basin offshore Canada. With robust validation on blind runs, proxy models trained on a small number of simulations (10–100) reproduce important risk indicators and probability maps. While NNM scales better with data, RSM delivers slightly more accuracy. Both techniques enhance well positioning and play ranking, providing usefulness for exploration workflows with uncertainty.