THE MONTE CARLO MYTH, OR WHY A RESERVOIR ISN’T A ROULETTE WHEEL

Investigations into reservoir uncertainty are becoming more commonplace, not just into reserves or static uncertainty but also into the behavioural uncertainty associated with the flow characteristics of the reservoir. Unfortunately many of these investigations rely on two assumptions; 1) that the input property distributions can be defined by normal distributions and 2) that Monte Carlo methods are applicable for the analysis of multivariate relationships (particularly those of spatial properties). These two assumptions have long been abandoned in other areas of uncertainty modeling and risk assessment, most noticeably in financial engineering. This paper will discuss the relevance of these practices to the evaluation of reservoir behaviour. If they show any validity in the petroleum industry it is important to understand under which conditions these assumptions can continue to be used. The paper goes on to identify those areas where conventional understanding of random behaviour is flawed and indicates courses of action to improve the effectiveness of uncertainty modeling and risk analysis of complex depositional systems. Good models, great models? Somewhere during the 1990s the process of reservoir characterisation was hijacked and veere