Richard Heinberg and David Fridley argue that coal reserves may be exhausted within decades (Nature 468, 367–369; 2010), basing much of their analysis on fits of cumulative coal production to logistic functions in the style of M. King Hubbert, who famously predicted peak oil supply. But this method is problematic — for example, fitting the decline in production of LP records to a logistic curve would incorrectly indicate that vinyl is a limited resource.
If scarcity were an important determinant of US coal-production history, prices should have increased. Yet they have stayed around US$34 per tonne for the past 50 years, irrespective of production trends. Alternative explanations could include changes in electricity demand and market structure. There would then be no justification for calculating the limit of coal resources from a logistic graph of production history.
The logistic fits that drive forecasts of coal exhaustion depend on which years are included in the analysis. Logistic fits using data up to 1989, 1999 or 2009 forecast an ultimate coal reserve of 52, 71 or 96 gigatonnes, respectively, and predict that production should have peaked in 1951, 1967 or 1986. In fact, coal production has increased since 1986 — highlighting the weakness of the scarcity-driven Hubbert model in explaining production.
An exponential fit explains as much of the variation in US production data as does a logistic fit. Yet the interpretation of the two models is different: the logistic model predicts the end of coal; the exponential fit predicts an infinite supply. Supply is obviously not infinite, but without a theoretical framework to support the choice of a logistic fit, its prediction may be just as wrong.
The end of easy oil is driving a shift towards carbon-intensive options, such as oil-sands mining or converting coal to liquid fuel. We must rely on policy changes to ensure a less carbon-intensive future, not the end of cheap coal.