Dole, R. M., 2000: Prospects for drought forecasts in the United States. In Drought: A Global Assessment (Volume I), D. A. Wilhite (Ed.), Routledge, 83-99.
In developing strategies to predict drought, certain fundamental challenges must be recognised. Perhaps the most basic question is, Precisely what is it that we wish to predict? Although there is broad agreement that drought is a prolonged period of anomalous moisture deficiency, identification of the onset, magnitude, and termination of drought varies substantially, depending on the user community or discipline (Wilhite and Glantz 1985). Indeed, even within a particular discipline, identification of drought onset and termination may be quite difficult, leading some to suggest that drought can be identified only retrospectively (Tannehill 1947). In addition, the term 'drought' is applied to phenomena ranging from abnormal dry spells persisting for less than a season to prolonged moisture deficits occurring on decadal or longer periods. This broad range of time scales suggests that an equally broad array of physical mechanisms must be considered in both understanding and predicting droughts.
Further compounding these difficulties is the fundamentally chaotic nature of mid-latitude weather variability, so that small differences in initial states eventually lead to large differences in the evolution of weather patterns. Because the time scales of drought are much longer than estimated limits of a few weeks for deterministic weather predictions (e.g., Lorenz 1983), the ability to predict the precise sequence of events associated with droughts is intrinsically limited. But does this mean that beyond some time period, say a month, all hope for useful predictions is lost?
Emphatically not, if instead of attempting to predict the precise sequence of events, the goal is changed to estimating how given initial and boundary conditions alter the risk of drought - i.e., its event probability, relative to climatological (or current) conditions. This predictability may result from either the presence of certain large-scale components of the atmospheric circulation, or through the effects of lower boundary forcings that evolve on much longer time scales than individual weather systems (e.g., Palmer and Anderson 1994). For operational purposes, the drought prediction problem becomes one of forecasting the probability distribution of some quantitative drought measure or index over a given region and time period. The remainder of this chapter examines the prospects for such predictions in the United States, first by considering factors that produce US droughts and second by discussing possible empirical and modelling approaches to drought predictions.