Ever since I can remember (let’s say 55 years or so) I’ve been awed, fascinated, inspired, passionate, delighted, infatuated, maniacal, and otherwise stir crazy about snow storms, especially the “big ones”. However, even today with state of the art observational, data assimilation, and modeling systems run routinely on super computers, the “big ones” more often than not are a surprise until the snow begins accumulating and red-faced forecasters and media types start eating humble pie and begin blaming the busts on THE models – whatever they are. WHY?

One of the first things I learned in Fred Sanders’ synoptic meteorology class at MIT (1967) was that forecast skill varied from one day to the next and forecasting is best approached in a probabilistic framework. This was before neither he nor I knew anything about Ed Lorenz’s early work on predictability and the “Butterfly Effect” which ultimately provided the scientific rational of what I’m sure to Fred was intuitive. Additionally, it became evident in that same course when studying Fred’s analytic model of cyclogenesis - accurate prediction of which is essential to forecasting big snow storms - how sensitive the development process was to small changes in the adjustable model parameters (e.g., wave length, phase relationships, vertical stability) and to, as later investigated by myself and Fred, the influence of latent heat release, especially that associated with convection.

Little did I know then how these influences would many years later (late 80’s) inspire me to pursue practical issues of predictability, i.e., the nature and consequence of small changes in initial conditions and physics parameterization in modern data assimilation and forecast systems. This in turn placed me (and, at first, only relatively few “true believers”) at the forefront in development and application of ensemble prediction systems (EPS). Today, it is generally accepted that EPS is “the wave of the future” in NWP – a capability that permits understanding the case dependent uncertainty of forecasts and provides a practical method for providing quantitative estimates of those uncertainties.

In this paper, I shall review the various elements of dealing with forecast uncertainty with focus on grappling with the “big ones”. Finally, I shall reveal the epiphany that, I believe, most dramatically demonstrated and “sold” the concept and operational applications of EPS. That epiphany dawned into my consciousness from seeds inadvertently planted by Fred many years ago and which even today he likely remains unaware.