The Sensitivity to Initial Conditions of Several High-Impact Winter Storm Events of Varying Levels of Predictability

Within the framework of numerical weather prediction (NWP), initial conditions play a pivotal role in the forecast of large-scale, high-impact weather events. Given the large population centers situated along the eastern seaboard of the United States, accurate forecasts of significant winter storms are paramount in protecting lives and property as well as ensuring the effective conduct of commerce in highly-populated regions such as New York City, Washington D.C. and Boston. It is with this in mind that the question is posed, what commonalities exist within the sensitivities to initial conditions of winter storms that are highly predictable and those that are highly unpredictable, and is there any relationship between predictability and sensitivity to initial conditions?

In order to answer this question, case studies of the structure and evolution of the sensitivity to the initial model state of these winter storms is investigated for several cyclones with varying degrees of predictability as described by GEFS ensemble model spread. Sensitivity to initial conditions is computed for a response function representing forecast storm intensity, using the GEOS-5 NWP and adjoint models. By using both the quantified ensemble uncertainty and adjoint-derived sensitivity to initial conditions, it is possible to identify precisely which uncertain features within the initial model ensemble (analysis) state most likely lead to uncertainties in the forecast intensity of the cyclone specifically. Relationships are explored between ensemble-derived uncertainty, adjoint-derived sensitivity to initial conditions, and deterministic forecast error.