Sources of uncertainty in ionospheric modeling: The neutral wind

The neutral wind is a critical input parameter for physics‐based ionospheric models, affecting both the height of the F layer and the total electron content. Unfortunately, the currently available models of the thermospheric wind do not seem to represent it very accurately, and this places a serious limitation on the effectiveness of ionospheric modeling and forecasting. We make use of several decades' worth of midlatitude ionosonde observations of the F region peak, in order to compare the effectiveness of several neutral wind models when used as drivers for an ionospheric model. We check the simulation results against the ground truth of the ionosonde observations using the technique of forecast skill scores. We find that with the ionospheric model in use here (the Utah State University Time Dependent Ionospheric Model (TDIM)), a very simple neutral wind pattern outperforms the more complex models. Increases in skill scores as high as 50% are obtained, relative to the reference case of zero wind; also, in some cases, there are similarly large decreases in skill scores. We view this as a sensitivity study, rather than an effort to identify the best wind model in an absolute sense, because any ionospheric model is an assemblage of algorithms, boundary conditions, and drivers that are themselves imperfect. We identify reasons for the large variability in skill scores with respect to season, longitude, and solar cycle level. We close with a brief discussion of other parameters in ionospheric modeling that are similarly uncertain, e.g., a downward electron flux and the Burnside factor.