Library

Notes: Abstract A physical model of the coupled thermosphere and ionosphere has been used to determine the accuracy of model predictions of the ionospheric response to geomagnetic activity, and assess our understanding of the physical processes. The physical model is driven by empirical descriptions of the high-latitude electric field and auroral precipitation, as measures of the strength of the magnetospheric sources of energy and momentum to the upper atmosphere. Both sources are keyed to the time-dependent TIROS/NOAA auroral power index. The output of the model is the departure of the ionospheric F region from the normal climatological mean. A 50-day interval towards the end of 1997 has been simulated with the model for two cases. The first simulation uses only the electric fields and auroral forcing from the empirical models, and the second has an additional source of random electric field variability. In both cases, output from the physical model is compared with F-region data from ionosonde stations. Quantitative model/data comparisons have been performed to move beyond the conventional “visual” scientific assessment, in order to determine the value of the predictions for operational use. For this study, the ionosphere at two ionosonde stations has been studied in depth, one each from the northern and southern mid-latitudes. The model clearly captures the seasonal dependence in the ionospheric response to geomagnetic activity at mid-latitude, reproducing the tendency for decreased ion density in the summer hemisphere and increased densities in winter. In contrast to the “visual” success of the model, the detailed quantitative comparisons, which are necessary for space weather applications, are less impressive. The accuracy, or value, of the model has been quantified by evaluating the daily standard deviation, the root-mean-square error, and the correlation coefficient between the data and model predictions. The modeled quiet-time variability, or standard deviation, and the increases during geomagnetic activity, agree well with the data in winter, but is low in summer. The RMS error of the physical model is about the same as the IRI empirical model during quiet times. During the storm events the RMS error of the model improves on IRI, but there are occasionally false-alarms. Using unsmoothed data over the full interval, the correlation coefficients between the model and data are low, between 0.3 and 0.4. Isolating the storm intervals increases the correlation to between 0.43 and 0.56, and by smoothing the data the values increases up to 0.65. The study illustrates the substantial difference between scientific success and a demonstration of value for space weather applications.