Institute of Terrestrial Magnetism, Ionosphere and Radio Wave Propagation, Russian Academy of Sciences, Troitsk, Moscow Region, Russia

Monthly median empirical models IRI-95 and NeUoG were compared with incoherent scatter EISCAT and Millstone Hill observations as well as with El Arenosillo Digisonde N e (h) bottomside profiles. A comparison was made for various seasons, levels of solar activity, daytime and night-time hours. The results on the topside comparison: 1) the IRI-95 model systematically and strongly overestimates the Ne (h) effective scale height both for daytime and night-time periods especially during maximum and middle solar activity both at EISCAT and Millstone Hill; 2) the NeUoG model on the contrary systematically underestimates the scale height at all levels of solar activity. But the NeUoG model provides much better overall agreement with SD being less by a factor of 1.5-1.7 in comparison with the IRI-95 model results. The results on the bottom-side comparison: 1) the IRI-95 accuracy is different for daytime and night-time hours, being much worse for the night-time; 2) the NeUoG model similar to IRI-95 demonstrates much worse accuracy for the night-time hours; 3) the NeUoG model demonstrates no advantages over the IRI-95 model in the bottomside N e (h) description. A new simple TopN e model for the N e (h) topside distribution based on the EISCAT and Millstone Hill observations is proposed. The model is supposed to be normalized by the observed hmF 2 and NmF 2 values and is valid below a 600 km height. The TopN e model provides good approximation accuracy over EISCAT and Millstone Hill observations. A comparison with the independent Intercosmos-19 topside sounder observations is given.

Ionosphere monitoring implies: observations, prediction and mapping of ionospheric parameters. A case with one available (El Arenosillo) ionosonde is considered. Some statistical methods for f0F2 short-term (1-24 h in advance) prediction are compared. The analysis of multi-dimensional regression for Df0F2 (relative deviation from running median) with Ap, F10.7 and previous Df0F2 observations has shown that inclusion of additional terms with Ap and F10.7 improves the prediction accuracy for lead time more than 15 h. For lead time 1-6 h a linear regression with earlier observed Df0F2 provides the f0F2 forecast with Relative Mean Deviation (RMD) 6-11%. This is acceptable from a practical point of view. A 24-h forecast can be done with RMD 10-11%. Multi-regressional methods provide better prediction accuracy than the usual 10-day running median or quasi-inertial method based on such median. Hourly f0F2 values may be used to calculate the effective index R12eff used as input to the ITU-R monthly median model. This allows the ITU-R model to "breathe" following hour-to-hour f0F2 variations. Then standard surfering methods may be applied for f0F2 mapping over the whole area. The f0F2 mapping accuracy based on the hourly R12eff index is shown to be 9-11% depending on solar activity level.

Seasonal variations of hmE and f0F2 are analyzed using El Arenosillo digisonde observations during solar minimum (1995-1996). Unlike some widely used empirical models daytime hmE show seasonal variations with winter hmE being higher than summer ones and seasonal differences increase with solar zenith angle. Model calculations enable us to reproduce the observed hmE seasonal variations but the calculated daytime f0E values are too low if conventional EUV fluxes and dissociative recombination rate constants are used. A reduction of a (NO+ ) by taking into account Te > Tn in the E-region as it follows from probe measurements seems to be a plausible solution. The E-region ion composition corresponding to rocket observations may be obtained in model calculations using an appropriate [NO] height distribution. Calculated summer concentrations of [NO] are by a factor of 3-4 larger than winter ones at the hmE-heights.

Examples are presented of the ray-tracing synthesis of multifrequency Oblique Sounding (OS) data on the Dourbes (Belgium) í Roquetes (Spain) path using electron density height profiles derived from Vertical Sounding (VS) measurements at both terminals. Comparison with the measured OS ionograms provides a means of assessing the accuracy of the VS true-height procedure POLAN. Particular attention was paid to a consideration of the E-F1 valley, that as concluded is often less deep than currently supposed, when derived using both ordinary and extraordinary components of the VS ionograms. Also, it was found that the peak of the F1 -layer should be expressed more distinctly (sometimes with a small valley between the F1 and F2 layers) though the corresponding VS ionograms may have no discontinuity in the region.