Data modeling and assimilation studies with the MU radar

We report initial results of data modeling and assimilation studies for several MU radar experiments. Various inputs to a one-dimensional ionospheric model are adjusted to provide agreement with observation and also to learn the sensitivity of the model to their variations. Certain observations are also used directly in the model to anchor or constrain its behavior. In particular, studies of the electron density from 100 to 500 km altitude in the ionosphere are carried out with the help of a theoretical model of O, NO, O2 and N + 2 densities and MU radar observations of the power, ion-drift and plasma-temperature pro®les. Four typical cases are selected to study quantitatively the e€ects of the (A) perpendicular-north component of the plasma drift (15 December 1986), (B) atmospheric composition (7 October 1986), (C) solar EUV ̄ux (2 August 1989) and (D) upper-boundary O density (5 October 1989) on the model NmF2, hmF2 and Ne pro®le, as well as on the neutral wind calculation from hmF2 and drift data. It is found that the measured vertical ion drift explains quantitatively well the measured hmF2 (particularly at low solar activity) while the model gives a better match with the measured Ne when it uses the hmF2-based wind rather than the measured plasma drift. Di€erent model values of the atmospheric O/N2 ratio and EUV ̄ux and di€erent values of the upper-bound O density may modify not only NmF2 markedly but also hmF2: a lower O/N2 ratio results in higher hmF2; the EUVAC model gives higher hmF2 at high solar activity than does the EUV91 model; with a smaller upper-bound O density, hmF2 is lower by day but little changed by night. We speci®cally note that the meridional wind needed by the model to reproduce the observed hmF2 di€ered according to how well the model reproduced the observed NmF2. The uncertainties in the MSIS86 and EUV model predictions are also discussed. It is found that if the MSIS and EUV91 models are used together, the model gives an NmF2 higher than that measured at high solar activity. Thus the O/N2 ratio needs to be reduced from the MSIS value if EUV91 is used. If EUVAC is used, no large modi®cation is required. At equinox for low solar activity, modeling with either EUV model produces NmF2 values lower than those measured, and so the true O/N2 ratio may be higher than that given by MSIS model. # 1999 Elsevier Science Ltd. All rights reserved.