The dynamics of the atmospheric meteor region have been studied using the data obtained with - - the Grahamstown (33°16'S, 26<>:30'E) meteor radar between the years 1987 and 1993 inclusive. Harmonic analysis and the maximum entropy method (MEM) , were used for the spectral characterization ofthe wind above Grahamstown. The prevailing wind, tidal (periods 12- and 24- h) and other (periods 8- and 6-h) oscillations were extracted from the data using the guidelines agreed upon by the A TMAP community. Above Grahamstown the zonal and meridional prevailing winds were found to be predominantly eastward and equatorward respectively. Tidal amplitudes are comparable to the magnitude of the prevailing wind vector, with the diurnal tide being stronger than the semidiurnal tide. The phase differences between ,the zonal and meridional components of the semidiurnal and diurnal tides are - 2 hand - 5 h respectively, which is in reasonable agreement with the corresponding expected values of3 hand 6 h. The tidal wind vectors are on average elliptically polarized with anticlockwise rotation. Longitudinal and day-to-day tidal variations were studied. From the longitudinal study, the semidiurnal tide was found to be dominated by migrating modes, while the diurnal tidal behaviour suggests the presence of nonmigrating modes with zonal wavenumber s = 4. Tides were found to be variable from day to day with little apparent correlation between the zonal and meridional components of the respective tides.

The International Reference Ionosphere (IRI) shape parameters B0, B1, and D1 provide a representation of the shape of the F2 layer, the thickness of the F2 layer and the shape of the F1 layer of the ionosphere respectively. The aim of this study was to examine the variability of these parameters using Grahamstown, South Africa (33.3±S, 26.5±E) ionosonde data and determine their predictability by the IRI-2001 model. A further aim of this study was to investigate developing an alternative model for predicting these parameters. These parameters can be determined from electron density profiles that are inverted from ionograms recorded with an ionosonde. Data representing the B0, B1 and D1 parameters, with half hourly or hourly intervals, were scaled and deduced from the digital pulse sounder (DPS) ionosonde for the period April 1996 to December 2006. An analysis of the diurnal, seasonal, and solar variations of the behaviour of these parameters was undertaken for the years 2000, 2004 and 2005 using monthly medians. Comparisons between the observational results and that of the IRI model (IRI 2001 version) indicate that the IRI-2001 model does not accurately represent the diurnal and seasonal variation of the parameters. A preliminary model was thus developed using the technique of Neural Networks (NNs). All available data from the Grahamstown ionosonde from 1996 to 2006 were used in the training of the NNs and the prediction of the variation of the shape parameters. Inputs to the model were the day number, the hour of day, the solar activity and the magnetic index. Comparisons between the preliminary NN model and the IRI-2001 model indicated that the preliminary model was more accurate at the prediction of the parameters than the IRI-2001 model. This analysis showed the need to improve the existing IRI model or develop a new model for the South African region. This thesis describes the results from this feasibility study which show the variability and predictability of the IRI shape parameters.