Comparison of the original and reanalyzed Chilton data from December 1993

A comparison of the two sets of December 1993 data illustrates the types of error that ADEP can introduce. The data that was reanalysed after several months experience with the ADEP package differs from the initial analysed data in several significant ways. For most parameters, there is approximately four percent of scaled values that are no longer considered valid. Experience with ADEP has enabled the scaler to over rule certain values that would previously have been accepted. This gives some indication of the percentage of a parameter's values that are incorrectly attributed by ADEP. When the common data are plotted against each other, the newly scaled parameters such as foF2, fxI, h'f and h'f2 are on the whole very similar to the original scaled values, with only the occasional rogue point. For some of the other parameters however, there are significant differences between the scaled parameters.

• M3000(F2) values show a significant spread with more values being revised upwards rather than downwards. This can be attributed to the sensitive nature of the M3000(F2) value to the shape of the virtual height profile, as detailed above. It is probable that a large proportion of the difference between Slough and Chilton values of M3000(F2) is a direct result of this sensitivity.
• With fmin, the biggest change between the original and revised scaled values, is that a number of erroneously low fmin values have been increased. This leads to a distribution of points that are more biased towards higher Chilton values.
• The fact that ADEP sets a minimum value of foE at all times where it cannot be measured means that there is still 100 percent common data for this parameter. The revised points are evenly distributed about the expected line. An interesting feature to notice is that the lower values (up to 1.4 Mhz) are in perfect agreement. This is probably a result of ADEP's predictive capabilities too.
• Most of the foEs values are in very good agreement, but there is a larger proportion of points (6%) that have subsequently been discarded. Of the common points there are five revised points that differ from the original values by more than one Mhz. These are most probably rogue foEs values that slipped through the checking process on the previous analysis. Out of a total of around 650 ionograms, this represents a tiny fraction of the data (0.8 per cent).
• The very small percentage of common foF1 data was reduced even further after reanalysis. 3.64 percent of the data from the initial analysis was discarded, while only 0.55 percent of the reanalysed data had newly attributed foF1 values. The three common points agree well.
• The most significant change to any one parameter's values was with h'E. Although the common data was over 95 percent, most of the new values were significantly greater than before. This is a consequence of ADEP scaling. In order to estimate all the other parameters, a value of foE must be chosen. If blanketing occurs, this value is set to any remaining E-region echoes. As these are around the critical frequency, the value of h'E that they give is anomalously large. Similarly, if blanketing is occurring, any E trace that is scaled has an h'E that is affected by the underlying ionisation. A significant proportion of h'E points that had previously been given values, were set to 90 km in the new analysis (the default value of ADEP), indicating that these values had been discarded. Values of h'f had a similar distribution ot foEs, in that there was a very good agreement between the initial and reanalysed points, with a few outliers.