The ability to obtain this figure from the ionosonde data appears to be extremely variable. In December, the percentage of common data was 72.9%, followed by a total in January of only 50%. Indeed, in April, the amount of common data dropped to only 41.5%. Such a situation is clearly worrying, but the reason for the lack of consistency is due to the superiority of the DPS at Chilton in comparison with the the old KEL at Slough. This can be seen by looking at the percentage occurrence of the parameter at each of the sites. For nearly all the data in this study, it was possible to derive M(3000)F2 from the Chilton site data. This is a significant improvement over the Slough data.
The ratio of values shows a definite variation between December 1993 and April 1994, as can be seen from the plot of gradient vs time. In December 1993, the best fit line through the points has a gradient of 0.91. This decreases steadily to a value of 0.69 by April 1994. This trend can also be seen in the correlation coefficients. These are completely different from the correlations between data sets for the related parameter, foF2, which are never below 97%. Clearly the correlation between the two sites is much less in March and April, despite the fact that the amount of common data seems unrelated to this trend.
For a given month, the scatter of points is quite large (~ 50) and it is interesting to note where this scatter originates. For the Chilton data, M(3000)F2 is scaled automatically by ADEP. This 'first guess' can be fine tuned by a scaler and the package then recalculates M(3000)F2. In contrast, the Slough data is scaled by hand, using a transparent 'slider' (as detailed in the U.R.S.I. handbook, report UAG 23, pp 21 to 23) to determine the value of M(3000)F2. This approach is much more coarse, and this is obvious from the plots of Chilton against Slough for M(3000)F2, where it can be seen that the Slough values fall into distinct bands approximately 0.05 apart.
Since the value calculated is sensitive to the shape of the ionogram trace as the curve approaches the F2 critical frequency, an automated calculation is likely to be much more sensitive to small changes in the gradient of the curve. It has been found that a small adjustment of just one pixel can alter the predicted M3000(F2) by four percent. Great care must be taken when scaling this parameter from ADEP in order to ensure that the accuracy to which M3000(F2) is determined is not spurious.