The mesosphere and thermosphere are predicted to cool by 10 K and 50 K, respectively, for doubled concentrations of mesospheric carbon dioxide and methane. Several papers have been published trying to reveal long-term trends in various parameters which could be due to an ionospheric greenhouse effect. Besides temperature measurements by sounding rockets and lidars, the altitude variation of the ionospheric F2-layer peak as observed by ionosondes seems to be one of the most promising indirect approaches: the layer is suggested to lower down due to shrinking of the underlying cooled atmosphere.

The F2-layer peak, the height of which is denoted by hmF2, is monitored by ionosondes since the 1930s. Since hmF2 is not usually scaled from ionograms, it can be estimated by means of empirical formulae from the maximum usable frequency factor and the critical frequencies of the E and F2 layer. A number of authors have observed a lowering of the F2 layer and speculatively connected their results to greenhouse cooling. However, recently down as well as up trends were found in studies of data of many ionosondes world-wide.

It is very difficult to compare the trends obtained by different authors for the same stations, because almost all authors use an individual way of data processing, i.e. of extracting the trend out of the noisy hmF2 time series. In the present work I will demonstrate how much the obtained trends depend on the choice of the data processing. I will suggest a common method for trend estimation, which allows for taking into account measurement errors: standard statistical significance tests might be insufficient to deal with these data sets.