Long‐term ionospheric trends based on ground‐based ionosonde observations at Kokubunji, Japan

<jats:p>Using monthly medians of ionosonde observations taken over a period including more than four solar cycles (1958–2003) at Kokubunji (35.71°N, 139.49°E), Japan and a linear regression model to eliminate solar and geomagnetic effects, we derive the long‐term trends of critical frequencies (<jats:italic>fo</jats:italic>E,<jats:italic>fo</jats:italic>F1, and<jats:italic>fo</jats:italic>F2), the E region height (<jats:italic>h</jats:italic>′E), and the F2 peak height (<jats:italic>hm</jats:italic>F2). The results show increasing trends for<jats:italic>fo</jats:italic>E (+0.002 MHz/yr),<jats:italic>h</jats:italic>′E (+0.189 km/yr), and<jats:italic>fo</jats:italic>F1 (+0.0107 MHz/yr) and decreasing trends for<jats:italic>fo</jats:italic>F2 (−0.0058 MHz/yr at noon and −0.0016 MHz/year at midnight) and<jats:italic>hm</jats:italic>F2 (−0.398 km/yr at noon and −0.505 km/year at midnight). We have also analyzed seasonal and diurnal trend variations. These trends differed for various times of day and months, even with altering signs (except for<jats:italic>hm</jats:italic>F2). An interesting phenomenon was observed for the first time; that is, the morphologies of the overall seasonal trend variations of<jats:italic>fo</jats:italic>E and<jats:italic>fo</jats:italic>F1 were opposite each other to some extent, although the trends remained positive. A similar phenomenon was found for the seasonal trend variation of<jats:italic>hm</jats:italic>F2 between local noon and midnight. In addition, comparison of the results from four regression models indicated that the differences induced by linear and quadratic models are slightly significant for<jats:italic>fo</jats:italic>F2 but not for other parameters. On the basis of the average of<jats:italic>fo</jats:italic>F2 data obtained over 5 hours around noon, we have assessed the performance of each of these regression models. The performance differed depending on the month. The results showed that the quadratic model and the linear model that took saturation into consideration performed better than the linear ones which did not consider saturation. In addition, it was indicated that the effect of geomagnetic activity was not significant in regression models at this station.</jats:p>