Application of Singular Spectrum Analysis to Solar 10.7cm Radio Flux 27-day Forecast

In this paper, singular spectrum analysis (SSA) method was applied to 27-day solar 10.7cm radio flux forecast for the first time. In order to focus on the F_10.7 prediction during the period of the low level solar activity, the test prediction range was chosen from April 30 to May 30, 2004 within the cycle 23 descendent phase. To ensure there were enough data to construct the sample series, both real time data and partial data in the cycle 20 descendent periods following the rule of similar cycle method were chosen, in this way sampling number increased and at the same time, effects of high solar activity are kept away so to get more precise F_10.7 forecast during the low solar activity. The forecasted results of these 31 days show that they mainly coincided with the general tendency of F_10.7 vaviations. The averaged relative error of each 27-day forecast was about 10%. The total average relative error of the 31 days was 10.5%. The minimum average relative error was 4.8% and the maximum was 16.7% during the test forecast period. Comparing with American Air Force (AAF) forecast in the same periods, the total average relative error of SSA is less than AAF, the former was 11.3% and the latter was 14.6%. Furthermore, each average relative error of SSA was less than AAF except two days. But to the short-term forecast within one day to 12 days, the result of AAF was superior to SSA.