Prediction of foF2 Disturbances Above Tokyo Using Solar Wind Input to a Neural Network

Abstract:

Neural network has the ability to learn the input-output relation from past data. It is often used to produce empirical prediction models of several space environmental parameters. One operational model (Nakamura, 2008) used K-index input to predict foF2 variations and storms above Tokyo. It was expected that the prediction at the disturbed situation would become more accurate when solar wind parameters are used to the inputs. Recently the availability of solar wind parameters from the Advanced Composition Explorer became longer enough to overlap one solar activity. In this study, solar wind proton velocity and IMF are used to the input to predict the foF2 disturbances above Tokyo (SW input model). The K-index input model (Nakamura, 2008) was also recreated using the same data term as the SW input model. The SW input model tends to predict better the negative disturbances, and it predicted daytime quick variations more accurate than the K-index input model. Statistical comparison of the prediction ability of those models will be discussed, and the contribution of the solar wind input parameters to the foF2 will be tested using an artificial input.