, Available online , doi: 10.11728/cjss2023-0121
Abstract:
The National Center for Space Weather (NCSW) has been providing space weather forecasts for the next 24h, 48h and 72h since July 1, 2004. In this paper, we use the average error, the average absolute error, the skill score, the median error and the interquartile range of error to verify the forecasts of F107 index and Ap index of NCSW from 2005 to 2022. It was found that the F107 index forecasted by NCSW for the next 24h, 48h, and 72h are usually smaller than the observed F107 index; the Ap index for the next 24h is usually higher than the observed Ap index, while Ap index forecasted for the next 48h and 72h are usually lower than the observed Ap index. The higher the level of solar activity, the greater the forecast error of the F107 index is. However, the maximum forecast error of the Ap index occurs in the declining period of solar activity. In addition, we compared the forecasts of NCSW with the simple numerical models such as the persistence model, 14 days recurrence model, 14 days corrected recurrence model, 27 days recurrence model, and 27 days corrected recurrence model, and found that the forecast performance of NCSW is usually better than that of five simple numerical models. For the F107 index, the forecast performance of NCSW is slightly better than that of the persistence model, and significantly better than that of the four recurrence models. However, when the solar activity level is high, the persistence model's performance of the F107 index for the next 72h is better than that of NCSW. For the Ap index, in most cases, the performance of NCSW is significantly better than that of numerical models. However, when geomagnetic disturbances are severe, the Ap index forecasted by the 27 days recurrence model is more accurate than that forecasted by NCSW.