Computer Science – Performance
Scientific paper
Sep 2010
adsabs.harvard.edu/cgi-bin/nph-data_query?bibcode=2010amos.confe..38w&link_type=abstract
Proceedings of the Advanced Maui Optical and Space Surveillance Technologies Conference, held in Wailea, Maui, Hawaii, September
Computer Science
Performance
Scientific paper
The planetary geomagnetic Kp index (3-hour average recorded every 3 hours) exhibits a high degree of correlation from one value to the next. In fact, a simple persistence model that forecasts the next 3-hr value as being equal to the current value shows a linear correlation coefficient of r = 0.797 and a root-mean square error of RMSE = 0.918 from actual values as calculated using historic Kp data from solar cycles 17 through 23. This simple persistence model can be used as a baseline for comparison to other forecast models and, when interpolation effects are taken into account, provides forecasts that are better correlated and have a smaller RMSE to the actual data than most existing neural network methods that use sentinel solar wind and interplanetary magnetic field data. A new forecast method based on the unscented Kalman filter (UKF) is developed to generate forecasts of Kp using previous values of this index to fully exploit persistence and sentinel solar wind interplanetary magnetic field data to provide a geomagnetic storm trigger. The resulting forecast model performs better than all existing Kp forecast models. Model performance is measured by calculating the linear correlation coefficient and the RMSE between the forecast value and the actual value. A new skill score that assesses how well the model predicts the onset of a geomagnetic storm is also introduced. The UKF-based model offers the opportunity for further improvement by adding new inputs and refining the state and measurement functions in the filter and can be used to forecast other geomagnetic indices as well.
Jah M.
Scro K.
Wetterer Charles
No associations
LandOfFree
Kp Forecast Model Using Unscented Kalman Filtering does not yet have a rating. At this time, there are no reviews or comments for this scientific paper.
If you have personal experience with Kp Forecast Model Using Unscented Kalman Filtering, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Kp Forecast Model Using Unscented Kalman Filtering will most certainly appreciate the feedback.
Profile ID: LFWR-SCP-O-1555882