The application of a hierarchical Bayesian spatiotemporal model for forecasting the SAA trapped particle flux distribution

作者：Suparta Wayan^*; Gusrizal

来源：Journal of Earth System Science, 2014, 123(6): 1287-1294.

• Forecasting
• SAA
• trapped particle
• hierarchical Bayesian
• spatiotemporal

We implement a hierarchical Bayesian spatiotemporal (HBST) model to forecast the daily trapped particle flux distribution over the South Atlantic Anomaly (SAA) region. The National Oceanic and Atmospheric Administration (NOAA)-15 data from 1-30 March 2008 with particle energies as >30 keV (mep0e1) and >300 keV (mep0e3) for electrons and 80-240 keV (mep0p2) and > 6900 keV (mep0p6) for protons were used as the model input to forecast the flux values on 31 March 2008. Data were transformed into logarithmic values and gridded in a 5 degrees x5 degrees longitude and latitude size to fulfill the modelling precondition. A Monte Carlo Markov chain (MCMC) was then performed to solve the HBST Gaussian Process (GP) model by using the Gibbs sampling method. The result for this model was interpolated by a Kriging technique to achieve the whole distribution figure over the SAA region. Statistical results of the root mean square error (RMSE), mean absolute percentage error (MAPE), and bias (BIAS) showed a good indicator of the HBST method. The statistical validation also indicated the high variability of particle flux values in the SAA core area. The visual validation showed a powerful combination of HBST-GP model with Kriging interpolation technique. The Kriging also produced a good quality of the distribution map of particle flux over the SAA region as indicated by its small variance value. This suggests that the model can be applied in the development of a Low Earth Orbit (LEO)-Equatorial satellite for monitoring trapped particle radiation hazard.

• 出版日期2014-8