LANL*: Radiation belt drift shell modeling
Abstract
LANL* calculates the magnetic drift invariant L*, used for modeling radiation belt dynamics and other space weather applications, six orders of magnitude (~ one million times) faster than convectional approaches that require global numerical field lines tracing and integration. It is based on a modern machine learning technique (feedforward artificial neural network) by supervising a large data pool obtained from the IRBEM library, which is the traditional source for numerically calculating the L* values. The pool consists of about 100,000 samples randomly distributed within the magnetosphere (r: [1.03, 11.5] Re) and within a whole solar cycle from 1/1/1994 to 1/1/2005. There are seven LANL* models, each corresponding to its underlying magnetic field configuration that is used to create the data sample pool. This model has applications to realtime radiation belt forecasting, analysis of data sets involving tens of satelliteyears of observations, and other problems in space weather.
 Publication:

Astrophysics Source Code Library
 Pub Date:
 September 2014
 Bibcode:
 2014ascl.soft09003Y
 Keywords:

 Software;
 NASA