HigherOrder Statistical Method for Extracting Dependencies in Geospace Data Sets

1
Higher-Order Statistical Method for Extracting
Dependencies in Geospace Data Sets

• Jay R. Johnson
• jrj_at_pppl.gov
• Project Summary
• The goal of this project is to develop
  higher-order statistical techniques to identify
  nonlinear dependencies and couplings in geospace
  data sets
• Plans for Upcoming Year
• Build a database of direct measures of
  magnetospheric state
• Develop MI/Cumulant analysis to
• characterize the underlying dynamics
• discover the most important nonlinearities
• determine information horizon
• obtain a coupling function
• investigate dimensionality
• compress data stream through dimensional
  reduction
• Publish paper on the cumulant technique and
  Applications to high speed streams

• Last Years Highlights
• Identified nonlinearity in magnetospheric
  dynamics during the declining phase of the solar
  cycle
• Identified a timescale (information horizon) of
  the nonlinearity
• Determined that the nonlinear response is an
  internal magnetospheric response to solar wind
  velocity enhancements
• Suggested an important nonlinear coupling
  sensitive to solar wind velocity
• Presented invited talk at ISSC2 meeting and
  contributed presentations at ISROSES 2006 and AGU
  meetings

Impact on Science Information-theoretical
approach provides a significant advance over
traditional linear methods of time-series
analysis used in geospace science by accounting
for nonlinear dependency. The approach
provides a systematic method to determine how far
ahead geospace events The approach can
be used to characterize the solar-wind-magnetosphe
re coupling function
Mutual Information Cumulants
AISRP 2006-2009
http//w3.pppl.gov/jrj/cumulant.html