| Summary: | Using artificial data sets it was earlier demonstrated that noise in solar-wind variables alters the functional form of best-fit solar-wind driver functions (coupling functions) of geomagnetic activity. Using real solar-wind data that noise effect is further explored here with an aim at obtaining better best-fit formulas by removing noise in the real solar-wind data. Trends in the changes to best-fit solar-wind formulas are examined when Gaussian random noise is added to the solar-wind variables in a controlled fashion. Extrapolating those trends backward toward lower noise makes predictions for improved solar-wind driver formulas. Some of the error (noise) in solar-wind data comes from using distant L1 monitors for measuring the solar wind at Earth. An attempt is made to confirm the improvements in the solar-wind driver formulas by comparing results of best-fit formulas using L1 spacecraft measurements with best-fit formulas obtained from near-Earth spacecraft measurements from the IMP-8 spacecraft. However, testing this methodology fails owing to observed large variations in the best-fit-formula parameters from year-to-year and spacecraft-to-spacecraft, with these variations probably overwhelming the noise-correction variations. As an alternative to adding Gaussian random noise to the solar-wind variables, replacing a fraction of the values of the variables with other values was explored, yielding essentially the same noise trends as adding Gaussian noise.
|
|---|