FRONTIER DEVELOPMENT LAB 2018: SPACE WEATHER Extreme UV Spectral Irradiance Richard Galvez, Alexandre Szenicer, Rajat Thomas, & Paul Wright Mentors: Mark Cheung, David Fouhey, Meng Jin, Graham Mackintosh, & Andrés Muñoz-Jaramillo PROBLEM SOLUTIONS Solar...
More
FRONTIER DEVELOPMENT LAB 2018: SPACE WEATHER Extreme UV Spectral Irradiance Richard Galvez, Alexandre Szenicer, Rajat Thomas, & Paul Wright Mentors: Mark Cheung, David Fouhey, Meng Jin, Graham Mackintosh, & Andrés Muñoz-Jaramillo PROBLEM SOLUTIONS Solar Extreme Ultraviolet (EUV) radiation: The main driver of upper atmospheric variability Physics Irradiance Model Full ML Irradiance Model SOHO EIT 195 A 1996 2006 DeepEM (MLP temperature inversion) transforms AIA images into We use a Residual Neural Network (ResNet) to transform AIA 1997 2005 emission measure (EM) cubes in temperature space. This is the images into EVE spectra. The ResNet has 18 convolutional layers, common latent space between AIA and EVE data. Convolution with and is augmented with the means of the AIA images at the last 1998 2004 atomic response functions yields synthetic EVE emission. layer before output to help improve prediction. SDO/AIA 1999 2003 SDO SDO SDO/AIA DeepEM Physics-based EVE Full Machine Learning EVE (t
Less
