Specifying High Altitude Electrons Using Low-Altitude LEO Systems (SHELLS)

Accompanying archive for Specifying High Altitude Electrons Using Low-Altitude LEO Systems: Updates to the SHELLS Model. All the files and data necessary to run the SHELLS model are provided here.

Files contained in this archive:

POES data files (allsats_wsarsV6YYYYMM.p)

• Monthly pickle files for the POES input data from 2013-2018 that was used to train and run the SHELLS model. The files contain the column names, time tags, and log₁₀ of the POES flux data (#/cm2-sr-s-keV) for 4 energy values (e1,e2,e3,e4) at 29 L-shells (every 0.25 from L=1 to L=8). The code for generating additional SAR-corrected POES input data can be found here: https://github.com/poes-metop-sem/poes_metop_sem

Trained model file (model_shells_v1.h5)

• Trained neural network file containing the model architecture and weights. Can be loaded using the TensorFlow/Keras load_model function. Model was trained using TensorFlow version 2.6.2.

Input/Output Scaling Arrays (in_scale_v1.bin and out_scale_v1.bin)

• Binary files that provide the scaling transforms to standardize the 1x121 input vector and 1x3 output vector. Scalers were fit using scikit-learn version 0.22.1.

Sample run script (run_shells.py)

• Simple python script showing how to run the model with nominal inputs (L, B_mirror, Energy, Kp) along with the expected outputs.