Data for "Addressing the Nonlinear Effects in the Albedo Feedback Using the Neural Network Method"

This dataset contains a neural network model for the shortwave radiation prediction, scripts to generate data for the radiative feedback quantification in the Arctic, RTM simulated radiative feedbacks, and results from the kernel method.
I. File list
------------
Figures/ scripts to generate plots
era5_grid1_data/ Sept. 1992 and Sept. 2012 ERA5 data
NN/nn_tsr.mat NN model for TSR flux prediction
NN/nn_ssr.mat NN model for SSR flux prediction
NN/nn_toa.m function for the TSR prediction
NN/nn_sfc.m function for the SSR prediction
NN/calculate_toa_feedbacks.m radiative feedback quantification at the TOA
NN/calculate_sfc_feedbacks.m radiative feedback quantification at the surface
NN/train_nn_toa.m script to train nn model for TSR flux prediction

Note: the folder NN contains scripts for the shortwave radiative feedback quantification in the Arctic:
1. calculate_toa_feedbacks.m
2. calculate_sfc_feedbacks.m
The results generated by these scripts are presented in Figure 6 and Figure S7.

II. NN design
-------------

NN model for the top net solar radiation (TSR) prediction. Input variables are:
1. TOA incident solar radiation (W*m**-2), downward positive
2. Total column cloud ice water (kg*m**-2)
3. Total column cloud liquid water (kg*m**-2)
4. Total column water vapour (kg*m**-2)
5. High cloud cover (0-1)
6. Medium cloud cover (0-1)
7. Low cloud cover (0-1)
8. Surface pressure (Pa)
9. Total column ozone (kg*m**-2)
10. Forecast albedo (0-1)

Output: TSR (W*m**-2), downward positive
All variables are monthly averaged values. Radiation variable is for all-sky conditions.