AnneBF / snownet

This repository contains neural network files and sample code to estimate snow depth on sea ice as presented in A. Braakmann-Folgmann and C. Donlon: Estimating Snow Depth on Arctic Sea Ice using Satellite Microwave Radiometry and a Neural Network, The Cryosphere, 2019.

The sample code 'SNOWNET_sample_code.py' demonstartes how to use the AMSR2-only ('snownet.h5') 
and the AMSR2+SMOS ('snownet_smos.h5') neural networks for snow depth estimation.
Since July the new version of 'SNOWNET_sample_code.py' also includes an uncertainty estimation.

To run the code you need a few packages installed including Tensorflow and Keras.
To make sure the model can be loaded and to reproduce the same results as presented 
in the paper, please install Tensorflow version 1.10.0 and Keras version 2.2.2.

When using the neural networks or the code, please cite:
'A. Braakmann-Folgmann and C. Donlon: Estimating Snow Depth on Arctic Sea Ice using 
Satellite Microwave Radiometry and a Neural Network, The Cryosphere, (2019)'