Test Data Mean prediction interval - Single Point downscaling.

  • CNN-LSTM produced lower MPI on 95% coverage probability,
    • \(\pm 101.69 Wm^{-2}\) or \(\pm 8.79MJm^{-2}day^{-1}\)

  • GLM and CNN-LSTM-DENSE exhibited very similar 95% MPI.

    • GLM exhibited 95% MPI of
      • \(\pm 107.6 Wm^{-2}\) or \(9.3 MJm^{-2}day^{-1}\)
      • CNN-LSTM-DENSE had 95% MPI of
        • \(\pm 106.51 Wm^{-2}\) or \(9.2 MJm^{-2}day^{-1}\)

      Mean prediction interval on test set for selected models

Creative Commons License
Downscaling Global Climate Models with Convolutional and Long-Short-Term Memory Networks for Solar Energy Applications by C.P. Davey is licensed under a Creative Commons Attribution-NonCommercial-ShareAlike 4.0 International License.