loading page

Subseasonal Forecasts of Opportunity Identified by an Interpretable Neural Network
  • Kirsten J Mayer,
  • Elizabeth A. Barnes
Kirsten J Mayer
Colorado State University - Department of Atmospheric Science, Colorado State University - Department of Atmospheric Science

Corresponding Author:kjmayer@rams.colostate.edu

Author Profile
Elizabeth A. Barnes
Colorado State University, Colorado State University
Author Profile


Midlatitude prediction on subseasonal timescales is difficult due to the chaotic nature of the atmosphere and often requires the identification of favorable atmospheric conditions that may lead to enhanced skill (“forecasts of opportunity”). Here, we demonstrate that an artificial neural network can identify such opportunities for tropical-extratropical teleconnections to the North Atlantic circulation at a lead of 22 days using the network’s confidence in a given prediction. Furthermore, layer-wise relevance propagation, an ANN interpretability technique, pinpoints the relevant tropical features the ANN uses to make accurate predictions. We find that layer-wise relevance propagation identifies tropical hot spots that correspond to known favorable regions for midlatitude teleconnections and reveals a potential new pattern for prediction over the North Atlantic on subseasonal timescales.
28 May 2021Published in Geophysical Research Letters volume 48 issue 10. 10.1029/2020GL092092