This web application contains interactive visualisations that I have developed as part of my research on ENSO predictability and interpretabilty of entropic learning forecasts of ENSO (to be published in an upcoming paper). The app takes static SVG images of the Markov Chains and Directed Acyclic Graphs (DAGs) from the paper and allows users to interact with them dynamically by either hovering over nodes/edges to see summary tooltips containing information such as transition probabilities, or by clicking on nodes to bring up modals with more detailed information including a video of the spatio-temporal composite pattern that is associated with each node. Sequences of composite patterns for specific forecasts, as well as feature importance maps and more are also available to view as videos on the Case Studies and Feature Importance pages.
The details of the ENSO forecasting system that was used to generate these results can be found in this paper. For more information about the development of this application, see this blog post on my website.