Group Meeting: ENSO Dynamics, Trends, and Prediction Using Machine Learning (Santiago)

Presentation Date: 

Friday, October 30, 2020

Presentation Slides: 

The authors of this paper used a non-homogenous Hidden Markov model (NHMM) to study ENSO dynamics and its trends within the last century as well as the capacity of the model to make predictions for the 2015-2016 El Nino. The NHMM produced 6 different hidden states that represent different well-known spatial patterns within the tropical Pacific basin from central Pacific El Nino and La Nina patterns to events confined at the eastern portion of the basin. Temporal trend for the mode that corresponded to a well defined central Pacific El Nino (often referred to as El Nino Modoki), an event that is more frequent in observations for the last few decades, seems to correspond to the global surface temperature trend. While the latter suggests an influence of climate change in the frequency of such event, the author did not rush into making any conclusions as the methodology that was used was not the best for explaining relation. On the other hand, the ability of the model to predict future events with data that was not used during the training process, predicted satisfactory past events with an accuracy that is similar and even better to some of the different models that are currently used to predict ENSO.