Using a transfer matrix based on 10 CMIP5 models for ENSO prediction

Using a transfer matrix based on 10 CMIP5 models for ENSO prediction

Jongedijk, C.E.

Katsman, C. (mentor)
Sévellec, F. (mentor)
de Vries, S. (mentor)
Drijfhout, S. (mentor)

Civil Engineering and Geosciences

Hydraulic Engineering

Environmental Fluid Mechanics

2017-05-09

Additional Thesis work performed as part of the master Hydraulic Engineering at Delft University of Technology in collaboration with University of Southampton, Ocean and Earth Sciences, Physical Oceanography - In this research the outcome of several realisations of 10 different models from the CMIP5 program is used to predict the El Niño/Southern Oscillation (ENSO).With her extreme phases ’El Niño’ and ’La Niña’, ENSO is probably the most infamous oceanographic and climate phenomenon in the physical climate system with a great influence on the world’s climate and weather systems. Despite the fame amongst public and the high importance in research programmes throughout the last decades, the potential causes, cyclicity patterns and the possible forcing of ENSO are still not fully understood. Even with state of the art models its predictability (up to 12 months ahead) remains quite low compared to the time scale of the phenomenon (2-5 years). In this study a statistical model is developed to predict ENSO with a transfer operator framework based on model output from 10 models. This method is based on a recent study by Sévellec [19] and transforms a deterministic single time series evolution of sea surface temperature in the Tropical Pacific, extracted from existing coupled atmosphere-ocean model data, into a probabilistic method to determine the evolution of an observational initial condition in time. The main goal of this research is to explore and define by means of hindcasting (predicting the past) the predictive skills and reliability of this statistical method. The results show that the reliability is similar to previous studies. Where the most computational extensive coupled ocean-atmosphere models show a good prediction skill up to 9 months, with this quick model a similar skill is sustained up to a prediction window of 5 months. Since this is the first time this method applied on ENSO, recommendations are done for further development of the model as well as for the application on ENSO. Suggestions for potential other systems this model could be applied to are made.

physical oceanography
numerical modeling
climate variability
ENSO

http://resolver.tudelft.nl/uuid:14af7c9f-b1c0-4af5-9a91-20b261493b15

Student theses

student report

(c) 2017 Jongedijk, C.E.