Back to search

University of British Columbia

Forecasts of tropical Pacific sea surface temperatures by neural networks and support vector regression

Abstract

dc:description

Nonlinear and linear regression models were developed to forecast the sea surface temperature anomalies (SSTA) across the tropical Pacific ocean. The methods used were, Bayesian neural networks (BNN), support vector machines for regression (SVR) and linear regression (LR). The predictors of the models were a particular combination of the principal components of sea level pressure (SLP) and sea surface temperature (SST) data at lead times ranging from 3 to 15 months. Two data sets corresponding to the time periods, 1950-2005 and 1980-2005 were independently studied. The later data set contained the standardized volume of warm water (WWV) integrated across the tropical Pacific basin as an extra predictor. The seasonal cycle was removed from both data sets prior to their introduction into the regression models. Various graphical and statistical tools were used to compare the performance of these models, in particular the correlation scores. The results reveal that although there appeared to be nonlinear elements of the SSTA response to the previous SLP and SST conditions, this response is of second order. The large amplitude variance represented by the first principal component (PC) of the SST field can be adequately expressed as a linear response. Consistently in both data sets, the advantages in the nonlinear modelling manifested mainly through the second PC of the SST variability. The spacial distribution of correlation skills between the predicted and observed SST fields corresponding to the BNN and SVR models show some nonlinear structures in various parts of the domain. Despite some differences in the temporal and spacial distribution of these correlation skills, there was no decisive advantage between using BNN or SVR. Adding the WWV to the predictor set reduced the marginal advantage of the nonlinear methods over the eastern region of the tropical Pacific, in particular at lead times of 9-15 months. These results generally agree with linear recharge oscillator models, and the hypothesis of a linear relationship between thermocline depth and SST in the eastern region.

Degree

thesis:*
Name thesis:degree_name
Master of Science - MSc
Level thesis:degree_level
master's
Discipline thesis:degree_discipline
Physics
Grantor dc:publisher
University of British Columbia
Year dc:date
2008

Author and committee

dc:creator, dc:contributor.*
Author dc:creator
  • Aguilar-Martinez, Silvestre

Rights

dc:rights
Statement dc:rights
  • Attribution-NonCommercial-NoDerivatives 4.0 International
Language dc:language
eng

Identifiers

dc:identifier.*
Handle dc:identifier
http://hdl.handle.net/2429/7577

Chain of custody

source
Harvested from
University of British Columbia
Base URL
circle.library.ubc.ca/oai/request
Last updated
2026-11-19
Source record
OAI-PMH GetRecord
related terms
citation

Aguilar-Martinez, Silvestre. Forecasts of tropical Pacific sea surface temperatures by neural networks and support vector regression. master's thesis, University of British Columbia, 2008. http://hdl.handle.net/2429/7577