Forecasting northern polar stratospheric variability using a hierarchy of statistical models
| dc.contributor.author | Minokhin, Ivan | |
| dc.date.accessioned | 2015-10-02T18:54:11Z | |
| dc.date.available | 2015-10-02T18:54:11Z | |
| dc.date.issued | 2015-10-02 | |
| dc.date.submitted | 2015 | |
| dc.description.abstract | The northern polar stratosphere plays an important role in modulating the wintertime near-surface temperature conditions in midlatitudes. Forecasting northern polar stratospheric variability will have the potential to extend the winter weather forecasts in midlatitudes. As such, this research seeks to explore a novel approach of forecasting short-term northern polar stratospheric variability using a hierarchy of linear and non-linear statistical models. In addition to El NiƱo Southern Oscillation, the Quasi-biennial Oscillation, and the 11-year solar cycle indices, this research uses the upward flux of wave activity from the troposphere into the stratosphere as a predictor for modeling and forecasting northern polar stratospheric temperature and geopotential height anomalies. The upward flux of wave activity entering the stratosphere is the primary source of intraseasonal variability in the wintertime stratospheric polar vortex. Multiple linear regression and machine learning models were trained over the 1980-2005 time period, and the 10-day and 20-day northern polar stratospheric temperature and geopotential height forecasts were generated over the 2005-2011 time period. The importance of each predictor for modeling northern polar stratospheric variability was assessed using a permutation-based method. The study has found that the use of the meridional wave heat flux predictors improves the accuracy of short-term northern polar stratospheric geopotential height forecasts as demonstrated by the correlation coefficient of 0.48 over the 2005-2011 time period. In contrast to previous studies, multiple linear regression shows better predictive performance than the machine learning models over the 2005-2011 time period. A better predictive performance of multiple linear regression in comparison to machine learning models is due to a much higher contribution of the upward flux of wave activity than other predictors to forecast skill. | en |
| dc.identifier.uri | http://hdl.handle.net/10012/9789 | |
| dc.language.iso | en | en |
| dc.pending | false | |
| dc.publisher | University of Waterloo | |
| dc.subject | northern polar stratospheric variability | en |
| dc.subject | sudden stratospheric warming | en |
| dc.subject | planetary waves | en |
| dc.subject | meridional wave heat flux | en |
| dc.subject | machine learning | en |
| dc.subject.program | Geography | en |
| dc.title | Forecasting northern polar stratospheric variability using a hierarchy of statistical models | en |
| dc.type | Master Thesis | en |
| uws-etd.degree | Master of Science | en |
| uws-etd.degree.department | Geography | en |
| uws.peerReviewStatus | Unreviewed | en |
| uws.scholarLevel | Graduate | en |
| uws.typeOfResource | Text | en |