Please use this identifier to cite or link to this item:
Title: The predictability of UK drought using European weather patterns
Authors: Richardson, Douglas
Issue Date: 2019
Publisher: Newcastle University
Abstract: This thesis explores the use of a 167-year daily weather pattern (WP) classification (MO-30) in UK meteorological drought prediction. As MO-30 was recently introduced, necessary analyses as a precursor to building a forecast model are conducted. First, an exploratory analysis of MO30’s fundamental characteristics and its relation to UK precipitation and drought climatology is carried out. Second, two novel methods to find weekly to seasonal persistence in MO-30 are used in order to assess if there is any inherent predictability within MO-30. Third, a statistical model based on historical analogues for predicting 30-day periods of WPs is constructed, from which precipitation forecasts are derived. Finally, a dynamical ensemble prediction system is applied to forecast WPs, with resultant precipitation estimated in the same way as for the statistical method. MO-30 is shown to be suitable for precipitation-based analyses in the UK. Furthermore, intraWP precipitation variability, defined by the interquartile range, is lower in MO-30 compared to another commonly used WP classification. Six WPs are associated with nationwide drought, with several other WPs linked to regional drought. Results from the persistence analysis show that there are multi-month periods when small sets of four to six WPs dominate, and some of these periods coincide with notable meteorological events, including droughts and storms. Some WPs also behave as ‘attractors’, showing increased probability of reoccurrence despite other WPs occurring in-between. The statistical method for WP and precipitation forecasts is no more skilful than climatology, suggesting that the model did not adequately exploit the persistence identified previously. However, WPs are shown to be potentially useful for drought forecasting, as an idealised, perfect prognostic model (with WP observations as inputs rather than predictions) substantially improves skill, with a skill score of almost 0.5 (out of one) for north-eastern regions. Using a dynamical model to predict WPs, while keeping the precipitation estimation procedure the same as for the purely statistical method, yields overall higher skill compared to a benchmark statistical method for predicting droughts. The model also outperforms direct (modelled) dynamical precipitation forecasts for lead-times greater than 16 days during winter and autumn, with the greatest skill advantage for western regions. This is despite the relatively modest skill scores of all forecast models (rarely above 0.4). Again, high skill scores, of almost 0.8 on occasions, are achieved by the perfect prognostic model, demonstrating the potential for incorporating WPs into precipitation and drought forecast systems.
Description: PhD thesis
Appears in Collections:School of Engineering

Files in This Item:
File Description SizeFormat 
Richardson D 2019.pdfThesis8.37 MBAdobe PDFView/Open
dspacelicence.pdfLicence43.82 kBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.