UCL Discovery
UCL home » Library Services » Electronic resources » UCL Discovery

Statistical downscaling of air quality models using Principal Fitted Components

Alkuwari, Farha Ahmad Z. A.; (2018) Statistical downscaling of air quality models using Principal Fitted Components. Masters thesis (M.Phil), UCL (University College London). Green open access

[thumbnail of Al-Kuwari_10044968_thesis_corrected.pdf]
Preview
Text
Al-Kuwari_10044968_thesis_corrected.pdf

Download (3MB) | Preview

Abstract

Statistical downscaling is a technique that is used to extract high-resolution information from regional scale variables produced by Chemical Transport Models (CTMs). The aim of this thesis is to shade light on the advantages of statistical downscaling in improving the forecasting ability of air quality models. Many statistical downscaling methods in geophysics often rely on dimension reduction techniques to reduce the spatial dimension of gridded model outputs without loss of essential spatial information. In this thesis we developed a new downscaling methodology that relies on using Principal Fitted Components (PFCs) to downscale an air quality model. The main advantage of employing PFCs in downscaling relies in the fact that PFCs represent space-time variations associated with a particular location through the use of inverse regression. This means that PFCs will emphasize on location related regional information. We illustrate our proposed method by both simulation and application on ground level ozone over southeastern U.S region to downscale the Regional ChEmical TrAnsport Model (REAM). Both simulation and applications results indicate that PFC downscaling appears to yield more accurate forecasts. Moreover, we accommodate the fact that covariance matrices that are used to compute PFCs might be unstable due to the fact that they have a relatively large dimension. This issue has motivated us to regularize the covariance matrices by thresholding prior to computing the PFCs and then proceed with the downscaling using thresholded PFCs. We illustrate the modified downscaling approach by simulation and application to ground level ozone. Simulation results suggest that employing thresholded PFCs in downscaling have improved the downscaling results, however, the application results do not agree with the simulation results. Finally, we extend our PFC downscaling method to downscale an ensemble of air quality models. We propose a new two-stage dimension reduction approach to reduce the dimension of an ensemble. The proposed methodology reduces the spatial dimension in each ensemble member, and then the reduced variables are reduced further across the ensemble models. We illustrate our proposed methodology by simulation and application to downscale ground level ozone ensemble outputs in France. Both simulation and application results suggest that our proposed technique seem to show an adequate predictive performance.

Type: Thesis (Masters)
Qualification: M.Phil
Title: Statistical downscaling of air quality models using Principal Fitted Components
Event: UCL (University College London)
Open access status: An open access version is available from UCL Discovery
Language: English
UCL classification: UCL > Provost and Vice Provost Offices
UCL > Provost and Vice Provost Offices > UCL BEAMS
UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Maths and Physical Sciences
UCL > Provost and Vice Provost Offices > UCL BEAMS > Faculty of Maths and Physical Sciences > Dept of Statistical Science
URI: https://discovery.ucl.ac.uk/id/eprint/10044968
Downloads since deposit
231Downloads
Download activity - last month
Download activity - last 12 months
Downloads by country - last 12 months

Archive Staff Only

View Item View Item