Murdoch University Research Repository

Welcome to the Murdoch University Research Repository

The Murdoch University Research Repository is an open access digital collection of research
created by Murdoch University staff, researchers and postgraduate students.

Learn more

Statistical downscaling from numerical climate models

Charles, Stephen Philip (2002) Statistical downscaling from numerical climate models. PhD thesis, Murdoch University.

[img]
PDF - Whole Thesis
Available Upon Request

Abstract

Statistical downscaling techniques address the disparity between the coarse spatial scales of numerical climate models (NCMs), typically 100-500 km, and point meteorological observations. However, there has been limited success in developing statistical downscaling techniques that can reproduce important properties of daily precipitation such as long runs of wet- or dry-spells. There has also been a lack of techniques applicable to multi-site networks, where there are strong inter-site correlations in daily precipitation. When used in climate change investigations, there has been limited success in producing downscaled precipitation projections in accordance with the general trends indicated by the host NCM.

In this thesis, an extended nonhomogeneous hidden Markov model (extended-NHMM) for multi-site, daily precipitation occurrence and amounts is developed. Its performance is assessed according to:
•its physical realism;
•its ability to reproduce the multi-site, daily precipitation statistics of a moderately dense site network;
•its ability to successfully downscale NCM simulations of present day climate; and,
•its ability, when used for climate change projection, to produce daily precipitation projections for the site network in accordance with the trends indicated by the host NCM.

The model is applied to a moderately dense network of 30 rain gauge stations in southwest Western Australia (SWA) using 15 years (1978 to 1992) of historical ‘winter’ (May-October) daily precipitation and atmospheric data. The extended-NHMM assumes that multi-site, daily precipitation occurrence patterns are driven by a finite number of unobserved weather states that evolve temporally according to a first order Markov chain. The weather state transition probabilities are a function of observed or modelled synoptic-scale atmospheric predictors such as mean sea level pressure. Within each weather state, the site daily precipitation amounts are modelled as regressions of transformed amounts at a given site on precipitation occurrence at neighbouring sites.

Results indicate that the extended-NHMM successfully reproduces the at-site and inter-site statistics of daily precipitation (frequency of wet-days, dry- and wet-spell length distributions, amount distributions, and inter-site correlations in occurrence and amounts). The weather states provide a regional hydroclimatology of the study region. They represent the dominant spatial patterns of daily precipitation occurrence that are related to synoptic conditions, and thus climate variability, via the optimum selection of a small set of atmospheric predictors.

The extended-NHMM fitted to observed SWA data has been driven with atmospheric predictor sets extracted from General Circulation Model (GCM) and Limited Area Model (LAM) present day climate runs, an atmospheric GCM hindcast run forced by observed SSTs, and a climate change (2xC02) LAM run. Downscaling from the GCM and LAM present day climate predictors reproduces the observed statistics of daily precipitation. Downscaling from the SST-forced GCM hindcast only reproduces the statistics of the recent period, with poor performance for earlier periods attributed to inadequacies in the forcing SST data. Climate change (2xC02) precipitation occurrence projections in accord with the trend indicated by the LAM were only obtained from an NHMM that included a predictor representing relative moisture. Thus assessing predictor set selection for climate change downscaling is critically important.

Item Type: Thesis (PhD)
Murdoch Affiliation: Division of Science and Engineering
Notes: Note to the author: If you would like to make your thesis openly available on Murdoch University Library's Research Repository, please contact: repository@murdoch.edu.au. Thank you.
Supervisor(s): Bates, B. and Lyons, Tom
URI: http://researchrepository.murdoch.edu.au/id/eprint/51653
Item Control Page Item Control Page