Use this URL to cite or link to this record in EThOS:
Title: Modelling intense rainfall in a changing climate
Author: Cross, David
Awarding Body: Imperial College London
Current Institution: Imperial College London
Date of Award: 2019
Availability of Full Text:
Access from EThOS:
Access from Institution:
Anthropogenic climate change is unequivocal, with serious implications for society. Decades of atmospheric pollution have precipitated rapid non-stationarity in the hydrosphere, changing the frequency and intensity of storms in space and time. Utilities and civil infrastructure span generations, requiring practitioners to assess the local impacts of hydro-meteorological change. Rainfall simulation and extreme value theory are necessary for water resources planning and hazard mitigation. However, purely statistical techniques lack physical realism and the estimation of larger extremes can be highly uncertain. This thesis presents a new approach for estimating short duration rainfall extremes in a changing climate with mechanistic stochastic rainfall models. Mechanistic stochastic models simulate rainfall with rectangular pulses which conceptualise the phenomenology of rainfall generation in storms. But, since their inception over 30 years ago, they have tended to under estimate rainfall extremes at fine temporal scales. Motivated by industry to improve the physical realism of extreme rainfall estimation at sub-hourly scales, a censored modelling approach is presented with Bartlett-Lewis rectangular pulse models to simulate the intense rainfall profile. With censored rainfall simulation, intense storm profiles are constructed from the superposition of cells, from which extremes are sampled. The approach is applied to two test sites in the UK and Germany and used to estimate rainfall extremes in the present and hypothesised future climates at the end of this century. A new downscaling methodology is developed in which the rainfall models are conditioned on an ensemble of CMIP5 climate model outputs for moderate and severe climate forcing. Using K-nearest neighbour sampling to identify the training data for calibration, model parameter estimators are approximated using multivariate linear regression to enable estimation outside the covariate range. The approach is introduced with conditioning on mean monthly near surface air temperature and verified with further conditioning on relative humidity.
Supervisor: Onof, Christian Sponsor: Engineering and Physical Sciences Research Council ; EDF Energy
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral