Stochastic modelling of single-site rainfall in a nonstationary climate - recent model developments and local estimation.
Speaker: Valerie Isham (University College London)
Time: 3.00 PM
Date: Thursday 4th April 2013
Location: A109 Newman building.
Abstract:
Simulations of rainfall at sub-daily time scales are required as input into various models, for example the rainfall-runoff models used for hydrological design. Stochastic, mechanistic models based on an underlying point process structure have been in widespread use for this purpose for more than twenty years and perform well in representing rainfall at hourly and daily timescales. However, for urban drainage and other applications, sub-hourly timescales are needed. Recent work, fitting these models and some new extensions to 5-minute rainfall data, has provided new insights and clarified the complex interactions between the models, the method of fitting, and the resolution of the data used.These models are stationary by construction, whereas rainfall is known to be seasonal. Current practice is to fit a separate model for each calendar month or season. However, under climate change scenarios, calendar month is likely to become increasingly less reliable as an indicator of rainfall behaviour, and is in any case just a proxy for the real drivers, atmospheric variables such as temperature and pressure. The current approach is therefore extended by replacing the discrete covariate, calendar month, with continuous covariates which are more directly related to the incidence and nature of rainfall. The covariate-dependent model parameters are estimated for each time interval using a kernel-based nonparametric approach within a Generalised Method of Moments framework (“local GMM”).
Series: Statistics Seminar Series
Social Media Links