Skip navigation

Precipitation modelling using doubly stochastic point process models at fine time-scale

Precipitation modelling using doubly stochastic point process models at fine time-scale

Ramesh, N. I. ORCID: 0000-0001-6373-2557, Onof, C. and Xie, D. (2010) Precipitation modelling using doubly stochastic point process models at fine time-scale. In: 10th International Precipitation Conference (IPC10), 23-25 June 2010, Coimbra, Portugal.

[img]
Preview
PDF (Author Accepted Abstract)
17984 RAMESH_Precipitation_Modelling_2010.pdf - Presentation

Download (9kB) | Preview

Abstract

This paper explores the use of a class of stochastic point process models, based on doubly stochastic Poisson processes, in the modelling of rainfall. We examine the application of this class of models, a neglected alternative to the widely-known Poisson cluster models, in the analysis of fine time-scale rainfall using tipping-bucket raingauge tip times. These models are used to analyse data from a single and multiple sites to study the temporal and spatial variability of precipitation at fine time-scales.

The arrival pattern of raingauge bucket tip times is viewed as a Poisson process whose rate of occurrence varies according to an unobserved finite state irreducible Markov chain. Since the likelihood function can be expressed in a tractable form, the models are fitted with maximum likelihood methods, thus enabling a more straightforward estimation of parameter uncertainty than for Poisson-cluster models. One of the advantages of this modelling approach is that it allows us to fit various sub-models of interest for the data, and also to test hypotheses on model parameters. As the tip-times from raingauges are modelled directly, this approach overcomes the usual need to convert tip-times into rainfall depths.

We study precipitation patterns at a network of gauges in the South-West of England (HYREX data) using univariate models and the extension of these models to a multivariate framework is examined. The multivariate version of our proposed model can provide the basis for a proper spatial-temporal model for which the parameters are well identified using maximum likelihood estimation. Results on fitting 3-state and 4-state models are presented. A small scale simulation study is carried out to compare some of the rainfall characteristics of accumulated rainfall.

Item Type: Conference or Conference Paper (Paper)
Uncontrolled Keywords: Point process models, Doubly stochastic, Precipitation, Rainfall, Maximum likelihood
Subjects: G Geography. Anthropology. Recreation > GE Environmental Sciences
Faculty / School / Research Centre / Research Group: Faculty of Engineering & Science > School of Computing & Mathematical Sciences (CMS)
Faculty of Engineering & Science
Last Modified: 04 Mar 2022 13:08
URI: http://gala.gre.ac.uk/id/eprint/17984

Actions (login required)

View Item View Item

Downloads

Downloads per month over past year

View more statistics