Probabilistic analysis of coincident sums of precipitation at two measurement stations. Introduction to the method and an example

Gabriela Biel,W Zielinski,A Michalski,L Kuchar,B Kazmierczak

doi:10.1051/itmconf/20182300002

Gabriela Biel, W Zielinski + Show 3 more

Open Access

https://doi.org/10.1051/itmconf/20182300002

Copy DOI

Abstract

This article proposes the use of copula (copula function) for the purpose of two-dimensional analysis of the sums of precipitation as measured with a Hellman rain-gauge. The sums of precipitation are characterized by a two-dimensional random variable: the sum of uninterrupted sequence of rainfalls which were measured in Jelcz-Laskowice and the corresponding (coincident) sum of precipitation at the Botanical Garden in Wrocław. Several problems occur from the very start: debonding from time and lack of precipitation on one of stations. For the purpose of greater precision and correction it should be stated that in order to apply the two-dimensional copula functions we will use a random vector determining the sum of uninterrupted sequences of rainfalls at two simultaneous stations. In that way, this will not be a characteristics of the phenomenon, but rather the definition of two-dimensional random variable under analysis. Data for analysis has been derived from observational logs of the Institute of Meteorology and Water Management, branch in Wrocław. The results obtained in years 1980-2014 were subject to analysis. The aim of the work was to find the best two-dimensional probability distribution of a random variable (OpadJelcz, OpadOgród). The following were analysed from among the known copulas: the Archimedean copulas (the Gumbel copula, the Frank copula and the Clayton copula) and the Gaussian elliptical copula. The study of fitting of copulas to observed variables was carried out using the Spearmann's rank correlation coefficient and the best fitting was obtained for the Frank's copula.

Highlights

A multidimensional analysis of the amount of rainfall and the idea to use the copula function have defined a sequence of steps to be followed [3, 6]
The copula function has been chosen as a method of statistical estimation parameters of multivariate probability distributions because it enables: - the combination of any boundary distributions, - the compilation of building multivariate statistics based on estimated marginal distributions [3]
Data for analysis comes from the logs of atmospheric precipitation observations from the last 30 or 40 years. It was collected using Hellmann rain gauges from the measuring stations located at a similar height above the sea level [2, 15]

Summary

Introduction

A multidimensional analysis of the amount of rainfall and the idea to use the copula function have defined a sequence of steps to be followed [3, 6]. The use of the copula function is one of the key elements of this methodology Correlation coefficients such as, for example, Pearson, Kendall or Spearman coefficients, are usually used for evaluating the relationship between the features under analysis [3]. The expected result is to obtain the best compatibility with one of the copulas functions, to use this copula function to estimate the multivariate probability distribution of random variable and use the multivariate cumulative distribution function to determine the degree of bulk tails out and the interdependence of marginal distribution's tails. These aspects are rarely given attention and found in traditional engineering methods

Objectives

Methods

Conclusion