Table of Contents
Journal of Climatology
Volume 2013 (2013), Article ID 390945, 15 pages
http://dx.doi.org/10.1155/2013/390945
Research Article

Efficiencies of Inhomogeneity-Detection Algorithms: Comparison of Different Detection Methods and Efficiency Measures

Centre for Climate Change, University of Rovira i Virgili, Campus Terres de l’Ebre, Avenue Remolins 13-15, 43500 Tortosa Tarragona, Spain

Received 18 March 2013; Accepted 2 September 2013

Academic Editors: S. Feng, L. Makra, and A. P. Trishchenko

Copyright © 2013 Peter Domonkos. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

Abstract

Efficiency evaluations for change point Detection methods used in nine major Objective Homogenization Methods (DOHMs) are presented. The evaluations are conducted using ten different simulated datasets and four efficiency measures: detection skill, skill of linear trend estimation, sum of squared error, and a combined efficiency measure. Test datasets applied have a diverse set of inhomogeneity (IH) characteristics and include one dataset that is similar to the monthly benchmark temperature dataset of the European benchmarking effort known by the acronym COST HOME. The performance of DOHMs is highly dependent on the characteristics of test datasets and efficiency measures. Measures of skills differ markedly according to the frequency and mean duration of inhomogeneities and vary with the ratio of IH-magnitudes and background noise. The study focuses on cases when high quality relative time series (i.e., the difference between a candidate and reference series) can be created, but the frequency and intensity of inhomogeneities are high. Results show that in these cases the Caussinus-Mestre method is the most effective, although appreciably good results can also be achieved by the use of several other DOHMs, such as the Multiple Analysis of Series for Homogenisation, Bayes method, Multiple Linear Regression, and the Standard Normal Homogeneity Test.