Home // International Journal On Advances in Intelligent Systems, volume 6, numbers 1 and 2, 2013 // View article
Authors:
Jorge Gorricha
Victor Lobo
Ana Costa
Keywords: Kriging; Precipitation patterns; Self-Organizing Map; 3D Self-Organizing Map
Abstract:
Extreme weather events such as heavy precipitation can be analyzed from multiple perspectives such diverse as the daily intensity or the number of consecutive wet days. Thus, it is necessary to get an overall view of the problem in order to characterize the extreme precipitation occurrence along time and space. Extreme precipitation indices, estimated from the empirical distribution of the daily observations, are increasingly being used not only to investigate trends in observed precipitation records, but also to examine scenarios of future climate changes. However, each of the indices, by itself, shows only a part of the phenomenon and there are multiple examples where one single index is not sufficient to characterize the occurrence of extreme precipitation. Therefore, a high dimensional approach should be considered. In this paper, we propose a framework for characterizing the spatial patterns of extreme precipitation that is based on two types of visualization approaches. The first one uses linear models, such as Ordinary Kriging and Ordinary Cokriging, to produce continuous surfaces of five extreme precipitation indices. The second one uses a three-dimensional Self-Organizing Map to visualize the phenomenon from a global perspective, allowing identification and characterization of spatial patterns and homogeneous areas. Also, to allow an easy interpretation of spatial patterns, a pattern matrix is proposed, where variables and color patterns are ordered using a one-dimensional Self-Organizing Map. The proposed framework was applied to a set of precipitation indices, which were computed using daily precipitation data from 1998 to 2000 measured at nineteen meteorological stations located in Madeira Island. Results show that the island has distinct climatic areas in relation to extreme precipitation events. The northern part of the island and the higher locations are characterized by heavy precipitation events, whereas the south and northwest parts of the island exhibit low values in all indices. The promising results from this study indicate that the proposed framework, which combines linear and nonlinear approaches, is a valuable tool to deepen the knowledge on local spatial patterns of extreme precipitation.
Pages: 16 to 26
Copyright: Copyright (c) to authors, 2013. Used with permission.
Publication date: June 30, 2013
Published in: journal
ISSN: 1942-2679