Discovery of hydrometeorological patterns

Authors : Mete ÇELİK, Filiz DADAŞERÇELİK, Ahmet Şakir DOKUZ

Pages : 840-857

Doi:10.3906/elk-1210-20

View : 16 | Download : 11

Publication Date : 0000-00-00

Article Type : Research Paper

Abstract :Hydrometeorological patterns can be defined as meaningful and nontrivial associations between hydrological and meteorological parameters over a region. Discovering hydrometeorological patterns is important for many applications, including forecasting hydrometeorological hazards insert ignore into journalissuearticles values(floods and droughts);, predicting the hydrological responses of ungauged basins, and filling in missing hydrological or meteorological records. However, discovering these patterns is challenging due to the special characteristics of hydrological and meteorological data, and is computationally complex due to the archival history of the datasets. Moreover, defining monotonic interest measures to quantify these patterns is difficult. In this study, we propose a new monotonic interest measure, called the hydrometeorological prevalence index, and a novel algorithm for mining hydrometeorological patterns insert ignore into journalissuearticles values(HMP-Miner); out of large hydrological and meteorological datasets. Experimental evaluations using real datasets show that our proposed algorithm outperforms the naïve alternative in discovering hydrometeorological patterns efficiently.
Keywords : Data mining, hydrometeorological pattern, association rule mining, ydrological databases, meteorological databases