Akademik Veri Yönetim Sistemi
STOCHASTIC ENVIRONMENTAL RESEARCH AND RISK ASSESSMENT, cilt.35, sa.3, ss.535-548, 2021 (SCI-Expanded)
Adequate understanding of the temporal connections in rainfall is important for reliable predictions of rainfall and, hence, for water resources planning and management. This research aims to study the temporal connections in rainfall using complex networks concepts. First, the single-variable rainfall time series is represented in a multi-dimensional phase space using delay embedding (i.e. phase-space reconstruction), where the appropriate delay time and optimal embedding dimension of the time series are determined by using average mutual information and false nearest neighbors methods, respectively. Then, this reconstructed phase space is treated as a 'network,' with the reconstructed vectors serving as 'nodes' and the connections between them serving as 'links'. Finally, the strength of the nodes are calculated to identify some key properties of the temporal rainfall network. The approach is employed independently to monthly rainfall data observed over a period of 38 years (1979-2016) from 14 rain gauge stations in the Vu Gia Thu Bon River basin in central Vietnam. Moreover, entropy values of the original rainfall time series are calculated for obtaining additional information on the properties of the rainfall dynamics. The average node strengths are also examined in terms of the mean annual rainfall, entropy of the time series, and elevation of the rain gauge station. The results indicate that: (1) while some adjacent stations (i.e. networks) have somewhat similar strength (average node strength) values, several others that are geographically close show significantly different network strengths; (2) similar entropies for adjacent stations are found more frequently than similar average node strengths; (3) there is generally a positive and proportional relationship between average strengths of nodes and entropies; and (4) the average node strengths of different months have some distinct temporal patterns (3-month, 4-month, and 6-month patterns) in rainfall dynamics, depending upon the specific region of the study area. These results have important implications for prediction, interpolation, and extrapolation of rainfall data.