Fluctuation Analysis of Runoff Time Series Under Coarse-Graining Network Modeling Method

Abstract

Based on coarse-graining complex networks modeling method, the fluctuation characteristic of the daily runoff series of the Yangtze River in China and the Ocumlgee River in America are investigated in this paper, respectively. First, the related runoff time series were transformed into discrete symbolic sequences by coarse-graining preprocessing method, and then, related complex networks were created. Dynamic statistical features and topology parameters of the two fluctuation networks, such as clustering coefficient (CC), characteristic path length (CPL), and betweenness centrality example, from 2-June to 19-June in (BC) of nodes, are calculated and compared with each other. It can be found that, the clustering coefficients of both networks are much larger than random networks with the same scale, and both characteristic path lengths are as small as random networks with the same scale. It indicated that the short-range correlation exists in different fluctuation patterns of the discrete symbolic sequences obtained by the coarse-graining method. Furthermore, it can be found that, the betweenness centrality of different node is obvious different from each other in both networks, which means that some fluctuation patterns have important significance and can be seen as a conversion precursor between the various fluctuation patterns. Also, the community analysis are also carried out and closer connections are revealed for further prediction research. These results indicates potential value for research on short-term prediction of runoff process.