Streamflow Connectivity in a Large-Scale River Basin

Abstract

Large-scale river basins play a key role in the development of many regions around the world. However, their enormous sizes and the associated complexities also make streamflow modeling and, hence, water planning and management a tremendous challenge. The present study examines the spatial connections in streamflow in a large-scale river basin using the concepts of complex networks. The Mississippi River basin in the United States is considered as a representative large-scale river basin, and daily streamflow data from 1663 stations across the basin are analyzed. Three network-based methods are employed to examine the spatial streamflow connections: clustering coefficient, degree distribution, and shortest path length. The influence of streamflow correlation threshold (i.e., correlations in streamflow between stations) on the outcomes of these methods is also investigated by considering four different threshold levels: T = 0.70, 0.75, 0.80, and 0.85. The results indicate that: (1) streamflow stations in some regions (Missouri River, Upper Mississippi River, Ohio River, and Tennessee River regions) of the Mississippi River basin have generally higher clustering coefficients (i.e., high connectivity with the rest of the stations in the entire network) when compared to streamflow stations in some other regions (Arkansas-Red-White River and Lower Mississippi River regions); (2) the streamflow network is not a random graph, but one that exhibits a combination of scale-free and exponential distribution; and (3) the streamflow network is inefficient, with even the most-connected station is connected to only about 6% of the 1663 stations in the network.