Abstract
Improving climate models to better forecast water-related hazards, such as floods, and managing water resources require understanding how extreme precipitation events (EPEs) are propagated across spatial and temporal scales. Complex network analysis is a promising approach to study spatial and temporal dependencies in large unstructured data, such as those in climate science. Here, we studied EPEs and their synchronization in North America using complex network analysis. We constructed the EPEs networks based on the gauge-based daily precipitation data from the Climate Prediction Center database. The nodes within the networks were the geographical grid points connected through the links determined by applying a non-linear method called event synchronization. We analyzed the spatiotemporal patterns within the complex networks of EPEs via network measures, such as degree centrality (DC), mean geographic distance (MGD), clustering coefficient (CC), betweenness centrality (BC) and long-ranged directedness (LD). We first found hubs—locations important in propagating EPEs and teleconnections—in areas such as Montana, Wyoming, Alberta, and Saskatchewan in summers and the West Coast and eastern North America in winters. We demonstrated that the EPEs are more spatially coherent during winters than summers across the continent. We then identified certain locations in Utah, Colorado, South Dakota, northern Mexico, southeast British Columbia, and northern Quebec that play a crucial role in the long spatial propagation of the EPEs from one region to another throughout summers, while regions in the West Coast, southern Colorado, New Mexico, Wisconsin, and central Alberta are dominant in winters. Additionally, we uncovered atmospheric moisture pathways for the EPEs in both seasons. Finally, we showed that the EPEs complex networks are sensitive to the lag time Tmax, a parameter in the ES method. As such, it should be chosen depending on the atmospheric phenomena of interest and its temporal scale.
