| We introduce a method for characterizing extreme events in spatially extended systems, which is based on a quantitative evaluation of linear and nonlinear interdependencies between observations from different locations. To illustrate the potentials of our approach for real-world applications, we present two case studies from different scientific disciplines which are based on observational time series: (i) river runoffs in a regional catchment during a flood event (hydrology), (ii) the formation and propagation of a traffic jam on a road network (transportation sciences). The presented examples suggests that our approach might be of particular use for studying and early detecting extreme events in networks, whose initial phases are often characterized by collective dynamics in different parts of the system. |