hide
Free keywords:
-
Abstract:
In the past years, there has been an increasing number of
applications of functional climate networks to studying the spatiotemporal organization of heavy rainfall events or similar types of extreme behavior in some climate variable of interest. Nearly all existing
studies have employed the concept of event synchronization (ES) to
statistically measure similarity in the timing of events at different grid
points. Recently, it has been pointed out that this measure can however lead to biases in the presence of events that are heavily clustered in
time. Here, we present an analysis of the effects of event declustering on
the resulting functional climate network properties describing spatio-
temporal patterns of heavy rainfall events during the South American
monsoon season based on ES and a conceptually similar method, event
coincidence analysis (ECA). As examples for widely employed local
(per-node) network characteristics of different type, we study the degree, local clustering coefficient and average link distance patterns, as
well as their mutual interdependency, for three different values of the
link density. Our results demonstrate that the link density can markedly
affect the resulting spatial patterns. Specifically, we find the qualitative
inversion of the degree pattern with rising link density in one of the
studied settings. To our best knowledge, such crossover behavior has
not been described before in event synchrony based networks. In addition, declustering relieves differences between ES and ECA based network properties in some measures while not in others. This underlines
the need for a careful choice of the methodological settings in functional
climate network studies of extreme events and associated interpretation
of the obtained results, especially when higher-order network properties
are considered.
