Potsdam Institute for Climate Impact Research (PIK)-Member of the Leibniz Association, Telegrafenberg A56, 14473 Potsdam, Germany.
Institute for Astrophysics, Georg-August-University, Friedrich-Hund-Platz 1, 37077 Göttingen, Germany.
Chaos. 2020 Mar;30(3):033102. doi: 10.1063/1.5134012.
Understanding spatiotemporal patterns of climate extremes has gained considerable relevance in the context of ongoing climate change. With enhanced computational capacity, data driven methods such as functional climate networks have been proposed and have already contributed to significant advances in understanding and predicting extreme events, as well as identifying interrelations between the occurrences of various climatic phenomena. While the (in its basic setting) parameter free event synchronization (ES) method has been widely applied to construct functional climate networks from extreme event series, its original definition has been realized to exhibit problems in handling events occurring at subsequent time steps, which need to be accounted for. Along with the study of this conceptual limitation of the original ES approach, event coincidence analysis (ECA) has been suggested as an alternative approach that incorporates an additional parameter for selecting certain time scales of event synchrony. In this work, we compare selected features of functional climate network representations of South American heavy precipitation events obtained using ES and ECA without and with the correction for temporal event clustering. We find that both measures exhibit different types of biases, which have profound impacts on the resulting network structures. By combining the complementary information captured by ES and ECA, we revisit the spatiotemporal organization of extreme events during the South American Monsoon season. While the corrected version of ES captures multiple time scales of heavy rainfall cascades at once, ECA allows disentangling those scales and thereby tracing the spatiotemporal propagation more explicitly.
理解气候极端事件的时空模式在当前气候变化背景下具有相当重要的意义。随着计算能力的提高，基于数据驱动的方法，如功能气候网络，已经被提出，并已经为理解和预测极端事件以及识别各种气候现象的发生之间的相互关系做出了重要贡献。虽然（在其基本设置中）无参数事件同步（ES）方法已被广泛应用于从极端事件序列构建功能气候网络，但已发现其原始定义在处理后续时间步的事件时存在问题，需要加以考虑。随着对原始 ES 方法的这一概念限制的研究，事件一致性分析（ECA）已经被提出作为一种替代方法，该方法包含一个额外的参数，用于选择事件同步的某些时间尺度。在这项工作中，我们比较了使用 ES 和 ECA 以及不使用和使用时间事件聚类校正获得的南美的强降水事件的功能气候网络表示的选定特征。我们发现这两种方法都表现出不同类型的偏差，这对生成的网络结构有深远的影响。通过结合 ES 和 ECA 捕获的互补信息，我们重新审视了南美季风季节极端事件的时空组织。虽然经过修正的 ES 可以同时捕捉多个时间尺度的强降雨级联，但 ECA 可以将这些尺度分开，从而更明确地追踪时空传播。
