基于限定motif关联规则的海洋热浪事件模式发现研究

Discovery of marine heatwave event patterns based on restricted motif association rules

海表温度作为近岸海洋热浪事件的重要孕灾环境要素,揭示其变化规律有助于预测未来热浪事件的发生。本文将用于表达热浪事件发生前和发生时海表温度变化规律的关联规则定义为事件模式,提出了基于限定motif关联规则挖掘的模式发现方法。方法通过STAMP算法挖掘出motif,运用MDL评分策略分割motif形成候选关联规则,结合热浪事件发生的约束条件,实现对海洋热浪事件模式的提取。本文利用中国近海3个站点的海表温度数据进行海洋热浪事件模式发现实验,研究结果表明,中国近海的海洋热浪呈季节特征,主要发生于夏季,且持续时间达20天以上,海洋热浪事件模式呈现有规则的升温和下降趋势,升温速率和降温速率呈对称特性,且在热浪发生前,海表温度有一个短暂的升温间歇期。

Sea surface temperature is an important disaster-promoting environmental factor for nearshore marine heatwave events,and revealing its variation pattern will help predict the occurrence of future heatwave events.In this paper,the association rules used to express the changes of sea surface temperature before and during heat wave events are defined as event patterns,and a pattern discovery method based on restricted motif association rule mining was proposed.The STAMP algorithm was used to mine the motifs,and the MDL scoring strategy was used to segment the motifs to form candidate association rules.Combined with the constraints of the occurrence of heat wave events,the pattern of marine heat wave events was extracted.In this paper,the sea surface temperature data from three stations in China’s coastal waters are used to conduct a marine heat wave event pattern discovery experiment.The results show that the marine heat wave in China’s coastal waters is seasonal,mainly occurs in summer,and lasts for more than 20 days.The marine heat wave event pattern has a regular heating and falling trend,the heating rate and cooling rate are symmetrical,and before the heat wave occurs,the sea surface temperature has a short heating interval.