Cluster Extreme Events Based on Point Process Theory

Extreme weather and climate events have attracted more attention in the last few years due to the often large loss of human life and exponentially increasing costs coping with them. The assessment of extreme climate events has found that extreme events are showing the characteristic of clustering. Many natural phenomena manifest themselves as spatial point processes which produce numerous events in space, such as extreme events. For clarification, the occurrence of a phenomenon located at a single point is defined as an event in contrast to a simple geometric point. Some events assemble in a restricted region while other events are dispersed. Regarding extreme events as spatial point processes and the cluster events as the clustered points, a method is developed to delineate the cluster events from a region which is called k-th order nearest distance spatial point method. By combining spatial point theory with cluster events, the algorithm is validated with a new method which consists of four major steps. The first step is to define different kinds of extreme events including 12 kinds of extreme temperature events and 6 kinds of extreme precipitation events. The second step is to calculate the statistically weights of every station based on the frequency and strength. The third step is to give the spatial distributions of all the stations that extreme events happens and the last step is to pick out the clustered events combining the spatial theory. Based on this method, extreme events research in eastern and middle areas of China have been accomplished, including extreme temperature events and extreme precipitation events. Results of k-th order nearest distance, the material definitions of weights and the distribution charts are given, which can detect cluster extreme events effectively. Based on the given charts and two efficient indexes (one is called average of ratio which is represented by R, the other is called efficiency represented by η), all kinds of extreme climate events are tested to validate the method, finding out that the algorithm is fit for cluster extreme climate events, especially for heavier extreme events. The method performs better on temperature extreme events than precipitation extreme events. So it's concluded that cluster extreme events can be delineated not only by qualitative research, but also by quantitative research, which offers a new idea in extreme events researches.