Effect of extreme value loss on long-term correlated time series

Effects of extreme value loss on long-term correlated time series are analyzed by means of detrended fluctuation analysis (DFA) and power spectral density analysis. Weaker memory can be detected after removing of extreme values for the artificial long-term correlated data, indicating the emergence of extreme events may be closely related to long-term memory. For observational temperature records, similar results are obtained, but not in all stations. For example, in some stations, only extending of scaling range to smaller time scales occurs, which may be due to the asymmetric distribution of values in the record. By comparing our findings with previous works, clustered positions of the extreme events are recognized as an important property in long-term correlated records. Through a simple numerical test, close relations between extreme events and long-term memory are discovered, which is helpful for our understanding of the effects of extreme value loss on long-term correlated records.