Identifying climate memory impacts on climate network analysis: a new approach based on fractional integral techniques

Climate network (CN) analysis has shown great potential in detecting early warning signals for major climate events, but it is still challenging to reveal the underlying mechanisms. One possible reason for this issue is related to the ubiquitous climate memory, which may affect the calculations of links in climate networks, and further hinder us from a clear judgment on the sources of the early warning signal. Here in this study, we aim at identifying the climate memory impacts on the CN analysis. Combining with the Fractional Integral Statistical Model (FISM), we proposed a new approach named as CN-FISM. With FISM, one can extract the climate memory component and modify the considered time series into new series with a given length of memory preserved. By repeating the CN analysis, one thus can quantify the impacts of climate memory. We employed this approach to a recent CN analysis on the Pacific Decadal Oscillation (PDO) phase change. By comparing the CN results based on data with different memory lengths preserved, we found the climate memory within timescale of 2 years plays an important role in the arising of the early warning signal for the PDO phase change. This finding suggests that some physical processes on timescale of 2 years may be crucial for the PDO phase change, according to which one may better understand its underlying mechanisms. Compared with the current Pearson correlation-based CN approach, the CN-FISM offers the potential of improved interpretability of the CN results.