Long-range correlations identified in time-series of volcano seismicity during dome-forming eruptions using detrended fluctuation analysis

Understanding the underlying structure of data from volcano monitoring is essential to identify precursors to changes in eruptive activity and to comprehend volcanic processes. However, effective analysis of longer-term trends in these signals is challenging as volcanic data are not necessarily statistically stationary or linear, particularly those from lava dome-forming volcanoes, which are commonly characterised by pulsatory eruptive activity. Here, we use detrended fluctuation analysis (DFA), a statistical technique previously applied to nonstationary data, to identify long-range (slowly decaying, e.g. power-law) correlations in a number of time-series of volcano seismicity recorded during the recent dome-forming eruptions of Volcلn de Colima, Mexico, and Soufrière Hills Volcano, Montserrat. For all the time-series analysed, correlation strength varies through time and/or on different timescales; in some cases, this variation is periodic, seasonal, and/or related to activity. These results may provide new insights into eruptive processes and possibly further constrain the generation mechanisms of a number of the volcano-seismic event classes analysed. Furthermore, the correlation properties of real-time seismic measurements are shown (retrospectively) to contain information valuable to real-time volcano monitoring that is not identifiable by conventional analysis techniques. This study therefore demonstrates that long-range correlation analysis may be useful for extracting additional information from monitoring data at dome-forming or similar volcanoes.