Non-linear analysis of geomagnetic time series from Etna volcano

Abstract. An intensive nonlinear analysis of geomagnetic time series from the magnetic network on Etna volcano was carried out to investigate the dynamical behavior of magnetic anomalies in volcanic areas. The short-term predictability of the geomagnetic time series was evaluated to establish a possible low-dimensional deterministic dynamics. We estimated the predictive ability of both a nonlinear forecasting technique and a global autoregressive model by comparing the prediction errors. Our findings highlight that volcanomagnetic signals are the result of complex processes that cannot easily be predicted. There is slight evidence based on nonlinear predictions, that the geomagnetic time series are to be governed by many variables, whose time evolution could be better regarded as arising from complex high dimensional processes.