| Summary: | To understand the nature and cause of natural climate variability, it is important to possess an accurate estimate of past climate forcings. Direct measurements that are reliable only exist for the past few decades. Therefore knowledge of prior variations has to be established based on indirect information derived from natural archives. The challenge has always been to find a strict objective method that can identify volcanic events and offer sound amplitude estimates in these noisy records. An automatic procedure is introduced here to estimate the magnitude of strong, but short-lived, volcanic signals from a suite of polar ice core series. Rather than treating records from individual ice cores separately and then averaging their respective magnitudes, our extraction algorithm jointly handles multiple time series to identify their common, but hidden, volcanic pulses. The statistical procedure is based on a multivariate multi-state space model. Exploiting the joint fluctuations, it provides an accurate estimator of the timing, peak magnitude and duration of individual pulse-like deposition events within a set of different series. This ensures a more effective separation of the real signals from spurious noise that can occur in any individual time series, and thus a higher sensitivity to identify smaller scale events. At the same time, it provides a measure of confidence through the posterior probability for each pulse-like event, indicating how well a pulse can be recognized against the background noise. The flexibility and robustness of our approach, as well as important underlying assumptions and remaining limitations, are discussed by applying our method to first simulated and then real world ice core time series. Signal extraction; Multiprocess Kalman filter; Volcanic eruptions; Pulse-like signals; Climate forcing;
|
|---|