Editorial: Advanced Time Series Analysis in Geosciences

A time series is an ordered sequence of data indexed by time. In other words, it is a sequence of discrete-time data, usually obtained at equally spaced points in time. Time series analysis is the attempt of extracting meaningful characteristics and statistical information from data organized in chronological order. Nowadays, there are numerous types of data analysis approaches available for time series which are suitable for different purposes: diagnosing past behavior, prediction and forecasting, curve fitting, interpolation and extrapolation, classification and clustering, segmentation and decomposition, frequency characterization, etc. The theoretical advances in time series analysis started early at the beginning of the last century with new developments in the field of stochastic processes. The first actual application of autoregressive models to time series can be identified in the work of Yule (1927) and Walker (1931). But it is since the pioneering book “Time Series Analysis” by Box and Jenkins in the 1970s (Box and Jenkins, 1970), that many lines of study in time series analysis have been developed. Today we are witnessing a rapid increase in quantity, quality and importance of time series data in Earth Sciences. Across its vast number of subdisciplines, the massive production of data, e.g., through the growth of continuous monitoring networks and the availability of abundant remote sensing data, is making increasingly important the use of analysis tools capable of synthesizing information contained in large time series. To deal with the increasing amount of available data in an automated way, the first emerging approaches of machine learning in time series analysis date back to the early 1980s (Nielsen, 2019). At present, although classical methods are still dominant, machine learning is rapidly emerging as a valid alternative approach to time series analysis, finding effectiveness especially in multivariate time series.