A Generic Procedure for Simultaneous Estimation of Monotone Trends and Seasonal Patterns in Time Series of Environmental Data

Abstract

Most of the currently used methods for time series analysis have been developed for official statistics and economic forecasting. This article is devoted to new non-parametric procedures that take into account some very widespread features of environmental data. In particular, we propose an algorithm for simultaneous estimation of monotone temporal trends and smooth seasonal patterns. Similar algorithms may also be used to combine the extraction of trend and seasonal components with removal of natural fluctuations that can be attributed to meteorological and other physical forcings of the environmental system under consideration.