Extraction of Nonlinear Rainfall Trends Using Singular Spectrum Analysis

Abstract

Rainfall plays a crucial role in the socioeconomic development of a country. Knowledge of both the amount of rainfall and its pattern of distribution are equally important for proper management of water resources systems. In the present study, trends of two rainfall series from different locations having different time periods and time steps have been extracted using the singular spectrum analysis (SSA) method. Data analyzed include monthly data of England and Wales precipitation (EWP) from 1766 to 2002 in which no periodic component is prevailing and daily rainfall data of Koyna watershed, Maharashtra, India from 1961 to 2009, which shows a strong periodic component of 365 days. Method of periodogram analysis has been used in order to select the components corresponding to trend in the grouping stage of SSA. The Mann–Kendall (MK) test is also used to detect trends in EWP monthly series and the performance of SSA and MK test is compared. The result showed that the MK test could detect the presence of a positive or negative trend at a significant level, whereas the proposed SSA method could extract the nonlinear trend present in the series along with its shape. Trends extracted from the England and Wales precipitation are compared with a previously published EWP trend extraction study. The comparison shows that the method of SSA in trend extraction could extract nonlinear trends along with its shape whereas the previous study extracted linear trends. The EWP monthly rainfall series showed an increasing trend during the winter season and a decreasing trend during summer. The trend extracted for the Koyna series has very small values (almost constant), implying that the rainfall series is almost stationary. The study proves the applicability of SSA for extracting nonlinear trends that provide more insight into the observed time series.