Extraction of Nonlinear Trends in Time Series of Rainfall Using Singular Spectrum Analysis

Abstract

Characterization of nonlinear trends in time series of hydroclimatic variables exhibiting nonstationarity is necessary for more realistic projections of climate change and for optimal design of hydraulic structures. The present study was conducted to demonstrate the applicability of a novel Monte-Carlo-based singular spectrum analysis (SSA) to characterize nonlinear trends in historical time series of rainfall characteristics. Long-term (1960–2015) rainfall records for 17 gauges located in the Malaprabha River Basin, India, were used to analyze spatiotemporal variabilities of trends in rainfall totals and number of rainy days for annual and seasonal time periods. While the traditional Sen’s Slope and Mann–Kendall (MK) trend tests indicated statistically nonsignificant decreasing monotonic trends at most gauge stations, SSA revealed the existence of steep nonlinear trends and distinct change points in the direction of the trend over the period of record for both rainfall and rainy days. Results of this study demonstrate the potential for SSA to extract crucial information on the trajectories of nonlinear trends and change points in time series of hydroclimatic variables that exhibit nonstationarity.