Testing Hydrologic Time Series for Stationarity

Abstract

The main objective of this study is to examine if hydrologic monthly series are stationary. A segmentation algorithm is applied so that nonstationary series are identified and partitioned into stationary segments. Four sets of hydrologic data are analyzed, including monthly streamflow, temperature, precipitation, and Palmer’s drought severity index series. The first-order differenced standardized monthly series are also analyzed, in addition to the standardized monthly series. The results indicate that more series of monthly streamflow and Palmer’s drought severity index are identified as nonstationary than stationary, while more series of monthly temperature and precipitation are stationary. The change points are commonly observed during two periods, one between 1960 and 1970, and the other between 1930 and 1940. In general, standardized hydrologic monthly series, either differenced or not, are nonstationary.