Abstract: Studying the evolution trends of river runoff is of great significance for hydrological forecasting and rational allocation of water resources. In view of the multi-time scales and non-linear trend aliasing in runoff sequences, this paper presents an application of extreme-point symmetric mode decomposition (ESMD) to analysis of the spatiotemporal evolution of runoffs for the upper reach of the Yangtze River. Using ESMD, we decompose runoff sequences into various modes, including the annual, monthly, and daily runoff sequences for eight hydrological stations on the mainstream and tributaries of the reach. Then, we identify large-scale cyclic and non-linear trends using state components and trend residuals; obtain the sequences’ time periods, variation trends, and abrupt changes though examining the cycle diagrams, trend residuals, and time frequency analysis; and compare with previous methods. The results show that temporally, two types of dominant cycles are observable: annual cycles of 6 months, 1 year, 2 years, 4 years, and 22-23 years; and inter-annual cycles of 2 years, 6 years, and 10-11 years. Spatially, all the stations show a decreasing trend in the annual and inter-annual runoffs, except for the Zhutuo station featured with an increasing trend. The abrupt change patterns at the eight stations are not the same. This study demonstrates ESMD can excavate more hydrological characteristics of runoff through identifying the cycles, trends and mutations, opening up new ideas for spatiotemporal evolution analysis of river runoff.

Key words: extreme-point symmetric mode decomposition, spatiotemporal evolution, upper reaches of Yangtze River, non-stationary sequence, multi-time scales