Temporal variations of runoff and sediment load in the upper Yellow River, China

•This study analyzed the intensive fluctuations of runoff and sediment load at the upper Yellow River, which has not yet been studied clearly before.•Temporal variations in runoff and sediment load were discerned in terms of changing trend, abrupt change year, and multi-periods from the full measured datasets of all hydrological stations.•Impact of human activities on the hydrological variations was evaluated based on comparison of the measured and the reconstructed time series. Operations of amount of large and medium reservoirs have remarkably changed runoff and sediment loads, and consequently affected the fluvial processes intensively on the upper Yellow River. In this study, time series of runoff and sediment transport rates of four major hydrological stations on the upper Yellow River were used to analyze temporal and spatial variations of runoff and sediment loads. Mann-Kendall test and wavelet analysis were applied to analyze the datasets. Temporal variations in runoff and sediment loads were discerned in terms of changing trend, abrupt change year, and multi-periods from the time series. Additionally, a time series of runoffs was reconstructed based on a modeling scenario removing impact of reservoir impoundment and water diversion to indicate the status without human disturbances. Comparison of the measured and reconstructed datasets was conducted to evaluate the impact of reservoir impoundment and water diversion on hydrological variations. The measured runoff and sediment loads showed obvious decreasing trends due to impoundment and water diversion. However, the reconstructed runoffs regarded as only affected by climate change showed no significant changing trend at a 95% confidence level, confirming that, instead of climate change, human activities played the most important roles in variations. The measured datasets indicated that the reservoir construction and operation caused an abrupt change in runoff in 1986, while the reconstructed datasets indicated another abrupt change in 1995. Wavelet transform analysis distinguished the multi-periodical features of the time series. The first, second, and third dominant periodicities of the measured runoffs were 25–27, 16, and 8 years, respectively, whereas the first three dominant periodicities of the reconstructed runoffs were 26–31, 12, and 7 years, respectively. It was indicated that the high-frequency periodicities of runoffs decreased while the low-frequency periodicities increased, owing to res