Temporal trend analysis of synthetic and real hydroclimate time series and impacts of long term quasi-periodic components on trend tests

Abstract

Studies of climate and hydro-meteorological time series have found the presence of decadal and inter-decadal oscillations of quasi-periodic components (broader band signals) as part of long term natural variations in the data. If the oscillation of the quasi-periodic component is prominent, the impacts on the Mann-Kendall (M-K) and Thiel-Sen (T-S) trend tests could lead to a biased estimate and affect the prediction of future trends. This study performed the Thiel-Sen test on simulated time series of periodic components with and without trends as well as real time series of climate and river discharge data with long records in the Canadian Prairies. The results of tests suggest that the T-S test is sensitive to the presence of oscillation of long term quasi-periodic components. Data record length, magnitude of cyclic components and phase are the three most important factors affecting the M-K type of tests. The conclusion derived for the T-S test in this study can also be applied to the M-K test.