Detectability of changes in hydrological records / Possibilité de détecter les changements dans les chroniques hydrologiques

Abstract

Abstract One of the important practical problems in the analysis of long time series of hydrological records is the evaluation of detectability of trends. This is not a trivial task, especially where the change is weak, while the natural variability is considerable (e.g. in the case of river flow records). The results of a study of a run-up effect allow experts to assess how strong a change (gradual trend or abrupt jump) must be and how long it must take in order to be detected by statistical tests. A set of generated, trend-free data mimicking river flow records, has been contaminated by the controlled addition of a gradual change or an abrupt jump. Analysis of detectability of change for different conditions of natural variability of the river flow process and for different properties of the contaminating component (its amplitude, intensity and run-up time) has been made, and the performance of different tests has been compared. The results are of broad applicability in the search for a climate change signature in hydrological data. Tests are not able to detect weak changes or changes which have not lasted long, but this cannot be interpreted as a demonstration of absence of a change. With the enhanced climate change, the changes of hydrological processes may be stronger and last longer, so that the likelihood of change detection may grow.