A global-scale investigation of trends in annual maximum streamflow

This study investigates the presence of trends in annual maximum daily streamflow data from the Global Runoff Data Centre database, which holds records of 9213 stations across the globe. The records were divided into three reference datasets representing different compromises between spatial coverage and minimum record length, followed by further filtering based on continent, Koppen-Weiger climate classification, presence of dams, forest cover changes and catchment size. Trends were evaluated using the Mann-Kendall nonparametric trend test at the 10% significance level, combined with a field significance test. The analysis found substantial differences between reference datasets in terms of the specific stations that exhibited significant increasing or decreasing trends, showing the need for careful construction of statistical methods. The results were more consistent at the continental scale, with decreasing trends for a large number of stations in western North America and the data-covered regions of Australia, and increasing trends in parts of Europe, eastern North America, parts of South America and southern Africa. Interestingly, neither the presence of dams nor changes in forest cover had a large effect on the trend results, but the catchment size was important, as catchments exhibiting increasing (decreasing) trends tended to be smaller (larger). Finally, there were more stations with significant decreasing trends than significant increasing trends across all the datasets analysed, indicating that limited evidence exists for the hypothesis that flood hazard is increasing when averaged across the data-covered regions of the globe. (C) 2017 Elsevier B.V. All rights reserved.