This paper reviews the results of a number of studies that have investigated streamflow data for the existence of trend. These studies provide evidence that trends in various, but not all, streamflow regimes are occurring at rates that are higher than one might attribute to chance alone. Results of different studies using different approaches were compared and were shown, at times, to have dramatic differences. These differences might potentially be due to pre-conditioning of data prior to trend detection in attempts to minimize the impacts of serial correlation on testing procedures. It was also evident that patterns of trend can vary over small spatial scales and that a relatively high-density network is required to effectively comprehend trend and how it might be altering across an area. A global network of streamflow sites representing pristine or stable conditions is needed to assess patterns of change. Selection criteria for sites within such a network are provided, and it is highlighted that local knowledge is required to perform this selection.

This content is only available as a PDF.