This study emphasizes the differences to be found in using a variety of statistical techniques. None of the analyses gave identical results, although a broad picture
can be inferred by comparing results from a variety of techniques. It is difficult to ensure the quality of sets of data collected over long periods of times as methods may have changed, but not been documented within the database, over the long time-scale over which data were collected. An example is phosphate. Water samples were only collected once a day, always from the same place (where there was access to the water from, e.g. a bridge), and usually in the morning. DO, and likely other variables, will not be at concentrations representative of an entire day at that time, as they will be influenced by biological activity. This emphasizes a general problem of using existing databases that were not developed with clear aims, as found also by Sliva and Williams [27]. Nevertheless, such long runs of data on rivers are rare and their analysis is essential in
providing invaluable insights into the changes that are undoubtedly occurring in these ecosystems, as well as assisting in the formulation of testable hypotheses to explain the causes of change.