There have been many published studies aiming to identify temporal changesin river flow time series, most of which use monotonic trend tests such asthe Mann–Kendall test. Although robust to both the distribution of the dataand incomplete records, these tests have important limitations and provideno information as to whether a change in variability mirrors a change inmagnitude. This study develops a new method for detecting periods of changein a river flow time series, using temporally shifting variograms (TSVs) basedon applying variograms to moving windows in a time series and comparingthese to the long-term average variogram, which characterises the temporaldependence structure in the river flow time series. Variogram properties ineach moving window can also be related to potential meteorological drivers.The method is applied to 91 UK catchments which were chosen to have minimalanthropogenic influences and good quality data between 1980 and 2012inclusive. Each of the four variogram parameters (range, sill and twomeasures of semi-variance) characterise different aspects of theriver flow regime, and have a different relationship with the precipitationcharacteristics. Three variogram parameters (the sill and the two measuresof semi-variance) are related to variability (either day-to-day or over thetime series) and have the largest correlations with indicators describingthe magnitude and variability of precipitation. The fourth (the range) isdependent on the relationship between the river flow on successive days andis most correlated with the length of wet and dry periods. Two prominentperiods of change were identified: 1995–2001 and 2004–2012. The firstperiod of change is attributed to an increase in the magnitude of rainfallwhilst the second period is attributed to an increase in variability of therainfall. The study demonstrates that variograms have considerable potentialfor application in the detection and attribution of temporal variability andchange in hydrological systems.
展开▼
