Have trends changed over time? A study of UK peak flow data and sensitivity to observation period

Abstract

Abstract. Classical statistical methods for flood frequency estimation assume stationarity in the gauged data. However, recent focus on climate change and, within UK hydrology, severe floods in 2009 and 2015 has raised the profile of statistical analyses that include trends. This paper considers how parameter estimates for the generalised logistic distribution vary through time in the UK. The UK Benchmark Network (UKBN2) is used to allow focus on climate change separate from the effects of land-use change. We focus on the sensitivity of parameter estimates to adding data, through fixed-width moving window and fixed-start extending window approaches, and on whether parameter trends are more prominent in specific geographical regions. Under stationary assumptions, the addition of new data tends to further the convergence of parameters to some final value. However, addition of a single data point can vastly change non-stationary parameter estimates. Little spatial correlation is seen in the magnitude of trends in peak flow data, potentially due to the spatial clustering of catchments in the UKBN2. In many places, the ratio between the 50-year and 100-year flood is decreasing, whereas the ratio between the 2-year and 30-year flood is increasing, presenting as a flattening of the flood frequency curve.

Stationary vs non-stationary modelling of flood frequency distribution across northwest England

“…7(b). The importance of evaluating different windows of records was also emphasized by Griffin et al (2019).…”

The effect of depth‐duration‐frequency model recalibration on rainfall return period estimates

“…However, fitting a non‐stationary model requires assumptions to be made about how the effects of climate change translate into specific model parameterisation and can greatly increase the complexity of model fitting and risks of equifinality, which may lead to wildly different results when extrapolating. Even assuming stationarity, Griffin, Vesuviano, and Stewart (2019) highlight the large impact that one additional year of data collection can have on flood frequency analysis, and this finding is equally applicable to rainfall frequency analysis. Taken together with the emerging evidence on regional climate change, we demonstrate the importance of routinely updating the calibration of the FEH13 DDF model to ensure that rainfall DDF estimates always use the most up‐to‐date and reliable data.…”

The 2019/2020 floods in the UK: a hydrological appraisal

“…Of the 35 catchments with sufficient record lengths to provide robust trend assessments (at least 1969‐present) which also recorded one of their highest three AMAX on record, the vast majority (33 catchments, 94%) demonstrated a positive trend (12 of which were statistically significant at the 10% level). While any trend assessment is sensitive to the record length, as well as the start and end dates over which analysis is undertaken (Griffin et al ., 2019), 28 of the 35 catchments also exhibited positive trends over a shorter timeframe (1987‐present). Figure 6 illustrates these findings for two catchments that were particularly impacted by flooding in 2019/2020.…”