In the past few months, Richard, a postdoctoral researcher in our Bangor team, has been revisiting some analysis completed during his PhD with Dŵr Uisce. This work aims to improve our understanding of changes in the characteristics of high flow events in Wales under future climate change. It is important to understand such events, as very large streamflows can results in inundation and flooding, bringing economic and social costs, implications, and distress.
The analysis is based on the results of an extensive round of hydrological modelling completed for five catchments in Wales (Clwyd, Conwy, Dyfi, Teifi, Tywi) under a worst-case future climate change scenario (Representative Concentration Pathway 8.5 [RCP8.5]). Future climate scenarios were taken from the 2018 UK Climate Projections (UKCP18) data from the UK Met Office’s Hadley Centre, the most recent projections available. These were implemented at a daily timestep into the Soil and Water Assessment Tool (SWAT) hydrological model, in order to project future streamflows up to 2080.
Previous work focussed on analysis of change in the number of days per year and season where streamflow is greater than the 95th percentile value of the whole dataset, as well as change in seasonal and annual 1-day maximum flows. While these particular factors do give an indication of the future character of high flow days (frequency and severity respectively), it is not possible to take account of or characterise changes in prolonged high flow events, such as event severity and length. The graph in Figure 1 demonstrates the factors analysed in the original research, using an example dataset.
Figure 1. Example of previous analysis factors.
In order to better characterise high flow events observed in future streamflow projections, as opposed to just high flow days as previously studied, it is first necessary to separate one event from another, so that each event is hydrological independent. To do this, the method of baseflow separation was adopted, whereby the digital filter method is first applied to the streamflow timeseries to separate the baseflow (shown by the red line in Figure 2) from the overall streamflow (shown by the blue line). Following this, individual events with at least one day of flow above the dataset 95th percentile value can be identified, with the length of these being defined as starting when streamflow rises above baseflow levels and ending when the two are equal again (highlighted in Figure 2). This method therefore allows for the identification of multi-peak events, and marks a distinction from the previous analysis by not conflating an increased number of high flow days, with an increased number of high flow events, as the former maybe a product of longer and more severe events, rather than a larger number of events. By identifying individual events, it is therefore possible to gain a better understanding of how high flow events are changing. Changes in the frequency of high flow events, for example, can be calculated by totalling the number of individual events each year and season, and changes in the duration of events can be computed by taking the mean number of days between event start and end. Changes in the severity of events can also be studied, in two ways in fact, first, by calculation of the number of days per event where streamflow peaks above the dataset 95th percentile, and second, as in the previous analysis, 1-day maximum flow. Figure 2 provides a visualisation of the factors analysed, following the new methodology.
Figure 2. Example of the new analysis factors.
Figure 3. Seasonal and annual Mann-Kendall trend analysis results for Welsh study catchments.
As can be seen, the new analysis has the potential to provide a much clearer picture of changes in the severity, frequency, and duration of future high flow events. Indeed, the results of a Mann-Kendall trend analysis of this work (Figure 3) suggest that over the study period (2021-2080) several changes are projected, with varying patterns seen both between seasons and catchments. One of the clearest trends, with most agreement between catchments, is an increase in the severity (maximum flows) and frequency (number of high flow events) of high flow events in the winter and spring seasons. The least change is seen in the mean high flow event duration, with it generally decreasing across all seasons and annually, but only to a relatively small degree, and with the least statistical significance of all of the factors presented. As can be seen in all three factors presented, seasonal change is generally more substantial and significant than annual change, with large seasonal shifts often cancelling each other out in the annual picture. No statistically significant trends were found in the analysis of number of peaks above the 95th percentile per event, these results are therefore not included in Figure 3.
The results of this analysis have important implications for a variety of sectors, indeed anyone who abstracts from river systems or relies on their flow for other means could be impacted, as well as communities or individuals at risk of flooding. In terms of water resource use, hydropower operators in particular could be affected by the results presented, with two of the three catchments displaying declines in annual maximum flows. In addition, the increases seen in spring in terms of maximum flow volumes, and number of events, could also be worrisome for flood planners, as this has important implication for mitigation plans and measures. The impacts of changing high flow and flood events have far reaching consequences, it is for this reason that getting the best characterisation of future changes is important and is why this analysis has been strengthened.
We are grateful to Professor Neil Macdonald of the University of Liverpool for his suggestions for strengthening the analysis conducted. The improved analysis has been included in an article submitted to the Hydrological Sciences Journal and greatly enhances the results of this paper. Look out for news of this publication later this year, both here on our website, as well as on the Dŵr Uisce Twitter feed, @Dwr_Uisce.
