Flood forecasting in the UK: what should we learn from the winter 2015 floods? Interview with Hannah Cloke and David Lavers
Contributed by Louise Arnal
In November 2014, a HEPEX post entitled “Flood forecasting in the UK: what should we learn from the Winter 2013/14 floods?” was written by Liz Stephens and Hannah Cloke (it can be found here). This post presented the forecast lessons learnt after the winter 2013/14 floods in the UK. Two years later, progress has been made in flood forecasting, early warning and emergency response, but some challenges remain.
In September 2016, the UK government Department for Environment, Food & Rural Affairs (Defra) published a National Flood Resilience Review (NFRR). This review aimed at:
- understanding the risks of river and coastal flooding from extreme weather over the next 10 years in the UK;
- assessing and finding ways to improve the resilience of key local infrastructure in the UK (i.e., energy, water, transport and communications);
- improving the response to flood incidents in the UK, focusing on the installation of new temporary flood defences.
Hannah Cloke (University of Reading) and David Lavers (ECMWF) were both sitting on the Scientific Advisory Group, whose role was to review and validate the science in the NFRR. They have kindly agreed to answer some questions about this review.
Louise Arnal: Why was the NFRR written?
Hannah Cloke: The NFRR was written in response to the Cumbria floods of winter 2015. This event has shown that there is a need to reassess the current risk of flooding in the country and find ways for the country to be better prepared for future flooding and extreme weather events. For this review, the Met Office and the Environment Agency (EA) have stress tested the EA’s models and flood risk maps (for river and coastal flooding), using extreme rainfall and tidal scenarios produced by the Met Office. This was really important because the EA have so far based their assessments of the risk of fluvial floods on historical records of river levels during previous floods, rather than on Met Office extreme rainfall projections.
Louise Arnal: How were those extreme rainfall scenarios generated?
David Lavers: Possible extreme rainfall scenarios were developed by considering “ensemble forecasts and projections” which gives a range of counter-factual worlds. The difference between these scenarios and what observed rainfall has already occurred was taken to be an indicator of how much worse the rainfall could be. The Met Office ran 11,000 months of weather simulations, from which they estimated that a ‘plausible’ uplift would be between 20-30% for winter monthly rainfall totals, although this figure is different for each region across England and Wales. These possible rainfall differences are similar to those calculated by the ECMWF who ran a similar experiment. The Met Office then applied these uplifts to simulations of recent extreme rainfall events (2km resolution every fifteen minutes). These new simulations were later used as input to the EA hydrological models.
Note from the review: ‘Plausible’ extreme tidal scenarios were produced by combining a recent storm surge with the highest recorded astronomical tide.
Louise Arnal: What hydrological models does the EA use for flood forecasting?
Hannah Cloke: The EA have around 2000 local detailed models, with a spatial resolution ranging from a few km to a whole catchment). These models combine three main components:
- Survey information: the river channel shape and surrounding landscape.
- A hydrological model.
- A hydraulic model: which enables flood extent and depth to be mapped.
The EA combines outputs from their local detailed models with less detailed broad scale modelling outputs and observed flood data to create the national Extreme Flood Outlines (EFO) maps. These maps show the extent of extreme floods (from rivers and sea) at any specific location in the UK, taking into account flood defences. On these maps, the outer boundary of an area is called the Extreme Flood Outline and shows the extent of a fluvial or tidal flood with a 0.1% chance of happening in any year at this location.
Louise Arnal: So, how were these maps used in this review?
David Lavers: A stress test was performed for the review by forcing several local detailed models from the EA with the extreme scenarios produced by the Met Office. This was done to see whether these floods would extend beyond the areas shown in the EFO maps. In modelling the predicted floods, a ‘worst case’ approach to other parameters of the model was adopted (e.g., prior soil saturation). For the review, the EA compared their current EFO maps with the extreme scenarios EFO maps for six case studies, four inland and two coastal areas. They observed that the flood extents and depths lay within, or very close to, the current EFOs. Figure 1 is an example of the EFO map for Oxford.
Louise Arnal: How confident can we be in those results?
David Lavers: The assessment of flood risk is dependent on observed records of river flows which typically go back only 30-40 years, the longer the timeseries, the more accurate the results. We advised the Met Office and the EA to extend their flood records by using information from historic sources, for example, by using old newspapers or photographs.
Hannah Cloke: Although we judged the EFO maps to have passed a reasonable stress test for this review, the extent of flooding is impossible to forecast precisely and the possibility of floods extending beyond the EFO cannot be excluded. A source of uncertainty in this review is for example changes in the catchment response over time, which were not accounted for, like a change in the capacity of the catchment to absorb water. We concluded that the results were indicative and that the statistical methods to reduce uncertainties in flood estimation should be developed further. This work should also be extended to surface water and groundwater flood risk.
Louise Arnal: One major theme of this review was the communication of flood risk. What do you think are the current challenges in the UK and what advice was given to Defra?
Hannah Cloke: According to a survey from the EA, “although nearly half the population surveyed in recent research reported being aware of a local flood risk, only 7% felt this risk applied to their own property.” There is clearly an awful lot of work left to do to make sure that everyone is aware of their flood risk and knows how to prepare for flooding (see EA’s #floodaware campaign). One of the most important conclusions about flood risk communication was that widespread scientific descriptions of a ‘1 in x year’ flood risk are confusing to the public and alternative descriptions must be developed.
Louise Arnal: In November 2016, just two months after the NFRR was published, the UK House of Commons Environment, Food and Rural Affairs Committee (EFRA; committee appointed by the House of Commons to examine the expenditure, administration and policy of the Defra) published a Future Flood Prevention Report. This report was written as a response to the NFRR, which according to EFRA provided only limited solutions, insufficient to tackle “fundamental structural problems”. What do they mean by that?
David Lavers: One main criticism that they make is the fact that the EA currently relies too much on preventing floods by constructing defences at the point of impact, for example, in town centres. Although the NFRR mentions that engineered hard defences are not the only part of the solution, alternative methods were not addressed in this review. According to the report by the EFRA committee, there should be more emphasis made on catchment measures, which can help prevent flood waters from the source and along the river path to the impact point.
Louise Arnal: What are catchment measures exactly? and how good are they?
Hannah Cloke: Catchment measures are flood management measures that can be used to reduce the risk of flooding across a river catchment. They include natural flood storage measures like afforestation to increase water infiltration upstream close to the source of runoff, a combination of natural flood storage measures and ‘soft engineering’ defences like dykes along the river path, and measures to increase the resilience of communities in settlements at the point of impact (see Figure 2). These measures were set up for many smaller catchments in response to the Pitt Review of the 2007 summer floods (an example is the Pontbren project). These measures have been trialled only on small catchments so far, where they have shown positive impact on flood risk alleviation. But there is limited evidence as to their effectiveness on larger and more lowland catchments and for extreme events such as the Cumbria floods. So the report recommends further tests of those methods on a wider set of catchments.
Louise Arnal: What are the next steps towards improving flood forecasting in the UK?
David Lavers: The report mentions the need for more accurate severe weather forecasts, both in time and space. Furthermore, although the NFRR have reduced some uncertainties in projecting the near-term impact on rainfall and flooding, more remain which should be tackled, for example, by extending the flood events time series. And last but not least, modelling and forecasting practices should move towards fully integrated flood risk modelling, from weather forecasting to impact assessment.
Louise Arnal: Thank you both for your time!
Where you can find the reports on which this interview was based: