Simple measures to slow down the outflow from river catchments can take the pressure off flood-prone towns. Computer models hold the key to getting it right.
While unprecedented rainfall in the UK has inundated parts of Cumbria, Lancashire, Yorkshire and Scotland this winter, the flood-prone town of Pickering in North Yorkshire has (so far) stayed dry.Pickering’s residents, it seems, were protected by relatively low-cost flood defences developed in a project involving Forest Research and Durham, Oxford and Newcastle universities, funded by Defra, Ryedale District Council, North Yorkshire County Council and the Flood Levy.
Instead of building a £10m concrete floodwall through the town centre, they spent around £2m on a carefully engineered concrete bund that could store up to 120,000 cubic metres of floodwater upstream of the town. In the catchments of Pickering Beck (draining through Pickering) and the neighbouring catchment of the River Seven (draining through the village of Sinnington) they planted 29 hectares of forest and installed 167 leaky dams of logs and nearly 200 smaller obstructions, made of heather, to slow the water flow into the main river of Pickering Beck.
While it is too early to conclude that the town escaped flooding because of these measures and not because rainfall was less severe than in other parts of the country, the approach clearly merits attention.
Designing this kind of distributed flood management scheme across a catchment has only become feasible with new computer models that can efficiently simulate water flow in the river catchment upstream of the location to be protected. This may be across large areas (hundreds or thousands of square kilometres) for a big catchment. The Pickering project team used a modelling tool called Overflow, developed at the University of Durham.
Such models have to take into account river dimensions and water resistance, water run-off from land and also whether it’s farmland, forest, roads, limestone, chalk and so on, all of which influence how quickly water flows.
The latest figures from the UK Met Office suggest there is a growing need to use such approaches to manage and model water in our landscape more intelligently. December 2015 was the wettest ever recorded in the UK and also the warmest, with an average temperature of 7.9°C (4.1degrees higher than the long-term average). The government’s 2012 Climate Change Risk Assessment said that climate change would significantly increase flood risk in the UK.
Big storms in Britain are a function of moisture in the atmosphere and the position of the circumpolar vortex or ‘jet stream’. This winter has been unusual. Normally, the jet stream becomes more vigorous, mobile and sometimes sinuous in winter. Whilst vigorous, the amplitude has been relatively flat and fixed over the North Atlantic, and this has caused the deep low-pressure systems to track across the UK and often to slow down.
Whether this arises from global warming or is part of a natural cycle (or both) is another matter. Commenting on the storms in the Guardian in January 2016, Myles Allen, professor of geosystem science at the University of Oxford, was blunt: “Normal weather, unchanged over generations, is a thing of the past. You are not meant to beat records by those margins and if you do so, just like in athletics, it is a sign something has changed... The physics is relatively simple: as we warm the atmosphere, the weather systems that move in from the Atlantic contain more moisture, so they dump more rain.”
When a storm causes a flood in a town like Pickering (which sits downstream of a steep valley running down from the North York Moors), the flood is the sum of waves from multiple streams feeding into a main river, explains Dr Nick Odoni of the University of Durham, who worked on the project with Durham colleague Stuart Lane, now professor of geomorphology at the University of Lausanne, Switzerland.
Placing debris dams and planting trees upstream of a location at flood risk may have the effect of ‘stretching’ or attenuating flood waves, causing a lower and later flow peak. This is a form of ‘natural flood management’ based on slowing the river flow. Yet modelling them on a computer over large areas is essential because a lower and later flood peak at one location may then become coincident with the flood wave from a sub-catchment downstream, thus increasing flood magnitude even further downstream at the location we are trying to protect.
Discovering what works best in a catchment can mean trying hundreds or thousands of combinations of interventions. Take, for example, a simple river with two branches (a Y shape). A debris dam that slows water down could be used across one, two or three (or none) of the reaches: that’s 2 to the power of 3 (eight) options to try. Adding a buffer strip of trees and/or vegetation to a floodplain increases that to 64 variations (4 to the power of 3). Trying out meanders in the reaches as a third option gives a total of 512 (8 to the power of 3) scenarios.
Pickering Beck has around 95 reaches, which (theoretically) would mean simulating 8 to the power of 95 scenarios, i.e. trillions of simulation runs if you were to try debris dams, buffer strips and meanders in every possible place and configuration. Using the Overflow tool, the researchers simulated adding buffer strips and debris dams around the catchment with only around 2000 runs.
Overflow combines a statistical treatment for runoff generation with the more tightly constrained short timescales of hydraulics. It also uses techniques like power sampling and fractional factorials to cut down the number of simulation runs. “It still needs a lot of computing capacity, but we can do it in an automated way over a weekend using a university supercomputing facility to batch the job,” explains Odoni.
The Pickering project calibrated the model with rainfall data and hydrographs (flow rate versus time measured at specific points in rivers or streams) from known flood events in June 2007 and a smaller one in November 2000. Using these, the researchers could then see if the interventions would slow floods during both scales of event.
They found that peak flow reductions with the natural landscape measures ranged from 3 to 0.8 cubic metres per second for the larger 2007 event, which would see reductions of flood volume between 99,000 and 19,000 cubic metres. The figures were lower for the small flood in 2000, but the simulations did show that at some sites the interventions would have marginally increased the flood risk. “Trying out multiple interventions gives engineers a steer as to where they should apply their hydraulic models in much greater detail,” Odoni adds. Hydraulic models use precise dimensions of the water channels and require the boundary conditions of each channel across the area described. They can take into account of water flow changes over seconds and even fractions of second.
How applicable is this approach elsewhere? Odoni thinks that managing water flow in upstream landscapes is going to be an important way forward. “It will mean growing a lot more trees as a start. We need spatially distributed modelling like Overflow so we can assess what to do where.”
Such interventions would not be instead of hard-engineered defences like floodwalls and dams, but they might allow far less costly solutions such as the bund at Pickering.
The problem with seeing this kind of scheme as a solution for some of the 2015 floods in places like York or Carlisle is that these events covered huge catchment areas of thousands of square kilometres, explains Lane. Even if it was technically feasible to model at this scale and for some subcatchments to be desynchronised for a given storm event, this method would only work if the rain were to move across the catchment in the same way in every storm and, unfortunately, “we can’t play God with rainfall,” says Lane.
Living with water
The Intergovernmental Panel on Climate Change’s most recent report (IPCC 5th Assessment, 2013) states that climate warming is unequivocal and extreme precipitation events are likely to be more intense and more frequent. Many regions including the UK may need to live with more water. That means taking a longer-term view of managing not just river flooding but off-field, storm-drain and groundwater flooding, and also drought.
“If the storms had hit the south of the UK where many places are built on or close to chalk, there would have been a huge amount of groundwater flooding that could have affected tens or hundreds of thousands of people,” notes Odoni. “Flood walls do nothing in those circumstances, so being able to model these effects in a coupled hydraulic/hydrological way is increasingly pressing.”
Indonesia has managed living with water for thousands of years by building houses on stilts, points out Lane, only half-joking. While he is not seriously suggesting stilts as the solution for Britain, building a more flood-resilient society is key. “You can do that individually by flood-proofing your property, or through a stronger planning system, and you may do it by some investment in defences as well as natural flood management, “ he says. “But some of these are politically challenging and sensitive issues and as a society we will need to expect less engineering and to be more intelligent.”
“For instance, people who wish to develop their basements and live in a flood risk area, should not be given planning permission, because society will eventually have to pay for it, whether through new flood defences or higher insurance premiums. Floodplains should be strictly kept as floodplains when evaluating new development.
“Insurance companies should not insist on like-for-like replacement, but allow for investments that make flooded houses less vulnerable if flooding happens again such as requiring waterproof plaster or moving electricity supplies and so on.”
Some countries already take a far more serious landscape management approach to flood alleviation and flood protection. “In Switzerland, floods aren’t just about damage to property, they are about loss of life, because we are looking at run-off from steep and dangerous Alpine catchments. In the town of Brig about a decade ago, 2–3m of sediment was deposited,” says Lane. “They have also been much more intelligent about vulnerability to flood risk. So they are very much stricter about where you can build.”
The next stage for the Pickering researchers is to apply to the Environment Agency for the local rainfall and stream gauge data collected during the storms this winter. With this dataset, the team will be in a position to asses whether the interventions at Pickering were what protected the town. “To do this, we would need to use precise hydraulic models of the bund and some of the natural flood management interventions and run the model with the latest data with and without the flood defences,” explains Odoni. “But the local knowledge from residents is that the bund is working. Photographs, taken at the bund site on Boxing Day, clearly show water backing up on the upstream side of the bund and starting to collect in the storage zone, just as intended during higher flow events”.
There is a growing argument that public engagement is essential to the revolution required in how we manage flooding and water resources in the UK, including drought. The Pickering project was founded on academics working with local people to identify new flood risk reduction solutions and to try them out by developing and using mathematical models together.
The science of flood risk is uncertain and there are no simple answers. Developing flood knowledge together is vital for the public to understand this uncertainty and hence the limits to different solutions. Part of increasing public involvement, says Odoni, may be to make flood and rainfall data more widely available to the public who fund its collection through their taxes.
But it will also require flood schemes to be developed with communities and not just for them.
Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.