Flood Risk Management in Cumbria: Events, Methods and Schemes (Postgraduate Summary)

Scope and sources

This summary synthesises five study materials on flood risk management (FRM) in Cumbria, north-west England: the Cumbria Floods Technical Report on the January 2005 floods (Environment Agency, Black & Veatch and Met Office, November 2006); an MSc dissertation using the CityCAT hydrodynamic model in Kendal (Newcastle University, 2019); Faulkner et al. (2020) on non-stationary flood frequency analysis (Journal of Flood Risk Management); the PERC report Flooding after Storm Desmond (Zurich and JBA Trust, August 2016); and the Kendal Flood Risk Management Scheme catchment map. Together they trace a sequence of extreme floods (2005, 2009, 2015), the scientific tools used to understand them, and the engineering and social responses.

1. The January 2005 Cumbria floods (Cumbria Floods Technical Report)

Meteorology

• Heavy rain was caused by a strong airflow of unusually warm, moist tropical air forced northwards ahead of an Atlantic cold front, with orographic enhancement over the Lake District, Howgills and Yorkshire Dales (a seeder-feeder mechanism) and strong frontal uplift and convection at the end of the storm.
• Rain fell mainly between midnight on 6/7 January and midday on 8 January 2005. Totals exceeded 200 mm at some gauges (e.g. about 213 mm in 36 hours at Honister; 206.8 mm at Wet Sleddale).
• The rarity came from the storm's duration rather than its intensity. The overall return period was about 100 years (1% annual chance); at some gauges it reached about 175 years.
• Hurricane-force winds, probably strengthened by a sting jet, caused widespread damage (over a million trees down) but had little or no effect on rainfall amounts.
• Tipping-bucket gauges probably under-recorded in the strong winds (by about 8%), and radar from Hameldon Hill gave poor quantitative rainfall over much of Cumbria because of distance and high ground.

Hydrology

• Soils were close to field capacity beforehand, so runoff was very high (runoff ratio about 73% for the Eden at Sheepmount).
• Flows on the Eden, Kent and Derwent were the highest on record. The Eden at Sheepmount (Carlisle) peaked at an estimated 1520 m³/s at 14:30 on 8 January, a return period of about 175–200 years; Carlisle levels were at least 1 m above any event since 1771.
• The upper Eden, Derwent and the Kent in Kendal peaked in the early hours of 8 January.
• Flooding often began with surface water as drainage networks were overwhelmed (e.g. Keswick, Carlisle), plus some groundwater flooding.
• The high tide in the Solway Firth had no effect on peak river levels in Carlisle. Haweswater did not spill until after the Lowther peak, which reduced the peak.
• Rating curves did not extend to such high flows, so peak flow estimates were revised significantly after the event.

Impacts

• More than 2,500 properties flooded (2,544 in the county table), about 75% in Carlisle (1,865 properties). Two elderly residents drowned in their homes in Warwick Road, Carlisle.
• Total damages were likely around £500 million; Carlisle City Council estimated £350 million in its area.
• More than 250,000 homes and businesses in Cumbria and north Lancashire lost power; the Willowholme substation (the Carlisle Bulk Supply Point) was under 1.5 m of water, cutting power to 60,000 properties.
• Carlisle police station flooded to 2.5 m: the first closure of a major police station in peacetime. The civic centre flooded, so Gold command moved to police headquarters in Penrith.
• About 90% of the county's schools were closed on the Monday. By May 2006, about 320 homes were still uninhabitable, and victims reported many health problems.
• In Kendal, 107 properties flooded (worst at Mintsfeet, about 60 properties). The 1970s central Kendal scheme coped well, though the Aynam Road defences were overtopped for the first time.

Warning and forecasting

• About 50% of properties in flood warning areas were not registered for the AVM warning service. Its successor, Floodline Warnings Direct, raised registration to up to 75% across Cumbria.
• Simple trigger models performed well. Correlation models under-predicted levels on the Eden and at Keswick. Of the transfer function models, only Kendal's performed well. The ISIS real-time Eden model under-predicted the peak, largely because stage-discharge relationships at upstream gauges under-estimated input flows.
• Key lesson: extreme floods can behave differently from more frequent ones, so models calibrated on ordinary events may fail.
• After 2005: new and improved flood warning areas, quadratic rather than linear correlations, ISIS improvements, and works in Carlisle, Penrith, Keswick, Appleby and Kendal (improved Mintsfeet embankments).

2. Storm Desmond, December 2015 (Flooding after Storm Desmond)

• Storm Desmond (4–6 December 2015) followed storms Abigail, Barney and Clodagh on already saturated ground. It was fed by an atmospheric river of moist subtropical air.
• Records: 341.4 mm in 24 hours at Honister Pass (UK record) and 405.0 mm in 48 hours at Thirlmere.
• Peak flows on the Eden, Lune and Tyne were each about 1,700 m³/s. The Eden at Carlisle reached 1680 m³/s (previous record 1516 m³/s in January 2005), and the Kent at Victoria Bridge 403 m³/s (previous 286 m³/s in 2005).
• Defences were overtopped in Carlisle (£38m scheme, 0.5% standard of protection), Keswick and Cockermouth, although there were no structural failures, and the EA estimated that defences protected 8,600 homes.
• Insured losses were about £520–662 million, and the ABI estimated economic losses of about £1.3 billion for Desmond, Eva and Frank.
• The report applies the 5C-4R resilience framework: five capitals (human, social, physical, natural, financial) and four Rs (robustness, redundancy, resourcefulness, rapidity).
• Key insights:
  • Residual risk always remains.
  • The "100-year flood" label is misleading: a 2% annual chance flood has about a 64% chance of occurring at least once in 50 years.
  • Collective memory is short.
  • Uptake of resilience grants was low.
  • The report recommends second and third lines of defence, alternatives to sandbags, and "building back better".
• Institutions discussed include:
  • the Environment Agency, Lead Local Flood Authorities under the Flood and Water Management Act 2010, the Cumbria Local Resilience Forum and the Civil Contingencies Act 2004;
  • Flood Re, a reinsurance scheme launched in April 2016 that excludes businesses and homes built after 2010.

3. Non-stationary flood frequency analysis (Faulkner et al., 2020)

• Standard UK flood estimation assumes stationarity. Milly et al. (2008) declared "stationarity is dead", while others argued it is "undead" or "immortal".
• Faulkner et al. tested 509 UK gauging stations (at least 40 years of record). Mann-Kendall found increasing trends at 117 stations (23%), and a non-stationary GEV was preferred at 166 (33%). There was an abrupt increase in the late 1990s, and trends were concentrated in wet north-western catchments (60% of Cumbrian stations had trends at p<0.05).
• Method: a GEV distribution fitted by maximum likelihood, with location (and optionally scale) varying linearly with a covariate (time), shape held constant, and models compared by AIC or likelihood ratio.
• In the North West case study (33 stations commissioned after December 2015), non-stationary models were preferred at 24 stations and gave present-day estimates up to 55% higher than stationary ones. At the Lune at Killington, including 2015/16 increased the non-stationary 1% AEP flow by 28% relative to the stationary estimate to 2014/15.
• Autumn and winter rainfall was the best covariate at four Cumbrian stations, but using rainfall creates a "double stochastic" problem.
• Recommendation: use non-stationary analysis to estimate present-day flows, alongside conventional methods where a trend is apparent, and apply climate change allowances separately. The method has no predictive power for the future.

4. Hydrodynamic modelling of Kendal with CityCAT (MSc dissertation)

• During Desmond, Kendal received 183 mm of rain in 48 hours, and about 2,150 properties were affected. The Kent exceeded the roughly 280 m³/s capacity of the 1970s Kendal Flood Relief Scheme (just over 400 m³/s at Victoria Bridge).
• CityCAT solves the full 2D shallow water equations with a finite volume, shock-capturing scheme and Green-Ampt infiltration, and removes buildings from the grid. The study used a 2 m LiDAR DEM (2.9 million cells) with inflows from the Bowston, Sprint Mill and Mint Bridge gauges.
• Validation using the Critical Success Index (CSI = A/(A+B+C)) gave 0.63 for CityCAT versus 0.55 for the EA flood map (weighted CSI 0.71). Simulated depths averaged about 32% above reported depths, and flood mechanisms occurred earlier in the model.
• Limitations: no storm drains (effectively a complete drainage failure scenario), no bridges, and LiDAR under-represents channel bathymetry.
• The dissertation discusses the Stock Beck culvert, Birds Park Reservoir (only about 1,000 m³ storage) and the fact that Phase 1 of the FRMS alone (£16m) gives limited benefit.

5. The Kendal Flood Risk Management Scheme (scheme map and CityCAT dissertation)

• Overall, the scheme aims to protect about 1,437 homes and 1,151 businesses to a 1-in-100 standard, at a cost of about £56m (CityCAT dissertation).
• Phase 1: about 6 km of linear defences through Kendal (average 1.1 m, maximum 2 m high, with 14 flood gates) and a 1 m³/s Stock Beck pumping station.
• Phase 2: linear defences at Burneside (0.6 km) and Staveley and Ings (0.7 km).
• Phase 3: upstream storage.
  • Kentmere Tarn: about 1.2 million m³, with pass-forward flow restricted to about 41 m³/s.
  • Kentrigg: about 1.7 million m³, with pass-forward flow of about 183 m³/s and embankments up to 6.5 m high.
  • A Stock Beck catchment drain up to 2 km long into the River Mint.
• The CityCAT dissertation notes there is no guarantee that all phases will be built.

6. Cross-cutting themes

• Repeated overtopping: defences built after 2005 and 2009 were overtopped in 2015, which raises questions about design standards under non-stationarity.
• Data limitations: rating curves at extreme flows, gauge under-reading, radar quality and short records all affect flood estimates.
• Integrated FRM: defences, storage, warning, insurance, recovery funding and community resilience all play a part, and residual risk never disappears.
• Climate change: the EA projects peak flow increases by the 2080s of about 30% (central estimate) and up to 70%; Otto et al. (2017) estimated that December 2015 rainfall was about 60% more likely because of climate change.