Hello everyone My presentation will be on rainfall forecast alarms for high priority rapid response catchments. My name is Oliver Pollard. I have over 20 years hydrological experience with the Environment Agency and have spent the last 15 years developing operational flood forecasting models in South West England. 1
Flash flood events on rapid response catchments can present an extreme danger to life because of the suddenness in the rise in water level, the flow velocity and debris. This iconic image is of the serious flash flood event at Boscastle in August 2004. It is amazing that no one died in this flood; it would have been worse if it happened at night. This desire to minimise the loss of life risk has strongly motivated me to develop a way to improve our detection and forecasting capability in these catchments where it is very difficult to provide a flood warning. Nationally, there are nearly 300 high risk rapid response catchments. South West Region is particularly vulnerable from serious flash flooding as it has a relatively high proportion of steep fast responding catchments. 2
What we have learnt from Boscastle and other other similar floods is that we need to improve both the forecasting and detection of flash flood events. This was again identified in the recent Environment Agency report of the flooding in Wales during summer 2012. John Curtin, head of the Environment Agency s Incident Management team, made a recommendation that the technique for linking radar rainfall products with data held on rapid response catchments within the National Flood Forecasting System represents best practice and could now be rolled out to all parts of the Environment Agency. 3
It is also recognised externally, that we need to improve both the detection and forecasting of flash flood events. After the severe and rapid flooding in Cornwall in 2010, the Prime Minster issued a challenge to share more information to communities at risk of flash flooding that can fall outside the traditional flood warning service. He said the warnings work according to the river levels whereas actually we ought to maybe be giving people warning when we re expecting excessive rainfall 4
South West Region has the highest proportion of extreme rainfall events (>200mm) in the UK. To compound this issue there has been gap in forecasting and detection of flash floods from extreme rainfall. This graph of extreme rainfall includes a critical threshold associated with serious flash flooding and underpins the rainfall forecast alarm technique. It is based on the UK Flood Studies Report 100 year return period rainfall depth duration line. More information: A study of twentieth century extreme rainfall events in the United Kingdom with implications for forecasting by W. Hand et al, Meteorol. Appl. 11, 15 31 (2004). Available from internet: www.onlinelibrary.wiley.com/doi/10.1017/s1350482703001117/pdf Presentation An Overview of Extreme Rainfall Events from the British Rainfall Archive 1866 1968 by Rodda et al, Jan 2008, available from the internet: www.badc.nerc.ac.uk/data/free/oxjan08 RoddaHarvey.pdf Forecasting extreme rainfall leading to flash flood events in the UK by Collier et al, available from the internet: https://ams.confex.com/ams/pdfpapers/47398.pdf (all accessed on 20 th Nov 2013) 5
The rainfall thresholds 80mm in 1 hour to 120mm in 6 hours are based on the 100 year return period line in the previous slide. The rainfall forecast alarms used in South West Region also take account of catchment state, using the assumption that the lower rainfall thresholds on wet catchments can generate the same impact as higher thresholds on dry catchments. The catchment is defined as wet if the catchment wetness index (CWI) > 125mm (defined in the UK Flood Estimation Handbook). In one hour for example, 40mm rather than 80mm of rain is assumed to be required to generate a flash flood event on a previously wet catchment; this is based on data from one flash flood event in Cornwall in November 2010. The rainfall alarm thresholds are the same for every rapid response catchment due in part to a lack of data correlating extreme rainfall to the damage caused by flash flood events. The use of a restricted set of extreme rainfall thresholds means that it will be easier for the National flood forecasting centre in England and Wales to assess the accuracy of the 6 hour radar rainfall forecast. 6
The rainfall forecast used operationally by the Environment Agency detects organised bands of rain such as frontal rainfall in the right place; an example of this is shown in the red area. This data product is unreliable for accurately detecting the location of small scale features such as embedded convection or thunderstorms often responsible for producing flash floods on rapid response catchments; an example of this is shown in the green area. The Met Office solution to this problem is to ensure that any small scale rainfall llforecast data is presented probabilistically. As the Environment Agency Regional duty teams do not use probabilistic rainfall data an innovative solution was required that made best use of the expected rainfall forecast product. 7
(top left image ) The technique is using the available rainfall data but in a different way. We are using radar rainfall data because it provides coverage everywhere. We have changed the data feed to provide us with the catchment maximum rainfall based on 2km grid 6 hour radar rainfall forecasts rather than the usual catchment th taverage as this better detects t localised li intense rainfall. This image, taken from the National Flood Forecasting System, shows radar rainfall near the Dawlish Warren high priority rapid response catchment; note that this catchment is shown in dark grey and is only 8 km2. Also note the proximity of intense rain which is coloured white and pink. (central image) We know that there are some limitations with the radar rainfall forecasts. Even with the recent improvements in the resolution of the data there is a limitation to getting the rain in the right place. As a result we apply a 5km buffer around the catchment to improve the detection rate and overcome this limitation. This image shows the extent of the buffered Dawlish Warren rapid response catchment in dark grey. (bottom right image) When the radar rainfall forecast exceeds the critical threshold mentioned earlier, the National flood forecasting system generates an alarm via the regional telemetry system to the Environment Agency s monitoring & forecasting duty officer. This image shows an alert summary table for high priority rapid response catchments; the red lines show when the critical threshold will be exceeded in the next 6 hours and the yellow lines show when the threshold is nearly crossed. 8
The rainfall forecast alarms technique has been trialled since May 2012 in South West Region. It has identified four out the five severe floods in rapid response catchments in 2012. The one miss was due to the radar at that time underestimating the rain rate. The example shown is at Dawlish Water in August 2012. The river is usually a small stream on the right of the wall in the picture but here covers the road as well. This was a result of 100 mm of rainfall with 70mm in 2 hours recorded at the rain gauge. Houses were evacuated at Kingsteignton as a result. 9
The Environment Agency s monitoring and forecasting duty officers (M&FDO) have an important role in reducing the false alarm rate of the rainfall forecast alarms. The duty officers have procedures to contact the National flood forecasting centre (England & Wales) to check the accuracy of the rainfall forecast; they use a wider range of rainfall forecast products than available to the Environment Agency s duty officers. The National flood forecasting system has been configured to allow the quick comparison of the radar rainfall with raingauge rainfall. This is important because the radar data initialises the 6 hour rainfall forecast product. In this example, the grid radar and raingauge derived rainfall are overlain on a map display of the high priority rapid response catchments. Note that the heaviest rainfall is shown in purple and red. There is wide spatial variability in the accuracy of the weather radar with it overestimating the rain by 50% at one raingauge and being in good agreement at the other selected raingauge. 10
As part of the trial in South West Region, the rainfall forecast alarms have been provided to a number of the Cornwall Flood communities from 2010 to trigger actions in their community flood plans. As more community flood plans are produced around the region we encourage the use of the rainfall forecast alarms to trigger their effective action. The Environment Agency in South West Region has strong links to the national rapid response catchment programme so that other Regions are aware of the trial, as part of a suite of triggers that can be offered to communities at risk of severe and rapid flooding. The technique has been peer reviewed by other Environment Agency Flood forecasting and Flood resilience teams across the country and has been identified as national best practice. 11
In summary It is an innovative technique that has filled a gap in our capability it has not been done beforebutis but reworking existing information. It is helping us learn our lessons from previous severe and rapid floods. It has been recognised as national best practice and is being adopted nationally. It was developed in partnership with the Environment Agency s flood resilence teams and the National flood forecasting centre. The community benefits are that it provides an appropriate trigger for their community flood plans and may help reduce the risk to life. There are a number of improvements in the pipeline, such as using high resolution probabilistic rainfall forecast products from the Met Office. 12
Further reading A digital archive of extreme rainfalls in the British Isles from 1866 to 1968 based on British Rainfall by Rodda et al, Weather, March 2009, Vol 64, No 3 www.maxlittle.net/publications/rodda_weather_09.pdf l / df (accessed on 20 th Nov 2013) Realism of Rainfall in a Very High Resolution Regional Climate Model by E Kendon et al, (2012), Journal of Climate, 25 (17); http://centaur.reading.ac.uk/31222/ (accessed on 20 th Nov 2013) Presentations from a joint BHS and ICE South West seminar on Flood Forecasting for Small Catchments, (17 th June 2011); internet links (accessed on 20 th Nov 2013) are appended: Flood Forecasting Challenges arising from the Cornwall Flash Flood Event of 17 Nov 2010 Mark Russell, Flood Forecasting Team Leader, Environment Agency SW Rapid Response Catchments Probabilistic Forecasting Options Murray Dale, Hydrometeorologist, Halcrow Group Flood Forecasting for Small Catchments NE experience Paul Wass, Hydrologist, JBA Probabilistic Flood Forecasting for Small Catchments using the G2G model Dr Steve Cole, Senior Hydrological Modeller, CEH Wallingford Addressing Rapid Response Catchments through land management techniques Nigel Hester, Projects & Grant Manager, National Trust Contact oliver.pollard@environment agency.gov.uk for more information 13
Develop a National GIS database of extreme rainfall events & catchment state; cross check against flash flood damage reports. Quantify the spatial accuracy of the 6 hour radar forecast product for different types of extreme rainfall and lead time in order to fine tune the size of the catchment buffer. Identify catchment specific rainfall forecast thresholds where there is local knowledge of flash flood events and special risk factors such as soil compaction. Base the catmax rainfall alarms on hourly maximum rainfall grid data rather than 15 minute maximum data which is more poorly correlated with the event total. Develop rainfall forecast alarms based on high resolution (2km grid) probabilistic (24 members) UKV7 rainfall forecast data used by the National flood forecasting centre (England & Wales). The alarm would be triggered when x out 24 members exceed the rainfall depth duration threshold. Develop a merged radar and raingauge rainfall data product to improve the initialisation of high resolution of rainfall forecast data. Collaborate with UK Met Office to assess the frequency and location of extreme rain events in Southern England using a 20 year simulation dataset of high resolution rain forecast data. 14