Following last winter’s flooding events the Government asked the Met Office to estimate the potential for record-breaking rainfall in the UK over the next 10 years. This work was part of the National Flood Resilience Review (NFRR).
The Met Office used an innovative approach which combined our expertise in both weather and climate science to estimate the scale of how much worse the next record breaking rainfall event is likely to be.
Using a new approach we assessed modelled weather patterns from our state of the art climate model, which generated over 11,000 monthly rainfall scenarios for 6 large regions covering England and Wales
This data included several hundred extreme rainfall events, which are meteorologically possible, but lie outside what has been experienced since our observational records began. From this data, the probability of extreme events like these was then estimated.
The results suggest there is a 1% probability in any winter of experiencing 15-35% more monthly rainfall than has been seen in the observations to date. Winter monthly rainfall totals could plausibly be 20% higher than recent past extremes in some parts of the country and in other areas up to 30% higher than recent past extremes. There is also around a 10% chance in any given year of existing monthly rainfall records, over any of the large regions, being matched or broken.
Stress Testing Flood Models
In order to translate these possible increases in regional rainfall into flood risk, we worked closely with the Environment Agency to identify large-scale river flooding events. These focused primarily on recent extreme flooding, so that any impact of climate change to date was accounted for. We also considered the effect which extreme river flooding, combined with an extreme tidal storm surge, would have.
Detailed rainfall accumulations are important for estimating hydrological response, but are very hard to measure: rain gauges only provide spot measurements and radar can have difficulty estimating some types of rainfall. However rainfall estimates produced by the Met Office UK kilometre-scale weather forecast model provided rainfall data on fine time (every 15 minutes) and space (every 2 km) scales for each of the case studies.
The rainfall data for each test case were increased by a certain percentage in order to simulate a plausible extreme case. We combined the evidence from the climate model, with expert judgement drawn from our understanding of model limitations and a detailed analysis of past rainfall records, to estimate rainfall increase for each case study. These ranged from 20% for North-West England to 30% for cases in Southern England. The estimated rainfall change represents a substantial increase in the volume of water entering river catchments. It can therefore be regarded as a robust test of the extreme flood outline.
These extreme rainfall data were passed to the Environment Agency, allowing them to feed these into their flood models and compute the likely flooding extent that would ensue. The results of these stress tests are presented in the NFRR report.
This new method, in which climate model simulations are used to explore the potential for meteorological ‘black swans’, is recognised as an innovative approach to estimating future environmental risks, such as heavy rainfall. These environmental risks are associated with weather and climate variability, as well as climate change.
Plans are now being considered for the development of a more integrated flood risk modelling approach which would allow simulations to be run linking meteorology, hydrology and flooding. This would make it easier and quicker in future to assess the probability of given levels of flooding, helping identify impacts and evaluate flood management measures.
“The rainfall data for each test case were increased by a certain percentage in order to simulate a plausible extreme case”.
I wouldn’t have thought that inserting fudge factors like this was particularly scientific, currently you are having problems forecasting where a developing low will track in just three days. I would expect that if you make the fudge factor 10% you would likely see a resulting 10% increase in the number of extreme events.
I would just like to add a further comment about when “your observational records began”, and that is (as far as I know) they began along time before 1910. Wouldn’t it be sensible to extend your gridded rainfall data series back as far back as possible by digitising the detailed British Rainfall Organisation data that you have? There surely would be extreme rainfall events that could make any further reports more accurate?
(“A fudge factor is an ad hoc quantity introduced into a calculation, formula or model in order to make it fit observations or expectations”)
I just wondered if there was some sort of problem with my comment or will I have to wait till Monday before its moderated?
Monday it is.