 |
| Rain radar for 2015 Cumbria floods |
Last week saw devastating floods in many parts of Cumbria and elsewhere in northern England - the latest in a series of big flood events to hit the region. So questions are naturally being asked about why this keeps happening. Probably the most important things to remember here are that hydrology is difficult, and that we should therefore expect weird things to happen.
What's so hard about hydrology? For flood protection, we're typically trying to build defences to a standard of "1 in 200 AEP". The AEP - Annual Exceedance Probability - is the probability that a flow or water level will be topped in any given year. So we're looking at rare events, and we simply don't have long enough records of hydrological events to estimate rare flows, rainfall, river levels etc accurately.
The longest records of river flows in the UK are at Kingston on the Thames and the Lee at Feildes Weir in Hertfordshire - both with ~130 years of record. The Eden at Carlisle has almost 50 years of record, meaning our estimates of extreme flows and probabilities are inherently uncertain - and when the next big flood comes along, our understanding of risk could change significantly.
Take flows on the Eden in Carlisle as an example. At the time, the 2005 event (with a peak flow of 1520 m3/s) was calculated (using the standard methods and data available at the time) as having an AEP of 1 in 150 - meaning that we'd expect a flow greater than this with a probability of 0.67% each year. This analysis used an approach called pooling - we combine data from other similar catchments to make up for our short record. But using pooling means we're mixing in data from other rivers which may not behave in the same way - what if Carlisle is unique? What if everywhere is unique?
I've estimated the AEP for a 1520 m3/s flow, using flow records up to, but not including, the 2005 event; including the 2005 event; and including an estimated flow for the 2015 event (I've used 1600 m3/s as the measured flow isn't available yet). Results are in the flood-frequency plots below - before the event, we would think a 2005 type event would happen once every 370 years. After the 2015 event, we think this will occur every 70 years - the probability for this event has increased by a factor of 5!
If we accept a figure of 1 in 70 as a reasonable estimate for the AEP of a 2005/2015 type event, then the probability of seeing two of these events in a 10 year period is approximately 1% - rare certainly, but not impossible. And given the number of gauges in the UK (around 1500) we might actually expect this kind of thing to happen at least somewhere every year or so.
There's also been a lot of talk about the new record rainfall (341mm over 24 hours) seen at Honister - a few km from Seathwaite where the previous record (314mm) was set in 2009. These are huge values - more like we'd expect in tropical countries like the Philippines. What's more, they're off the scale for the standard method for estimating extreme rainfall, the FEH DDF method, which only covers events no rarer than an AEP of 1 in 1000. It's difficult to unpick the significance of this - partly because rainfall data isn't openly available in the same way that river flow data is. But a quick back-of-an envelope calculation that if the AEP for 300mm is around 1 in 1000, then the probability of seeing 300mm twice in 10 years is around 1 in 20,000. So either this is a true freak event, or we need to reassess our extreme rainfall estimates.
So the science of hydrology says that any estimate of rare events is very uncertain - but as engineers we still need to be able to manage the flood risk. We need to acknowledge the uncertainty and build it into our decision making (we do this already, but perhaps not enough). And perhaps we also need to shift focus away from hard defences, whose performance depends so critically on estimating rare events, to other more robust measures (such as building resilient properties) that work no matter what the probability of flooding is.
