Using groundwater model results to estimate baseflow contribution to flooding in Hull

DSC_0507

While we* were developing the groundwater resource model for the Yorkshire Chalk aquifer, we were asked to provide estimates of extreme baseflows to the River Hull headwaters. Most of the material in this post was delivered, on behalf of ESI Ltd, by Dr Steve Buss at the British Hydrological Society Pennines Group conference on Groundwater Flooding, in Leeds, on 1 October 2013.

There is consensus in the groundwater industry that regional-scale groundwater models can’t (or, at least, shouldn’t) be used for groundwater flood risk prediction. But the Environment Agency has invested considerably in development of these models for the management of water resources. So are there some aspects of groundwater flooding that can be analysed with these models, to give added value to this investment?

Villages in the upper catchment of the River Hull, East Yorkshire, and along the Gyspey Race, suffer from flooding when groundwater levels rise and springs emerge upstream of the normal watercourse. Burton Fleming, Beverley, Kilham, Cherry Burton and Driffield all experienced groundwater flooding in early 2012. And groundwater flow from the aquifer, when it reaches Kingston-upon-Hull, contributes to flooding in the city, especially in Summer 2007.

Yorkshire aquifer location

Before this work started, the flood model for Kingston-upon-Hull used a single estimate of groundwater discharge to the River Hull, for all flood return periods. This was based on flow estimates for Hempholme Lock, a flow gauge about 17 km north of Hull (right), that does not measure all the baseflow discharge from the aquifer and especially misses flow from the Wolds to the west of the River Hull. Our calibrated groundwater model allows us to fill in that missing data, and make predictions of groundwater discharge for different return periods.

Modelled time series

The model is well-calibrated, especially for high flows at Snakeholme Lock (upstream of Hempholme Lock). We had flow data from 1989 to 2008, but the model allowed us to project this time series back to 1970. By applying an extreme value distribution to annual maximum flows we identified theoretical flows that corresponded to discrete return periods (1 in 200 years, to 1 in 2, right). Flows in November 2000 were most similar to the theoretical 1 in 200 year flows; January 1994 had about in 10 year flows, and April 1981 had about 1 in 5 year flows. Now these modelled flows could be used to construct accretion profiles along the River Hull downstream of the flow gauges, for each event (below).

discharge accretion

So there clearly is accretion downstream of Hempholme Lock: during periods of very high baseflow there is about 30-35% more baseflow contribution at Kingston-upon-Hull than at Hempholme Lock (we did not model drain flows on the Holderness Plain or runoff from impermeable surfaces). We also checked the mitigating effect of groundwater abstraction in the catchment by running a naturalised mode (‘nat’ in the figure above). Clearly, this was not large.

When coupled with an appropriate flood model, these results can be used to predict the effect of varying groundwater flows on flooding downstream.

* Main contributors to the model were Paul Daily and Victoria Price.  Model development was funded by the Environment Agency.