Groundwater flood hazard maps to improve community resilience to groundwater flooding.
Flood Modeller (v6.0) was used to route (2D) groundwater emergence flow over the ground surface.
Flood Modeller’s GPU-enabled 2D solver reduced runtimes by 80%, with further significant speeds ups expected if using an A40 and A100 NIVIDA card.
Creation of a web-based groundwater flood warning service.
What our client say
“Communities nationwide lack effective support due to insufficient monitoring and modelling of groundwater flood risk. These new groundwater hazard maps represent a significant step forward for groundwater flood management. This is long overdue, and Project Groundwater will advocate for this level of mapping to be produced and made freely available on a national scale, as these will be.”
Jed Ramsay Project Leader Project Groundwater, Buckinghamshire Council
Project Groundwater supports communities to be more resilient to groundwater flooding. Led by Buckinghamshire Council, the project is working with communities in 9 high-risk flood areas of the Chiltern Hills and Berkshire Downs, which spans six local authority areas. It is funded by Defra as part of the £150 million Flood and Coastal Resilience Innovation programme, which is managed by the Environment Agency to develop and test new approaches to resilience tailored to local communities.
Jacobs are a senior supplier on the project, leading the monitoring, modelling and alerts workstreams. Already work has been undertaken on groundwater flood modelling, with innovative solutions having been developed to enable the issue of timely alerts of groundwater flooding and help the design of resilience and place-making activities specifically suited to managing groundwater flooding.
Flood hazard mapping is widely available for rivers, the sea, surface water and reservoirs but, despite causing widespread flooding, groundwater flood maps are not available to a similar standard. Hence contributing to under-representation of its impact on flooding as well as limited management of the risk.
Initially MODFLOW was used to estimate the rate that groundwater emerges. Flood Modeller was then used to route (2D) the groundwater fluxes and produce flood depth results. The downstream limits of the Flood Modeller 2D domains were aligned to river gauging stations, facilitating model calibration. The gauging stations were checked to have reliable daily flow records for the dates of maximum groundwater level represented.
Six separate hydraulic models were setup, covering both the Chilterns Chalk and Permeable Superficial Deposits areas. For the Chalk areas, the resolution of the 2D domain was 4m for the smaller models (Pang and Kimpton) and 6m for the larger ones (Chalfont St Peter). The domains used resampled Integrated Height Model (IHM) data from the Environment Agency which has a horizontal resolution of 2m. Model roughness for each model was derived from Ordnance Survey MasterMap data.
To simplify the modelling, GIS analysis was first undertaken to identify groundwater emergence and then inflows of less than 0.005 m3/s were aggregated and applied as a single direct inflow boundary. Downstream boundaries assumed free flow conditions based on the slope of the local watercourses.
Preliminary simulations were run to check the position of the flow paths identified in the GIS analysis and rectify false obstructions to the flow in the IHM data caused by structures such as bridges or watercourses not properly defined and needing to be ‘burnt’ into the 2D domain.
Flood Modeller was run until a steady state between inflow and outflow was achieved. Longer, narrower catchments like the Misbourne lead to long travel times of the emerged groundwater to reach this steady state at the downstream boundary. These long travel times are exacerbated by the numerous very small emergence values filling and spilling over topographic features. Using the CPU processor, this resulted in lengthy run times.
To optimise run times, we tested artificially filling small topographic voids. Whilst this did not result in any significant change in the run times (CPU), filling larger bodies of water present in the IHM data as depressions did result in significant time savings to reach steady state.
Real-world run times depend on the size of the model, cell size and number of groundwater inflow points. By using Flood Modeller’s GPU-enabled 2D solver, run times could be cut by up to 80%, compared to running on the CPU. For example, the Kimpton model which took 62 hours on the CPU (Intel(R) Xeon(R) CPU E5-1620 v3), ran in 3.8 hours on the GPU (GeForce RTX 2080 Ti). We anticipate that run times would be less than an hour if using an NVIDA A100 GPU card.
To fully understand flood risk and manage it appropriately, groundwater flood mapping should be considered in the same way other sources of flooding already are. Combining, subsurface and surface hazard, is the most effective way to provide a comprehensive picture of flood risk.
Project Groundwater, with the aim of giving actionable information to communities, will see the development of a new web-based groundwater flood warning service using a groundwater map ‘library’. Working closely with communities to achieve local buy-in, raise awareness and build resilience, the mapping will also provide evidence and confidence to Risk Management Authorities to undertake flood risk assessments, develop business cases for management actions and design schemes. Project Groundwater’s vision “to forever transform how communities prepare and respond to groundwater flooding” heavily relies on robust, freely available, and relatable mapping.