The impacts of climate variability and change on groundwater resources in the UK

What are the factors that determine how sensitive groundwater resources are to climate variability and change?  How can process-based understanding and statistical analysis and machine learning be combined to assess the resilience of groundwater resources on a broad scale?

Background

Groundwater provides essential water supplies in Britain, in particular in areas in the south and east of the country where surface water resources are scarce. With growing attention to the resilience of water supplies in Britain, we need to understand better how groundwater resources respond to climatic variability and will respond to climate change in the future. This requires development of broad-scale understanding of the sensitivity of groundwater resources to climate variability and climate change.

Aims of the Project

To develop methodology and models to predict groundwater response to climatic variability and change on a broad scale (i.e. for groundwater resources in England and Wales)

Methods to be used

Statistical analysis and machine learning from groundwater data.

Physical characterisation of aquifers (based on existing secondary data sources)

Groundwater modelling using groundwater models, both existing (e.g. MODFLOW) and underdevelopment (e.g. HydroJULES)

New model development e.g. using Python programming

Project Description

Groundwater provides essential water supplies in Britain, in particular in areas in the south and east of the country where surface water resources are scarce. With growing attention to the resilience of water supplies in Britain, we need to understand better how groundwater resources respond to climatic variability and will respond to climate change in the future, as well as to increasing demand for water. A broad-scale, reduced complexity groundwater model, which could be used in groundwater yield assessment in given climate scenarios would be an extremely useful tool. Such a model does not yet exist, and constructing one will involve engaging with challenging scientific questions and statistical methodology. Methods from data science and machine learning may prove to be valuable in dealing with complex heterogenous datasets. The aim of the proposed research is to use a combination of data science methods and process-based insights to develop an efficient broad-scale model of groundwater yield for water supplies in a changing climate.

Groundwater levels are often modelled with distributed groundwater models e.g. MODFLOW.  Modelling is typically undertaken at the regional scale and can often result in lengthy runtimes.  There are a few large-scale groundwater models in England, but no model exists for all of England and Wales – and even if it did exist, it would be a rather unwieldy tool for assessing climate risks to groundwater yields, which requires the capacity to examine a large range of possible climate scenarios and associated uncertainties in climate downscaling and surface/groundwater hydrology.

There are also extensive datasets of groundwater withdrawals and abstraction licences in England and Wales, which record groundwater use and nominal limits on groundwater abstraction. These datasets contain valuable information about groundwater response to past climate variability and to groundwater abstractions. However, disentangling these various effects from historical observations is complex.

We believe it should be feasible to combine insights from physically based groundwater models with new methods from data science to develop a broad-scale understanding of potential groundwater response to climatic variability and change, and antecedent human water withdrawals. We expect that this would be based on a simplified representation of groundwater response (perhaps as a series of non-linear stores, as is typically used in conceptual hydrological models). Such an approach could be tuned to reproduce the response of more complex groundwater models and/or observations. Parameterisation will depend upon understanding of geology on a range of scales. We will be particularly interested in the spatial dependence between neighbouring boreholes, which will be investigated with a combination of spatial statistics and, where available, physically based groundwater modelling.

Subject to satisfactory progress, we are also interested in the relationship between groundwater levels and surface aquatic habitats (e.g. wetlands and streams). Ideally, we would like to develop a simplified representation of the response of these systems to changing groundwater levels, again using a combination of statistical analysis and physically based groundwater modelling.

The model (or models) that is developed will form a component assessment of the resilience of national water supplies to climate change and future demand. This national water resource systems model has been developed in the University of Oxford and is being used by the Environment Agency in England for strategic assessment of water resources, but so far the assumptions regarding groundwater yield are relatively crude. The proposed research will therefore fill an essential gap in our capacity to assess the resilience of the nation’s water supplies.

Scientific challenges will include:

  • identifying representations of groundwater dynamics that are appropriate for given hydrogeological settings;
  • statistical analysis of groundwater withdrawals and groundwater response;
  • construction of statistical or process-based emulator models of groundwater response;
  • estimating the effects of uncertainty in climatic, land surface and groundwater processes on borehole yields;
  • analysing the relationship between groundwater levels and surface aquatic habitats (e.g. wetlands and streams);
  • integration of this understanding in broad-scale water resources modelling to assess the impacts of climate change on the resilience of national water supplies.

The project will be conducted in collaboration with the British Geological Survey (BGS) and will involve collaboration with partners in the Environment Agency and the water industry. The will be the opportunity for a placement at BGS Wallingford and visits to BGS Keyworth to present results to groundwater modellers. BGS will support presentations at national and international conferences. Access will also be provided to BGS groundwater level and borehole databases and data on aquifer properties, as well as access to existing BGS groundwater models (about sixty lumped groundwater models across England and Wales). The student will be eligible to take BGS training courses in hydrogeology, statistics, numerical analysis, scientific writing etc.

Specialist skills required by the student

A student with a general quantitative background (e.g. physics, maths, engineering, quantitative environmental science) will be able to acquire the necessary statistical, modelling and programming skills.

A background in earth sciences or physical geography is desirable but not essential.

If you are interested in this project, please contact Jim Hall jim.hall@eci.ox.ac.uk