2023 – Groundwater in a digital landscape
We live in an increasingly digitised world, with a continuing rapid pace of change. 2023’s Ineson event on 15th November 2023 followed immediately after the Geological Society conference “Digital Geoscience: Unleashing the Power of Data and Technology in Earth Sciences” and looked at how this digital revolution affects the hydrogeological profession.
The aim was to address questions such as:
- What is the potential from new technological developments, what is available now, and how does this affect how we carry out our business?
- How quickly are things changing, and how will ‘game-changing’ or ‘disruptive’ technologies such as Artificial Intelligence (AI) or Machine Learning affect how we ‘do’ hydrogeology?
- Conversely, what are the risks and drawbacks behind these new technologies, and what are the barriers to achieving their potential?
The 2023 Ineson Lecture was delivered by Holger Kessler, a widely-known geoscientist at the British Geological Survey, discussing “30 years of digital transformation in the Geosciences – is the glass half full yet?”. He was worked closely with Government departments and bodies (including the Geospatial Commission, and Government Office for Science) on digital geospatial strategy for the environmental, water resource, engineering and infrastructure sectors (including the UK National Underground Asset Register, and Future of the Subsurface Foresight Programme).
Keynote speaker was Marian Scott, Professor of Environmental Statistics at the University of Glasgow, discussing “Digital earth systems: drawing intelligence from disparate data streams and models”. She is a fellow of the Royal Statistical Society and the Royal Society of Edinburgh and was awarded an OBE for services to science. Her main research interests are in the use of statistics and data analytics in understanding environmental systems.
Other speakers at the event explored a broad range of new technologies, including: digital data representation and visualisation (e.g. digital geology models), use of artificial Intelligence (AI) or Machine Learning in areas including combined remote sensing and ground data, integration of live sensors and digital twins in relation to different aspects of hydrogeology and groundwater management.
Speaker abstracts:
Marian Scott (University of Glasgow); Digital earth systems- drawing intelligence from disparate data streams and models. A digital earth is a term that is increasingly being used, simply put it may be described as” an interactive replica of the earth system”. If we imagine the digital earth as a model, then this model will need to be constantly updated with data. We may also think of the digital earth concept more broadly as a series of models (digital twins) for sub-systems. Developments in new sensor technology and innovations in monitoring are transforming our ability to observe the components of the earth system, and the digital earth concept provides us with a framework to develop our understanding. This lecture will consider where the challenges and opportunities lie.
Eleanor Blyth (CEH); Moving towards near-real-time soil moisture and groundwater data products for the UK.
Ivana Barisin (Terra Motus); Case study linking satellite aperture radar (SAR) and ground-truthed data to investigate soil moisture as a tool to monitor groundwater in chalk headwaters. Instruments on board of remote sensing platforms (satellites, airborne) have the ability to observe soil moisture. However, there are many obstacles to use remote sensing technologies to accurately measure the soil moisture’s concentrations in the soil. In the past, a number of approaches have been attempted with some limiting success where different types of data and different algorithms were used to improve on the detail and accuracy of this products. Although there are some global models available, they fail to provide a necessary detail for most applications. In this project we have used Machine Learning model that involves a number of datasets that can contribute to measuring the soil moisture remotely and also attempted to find the correlation with groundwater levels for the site of Gypsey Race. In this talk we are going to review the state-of-art remote sensing soil moisture measurement techniques and share our personal findings from this project.
Thomas Booth (WSP); Groundwater Numerical Modelling and Visualization Workflows in International Mining Projects. The presentation will describe state-of-the-art groundwater numerical modelling and visualization of workflows used by WSP Mining Teams across the world. Topics to be addressed include: 1) the use of Leapfrog Geo to develop geological models and to visualize and integrate geological, geophysical, and hydrogeological data and engineering designs; 2) the development of coupled surface water-groundwater models using HydroGeoSphere; 3) the development of Discrete Fracture Network models using FracMan; 4) development of coupled hydromechanical models using FEFLOW – FLAC3D; 5) development of multiphase heat and reactive transport models using PFLOTRAN; and 6) development of stochastic groundwater and water balance models using MODFLOW and GoldSim. Special attention will be put on the visualization component as a way to improve technical quality, efficiency, and interaction with clients and other disciplines.
Gerry Baker (Arup); The use of open-source 3D visualisation and groundwater modelling software. The presentation highlights the use of open source software (iMod, from Deltares) to help communicate complex geology and groundwater flow processes to clients and stakeholders. Two case studies are presented, the first a city wide scale groundwater model for Dublin city and the second a campus style development in the Netherlands. For the Dublin project Arup developed python script to extract 13,000 records from the Geological Survey Ireland (GSI) geotechnical database and import these into the iMod model which allowed for improved conceptualisation of the variability of thickness of the made ground and subsoil in the city. The groundwater model results were visualised with iMod 3D particle tracking tools which helped communicate groundwater flow pathways to the client. In the second example, the client had made a significant investment in ground investigation (over 400 CPTs), but this presented its own
challenge in how to make sense of such a significant database and communicate the complexity. Arup developed a soil classification based on the CPT results which allowed each CPT readings (at 0.1m intervals) to be assigned a soil type and iMod was then used to develop a voxel (Volume-Pixel) model at 1m intervals with inbuilt kriging tools to form the ground model. The resulting visualisations allowed for geotechnical engineers and hydrogeologists to develop more robust conceptual models and also provide the client with a graphical interface which readily communicated the complexity in the system and the value of the investment in the ground investigation.
Holger Kessler (BGS); 30 years of digital transformation in the Geosciences – is the glass half full yet? Holger will talk about his 30 year career in mapping and modelling the subsurface in order to make better environmental and societal decisions culminating with his recent roles at the heart of the UK Government. He will describe some of the successes and failures along the journey of digital transformation of the geoscience sector and attempt to give some outlooks to the future.
Poster abstracts:
AI needs the naked eye: Remote detection of rural water supplies in Ethiopia. Jack Brickell, Donald John MacAllister and Matthew Arran (BGS). Across Sub-Saharan Africa, limited knowledge of rural water supply infrastructure can prohibit access to safe water and undermine community resilience. Better understanding of rural water supply assets at a national level is required. This often relies on large-scale infrastructure asset surveys conducted every 5-10 years. However, recent work by BGS shows that routine monitoring of rural water supplies could reduce downtime, increase functionality, and build community resilience to external shocks (e.g.: droughts).
In this project, a neural network with U-NET architecture was designed to automatically detect handpumps in Ethiopia from aerial imagery. We compiled a large training dataset of freely available Google Earth aerial images and an existing rural waterpoint database, collected by UNICEF during the 2015-16 El Niño drought. Several challenges were encountered during this study. Manually collating the datasets was extremely time-consuming, and aerial image quality is highly variable with a bias towards Northern Ethiopia. The dataset is based on a historical survey and, in some cases, handpumps have been upgraded to solar pumps or decommissioned. Inaccuracies in waterpoint coordinates can make it time-consuming to visually locate the handpump’s position.
Although AI has potential to revolutionise how we work in geoscience, this project highlights current limitations with existing datasets. Planned development of the pilot study includes different water point types (solar pumps and motorised pumps) and expansion into other Sub-Saharan countries.
The Water AI partnership: recycling energy industry data to find new water resources. Marina Flores (University of Oxford). The ‘Water Al’ project was created by a multidisciplinary team within the Department of Earth Sciences and the Bodleian Library at the University of Oxford in partnership with Wood Mackenzie. The main motivation is to identify new freshwater resources in unconventional settings, such as under the seafloor, or in deep unexplored subsurface water accumulations on land. To achieve this, we repurposed deep and offshore Oil & Gas data for water resources, using a combination of Machine Learning (ML), Geographic Information Systems (GIS) and Virtual Reality (VR). We describe our methods, preliminary results and show how our engagement with industry enabled both researchers and external partners to share perspectives and lessons learned. The project results have potential for societal impact, by identifying additional, much needed water resources in drought-stricken countries worldwide. Also, they have potential for industry impact, by finding ways of reusing the existing global well and infrastructure databases traditionally destined to the Oil & Gas sector. In this project, we’ve combined salinity values from our industry partner and academic databases, to do a spatial analysis that includes a variety of parameters. GIS is used for the collection, storage, analysis, and visualisation of geographic or spatial data. Secondly, different Machine Learning algorithms were tested for predicting the distribution of fresh groundwater at global and basin scale. This allows us to identify and predict the spatial distribution of
potential unconventional groundwater resources. In addition, we are developing a VR pilot project that seeks to find uses of this technology for subsurface visualisation of geological models and regional aquifers, with focus on East Africa.
Training hydrogeologists in a digital landscape. Bethan Flynn (Environment Agency). The Environment Agency (EA) is a regulatory body which employs hydrogeologists across England. Our training offer is being converted to blended learning, where learners use self-led learning, virtual sessions and face to face sessions to cover hydrogeology topics. This mix of learning offers a more flexible approach to develop a career pathway in geoscience. Following a recent trailblazer, the new Geoscientist (integrated degree) apprenticeship is due to start in 2024.
iArsenic – Instant arsenic screening of hand pump tubewells in Bangladesh. Mo Hoque1, Jacek Kopecky1, Kazi Matin Ahmed2, and Adrian Butler3 (1University of Portsmouth, 2University of Dhaka, 3Imperial College London). Across the globe, waterborne pathogenic diseases pose significant health challenges, prompting action from governments and NGOs. In response to this widespread issue, Bangladesh, as a notable example, witnessed the promotion and installation of millions of low-cost mechanical hand-pump wells. This well-intended initiative, however, had unintended consequences, inadvertently exposing vast swathes of the population to arsenic, a naturally occurring element in the groundwater. Out of over 10 million wells, approximately 6 million remain untested, largely due to the need for specially trained personnel and chemical test kits—resources often out of reach for many in rural areas. Consequently, more than 20 million Bangladeshis remain at risk of arsenic exposure, illustrating the complex nature of addressing global water health issues.
Leveraging pre-existing research, we employed a dataset of around 2 million tubewells to create a web-based application. This tool estimates arsenic levels in untested or forgotten wells, relying on two key indicators: a geochemical cue, evident as staining on the tubewell platform, and a 3D location identifier, generated from the well’s depth and street address, as provided by the user. With these indicators, our model confidently predicts arsenic risks associated with a given tubewell, boasting an accuracy rate of over 80% in its alpha version evaluations.
The widespread mobile internet coverage in Bangladesh presents a powerful opportunity both for gathering data and for disseminating information on arsenic pollution directly to the users. We are actively working towards the full development and deployment of the iArsenic web application. Once fully operational, this app will not only allow users to make an informed decision about the continued use of their tubewell for drinking (thus potentially sparing them from ingesting arsenic), but will also assist the government in saving on screening costs and adopting risk-based mitigation strategies.
Application of Digital Twins and Water Treatment Optimisation. Hugh Lawrence (BIM Manager, Sweco). The concept of Digital Twins is already being applied in the water industry by Sweco Ltd.. This poster explores what value that bring and what data they can utilise, with a view to raising further thoughts and discussions on how the hydrogeological industry could also utilise Digital Twins. It aims to evaluate what synergies are there between the digital twins used in water treatment design and operation which are also beneficial to hydrogeological applications.
There are technologies that are on the market that have water industry application, however that do not necessarily utilise hydrogeological data from site instrumentation.
The different approaches used for operation and maintenance evaluation and optimisation between water treatment site specific design, environmental design and hydrogeological design are some way apart, and linking these all together would be of benefit to the owner operator, and then ultimately to customers.
Analysis of the technologies that Sweco use has shown that there are clear benefits from the use of digital twins, primarily for Owners or Operators, however there is also scope for Digital Twins to be used by consultants during design phases of projects to optimise design solutions early on in the design of the asset even before it has been built and in operation. There are technologies that are on the market that have water industry application, however that do not necessarily utilise hydrogeological data from site instrumentation.
Many of the benefits listed on this poster are likely to be key performance indicators for Water Companies and OFWAT, and therefore by merging the data collected from all of the different aspects of water design, we can further enhance digital twins to provide better value.
Techniques for the digitalisation of conceptual models. Kieran Topen & Joss Wilson (Sweco UK).
Geologists in both industry and academic fields depend on conceptual models to aid in the visualisation of information. Examples include geological facies, surface geomorphology, hydrogeological surfaces and water quality. These must be easily understood by both technical specialists and non-specialists. Traditionally, conceptual models represented by cross- or fence-sections relied on a hand-drawn process, but more recently digitalisation has changed the way we work entirely. We aim to highlight the main methods we can use, namely drawing software packages such as Inkscape, and 3D geological modelling software like Leapfrog, to produce the best geological conceptual imagery possible. We discuss the benefits and limitations of each method and how you can select and use the best one for your deliverables. As well as provide examples of how we have used the software in project work to produce both conceptual models but also a range of other visualisations. Then finally, highlight the benefit of taking a combined approach to software in general to improve deliverable outputs.
