School of Mathematics & Statistics

Analytics for Digital Earth - Workshop 3

Date:  26th October 2022

Time: 10am

Location: ARC/Online

Speakers: The subject for our second hybrid workshop was water with around 30 people attending in person and online. The first speaker was Prof Andrew Tyler (University of Stirling) who spoke about the digital revolution in water and how we can move to delivering net zero through digital technology. He gave a number of examples of new sensors being developed and the increasing use of earth observation. Next, we had Prof Claire Miller (University of Glasgow) and Dr Wayne Jones (Shell) gave a joint presentation around some recent digital water projects at Glasgow and the longstanding collaboration on the development of GWSDAT (groundwater monitoring). Finally, Prof Lindsay Beevers (University of Edinburgh) talked about the growing flood and drought research agenda, and the analytic tools being developed especially to quantify and visualise uncertainty. In addition, we had six lightning talks given by researchers from the James Hutton Institute and the University of Glasgow, focusing on different statistical questions and techniques in environmental studies, such as monitoring of biodiversity and groundwater quality, casual modelling, spatiotemporal modelling, data fusion and Bayesian optimization.  

Discussion Format: There were two breakout discussion sessions, with the group split into two in-person groups and one online group for each session. The first session looked at barriers to creating a digital twin, and the second session followed up on this by inviting each group to select one aspect of water and build a digital twin on it.  

Lightening Talk: Miriam Glendell spoke about causal reasoning for system-based decision making to address water related challenges. Bayesian Belief Networks are highlighted as powerful decision support tools for adaptive river basin management based on ‘smart data’. Yuan Liu presented her work on Bayesian Optimisation on complex constrained domain. The method has been applied to detect the hotspots of chlorophyll concentration in the Aral Sea.  Craig Wilkie talked about spatiotemporal modelling of water quality in the Ramganga river in India. He presented the statistical work on interpolating and validating the new satellite data of the highly polluted Ramganga river in northern India. Jafet Belmont Osuna spoke about Citizen Science and biodiversity monitoring programs in freshwaters. Zhaoyuan Zou presented the data fusion for the remote sensing reflectance data observed by two sensors at lake Garda. Peter Radvanyi demonstrated the Influence Analysis for Groundwater Monitoring Network Optimisation.  

Session 1: The discussion in session 1 focused on barriers to creating digital twins, with topics including: 

• What data are require / can be collected? 
• What are the technology gaps? 
• What are the analytics gaps? 
• How much does uncertainty focus in decisions? 
• How can we improve how we monitor / design monitoring networks? 
• What are the skills / training gaps? 

Discussion focus points: 

Challenges and/or considerations when building a digital twin: 

• Buy in to the resources needed and benefits of a digital twin are key. 
• Need to be very clear about the purpose of the twin, what questions would it address, recognising that some decisions can be made without a digital twin. Expert inputs should be included. 
• One challenge with water is integrating engineering system into natural system, also interacting with people. The digital twin might need to incorporate people and human behaviour. Agent based model can be used to model behaviour of people in social science.

Collecting data:  

• Need to think about data quality, assurance, anomaly detection, etc., some of which can be automated.  
• Need to think about data operability, data sharing and discovery. 

Design monitoring networks: 

• Think about the minimum requirements.  

Skills and training gaps:

• The large volume of data and expectation for computational efficiency require data and software engineers to be on board alongside analysts. The team should also include domain experts, and analyst in terms of co-design of visualisations.  

Session 2: The discussion in session 2 was focused on designing a digital twin for water. Each group chose a topic of their interest and discussed the following questions as part of the brainstorming:  

• What else is already out there? 
• What data are needed to create the digital twin? 
• What design features would it need to have? 
• How would it be used? 

Design of digital twin: 

Soil moisture 

• It might worth broadening the theme of soil moisture to include soil carbon, soil quality or soil health, linking to some of the discussions in the previous workshop. 
• Soil moisture can be influenced by soil type, meteorology, land use, vegetation, topography, tillage and irrigation, and drainage. Much of these data sources exist, both from sensors, field campaigns and earth observation, but they may vary in spatial and time resolution. Data on agriculture drainage may be harder to acquire. 
• The digital twin might be used by farmers/landowners, insurance companies, regulators, and businesses (e.g. financial sector).  
• While the framework for the digital twin could be generic, there could be important application differences specific to different catchments. 
• It is also beneficial to be agile, i.e. delivering something quickly, which interests senior management, and then develop from that.

The way forward for the next workshops: Our next workshop will focus on energy, specifically renewable energy. Future workshops will further the technical discussions around digital twins and discuss other environmental sector such as biodiversity.