School of Mathematics & Statistics

Analytics for Digital Earth - Workshop 1

Motivating and Idea Generation Workshop 

Date: 3rd February 2022

Time: 1pm

Location: Virtual

Speakers: There were 3 speakers given the task of providing motivating examples and questions. Firstly, Prof Mark Girolami (The Alan Turing Institute) gave a summary of digital twins and talked about some of the digital twin projects going on at the Turing Institute. Dr Kit Macleod (The James Hutton Institute) then gave 10 digital earth challenges and solutions. Finally, Gillian Dick (Glasgow City Council) talking about all aspects of planning and how they are using data and tools to come up with nature-based planning decisions. 

Discussion Format: The first workshop contained two specific breakout discussion sessions.  This draft note of the meeting includes the notes from the different breakout groups.  In total there were 6 breakout groups (3 sessions, 3 groups each).  The first session focussed on ‘themes for digital earth’, the second session was ‘themes and topics for analytic challenges’. Each group was asked: 

• What other themes should we include?​
• What are the key challenges within ​these themes?​
• Which challenges are common to ​multiple themes?​
• Is there an order of priority in these ​themes/challenges?​ 

Session 1: For session 1, a pre-workshop questionnaire had identified the digital earth themes of net zero, water quality and quantity, air quality, biodiversity, urban/rural, citizen science and clean energy. The discussion identified several themes that were thought to be missing, including soil and soil health, which may in turn be linked to land use, and carbon sequestration.  There was also a recognition that the digital twin may need to span anthropogenic and natural systems, that these may evolve independently but there is inevitably feedback between the two and are therefore not completely independent.  Linked to clean energy, wind farms (onshore and offshore) were specifically mentioned.  Regarding the urban/rural theme, there was discussion about natural and anthropogenic systems, but also linking the two within a digital twin. 

Session 2: In session 2, the same questions were used, but this time the focus was on analytic challenges.  In the same pre-workshop questionnaire, sensor networks, analytics generally, prediction and forecasting, visualisation and uncertainty were all identified. The break out group discussions focussed on a number of challenges and issues around data quality, how to design sensor networks, computing infrastructure (access to GPUs etc), working with digital earth twins, and visualisation where the link to gaming and augmented and virtual reality was made. 

Big picture challenges that cut across our themes include: 

• Scale of statistical inference: whether at the level of e.g. water plant, catchment, regional, national. This is becoming more important with new technologies where there is a mismatch in spatial (and temporal scale) between existing traditional sensors and new high-dimensional outputs. Can we combine or better use old sensors to improve new approaches? 
• Extreme events are becoming increasingly common - physical models are not designed to represent these. How do we parameterise these models better? How do we obtain representative datasets? Most models are not variable enough to account for this and our knowledge may not yet be good enough to do this well. 
• Combining/ integrating - Citizen science with traditional data 
  • challenges in standardising protocols and having appropriate meta data 
  • challenges in avoiding duplication of data with public data not valued as highly as academic (peer reviewed data) 
  • challenges considering appropriate bias 
  • challenges in quality assurance, secure data archiving, future proofing 
• Combining/integrating - low and high frequency data  
  • challenges in change of support, but even wider, data to data, model to model, model to data. 
• Data - how to manage data to be FAIR (findability, accessibility, interoperability and reusability)?  

Building the digital twin, what is the way forward?: 

• Need to plan in advance what is needed from a digital twin and be clear about the objectives.
• Need to think about system boundaries (especially key in th environmental/ecological context)
• Need to consider the different data streams that will be used and the coupling of them to the models that underpin the digital twin. Data fusion, data assimilation and the challenges of multiple scales.
• Need to consider how and who will use the digital twin- in terms of visualisation and communication, including uncertainty. 

The way forward for the next workshops: The overarching challenges for ‘A Digital Earth’ come under the broad application heading of ‘Net Zero’ and the broad analytic heading of ‘A Digital Twin’.  The future workshops will focus on specific topics within each of these two broad agendas. Discussing specific aspects of the net zero agenda identified above and the development of analytical tools required, and appropriate component parts of a digital twin framework. Firstly, workshop two will have a specific theme on ‘Soils’ and specifically ‘Carbon within Soils’, and will specifically consider the current state of play and developments required in data collection, modelling and forecasting.