VISION, which stands for Virtual Integration of Satellite and In Situ Observation Networks, is advancing atmospheric science by seamlessly integrating satellite and in-situ observations with atmospheric model data. One of the aims of the project – funded by the Natural Environment Research Council – was to create a digital twin demonstrator for environmental science using Earth Observation (EO) data and digital twinning technologies.

The VISION team has produced a toolkit, which is a set of python algorithms developed to enable efficient and accurate interfacing of models and observations. We’re able to co-locate model data from a few days of regional forecasts or several decades of Earth System simulations, to the same time and geographical location as the observations, making the comparison between models and observations easier and more meaningful.

If you work with models or if you are a provider of observational data then you could greatly benefit from knowing more about this new software!

We have the In-Situ Observations (ISO) Simulator, which samples model data at the same time and location as in-situ observations from flights, ships, and sondes. And we are also developing the Satellite Simulator, which samples model data in the same way that satellites sample the atmosphere. As well as a time and space colocation onto satellite swaths, there is an option to include information that mimics the satellite uncertainty in the vertical sampling through the atmosphere. We would do this by including the satellite’s averaging kernel and a-priori information.

Thanks to the VISION toolkit, researchers can more accurately and easily compare model predictions to real-world data, enhancing confidence in model performance and improving our understanding of atmospheric processes.

In the VISION project we have showcased two possible applications. The first one is to improve the UK Earth System Model, known as UKESM, and help with climate projections that feed into policy decisions around living with the impacts of climate change. Through better comparisons with a wider variety of observations, VISION has highlighted areas where UKESM is not doing as well as it could. Ultimately it’s led to improvements in model representation of specific atmospheric processes, including the modelled ozone response to emissions from lightning strikes!

In the second example we’ve combined VISION tools with satellite data and bespoke regional forecasts for the FAAM Airborne Laboratory’s research aircraft. We’ve been able to create a digital twin of the research aircraft and we hope it can be used to maximise research output and reduce carbon emissions when flying for science missions. This has been done alongside the FORCE framework, otherwise known as Forecasting Operations for Research Campaigns and Experiments.

In the context of atmospheric science, the “digital twin” is an extension of a “digital model” and a “digital shadow”. For example, a climate model has no real-time communication between the physical and digital systems. A digital shadow has one way communication from the physical to the digital systems, for example a weather forecast using data assimilation to improve forecast accuracy.

In preparation for a flight campaign, our digital twin of the research aircraft combines a weather forecast of the atmosphere and our ISO Simulator software to allow us to fly a virtual digital plane through our digital atmosphere. We can then differentiate between different potential routes for the research aircraft and identify the ones with the best chance of sampling the desired atmospheric conditions. This information can then be passed back to mission scientists and can influence flight plans.

After talking at the NCAS Staff Meeting in 2022, Luke and I started to develop some advanced software together in Cambridge, funded through the Archer-2 eCSE programme. We began by extracting ozone concentrations from historical data collected on missions made by the FAAM Airborne Laboratory’s research aircraft.

We turned it into a coherent dataset and made it easy to co-locate multiple years of atmospheric model data onto research aircraft flight tracks. We were then able to compare modelled quantities at the same spatial and temporal scale as the observations made by the aircraft.

We started thinking about how we could expand the flight simulator software so it could be used to seamlessly compare model data with other types of observations data. We wanted to use observation data from sondes, ships, buoys, and even satellites – which meant reaching out to Brian Kerridge’s research group at NCEO.

Following a presentation by Maria to NCAS staff in 2023, Alan Woolley who heads up the FAAM Airborne Laboratory was also interested in the potential of bringing back to life decades of archived research aircraft observations to support further model comparisons.

Summer 2023 came around, and these conversations continued and culminated in a successful funding proposal under the NERC TWINE programme for Digital Twin demonstrators. The project involves groups of NCAS staff at Cambridge, Reading, Leeds, NCAS’s FAAM Airborne Laboratory, and NCEO. The 15 month project started in January 2024 and came to an end in March 2025.

We love that different partners on the project bring very different expertise and knowledge. We have all really enjoyed working on VISION and have also learned a great deal from the other teams involved.

NCAS is a great organisation bringing together expertise in modelling, observations, and computational services. VISION is not only maximising the synergy between modelled and observational data but also between different parts of NCAS, as it involves research software engineers, weather scientists, climate modellers, and observational scientists at the FAAM Airborne Laboratory.

Our wishlist is to be able to convert existing historical data, collected by NCAS scientists over decades, into a VISION-compatible standard format. This would enable isolated historical observations to be used as part of something bigger, an ever growing database of NCAS observations to be used for model evaluation and to improve our understanding of atmospheric processes in years to come.

By developing an improved toolkit which builds upon tools and libraries developed in-house, namely cf-python and cf-plot, we no longer rely on software that we do not maintain, making these tools much simpler to install and use. We also discovered and implemented various improvements to cf-python and cf-plot by considering the science and computational challenges of the VISION remit, an unexpected benefit.

VISION builds on years of work in model evaluation of UKESM and applies this to operational systems used for supporting flight campaigns. This should have real-world benefits both to the science produced by the FAAM Airborne Laboratory, but also to the carbon emissions and flight costs involved with collecting data.