VISTA Open Day Highlights: Time Series Tools in Action
The second VISTA Open Day, held on 7 May 2025, marked another important step in the STELAR project’s final stages. The event brought together partners from the STELAR consortium to share project outcomes and demonstrate technical progress.
In research and innovation projects, Open Days serve a role beyond presentation. They create a space for exchanging critical feedback, aligning outputs with end-user needs, and ensuring the tools developed are practical and usable. They also strengthen links between research teams and those expected to apply the results.
Kicking Off the Open Day
The focus this time was on a topic central to STELAR’s innovation efforts: how satellite data is processed to support agricultural applications. In particular, the session highlighted the project’s technical progress in deriving consistent field-level time series from Earth observation sources. This was built on the groundwork laid during the first Open Day, which explored how early yield predictions were developed and validated through STELAR’s pilot operations.
The day opened with a welcome from Sandra Kolaric, Dissemination and Communication Manager at Foodscale Hub, followed by an overview by Dimitris Skoutas, STELAR’s coordinator from the Athena Research Centre. He provided a comprehensive introduction to the project’s goals, its consortium, including an introduction to the pilots and the current progress regarding the STELAR toolkit.
Before moving into the pilot results, the audience was briefly introduced to VISTA’s core competencies. Silke Migdall, Managing Director, outlined the company’s focus on crop monitoring, risk management, and sustainable farming practices. She also highlighted VISTA’s use of satellite data and simulation models to support early yield forecasting – one of the ambitions of STELAR.
Technical Insights from VISTA’s Pilot Work
Solveig Blöcher from VISTA led the technical session on pilot results, linking the second Open Day to themes explored in the first. Her presentation covered several tools developed under STELAR, including EO Data Fusion, Time Series Imputation, Field Segmentation, and Crop Type Classification.
She explained how these tools integrate satellite data – specifically Earth observation information – by combining multisensor, spectral, and time series inputs. This integration enables earlier and more complete predictions. Tools were tested across two diverse regions with differing topography and environmental patterns – Austria and France – demonstrating the toolkit’s adaptability to varied agricultural conditions.
Field Delineation for Targeted Predictions
Jens d’Hondt from Eindhoven University of Technology presented work on field delineation. Working in collaboration with VISTA, he introduced a pipeline designed to automatically define field boundaries based on satellite data. This task is crucial in improving the spatial accuracy of yield prediction.
His presentation detailed the motivations behind this process, the technical setup, and how early warning systems could benefit from clearer sub-field level analysis.
Time Series Imputation: Bridging the Data Gaps
George Marinos from Athena Research Centre shared progress on time series imputation. He introduced the TS-Impute tool, detailing its development, recent updates, and experimental results. The tool addresses data gaps in satellite observations and helps deliver more consistent input for downstream analysis.
This type of processing plays an essential role in real-time monitoring by improving data quality and reliability. George also offered practical pipeline examples to demonstrate how the tool fits within agricultural decision-making systems.
Bias-Aware Crop Classification
Chethan Krishnamurthy Ramanaik from the University of the Bundeswehr Munich presented an approach to crop type classification that accounts for data bias. His work included techniques such as scarcity-aware data slicing, model ensemble strategies, and the results achieved through this approach.
He stressed the need to accommodate data limitations common in agricultural datasets and showed how the STELAR approach aims to mitigate their effects using a carefully designed classification pipeline.
Next Steps for STELAR
The second VISTA Open Day offered more than a summary of project progress – it demonstrated how research can lead to tools that reflect real-world agricultural challenges. As the STELAR project moves toward completion, its continued engagement with the agricultural community will determine how its tools live on.
The next Open Day is just around the corner – taking place on 13 May – and will focus on the Abaco pilot on “Timely Precision Farming Interventions.” Stay tuned by following STELAR on LinkedIn and exploring our Blog page.
Catch up on everything you missed – watch the full recording of the event here.
