📅 Thursday 27th August 2026 | 🕐 10:30 –16:00
📍Hybrid: Cardiff University | Sir Martin Evans Building | BIOSCI – SMEB C124 Boardroom & Online
🔗Click here to register for the Earth Observation Knowledge Exchange Event
Many thanks for expressing an interest in attending a Planet Labs Earth Observation Data Knowledge exchange event. We hope you can attend on Thursday 27th August 2026.
We would welcome additional flash talks to find out more about how Earth Observation data is being used across Wales.
Proposed schedule
10:30 – 11:00 – Registration, coffee and networking
11:00 – 12:00 – Welcome.
- Brief introduction to Environment Platform Wales
- Planet Presentation. The future of the Welsh Government Earth Observation Program.
12:00 – 12:30 – What are you working on? Group discussions
12:30 – 13:00 – Detailed Presentation | Juan Suarez | Swansea University | Multitemporal Random Forest Classification of Forest Tree Species Using PlanetScope Imagery
13:00 – 14:00 – Lunch and Networking
14:00 – 15:00 – Flash presentations:
- Ben Clarke | Aberystwyth University | Identifying fire regime history and critical transitions in peatland functioning in Wales.
- Adetoun Afolabi | Swansea University | Nature-based solutions
- Mykola Kutia | Bangor University | Forest cover dynamics and ecosystem services quantitative analysis using high spatial resolution satellite imagery.
- Saman Sobhani | Aberystwyth University | Monitoring and Scenario Modelling for Sustainable Urban Green Space Management in Cardiff city.
15:00 – 15:30 – Use of Earth Observation data in education and teaching
- Aidan O’Donnell | Cardiff University | Use (current and potential) of data at Cardiff University
- Discussion on use of Earth Observation data in teaching
15:30 – 15:45 – Additional Flash Presentations welcome
15:45 – 16:00 – Wrap-up
| Juan Suarez presentation on: Multitemporal Random Forest Classification of Forest Tree Species Using PlanetScope Imagery This research investigates the use of multi-season PlanetScope surface-reflectance imagery and Random Forest classification to identify commercially important tree species in Thetford and Aberfoyle forests. This method of classification explores the phenological differences between tree species as the ground for classification. The method develops spectral-band and vegetation-index predictors, derives training data from field-verified pure stands, and applies forward feature selection to refine the model. To address spatial autocorrelation and avoid inflated accuracy estimates, the study employs spatial block cross-validation alongside conventional training and validation splits. The resulting classifications demonstrate strong species-level discrimination, indicating that multitemporal remote sensing, combined with carefully selected predictors and spatially robust evaluation, offers an effective approach for forest species mapping and sustainable management. |