Metrology
for Earth
Observation

Met4EO Training on Uncertainty Analysis for Earth Observation Datasets

• Introducing the principles of uncertainty analysis

  NPL hosted the Introducing the principles of uncertainty analysis workshop — a 1.5 day training course that introduced the principles of uncertainty analysis, supported by a suite of open-source software tools designed to help participants apply these concepts in practice.

  Kicking off with core fundamentals, participants were encouraged to complete a free online preparatory course.

  The training included guided, hands-on exercises using example datasets, and participants had the opportunity to explore their own data with support from NPL experts.

  The course was developed as part of the Metrology for Earth Observation (Met4EO) project, an ESA-funded initiative led by the National Physical Laboratory (NPL) that improved the reliability and interoperability of Earth Observation data through rigorous metrological methods. Met4EO delivered open-source tools, AI-supported sensor intercomparison techniques, practical training, and new frameworks for evaluating uncertainties in products such as land cover maps and radar-altimetry measurements. It also provided the CEOS Calibration Dashboard, enabling satellite operators to assess radiometric performance using international reference sites, supporting ESA’s drive toward high-quality, analysis-ready EO data.

• Training materials

  Links to presentations and exercises

  Foundational concepts

  • Welcome and context – Emma Woolliams

  Uncertainty analysis fundamentals

  • Key concepts in uncertainty analysis – Pieter De Vis

  • QA4EO 5 step approach – Emma Woolliams

  • Error correlation – Sam Hunt

  • Parameterising error correlation and UNC standard – Sam Hunt

  Practical workshops and exercises

  • Breakout workshop: Uncertainty tree diagrams – Ariadna Gonzalez

  • Breakout workshop: Effects tables with Measurement Model Workbook – Supporting workbook for effects tables practical

  CoMet tutorials: Uncertainty propagation

  • CoMet tutorial: Uncertainty propagation methods with punpy – Pieter De Vis & Sam Hunt

  • CoMet tutorial: Uncertainty propagation for spectrometer example – Pieter De Vis & Sam Hunt

  • CoMet tutorial: Error correlation in datasets with obsarray and UNC – Pieter De Vis & Sam Hunt

  Validation and comparisons

  • Comparisons of FRM to satellite – Pieter De Vis

  • Summary of Comparison Workshop – Emma Woolliams

  • Tutorial: Satellite validation using RadCalNet and HYPERNETS – Pieter De Vis

  Supporting tools

  • MetEOR toolkit – Sam Hunt

  ---

  Online training

  • CoMet toolkit introduction: A short introduction to key concepts, why they matter, and how CoMet helps with uncertainty handling.
  • An online training course has been developed in paralell; now accessible via NPL's training platform at: Uncertainty Analysis and Quality Assurance for Earth Observation Datasets.

  A self‑paced course providing in‑depth background material, practical examples, and supporting resources that underpin concepts covered in the workshop. Completion will enable participants to engage more fully with the in-person training sessions, follow technical discussions with confidence, and arrive prepared with informed questions.

  
• Related events

  To complement the workshop, a half-day FRM workshop takes place on the afternoon of 20 May (also at NPL), focused on the organisation/analysis of comparison exercises.

  Comparison with peers and or similar facilities are a key mechanism to evidence an achieved uncertainty from Cal/Val activities — and this workshop (via EO-related examples - will tillustrate and discuss comparisons and associated analysis/reporting processes.

  For details and about how to preprare slides for dicussions, see: Comparison Workshop, a space to share how comparisons between satellite datasets and Fiducial Reference Measurements are carried out in practice.

• Location

  IQST, National Physical Laboratory, Bushey Road, Teddington, TW11 0EB, UK

  
• Agenda

  Day	Time	Session
  Thu 21	08:30–09:00	Registration and tea / coffee
  Thu 21	09:00–10:30	Presentations and Q&A • Welcome and context (Emma Woolliams) • Key concepts in uncertainty analysis (Pieter De Vis) • QA4EO 5 step approach (Emma Woolliams) • Error correlation (Sam Hunt)
  Thu 21	10:30–11:00	Break
  Thu 21	11:00–11:30	Practical measurement example: MetEOC spectrometer system (spectral reflectance vs Spectralon tile)
  Thu 21	11:30–12:30	Breakout workshop: Uncertainty tree diagrams – spectrometer example, plus peer comparison and Q&A
  Thu 21	12:30–13:30	Lunch
  Thu 21	13:30–14:15	Breakout workshop: Effects tables – spectrometer example with Measurement Model Workbook, peer comparison and Q&A
  Thu 21	14:15–15:15	CoMet tutorial: Uncertainty propagation methods with punpy (Pieter De Vis & Sam Hunt)
  Thu 21	15:15–15:45	Break
  Thu 21	15:45–16:15	CoMet tutorial: Uncertainty propagation for spectrometer example
  Thu 21	16:15–16:45	Parameterising error correlation and UNC standard (Sam Hunt)
  Thu 21	16:45–17:30	CoMet tutorial: Error correlation in datasets with obsarray and UNC
  Thu 21	18:30	Non-hosted dinner at Anglers Pub
  Fri 22	09:00–09:45	Presentations and Q&A • Comparisons of FRM to satellite (Pieter De Vis) • Summary of Comparison Workshop (Emma Woolliams) • MetEOR toolkit (Sam Hunt)
  Fri 22	09:45–10:30	Tutorial: Satellite validation using RadCalNet and HYPERNETS
  Fri 22	10:30–11:00	Break
  Fri 22	11:00–12:00	Discussion session, general Q&A, feedback
  Fri 22	12:00	Finish

• Social

  Lunch and refreshments will be provided for participants Wednesday and Thursday. On Thursday evening there will be a non-hosted dinner at the Anglers Pub. For the latter we require an advance payment. More details on selecting menus and the payment of these will be shared with the participants in near future.

  For all NPL events we aim to accommodate specific needs and personal circumstances, but are reliant on individuals sharing this information with us. If you have any specific requirements or adjustments to support your attendance (whether in person or virtual), please do let us know by emailing events@npl.co.uk, so that we can discuss how we can best support you. Any information shared will be confidential and deleted after the event.