workshops.qmd

---
format:
  html:
    css: challenge.css
title: The I-ADOPT Variable Modeling Workshops
author:
  - name: Barbara Magagna
    orcid: 0000-0003-2195-3997
date: 01/27/2025
---


## Variable Modeling workshops using I-ADOPT and OGC OMS

The **RDA endorsed I-ADOPT Framework** is currently being discussed as a candidate model for extending the **OGC OMS (OGC Observations, Measurements, and Samples/ISO 19156) standard**. I-ADOPT could be implemented as an OMS extension for the observable property concept, providing richer and FAIRer representation of what is observed in an observation event. To be able to make use of I-ADOPT in OMS in a systematic approach it is crucial to instantiate the model integration with real data to represent the whole observation setting. 


We have planned two Modeling Workshops to be hold as in person events in Vienna, Austria:

1) **February 20/21, 2025** at the Natural History Museum with the focus on *biodiversity, surface and marine variables*

2) **March 10/11, 2025** at the Environment Agency Austria with the focus on *atmospheric variables* and the development of an LLM-based Annotation Service

Both workshops are sponsored by RDA TIGER that provides financial support to cover participants’ travel expenses.

[![](img/tiger_support_services_0.png){width="30%"}](https://www.rd-alliance.org/rda-tiger/){target="_blank"}

Participation is by invitation only and involves modeling experts for OMS and I-ADOPT and domain experts who will bring variables (and datasets) that challenge the representation with the models.

## Outcomes of the first I-ADOPT Variable Modeling Workshop

In the first I-ADOPT Variable Modeling Workshop (February 20&21 at the Museum of Natural History, Vienna) we introduced and discussed different relevant models that can be combined to represent observations: OMS, SensorThings API, a SensorThingsAPI generation service, and finally also the I-ADOPT model. One critical issue, heavily discussed, was how to connect the feature type of OMS with a specific description component in I-ADOPT. There is no easy answer and no easy pattern to follow although initially the ContextObject was identified to be the most appropriate. We clarified the notion of the Matrix to be an element rather than an object that might be measured in time and space. We have abandoned the idea that Matrix must be included in every variable which includes a ContextObject. A specific use case is how to handle distance variables where we identified symmetric and asymmetric scenarios. We discussed the potential modification of the I-ADOPT model to allow Constraints on Properties and identified a few us cases to be described before we can decide. Special attention was given to the concept of "product", which could complement I-ADOPT to allow the description of observations with information on how these variables should be properly monitored, i.e. adding spatial and temporal resolution triggered by the EBV case study. These and other patterns are being described in the [patterns repository](https://github.com/i-adopt/patterns). We decided to continue working on case studies around biodiversity, marine, and freshwater using [GitHub issues](https://github.com/i-adopt/OMSmodelling/issues) and a [template](https://github.com/i-adopt/i-adopt.github.io/raw/refs/heads/main/material/OMS-I-ADOPT_template.xlsx) that allows us to include annotations to address additional OMS concepts. A side activity was the upload of I-ADOPT aligned EBVs in EcoPortal (as they are also uploaded in OGC Rainbow server) and the small project to start a variable definition pipeline using nanopublications via EcoPortal. A key result is that I-ADOPT can be used to give the OMS observable property FAIR context information for data streams. However, I-ADOPT can also be used independently to enrich metadata.

### Participant list

Modeling subgroup:

* Barbara Magagna, GO FAIR Foundation, NL (RDA WG I-ADOPT, co-chair)
* Sirko Schindler, German Aerospace Center, DE (RDA WG I-ADOPT, technical lead)
* Katharina Schleidt, DataCove (OGC OMS - chair, TSML, conceptual modeller)
* Hylke van der Schaaf, Fraunhofer IOSB, DE (OGC - STA, chair)
* Rob Atkinson, OGC, AU (Spatial Data on the Web WG, co-chair)
* Raul Palma, PSNC, PL (AD4GD - EBV, FAIR2Adapt)
* Joan Masó, CREAF, ES (AD4GD - EBV)

Marine subgroup:

* Gwenaelle Moncoiffe, BODC, GB (RDA WG I-ADOPT, co-chair, content manager of NVS)
* Lorenzo Corgnati, CNR-ISMAR, IT (EuroGOOS, JERICO)
* Anne Fouilloux, Simula, NO (FAIR2Adapt - coordinator)

Biodiversity subgroup:

* Ivette Serral, CREAF, ES (AD4GD - EBV)
* Victoria Lush, Aston University, GB (AD4GD - EBV)
* Marina Pulieri, University of Salento, IT (LifeWatch Italy, EcoPortal)
* Andrea Tarallo, CNR, IT (LifeWatch Italy, ITINERIS)

Freshwater subgroup:

* Philipp Saile, Federal Institute of Hydrology, DE (UNEP GEMS/Water, WMO WHOS)
* Katarina Řiháčková, Masaryk University, SK (PARC)

Online participants:

* Sylvain Grellet, BRGM, FR (OGC OMS, HydroDWG, WMO WHOS, Water4all/OneWater, PARC)
* Charly Coussot, IRD, FR (French OZCAR Water Research Infrastructure)


## Outcomes of the second I-ADOPT Variable Modeling Workshop
In the second I-ADOPT Variable Modelling Workshop (March 10&11 at the Environment Agency Austria, Vienna) we presented the [outcomes and patterns](https://docs.google.com/presentation/d/1LC6nbm8Ps0aOdTtTi2cb7PAzc6pTyneD) we discovered during the first workshop and introduced an initial workflow for the development of an [LLM-enabled annotation service based on I-ADOPT](https://drive.google.com/file/d/1Ucl_k-cIILpsnNTiJ8h5sBn2wetHelI5/view). The variables in focus were provided by the meteorological and atmospheric domain. After a presentation on the current developments and efforts in WMO on the [WIGOS Metadata Representation](https://drive.google.com/file/d/1mZMYxrpibPQEyriPiKLISyk4hv8lTlbf/view?usp=sharing), we discussed specific complex variables and their decomposition. This process led to [Application Design Records (ADRs)](https://github.com/i-adopt/patterns/tree/main/adrs) which list options for each encountered design problem and on which the I-ADOPT core team will vote and agree to provide reusable patterns. We also identified [rules of alignment](https://docs.google.com/spreadsheets/d/1fclAwHtNP-d3NOWp62k7CkAAq4OqUakvpPEfHg_wipI/edit?usp=sharing) between the property type in I-ADOPT and the type of the Feature of Interest in OMS based on examples of the I-ADOPT Challenge analysed, but further analysis is required on a bigger set of examples. A key insight of the modelling discussion revealed the requirement to model a system of elements for any entity in I-ADOPT and these elements can be of any I-ADOPT nature (variables, properties, entities). In addition, we agreed on the requirement to include statistical modifiers (maximum, average, …) in the description. One issue was not yet addressed in detail but will need more attention later: the further formalization of constraints to allow better machine readability. 

A common practice of domain-specific vocabularies applying I-ADOPT (see [ACTRIS vocabulary](https://vocabulary.actris.nilu.no/skosmos/actris_vocab/en/) is to design I-ADOPT roles as classes which is only possible if the same concepts appear in the same position in the descriptions. This might be possible for a specific domain but requires clear modelling guidance for the domain with key demo examples provided by leading authorities like WMO. In addition, we further discussed properties required for the product description, where ISO 10131 (Data Product Specification - DPS) can give some directions but the real implementation will most likely not go via ISO but via OMS SWG. 

The I-ADOPT service developer team defined a harmonised [roadmap](https://docs.google.com/document/d/1eY9UGJv_YMGi1iYKPb-6SLo8GjZWMFy1oD4wWv8MsKY/edit?tab=t.0) identifying tasks needed to implement the service. The service should allow decomposing a human-readable variable description in I-ADOPT description elements, recognize existing I-ADOPT variables and propose them to the user or annotate the variable elements with concepts from semantic artefacts. Two approaches will be followed: i) using a LLM to recognise the variable concepts, and a mapping tool to export them to RDF; and ii) using a LLM to transform directly to RDF. The main challenge will be how to automatically evaluate the correctness of the results, including the reuse of adequate concepts from semantic artefacts. 

### Potential collaborations:

* I-ADOPT service development with developers from FAIR2Adapt, NFDI4Earth project and PARC 
* I-ADOPT variables modelled for WIGOS vocabulary and for the Atmospheric Composition Standard Variable Name Convention (ACSVNC) involving modellers from the WIGOS, NASA and CF Standard Names
* WMDR 2.0 proposal includes properties which are interesting for the product model in line with I-ADOPT-DPS-STA - a collaboration between modellers from WIGOS, AD4GD and OMS together with some RIs like eLTER, and projects like PARC and FAIR2Adapt would make it 

### Next steps:

* I-ADOPT Core team decides on the ADRs and publishes patterns
* I-ADOPT revises the challenge variables to provide a set of curated variables (L0) (deadline: April 10, 2025)
* The service developers will define on top of these variable definitions and the patterns the benchmark level L0 (April - May 2025)
* The domain communities revise their variable descriptions based on the variable modelling patterns using the template indicating also if the variable requires a sample (May 2025)
* The core I-ADOPT team provides a clean list of L1 variables as summary of the community input (June 2025)
* The developer team creates synthetic data via the service (L2) (July-August 2025)
* L2 will be validated by modelling and domain experts  (Sept 2025)

### Participant list

Modeling subgroup:

* Barbara Magagna, GO FAIR Foundation, NL (RDA WG I-ADOPT, co-chair)
* Sirko Schindler, German Aerospace Center, DE (RDA WG I-ADOPT, technical lead)
* Katharina Schleidt, DataCove (OGC OMS - chair, TSML, conceptual modeller)
* Ivette Serral, CREAF, ES (AD4GD - EBV)

Metereological subgroup: 

* Gao Chen, NASA, US (ACVSNC - Langley Research Center, lead)
* Morgan Silverman, NASA, US (ACSVNC - Langley Research Center, data scientist)
* Jörg Klausen, MeteoSwiss, CH (WMO TT - WIGOSMD, chair)
* Alison Pamment, Centre for Environmental Data Analysis/Sience and Technology Facilities Council, UK (CF Convention, I-ADOPT core team)
* Johannes Peterseil, Environment Agency Austria (eLTER RI, technical lead)

I-ADOPT service developer subgroup:

* José Manuel Gomez, Expert.AI, ES (FAIR2Adapt, developer)
* Cristian Berrio Aroca, Expert.AI, ES (FAIR2Adapt, developer)
* Èsteban Gonzalez, UPM, ES (FAIR2Adapt, developer)
* Arvin Rastegar, KIT, DE (NFDI4Earth, developer)
* Saurav Kumar, IISPV, ES (PARC, semantic expert)

Online participants:

* Iseult Lynch, University of Birmingham, UK (PARC - FAIR lead)
* Markus Fiebig, NILU, NO (ACTRIS - atmospheric scientist)
* Beth Huffer, NASA, US (Lingua Logica, semantic expert)

## Examples that illustrate the complexity character we are looking for 

* [mass flux of carbon into soil from vegetation due to senescence](https://i-adopt.github.io/examples/Challenge/6.ttl.html)
* [coarse woody debris abundance](https://i-adopt.github.io/examples/Challenge/8.ttl.html) 
* [height of atmospheric boundary layer](https://i-adopt.github.io/examples/Challenge/4.ttl.html) 

Other examples may be:

* Properties that relate to two objects and none fit the matrix or the context object definition like the distance between two habitats (see [GitHub issue 14](https://github.com/i-adopt/examples/issues/14)) 
* Measurements related to prey/predator or parasite/host relationships
* Fluxes of something across an interface between two systems 
* Variables that include statistical parameters or method related information 
* Variables that integrate procedural aspects

By the end of the modeling workshop, participants will be able to develop FAIR representations for each observation dataset or variable, based on OMS and I-ADOPT as an integrated model. An STA (Sensor Things API)  instance enriched with concepts from I-ADOPT will be made available, enabling experimentation with the enriched I-ADOPT extended OMS system.

One of the aims of this workshop is to clarify the modeling of complex variables to prepare clear guidelines on how to properly model your variable descriptions. Moreover, we hope that this will provide the basis for automatizing the modeling with LLMs. 

## Materials to study before the workshop

A [reference document](https://docs.google.com/document/d/1lV884X5lvFBA7OdE5fHUe4pIqi0-q7oX1tK728Lm5NQ/edit?usp=sharing) with important links for all related models and services.

For **OMS and SensorThings**:

1. [A Smiley OMS presentation](https://github.com/i-adopt/i-adopt.github.io/blob/main/material/Smiley%20-%20Observations%2C%20Measurement%20and%20Samples.pdf) explaining OMS kindly provided by Kathi Schleidt 

2. A detailed [description of the Observation of a Smiley Face](https://github.com/i-adopt/i-adopt.github.io/blob/main/material/Observation%20of%20a%20Smiley%20Face.pdf) kindly provided by Kathi Schleidt 

3. The [SensorThings API Data Model](https://datacoveeu.github.io/API4INSPIRE/sensorthingsapi/requestingData/STA-Data-Model.html) 

4. The [FROST-Server Plugin](https://github.com/hylkevds/FROST-Server.ObservableProperty) for implementing the I-ADOPT model

For **I-ADOPT**:

1. An introduction to the I-ADOPT Framework (<a href="https://www.youtube.com/watch?v=Q3thKyVVtCM" target="_blank">Video</a>, [Slides and Notes](material/I-ADOPT Framework - Slides and Notes.pdf){target="_blank"})

2. A step-by-step guide for creating an I-ADOPT compliant variable description ([Video](https://www.youtube.com/watch?v=43hAzmqBAfs){target="_blank"}, [Slides and Notes](material/I-ADOPT Step-by-Step - Slides and Notes.pdf){target="_blank"})

3. **Instructions for preparing variable models** to be discussed ([Presentation](https://github.com/i-adopt/i-adopt.github.io/blob/main/material/Preparing%20your%20variables.pdf){target="_blank"})