Enriching Earth observation datasets through semantics for climate change applications: The EIFFEL ontology

Abstract

Background: Earth Observation (EO) datasets have become vital for decision support applications, particularly from open satellite portals that provide extensive historical datasets. These datasets can be integrated with in-situ data to power artificial intelligence mechanisms for accurate forecasting and trend analysis. However, researchers and data scientists face challenges in finding appropriate EO datasets due to inconsistent metadata structures and varied keyword descriptions. This misalignment hinders the discoverability and usability of EO data.

Methods: To address this challenge, the EIFFEL ontology (EIFF-O) is proposed. EIFF-O introduces taxonomies and ontologies to provide (i) global classification of EO data and (ii) linkage between different datasets through common concepts. The taxonomies specified by the European Association of Remote Sensing Companies (EARSC) have been formalized and implemented in EIFF-O. Additionally, EIFF-O incorporates:1.An Essential Climate Variable (ECV) ontology, defined by the Global Climate Observing System (GCOS), is embedded and tailored for Climate Change (CC) applications.2.The Sustainable Development Goals (SDG) ontology is included to facilitate linking datasets to specific targets.3.The ontology extends schema.org vocabularies and promotes the use of JavaScript Object Notation for Linked Data (JSON-LD) formats for semantic web integration.

Results: EIFF-O provides a unified framework that enhances the discoverability, usability, and application of EO datasets. The implementation of EIFF-O allows data providers and users to bridge the gap between varied metadata descriptions and structured classification, thereby facilitating better linkage and integration of EO datasets.

Conclusions: The EIFFEL ontology represents a significant advancement in the organization and application of EO datasets. By embedding ECV and SDG ontologies and leveraging semantic web technologies, EIFF-O not only streamlines the data discovery process but also supports diverse applications, particularly in Climate Change monitoring and Sustainable Development Goals achievement. The open-source nature of the ontology and its associated tools promotes rapid adoption among developers.

Keywords: Climate change mitigation and adaptation; EO taxonomy; Earth Observation (EO); Essential Climate Variable; Ontology and semantics; Sustainable Development Goals.

Figure 1.. General methodology to build the ontology.

This methodology is a didactic and commonly used way to explain the different steps required to build an ontology.

Figure 2.. EO taxonomy - market view.

The EO taxonomy has two main views; the market view (consumer view) shows the perspective of the end-user, and how data consumers search for datasets in their own way. There are 26 different sectors which are grouped in 8 markets.

Figure 3.. EO taxonomy - provider view.

The EO taxonomy has two main views; the provider view shows the perspective of the data provider, and how they categorize their datasets. There are 31 areas grouped in 6 domains.

Figure 4.. ECV ontology.

The Essential Climate Variables (ECV) Ontology is a structured framework that categorizes and defines variables crucial for monitoring the Earth's climate system. Developed to support consistent and comprehensive climate data analysis, the ECV Ontology encompasses variables across atmospheric, oceanic, and terrestrial domains.

Figure 5.. SDG ontology.

Source: . The Sustainable Development Goals (SDG) Ontology provides a structured framework for representing and linking information related to the United Nations' 17 Sustainable Development Goals. It organizes and categorizes concepts, indicators, targets, and relationships associated with each SDG, enabling clear understanding and integration of diverse data sources.

Figure 6.. EIFF-O wrapper.

This wrapper merges the EO ontology, the ECV ontology and the SDG ontology, also including concepts from schema.org in order to support JSON-LD data formats. In addition, through the concept of Essential Variables other ontologies can be linked (e.g., SMURBS, Essential Agriculture Variables and Essential Biodiversity Variables).

Figure 7.. EIFF-O module (building blocks).

The EIFF-O module consists of a series of tools that facilitate the understanding and usage of the EIFF-O ontology. There is a front-end to access and view ontology files, documentation and code, but also a backend part providing Webvowl support, a REST API as well as a SPARQL endpoint.

Figure 8.. Ontology visualization tool.

The ontology viewer is based on Webvowl and allows visually loading an ontology, showing the relationships among the different entities (circles) through arrows (properties). It also includes informational text on the right.

Figure 9.. EIFF-O module (search for concept).

During the tagging of datasets, and considering the ontologies used derived from taxonomies, users can easily find and select through tree-based structures the related concept to be associated to a specific dataset.

Figure 10.. EIFF-O module (tagging visual table).

The tagging tool allows the user to easily list all semantic concepts from EIFF-O associated to a certain dataset. They are listed in a table describing the related concept and some actions to be performed (add concept, delete concept).