A data engineering approach to tracking global climate resilience

The adoption of 59 indicators for the Global Goal on Adaptation (GGA) at the 2025 United Nations Climate Change Conference (COP30) in Belém represents a significant development in measuring climate adaptation. After a decade in which international commitments were expressed largely through general formulations such as “enhancing resilience” and “reducing vulnerability”, Parties now have a defined set of measurable dimensions. These include water stress, crop yields, health system readiness, infrastructure resilience, and the protection of cultural heritage.

However, the value of any such indicator set depends entirely on whether the reporting entities measure the same phenomena using comparable methods. That condition is not yet satisfied, and meeting it constitutes the substantive work now ahead.

Adaptation is not a single dataset. It comprises numerous distinct data holdings distributed across national hydrological and meteorological services (NHMSs), sectoral ministries, national disaster management authorities (NDMAs), utilities, environmental and biodiversity monitoring programmes, and, of particular importance, the accumulated knowledge of local communities that have adapted to their environments over generations.

Each of these sources operates within its own technical conventions. For example, one country may define a "climate-resilient road" as a road designed to withstand a 1-in-50-year flood, while another may use a 1-in-100-year standard. As a result, data that appear similar may not be directly comparable unless the definitions, units, and evidence requirements are clearly documented..

This is the difficulty that Parties identified in the Baku Adaptation Roadmap. Operationalising the Belém adaptation indicators was acknowledged to require:

• Resolving outstanding methodological questions;
• Closing persistent data gaps;
• Strengthening capacity in countries whose statistical and observational systems remain limited; and
• Establishing a defensible basis for incorporating qualitative and Indigenous knowledge alongside quantitative evidence.

Identifying the gaps between indicators and evidence

It is useful to state the obstacles precisely, as most are technical or operational rather than political, notwithstanding the policy language in which they are ordinarily expressed.

• Fragmentation and institutional silos: Adaptation data is held separately by health ministries, NDMAs, meteorological services, and utilities, with each operating within its own technical conventions and systems of record.
• Methodological ambiguity: Definitions, scales, and units for the 59 Belém indicators remain unresolved. Consequently, terms such as vulnerability and resilient infrastructure continue to be interpreted inconsistently across reporting entities.
• Constraints on capacity and data availability: Many countries, particularly Least Developed Countries (LDCs) and Small Island Developing States (SIDS), lack the institutional capacity and data infrastructure required to populate the indicators reliably.
• Absence of interoperability: Systems are frequently unable to exchange information without intervention, obliging institutions to establish bespoke translation arrangements for each exchange. This creation of "many-to-many" linkages expands complexity with every additional partner.
• Integration of diverse knowledge systems: The legitimate incorporation of qualitative and Indigenous knowledge within a quantitative reporting framework remains unresolved. Treating Indigenous knowledge as a residual or secondary category is not a defensible outcome.
• Process reported instead of outcomes: Reporting continues to record whether a plan or policy instrument exists rather than whether vulnerability has been demonstrably reduced, although the latter is what the Global Stocktake (GST) under the Paris Agreement is required to measure.

The case for common engineering principles, not a single format

Policy processes at the international level commonly favour a single global standard: one indicator set, one definition, one taxonomy, and one dataset. The reasoning behind this preference is sound in principle, though it has proved difficult to realise. Forcing every country and sector into identical systems ignores real differences in capacity, context, and priorities. It also tends to produce brittle, resented compliance rather than durable data infrastructure.

A more workable model already exists in open science. Over the past decade, the Findable, Accessible, Interoperable, and Reusable (FAIR) principles have substantially reshaped the management of research data worldwide. The experience is instructive. The straightforward elements of FAIR are those concerned with making data findable and available for download. The demanding element—and the one that ultimately determines whether data is of any use—is interoperability: whether an entirely separate system can correctly interpret what the data means and apply it accordingly.

That is exactly the gap facing adaptation measurement. A downloadable table of flood exposure statistics is of limited value to a global adaptation progress  unless its units, geographic scale, reference period, uncertainty, and underlying definitions are documented in a form that both analysts and machines can interpret.

The Cross-Domain Interoperability Framework (CDIF) is a framework designed to enable data from different sectors, disciplines, and institutions to be discovered, understood, and reused consistently across systems . Rather than  requiring everyone to use the same standards, it works like a translator. Different communities do not need to change how they collect or manage data. Instead, CDIF helps make their information understandable to others. Climate science, public health, and disaster risk finance each retain their own standard, mapping them once to a shared set of common profiles that govern data discovery, variable description, controlled vocabularies, access rights, and packaging. Instead of creating separate data-sharing arrangements between every organisation, each organisation maps its data once to the common framework. This makes data exchange much simpler, more efficient, and easier to scale. 

The CDIF4EOSC initiative, a recently launched European project, is developing the corresponding toolkit. This includes artificial intelligence (AI)-assisted instruments that support institutions in making their data FAIR, alongside an implementation playbook for the European Open Science Cloud (EOSC). Climate adaptation is one of its three flagship use cases, alongside ocean science and sustainable materials research.

From publishing data to engineering it

This shift requires a change in orientation. For much of the past two decades, the open data agenda has been framed around publication, understood as depositing a file in an accessible location under an appropriate licence. Reporting against the Belém indicators requires something more demanding: data engineering. This encompasses pipelines, metadata profiles, controlled vocabularies, provenance records, and version control capable of withstanding simultaneous use by national statistical offices, hydrological services, and international aggregators operating on the same underlying figures.

CDIF is best understood in these terms. It does not constitute a competing indicator set, nor does it prejudge the methodological determinations that Parties will reach through their two-year process. Instead, it provides the engineering layer on which such determinations can be implemented.

Understood in this way, the framework offers several specific mechanisms:

• Functions as a lingua franca : Rather than requiring climate science, health, and finance to converge on a single global schema, CDIF mediates between their existing standards through a shared set of common profiles.
• Simplifies data exchange: It reduces a many-to-many exchange problem to a many-to-one relationship. An institution maps its data once to the CDIF profiles and thereby becomes intelligible to multiple international bodies.
• Applies FAIR principles by design: Metadata, provenance, and access rights are embedded at the point of data creation, removing the need to document a dataset retrospectively in advance of a reporting deadline.
• Standardises meaning through functional profiles: Specialised components, including Simple Knowledge Organization System (SKOS) for controlled vocabularies and Data Documentation Initiative Cross-Domain Integration (DDI-CDI) for data description, render a crop yield or water stress measurement recorded in one country comparable and machine-actionable in another.
• Reduces reporting burdens: Automating translation and formatting allows national systems with limited technical staffing to allocate effort to analysis rather than repetitive manual reformatting, while rendering capacity constraints visible to international funders.
• Supports accountability: Preserving the context and provenance of national data while structuring it for global aggregation establishes a precondition for a GST that commands confidence.

The principal argument for this approach concerns implementation costs. Adopting a new global reporting standard requires every participating institution to assimilate a new specification, reconstruct its systems against it, and maintain that setup indefinitely. An approach based on CDIF instead reuses the standards institutions already operate and requires only that a single mapping be established and maintained. The development and deployment effort associated with the second path is substantially lower.

Why this matters for the Global Stocktake and risk reduction

The next GST will ask a deceptively simple question: is the world adapting fast enough?

Answering it honestly requires evidence that is comparable across very different countries without flattening out the context that makes each country’s adaptation story meaningful. That is a genuinely difficult balance. It cannot be achieved by asking countries, after the fact, to retrofit their national data into a common format. It has to be built in from the start—embedding documentation, definitions, and metadata into datasets at the moment they are created, not scrambled together the week before a reporting deadline.

Interoperability infrastructure also carries a vital equity dimension. Reducing the burden of custom data translation disproportionately helps under-resourced national data systems—specifically Least Developed Countries (LDCs) and Small Island Developing States (SIDS) whose adaptation finance and capacity needs dominate climate negotiations. A shared interoperability layer does not fix the adaptation finance gap, but it can make invisible capacity gaps visible and free up scarce technical staff from repetitive formatting work so they can focus on critical risk reduction analysis.

A caution on governance and local ownership

None of this technical engineering substitutes for political agreement. The methodologies and metadata underpinning the 59 Belém indicators remain under negotiation through a formal two-year process. This is appropriate, as these are determinations for sovereign Parties rather than for technical standard bodies.

Interoperability frameworks are enabling infrastructure and nothing more. They cannot manufacture country ownership, nor can they displace sustained investment in national capacity. Crucially, they must be constructed in genuine accordance with Indigenous data governance frameworks—such as the CARE Principles for Indigenous Data Governance —rather than treating community knowledge as material to be forcibly fitted into an external schema.

Where these governance and capacity conditions are met, the achievable outcome is considerable. A flood risk assessment in Bangladesh, a coral reef monitoring programme in Fiji, and the harvest records of a farmer cooperative in Kenya could each contribute reliably to a common global measure of climate resilience while retaining the specificity that makes their evidence credible in the first place.

This is not a modest technical convenience; it is the difference between a Global Stocktake that merely reports what data could be hastily assembled and one that measures true risk reduction. Building an interoperability layer into the Belém indicators from the outset is substantially less costly than retrofitting it later, and it will ultimately determine whether the global adaptation scoreboard measures genuine progress or merely records the attempt to describe it.

Bapon Fakhruddin is Chair at CODATA TG-FAIR DRR and Climate Investment Principal at the Green Climate Fund.

Matti Heikkurinen is Project Portfolio Manager at CODATA.