CGIAR Climate Data Hub - System-Wide Climate Data Asset Mapping

Internal review draft

Published

July 9, 2026

Version 1.1.0-draft — 123 assets in this build. Interactive data access: CDH interactive dashboard.

Generated programmatically from the data — no numbers are hand-typed. The figures and tables below are interactive (hover, sort, filter) in this HTML edition. See Data Access, Feedback and Reproducibility near the end for the source code, steps to reproduce, and how to suggest corrections or additions.

Who This Report Is For

This report turns the first system-wide mapping of CGIAR’s climate data assets into something the Hub can act on. It is written for three audiences:

  • CDH leadership and the Core team — to see where the system is strong and where it is thin, and to steer Phase-1 priorities. (Start with the Executive Summary and Section 5.)
  • The CDH development and data team — to decide what to federate or ingest now, what to queue for later, and where effort is being duplicated. (Section 6.)
  • Contributing centres — to check how their assets are represented and flag corrections or additions. (Annex A and the feedback links in the Data Access section.)

The questions it answers

  • Where is CGIAR strong, and where are the gaps — by centre, theme, and geography?
  • Which assets should the Hub act on now, and which belong in the next cycle?
  • What is openly reusable, foundational, or nationally important — and what is locked behind an access conversation?
  • Where are centres independently reprocessing the same upstream climate inputs?

The needs it serves

The Hub exists to reduce fragmentation in CGIAR’s climate evidence base. This mapping is the evidence behind that effort: it prioritises a focused set of high-value assets for Phase-1 inclusion or federation, surfaces reusable and nationally-relevant datasets, flags duplicated preprocessing the Hub can do once instead of many times, and points to the gaps — and the centres — still to engage. It is deliberately strategic, not exhaustive (Section 1).

Executive Summary

CGIAR holds a wealth of climate data — but it is scattered across centres, programmes, and bilateral projects: hard to find, easy to duplicate, and difficult to reuse. The Climate Data Hub was created to change that, and this exercise is its first concrete step: a structured look at the strongest climate data assets the centres themselves put forward.

123 assets from 11 centres were catalogued, spanning hazard monitoring, adaptation analytics, exposure, mitigation accounting, and integrated multi-domain platforms; 70 (57%) come from the six Hub-funded centres. Coverage is deepest in Adaptation Analytics and Hazard, and concentrated in Africa and Global datasets.

Three messages stand out:

  • There are clear quick wins. 28 assets are openly accessible, technically ready, and high-value — they can be federated or ingested with little friction (Section 6.5).
  • The centres’ nominations provide a strong starting point. The 28 top-three nominations come with written justifications and are the primary starting set for immediate Hub consideration, but not the only assets that may enter Phase 1 review (Section 6.2; Section 8).
  • The gaps are specific and actionable. Adaptive Capacity is essentially absent, Latin America & Caribbean and Asia are thin, and two major centres — CIMMYT and ICARDA — have yet to submit (Section 5; Section 7).

The rest of the report makes each of these concrete: where the strengths and gaps sit (Section 5), and exactly what to do now and next (Section 6).

Work-in-progress caveat. This report is one piece of evidence to guide the Climate Data Hub, alongside other technical, strategic, and governance inputs; it does not by itself dictate exactly what the Hub will do. The mapping is iterative and will be updated as centres review their entries, flag corrections, and suggest additions. Centres and partners will have multiple opportunities across the annual CDH cycle to engage, refine priorities, and shape what comes next.

1. Background and Objectives

The CGIAR Climate Data Hub (CDH) is a CGIAR initiative under Area of Work 1 (AoW1), designed to surface, standardise, and federate climate-relevant data assets held across the CGIAR system. The Hub operates under a federation model: it points to data where it already exists, rather than duplicating it, and only ingests data where cloud-optimised or API-accessible formats are not available.

This report summarises the first system-wide asset mapping exercise, conducted in early 2026. Centres were asked to nominate climate data assets through a structured submission template covering identity, structure, spatiotemporal scope, thematic domain, context of use, and a readiness/nomination assessment.

Consistent with the mapping strategy, the exercise is deliberately focused and strategic, not exhaustive: each centre nominated a limited set (up to ~20) of its strongest, most decision-relevant assets, prioritising quality and reuse over volume. It is not an audit of past outputs, and not a scientific evaluation of data quality, and it does not alter data ownership or impose hosting requirements. It is a coordination and governance step to inform Phase 1 of the Hub.

Objectives of the mapping:

  1. Surface a focused, strategic set of high-value climate data assets across CGIAR (not an exhaustive inventory).
  2. Identify gaps and priorities for Hub curation and integration.
  3. Provide a foundation for cross-centre collaboration and data-sharing.
  4. Inform the CDH technical roadmap for ingestion and federation.

2. Methods

2.1 Submission process

Each centre received a standardised Excel template with six thematic sheets. Nominators were asked to complete one row per asset, covering:

  • Sheet A — Identity: asset name, nominator, organisation, asset type, short description.
  • Sheet B — Structure: file format, storage location, licence, API/access details.
  • Sheet C — SpatioTemp: spatial coverage, resolution, temporal type, update frequency.
  • Sheet D — Thematic: climate domain, farming system, commodity, output variable type.
  • Sheet E — Context & Use: decision relevance, reuse potential, existing use cases.
  • Sheet F&G — Assess & Nominate: technical readiness, contemporary validity, sustainability, asset rank, and intended Hub role.

Submissions were received from 11 centres.

The blank template, full submission guidelines, and every centre’s completed Excel workbook are version-controlled in the project repository: the raw submissions are at [data/submissions/](https://github.com/CGIAR-Climate-Data-Hub/cdh-asset-mapping/blob/main/data/submissions/) and the mapping strategy and template instructions at [docs/CDH-Asset-Mapping-Strategy.docx](https://github.com/CGIAR-Climate-Data-Hub/cdh-asset-mapping/blob/main/docs/CDH-Asset-Mapping-Strategy.docx).

2.2 Normalisation

Free-text fields for climate domain, asset type, and spatial coverage were normalised to controlled vocabularies using keyword matching. The normalisation rules are documented in src/ingest.py and applied reproducibly by the pipeline documented in the Data Access, Feedback and Reproducibility section.

2.3 Consolidation of duplicate entries

4 consolidations were applied where multiple submitted entries represented sub-components of a single dataset (e.g. model inputs and outputs from the same pipeline). This reduced the raw submission count of 129 to 123 catalogued assets. The full merge log is in Annex C.

2.4 Optional composite priority score

The mapping strategy deliberately avoids reducing the five qualitative criteria to a single rank (Section 6.1). For navigation only — to help sort long lists in the report and dashboard — a transparent composite score (0–100) is nonetheless computed. It is defined once in src/ingest.py and shared verbatim with the dashboard, so the two never diverge. It is not an official ranking; the authoritative signal remains each centre’s own top-three nomination.

Each component is mapped to a 0–1 sub-score, then combined as a weighted average:

  • Ordinal criteria (Decision Relevance, Technical Readiness, Reuse Potential, Contemporary Validity, Sustainability) map Very High = 1.0, High = 0.85, Medium-High = 0.75, Medium = 0.5, Medium-Low = 0.35, Low = 0.25.
  • Submitted rank maps to a centre-relative sub-score (max_rank_in_centre − rank + 1) / max_rank_in_centre, so rank 1 scores highest within each centre regardless of how many assets that centre ranked.
  • Intended Hub role maps Hub-native = 1.0, Derived = 0.8, Federation = 0.7, Reference/Operational = 0.6; unspecified is omitted.
Component Weight
Decision relevance 0.20
Technical readiness 0.20
Reuse potential 0.15
Submitted rank (within centre) 0.15
Contemporary validity 0.10
Sustainability 0.10
Intended Hub role specified 0.10

The score is the weighted mean of only the components that are present: a missing criterion is dropped from both numerator and denominator rather than scored as zero, so incomplete submissions are not unfairly penalised (they simply rest on less evidence). The portfolio mean is 73.

3. Domain Definitions

The following climate domain vocabulary is used throughout this report:

Domain Definition
Hazard Climate variables and indices that characterise physical hazard (e.g. rainfall, temperature, drought, flood extent).
Hazard / Climate Services Operationally processed hazard products delivered as services (e.g. seasonal forecasts, advisories).
Exposure Data on agricultural systems, land use, populations, and assets exposed to climate hazards.
Sensitivity Data on how exposed systems respond to climate stressors (e.g. crop yield sensitivity, disease risk models).
Adaptive Capacity Data on capacity of systems or communities to adjust to climate impacts.
Adaptation Analytics Integrated datasets and model outputs that assess adaptation options, impacts, or trade-offs (combines hazard, exposure, and response).
Mitigation GHG inventories, emission factors, and tools for quantifying emission reductions in agriculture.
Multi-domain Assets that span two or more domains without clear primary classification.
Hybrid labels Some assets are tagged with two adjacent domains (e.g. Sensitivity / Adaptation Analytics) where single label would be misleading.

These labels follow the mapping strategy’s domain vocabulary. The strategy lists Climate finance and Climate policy as separate domains; given the small number of assets in either, this report combines them under Climate Policy / Finance — they can be split if future submissions warrant.

4. Results

This section describes the portfolio as a whole — how much was submitted and by whom, which themes and geographies it covers, who owns the assets, and what they actually contain. Sections 5 and 6 then turn this into strengths, gaps, and actions.

4.1 Volume and coverage

A total of 123 assets were catalogued across 11 centres (Figure 1). All figures reported here reflect post-consolidation count (see Section 2.3 and Annex C).

Figure 1. Assets submitted per centre. Total catalogued assets per centre after consolidation (Section 2.3), ordered largest to smallest. Each bar is split into one coloured segment per nominating individual — a longer run of colours means more contributors behind a centre’s portfolio (e.g. AfricaRice and IITA drew on many nominators, while ILRI and IFPRI came through a single nominator). Segment colour is arbitrary and carries no meaning (hence no legend); the number is the centre total. Centres marked with an asterisk indicate cases where a single listed nominator may understate broader internal consultation or contribution. The interactive dashboard names each nominator and their count on hover.

Centre Assets Hub-funded
AfricaRice 20 No
IITA 17 Yes
Alliance 16 Yes
CIFOR-ICRAF 13 No
ILRI 11 Yes
IFPRI 10 Yes
CIP 9 No
IRRI 9 No
IWMI 8 Yes
WorldFish 8 Yes
ICRISAT 2 No
Total 123

Hub-funded centres account for 70 assets (57%) of portfolio.

Coverage caveat — centres not yet represented. Submissions were received from 11 centres. CIMMYT and ICARDA had not submitted at the time of this build and are therefore absent from all totals and figures below; their inclusion will materially change domain and geographic coverage (CIMMYT in particular for South Asian wheat/maize systems and adaptation analytics — see the box in Section 5.4). Targeted follow-up with both centres is underway.

4.2 Domain distribution

Figure 2. Distribution by climate domain. Number of assets in each normalised climate domain (definitions in Section 3); assets with no specified domain are excluded. Hover a bar for its share of the portfolio, the centres contributing it, and how much is openly accessible. Adaptation Analytics and Hazard are the largest domains; Adaptive Capacity and Climate Policy / Finance the thinnest.

Adaptation Analytics is most represented domain (32 assets, 26%), followed by Hazard (26 assets). Multi-domain assets (21 assets) reflect submissions where nominated asset spans two or more domains — common for integrated platforms and modelling frameworks.

High Multi-domain share from IITA reflects their submission labels (‘Agronomy and climate’, ‘Disease risk’) which span exposure, sensitivity, and adaptation. CIFOR-ICRAF’s multi-domain assets include food security and livelihoods datasets with indirect but significant climate relevance. See Figure 5 for centre-by-domain breakdown.

Cross-centre domain coverage

Figure 5. Centre × domain heatmap. Number of assets each centre holds in each single-label climate domain; darker cells hold more, blank cells none. Hover a cell for example assets and the open-access share of that centre–domain combination. Hybrid-domain assets are excluded here for readability but counted in Figure 2.

Heatmap shows number of assets per centre per domain. Hybrid domain labels (e.g. Sensitivity / Adaptation Analytics, Adaptation Analytics / Mitigation) are excluded from heatmap for readability; they are included in Figure 2.

A further interpretation caveat is that current centre profiles may reflect how assets were surfaced, not only the full thematic breadth held within each centre. In particular, IFPRI’s current submission sits entirely in Exposure and ILRI’s entirely in Hazard, so targeted next-round outreach would help test whether those centre-level profiles are complete or mainly reflect the first-round submission pathway.

4.3 Ownership and asset type

Figure 3. Asset type (ownership). Share of catalogued assets by ownership class — CGIAR-produced, co-produced with partners, or external datasets adopted into CGIAR workflows. Hover a bar for the centres behind it and its open-access share.

50 assets (41%) are CGIAR-produced; 29 (24%) are external datasets adopted into CGIAR workflows; and 29 (24%) are co-produced with external partners. External assets are included where centres have demonstrated active use in climate analytics and where Hub can add value through standardisation or linkage.

4.4 Geographic coverage

Figure 4. Geographic coverage. Number of assets per geographic grouping; assets with no specified coverage are excluded. Hover a bar for the dominant domains and centres in that region. Africa and Global dominate; Latin America & Caribbean, Asia, and Multi-regional are comparatively thin.

Africa (58 assets) and Global (46 assets) together represent 85% of portfolio. Asia and South/Southeast Asia (10 assets) is driven primarily by IRRI. Latin America & Caribbean (4 assets) is represented by Alliance and CIP submissions.

3 assets had no spatial coverage specified in submission and are excluded from Figure 4.

4.5 Priority nominations

Of 123 catalogued assets, 106 (86%) include Asset Rank (Section F&G of submission template) and 53 (43%) specify intended Hub Role.

Intended Hub Roles describe how asset should be integrated with Hub:

  • Hub Native — asset will be ingested and hosted directly by Hub.
  • Hub Reference — Hub will link to asset at its existing location, without ingesting copy.
  • Hub Derived — Hub will produce derived or value-added product from asset.

Hub preference is federation over ingestion. Datasets in existing platforms that are not cloud-optimised or available by API may require ingestion to enable interoperability; in such cases Hub will work with data owner to agree on approach. Where permissions already allow it, Hub may proceed without delay; otherwise, formal agreement with data owner is required before ingestion.

4.6 What the assets measure, and who uses them

Counts alone understate what was captured. The submissions also describe what each asset actually produces (template Sheet D — Thematic) and how it is used in practice (Sheet E — Context & Use). That detail is what turns an inventory into a basis for reuse, and it sharpens the picture of where the portfolio’s real value sits.

What they measure. 123 of 123 assets specify an output variable type, spanning raw climate variables and hazard indices, biophysical outputs (crop yields, biomass, soil moisture), greenhouse-gas emissions, and suitability or classification layers. The commodity focus tracks CGIAR’s mandate crops — led by Rice (32), Potato/sweetpotato (11), Livestock (11) — while farming systems are dominated by Cropping (60), Rice-based (15), Mixed (13). The recurring upstream climate inputs in Section 6.7 show how many of these outputs are, in turn, built on a small shared set of sources, which is exactly where the Hub can remove duplicated effort.

How they are used. Submitters most often describe their assets serving Modelling (44), Research (44), Policy (43) purposes — confirming a portfolio that skews decisively toward decision support rather than purely academic output. The user communities named most frequently are Researchers (77), Governments (76), Donors (52). Most tellingly for the Hub’s reuse mandate, 114 assets already name the CGIAR programmes using them, 68 are rated high or very-high national relevance (the datasets that underpin country engagement and policy dialogue), and 78 are flagged as foundational to ongoing work — assets whose withdrawal would break existing pipelines.

Two implications follow. First, the portfolio’s value is concentrated in a relatively small core of foundational, multi-programme, nationally-relevant datasets; these are the natural anchors for Phase 1, ahead of more peripheral or single-use submissions. Second, the strong policy, advisory, and modelling orientation means the Hub’s task is less to surface new science than to make assets that are already relied upon discoverable, interoperable, and durable — reducing the risk that critical datasets remain locked to the teams that happen to maintain them today.

5. Strength and Gap Analysis

This section is written for the CDH Core team: where the system-wide portfolio is strong, where it is thin, and where coverage depends on a single centre.

5.1 Domain × geography coverage

Figure 6. Where coverage is deep, thin, or absent (climate domain × geography). Each cell counts the assets in one domain (row) and geography (column); darker blue = more assets, a red dot = a true gap with none. Hover any cell for the exact count. Two patterns stand out: coverage concentrates heavily in Africa and Global, and Adaptive Capacity is empty across every region while Latin America & Caribbean and Multi-regional are thin throughout — the clearest targets for the next cycle. Hybrid-domain and unspecified-geography assets are excluded for legibility.

Coverage concentrates in two cells — Hazard × Africa (17 assets) and Adaptation Analytics × Africa (16 assets) — alongside a strong Global column. Of the 45 domain × geography combinations, 21 are empty.

The clearest thematic gap is Adaptive Capacity, with no catalogued assets in any region. Latin America & Caribbean and Multi-regional are the thinnest geographies across nearly every domain.

5.2 Concentration risk

2 domain(s) are currently represented by a single centre, making system-wide coverage dependent on one submitter:

Domain Sole centre
Adaptation Analytics / Mitigation IRRI
Climate Policy / Finance Alliance

These are priorities for cross-centre outreach: a second submitter would reduce single-point dependency and validate the domain’s representation.

5.3 Per-centre strength profile

Mean priority score summarises each centre’s portfolio on the shared 0–100 scale (see Section 6.1). ‘Domains’ counts the distinct climate domains a centre covers.

Table — Per-centre strength profile (mean priority is the optional composite of Section 6.1).

Centre Assets Mean priority Domains covered Hub-funded
IWMI 8 82 4 Yes
IRRI 9 81 3 No
AfricaRice 20 77 6 No
Alliance 16 74 5 Yes
IITA 17 73 4 Yes
CIFOR-ICRAF 13 72 3 No
CIP 9 70 3 No
IFPRI 10 69 1 Yes
ILRI 11 67 1 Yes
WorldFish 8 66 5 Yes
ICRISAT 2 63 1 No

5.4 Notable assets outside this mapping

Two strategically important assets are intentionally not ranked in the analysis above, and should be read alongside it:

AAA Atlas (CGIAR Adaptation Atlas). Already planned for integration into the Climate Data Hub from the outset, the AAA Atlas has been deliberately excluded from the nomination ranking so that other centre assets can surface on their own merit. Its inclusion in the Hub is assumed, not contingent on this exercise.

South Asia Adaptation Atlas (ACASA). Developed by CIMMYT / BISA (https://acasa-bisa.org/), ACASA is a major adaptation-analytics resource for South Asian food systems. CIMMYT has not yet submitted to this mapping, so ACASA does not appear in any totals or figures; it is flagged here as a high-priority asset to capture once CIMMYT engages.

6. Priorities and Actions

This section is written for the CDH development and data team: which assets to act on now, which to queue for the next cycle, and the suggested integration pathway for each.

6.1 How to read priority here

The mapping strategy is deliberate that the five qualitative criteria are a comparison aid, not a formal score or ranked list — collapsing them into a single number would oversimplify. This report follows that intent: the authoritative signal is each centre’s own ranking, and the top three assets per centre are the strategic nominations for immediate Hub consideration (Section 6.2).

As a navigation convenience only, each asset additionally carries an optional composite score (0–100) — a weighted blend of the five criteria, submitted rank, and hub-role specification, computed reproducibly in src/ingest.py and shared with the dashboard so the two never diverge. Treat it strictly as a sorting aid for long lists, not as an official quality verdict; absent criteria are dropped from the blend rather than penalised. Portfolio mean 73.

6.2 Strategic nominations — each centre’s top three

The 28 assets below are each centre’s three best-ranked submissions. Under the mapping strategy these are the unit for immediate Hub inclusion or federation; lower-ranked assets are candidates for later cycles. Each carries a centre-written justification (summarised below; full text in the asset record).

Data-quality note: 4 centre(s) (AfricaRice, ICRISAT, IITA, ILRI) submitted tied/duplicate ranks, so their nominations are capped at the three lowest-ranked assets (ties broken by name); IRRI submitted no ranks and so contribute no nominations; a centre with fewer than three ranked assets shows fewer. To be corrected with centres next cycle.

Table 1 — Strategic nominations: each centre’s top-ranked assets (sortable / filterable in HTML; asset names link to the data source).

Loading ITables v2.8.1 from the init_notebook_mode cell... (need help?)

6.3 Priority quadrant

Figure 7. Priority quadrant. Each point is an asset placed by its submitter-rated technical readiness (x-axis) and reuse potential (y-axis); colour shows access status (green = Open, orange = Restricted, grey = Unknown). Hover any point for the asset’s name, centre, domain, access, and priority score. The shaded top-right zone holds the natural ‘quick wins’ — assets that are both ready and broadly reusable. Reuse potential and access are independent criteria, so high-reuse Restricted (orange, upper) points are not a contradiction — they are the prime candidates for an access negotiation to unlock that value.

Why are several Restricted assets rated high reuse? The two are measured independently. Reuse potential is the submitter’s judgement of how broadly the asset’s content could serve other programmes, countries, or analyses; access status is whether it can be obtained today under current licensing or permissions. A dataset can be scientifically reusable while still gated — these are exactly the assets where a short access conversation converts latent value into usable value (see Section 6.6).

6.4 Suggested integration pathway

Figure 8. Suggested integration pathway. Each asset is classified by a heuristic combining its access status and file format into a starting pathway — federate (ready, or with light ingest), ingest candidate, negotiate access, or assess. Hover a bar for the file formats driving that pathway and example assets. The classification is advisory and should be verified per asset before committing.

A heuristic combining access status and file format suggests a starting integration pathway per asset. 68 assets look federation-ready or close to it; 32 are gated behind an access conversation; 12 are open but need ingestion to become interoperable. These labels are advisory — verify format and licence per asset before committing.

Of the 123 assets, 80 are Open access, 32 Restricted, and 11 unspecified. 78 are flagged foundational to ongoing CGIAR work and 80 are reported as actively maintained.

6.5 Act now — ready, open, high value

28 assets combine Open access, high technical readiness, and a priority score of 75+. These are the recommended near-term targets for federation or ingestion (all listed below; sort or filter the table in the HTML edition):

Table 2 — ‘Act now’ shortlist: Open-access, high-readiness assets (sortable / filterable in HTML; asset names link to the data source).

Loading ITables v2.8.1 from the init_notebook_mode cell... (need help?)

6.6 Next cycle — high value, currently blocked

26 assets score 70+ but are held back by restricted access or sub-‘High’ technical readiness. These warrant an access conversation or a readiness investment before the next mapping cycle (top 12 shown):

Table 3 — ‘Next cycle’ queue: high-value assets blocked by access or readiness. Asset names link to the data source.

Centre Asset Domain Blocker Rationale
AfricaRice AfricaRice-processed ENACTS climate datasets (enhanced and q Multi-domain Restricted access Essential input for most climate applications and models
AfricaRice Climate information service datasets used in digital advisor Multi-domain Restricted access Direct farmer-facing climate services
AfricaRice AfricaRice Data Sharing Tool (DST) for climate services Multi-domain Restricted access Enables institutional climate data sharing and integration
AfricaRice RiceAdvice Multi-domain Restricted access Widely deployed agronomic advisory platform
Alliance IPR Tool Adaptation Analytics Readiness below High The IMPACT model is a multimarket economic model for the agriculture sector widely used for assessing climate impacts on global food price, production, trade, a
AfricaRice AfricaRice AgDataHub climate database (integrated agrometeor Hazard / Climate Services Restricted access Backbone climate dataset enabling modelling, advisories, and risk analytics
AfricaRice Mali Disaster Risk Management (DRM) Maproom Multi-domain Restricted access Supports drought/flood risk management and response
AfricaRice Climate risk indicators datasets (drought, flood, heat stres Hazard Restricted access Key for hazard monitoring and resilience planning
AfricaRice RIICE satellite-derived rice monitoring and climate impact d Multi-domain Restricted access Supports production monitoring and policy decisions
IFPRI Global hamonized cropland Exposure Restricted access Advisory, Invesment, Decision making
AfricaRice Flood risk and submergence hazard datasets for inland valley Hazard Restricted access Critical for flood risk management in inland valleys
IFPRI High resolution national crop type maps Exposure Restricted access Research, Investment and policy

6.7 Shared dependencies and reuse signals

Duplication of climate inputs. Of the 19 assets that reported their underlying climate inputs, several inputs recur across multiple submissions — CHIRPS (8), Station / observational (7), CMIP6 (4), ERA5 (4). Shared upstream inputs are prime candidates for once-only Hub preprocessing (a stated aim of the mapping: detecting duplication in climate inputs and pipelines) rather than each centre reprocessing the same source independently.

Climate input Assets relying on it
CHIRPS 8
Station / observational 7
CMIP6 4
ERA5 4
WorldClim 2

Only 19 of 123 assets recorded their climate inputs; enforcing this field next cycle would complete the dependency picture.

Cross-programme reuse. 114 assets name the CGIAR programmes already using them — direct evidence of reuse beyond the originating team, which the strategy treats as a core signal for Hub inclusion.

National relevance. 68 assets are rated High or Very High for national relevance — the datasets most important for country engagement, policy dialogue, and partner buy-in. The strategy flags these as priorities even where they are not globally standardised or openly accessible.

7. Discussion

Stepping back from the numbers: what does the portfolio tell us, how far can we trust it, and what should happen next?

7.1 Coverage and gaps

This mapping captures a substantial and decision-relevant slice of CGIAR’s climate data portfolio, but by design it is neither exhaustive nor a complete census. The picture it paints should be read as what the engaged centres consider their strongest assets, not the full universe of CGIAR climate data. Two coverage caveats matter most when interpreting the gaps below.

First, two centres are absent: CIMMYT and ICARDA had not submitted at the time of this build (Section 4.1). This is consequential, not cosmetic — CIMMYT anchors much of CGIAR’s South Asian wheat and maize adaptation analytics (including the ACASA atlas, Section 5.4), and ICARDA anchors dryland and West Asia / North Africa systems. Several apparent gaps below will narrow once they engage.

Second, within the assets that were submitted, three gaps are robust enough to act on. Adaptive Capacity is essentially absent (0 assets) — the portfolio is strong on hazard and adaptation analytics but weak on the social and institutional capacity to respond, a recognised blind spot for climate targeting. Latin America & Caribbean and Asia / South & SE Asia are thin relative to Africa and Global coverage, concentrating geographic risk. And 2 domain(s) currently rest on a single centre (Section 5.2), so the system-wide picture in those areas is one withdrawal away from a hole. Each of these is an outreach target rather than a finding about CGIAR’s true capability — the mapping surfaces where to look next, not a verdict on what exists.

7.2 Data quality observations

The submissions are usable and rich, but several recurring data-quality issues shape how far the analysis can be pushed, and each has a concrete fix for the next cycle.

Incomplete assessment fields. Asset rank, intended Hub role, and the qualitative ratings were left blank for a non-trivial share of assets, and only 19 of 123 recorded their underlying climate inputs — which limits the duplication analysis (Section 6.7) more than any other gap. Mandating the Hub-role and climate-input fields would sharply increase the analytical value of the next round.

Inconsistent ranking. The strategy intends a clean 1..N ordering per centre, but AfricaRice, ICRISAT, IITA, ILRI submitted tied or duplicate ranks (e.g. several assets all ranked ‘1’ or ‘2’), so their strategic nominations were capped at the three lowest-ranked assets; IRRI submitted no ranks at all. Ranking is the single most important field for prioritisation, so a brief validation step with each focal point before the next submission would pay off directly.

Free-text variability and AI-drafted fields. Domain, type, and access were submitted as free text and normalised to controlled vocabularies (Section 2.2); a handful of entries also carried template example text or GPT-drafted descriptions that required scrubbing or validation. This is expected given the strategy’s allowance for GPT drafting, but it reinforces that externally-researchable fields must be expert-checked, and that internal-use and strategic-importance fields should never be GPT-generated.

Finally, 1 asset(s) had no climate domain specified and could not be classified from context; they are retained in the inventory but excluded from the domain figures.

7.3 Hub integration approach

CDH operates federation-first model. Many CGIAR data products already reside in well-maintained platforms (CGIAR Library, data.cgiar.org, institutional repositories). Hub preference is to register and link these rather than duplicate them.

Where data is not cloud-optimised, not accessible by API, or not in standardised format compatible with other CDH datasets, Hub may work with data owner to improve format or — where necessary — ingest copy. In all cases, CDH will notify data owner and, where permissions do not already allow federation, agree approach before proceeding.

Many CDH data products will be open access. CDH intends to link back into existing portals and platforms so that Hub amplifies rather than duplicates those investments.

What the mapping implies for sequencing is concrete. Roughly 68 assets look federation-ready or close to it and can be registered with little engineering; 32 are gated behind an access conversation and should enter a parallel, relationship-led track rather than block the technical work; and the shared upstream inputs identified in Section 6.7 (CHIRPS, station data, CMIP6, ERA5) argue for the Hub preprocessing these once, centrally, rather than each centre repeating the work. Federation-first keeps stewardship with the originating centres while still delivering cross-CGIAR discovery — the central tension the Hub is designed to resolve.

7.4 In short

Returning to the questions this report set out to answer: CGIAR’s catalogued climate data is strong on hazard and adaptation analytics, deepest in Africa and at global scale, and anchored by a core of foundational, multi-programme, nationally-relevant datasets — but thin on adaptive capacity, in Latin America and Asia, and dependent on single centres in several domains. For the Hub, the immediate move is clear: act on the 28 open, ready, high-value assets now; use the centres’ 28 strategic nominations as the primary starting set rather than the only filter; open access conversations for the high-value-but-restricted assets in parallel; preprocess shared climate inputs once; and run a targeted next-round outreach to close visible gaps, especially where current submissions likely under-represent important domains such as adaptive capacity. The detail sits in Sections 5 and 6; this is the throughline.

8. Next Steps

  1. Complete outstanding submissions — follow up with ICARDA and centres with incomplete assessment fields.
  2. Run targeted gap-filling outreach — use the Q2 workshop and other CGIAR channels to identify strong but currently under-represented assets in missing domains or geographies, especially adaptive-capacity assets and gaps in Latin America and Asia. A next round of outreach with ILRI and IFPRI could also usefully broaden thematic coverage, helping ensure that the current submissions are complemented by a wider range of climate-relevant assets from across each centre.
  3. Prioritise assets for Hub integration — use centre rank and Hub role to sequence technical work, but do not limit Phase 1 consideration strictly to the current top-three nominations where wider strategic value or obvious submission gaps suggest additional assets should be reviewed.
  4. Agree federation vs ingestion for each priority asset — work with data owners to determine appropriate integration pathway.
  5. Publish asset catalogue — make inventory available to CGIAR partners via CDH portal.
  6. Iterate mapping annually — re-run pipeline as new submissions arrive.

Acknowledgments

This mapping exists only because colleagues across the centres took the time to nominate, describe, rank, and justify their strongest climate data assets — a substantial effort on top of busy research agendas. We are sincerely grateful to every contributor, and in particular to the centre focal points and nominators listed below.

Coordinator names are shown in bold in the table below.

Centre Contributors
AfricaRice Aboubacar Diallo; Ali Ibrahim; Elliott Dossou-Yovo; Geoffrey Onaga; Mohamed Diallo; Paul Iboko
Alliance Alcade Segnon; Alejandro Ruden; Camilo Barrios; Carlos E. Navarro Racines; Carlos Gonzalez; Kaue De Sousa; Robert S. Andrade; Shalika Vyas; Tosin Akingbemisilu; Xiaojing Wei
CIFOR-ICRAF Ardhani, Trialaksita; Brockhaus, M.; Ickowitz, A.; Sufiet Erlita; Tamba, Yvonne; Winowiecki, Leigh Ann; Winowiecki, Leigh Ann (ICRAF)
CIP David Ramirez; Heidy Gamarra; Henry Juarez; Robert Hijmans
ICRISAT Mamta Sharma; Raman Babu
IFPRI Zhe Guo
IITA Eduardo Garcia Bendito; Francis Muthoni; Harun Murithi; James Legg; John Omondi; Julius Adewopo; Peter Neuenschwander; Rhys Manners; Siyabusa Mkuhlani; Thompson Ogunsanmi
ILRI Teferi Demissie
IRRI Anton Urfels; Emma Quicho-Diangkinay
IWMI Giriraj Amarnath; Niranga Alahacoon
WorldFish Michelle Tigchelaar; Thomas Kirina

IRRI note: Emma Quicho-Diangkinay coordinated the IRRI submission. The current submission records list Anton Urfels as nominator on the catalogued items, but broader IRRI contributor attribution is being checked and may be updated after review.

Coordination of the asset-mapping exercise is led by the Alliance of Bioversity International & CIAT under the Climate Action Program (Critical Capacity PoD2), in collaboration with the Climate Data Hub team and AoW1. Thanks also to the centre contacts who fielded follow-up questions on access, format, and provenance. Any errors of consolidation or normalisation are the compilers’, not the contributors’.

Data Access and Reproducibility

This report is generated programmatically: every figure, table, and statistic is computed from the normalised data, with no hand-typed numbers. The data and the code that produces this document are version-controlled and open.

Data

Code that generates this report

Reproduce this report

pip install -r requirements.txt
python src/ingest.py        # Excel submissions -> data/normalized/assets.json
python src/figures.py       # static figures -> outputs/figures/
quarto render report.qmd --to html   # or: --to docx

Feedback and Review

Feedback — corrections, additions, and questions

Spotted an error, or know of an asset that should be included? We want to hear it. No GitHub account is needed — use the feedback form; responses are routed automatically into the project’s GitHub issue tracker so every item is triaged and resolved (pipeline documented in FEEDBACK.md).

📝 Give feedback / suggest a correction or asset — 2-minute form, no login.

If you do use GitHub, you can instead open a pre-filled issue directly: correct a record, suggest a missing asset, or general feedback.

The HTML edition of this report also carries a comment thread at the bottom of the page (GitHub login required), backed by the repository’s GitHub Discussions — handy for quick remarks that don’t warrant a full issue. The same thread is embedded in the interactive dashboard.

Review participation

Review of this work is tracked as part of programme participation. To date: 5 reviewers across 4 review events, generating 14 tracked feedback items. The log is maintained in data/review_log.json and covers all channels (feedback form, GitHub, email, documents).

Date Reviewer(s) Affiliation Channel Scope Feedback items
2026-06-17 Brayden Youngberg Alliance of Bioversity International and CIAT GitHub issue Report #1
2026-06-18 Brayden Youngberg Alliance of Bioversity International and CIAT GitHub issue Report #4
2026-06-29 Michelle Tigchelaar Stanford University GitHub issue Report #5, #6
2026-07-08 Alcade Segnon, Franck Tonle, Robert Zougmore Alliance of Bioversity International and CIAT Email + PDF document Dashboard #7, #8, #9, #10, #11, #12, #13, #14, #15, #16
NoteCurrent review feedback themes

Last updated: 9 July 2026. This internal-review note summarises the current open feedback themes logged in GitHub. It is included for transparency and may change as comments are resolved; it is not part of the asset statistics above. Some points from this feedback have already been incorporated into Section 8 (Next Steps).

  • Resolved — the Alliance team’s systematic dashboard review (8 July 2026) logged ten issues covering navigation, filter consistency, coverage-map totals, responsiveness, and accessibility; eight are fixed and closed, and two data-completeness items (missing data links, unconfirmed contact emails) remain open pending centre outreach (#12, #13).
  • Implemented in the revised recommendations — targeted gap-filling outreach and a broader prioritisation approach that does not treat current top-three nominations as the only candidates for Phase 1 review (#5, #6).
  • Still open for discussion — complement this catalogue of assets produced with evidence on which climate datasets, boundaries, and crop or land-use maps people actually use across CGIAR (#4).
  • Still open for discussion — clarify governance for inclusion decisions, consider whether restricted-access assets should receive lower near-term priority, and distinguish underlying datasets from tools or catalogues (#1).

Interested readers can view the live GitHub discussion: all open feedback issues or #6, #5, #4, #1.

Annex A — Full asset list

All 123 catalogued assets, sortable and filterable in the HTML edition. Use the search box to find a centre, domain, or asset.

Loading ITables v2.8.1 from the init_notebook_mode cell... (need help?)

Annex B — Submission completeness

Centre Assets Has Rank Has Hub Role
AfricaRice 20 20/20 2/20
Alliance 16 16/16 3/16
CIFOR-ICRAF 13 13/13 1/13
CIP 9 9/9 1/9
ICRISAT 2 2/2 0/2
IFPRI 10 10/10 10/10
IITA 17 15/17 6/17
ILRI 11 11/11 11/11
IRRI 9 0/9 9/9
IWMI 8 2/8 2/8
WorldFish 8 8/8 8/8

Annex C — Merge log

C.1 Applied consolidations

Centre Original entries Consolidated name Entries removed Rationale
Alliance CCAFS-Climate data portal - statistical downscaling CIMP5; CCAFS-Climate data portal - statistical downscaling CIMP6; CCAFS-Climate data portal - MarkSim v2.0; CCAFS-Climate data portal -temporal bias correction of daily data CCAFS-Climate data portal (MarkSim, CMIP5/6 downscaling, bias correction) 3 Four sub-products of the same CCAFS-Climate platform submitted as separate entries
WorldFish Future fish biomass projections (DBEM); Future fish biomass projections (FishMIP) Future fish biomass projections (DBEM / FishMIP) 1 Two model runs of the same product submitted as separate entries
CIFOR-ICRAF Spatial Assessments of Changes in Soil Health Indicators in Benin; Spatial Assessments of Changes in Soil Health Indicators in Kenya Spatial Assessments of Changes in Soil Health Indicators (Benin, Kenya) 1 Same dataset applied in two countries submitted as separate entries
ILRI VECTRI Climate Forcing and Preprocessed Dataset; VECTRI Malaria Model Output Dataset for Ethiopia VECTRI Malaria Model Dataset (climate forcing inputs + model outputs, Ethiopia) 1 Model inputs and outputs from same pipeline submitted as separate entries