The UN Decade for Sustainable Transport (2026-2035): Deep Dive

1. Overview and Origins

The United Nations General Assembly proclaimed 2026-2035 as the United Nations Decade of Sustainable Transport through Resolution A/RES/79/225, adopted in late 2024. This followed years of advocacy, building on the outcomes of the Second UN Global Sustainable Transport Conference (Beijing, October 2024) and earlier milestones including the First Conference (Ashgabat, 2016) and decades of work by UNECE, UNCTAD, and other UN agencies.

The Decade represents the first time the international community has designated a full decade-long focus on transport, reflecting the sector's cross-cutting importance to virtually all 17 Sustainable Development Goals.

2. Key Resolutions and Mandates

General Assembly Resolution A/RES/79/225

The resolution establishing the Decade calls upon member states, international organisations, and stakeholders to:

• Accelerate the transition to sustainable, low-carbon, resilient transport systems
• Strengthen data collection, sharing, and analysis on transport at national and global levels
• Promote interoperability and standardisation of transport systems and data
• Enhance capacity building in developing countries, particularly Least Developed Countries (LDCs), Landlocked Developing Countries (LLDCs), and Small Island Developing States (SIDS)
• Leverage digital technologies and innovation, including AI and data analytics, to improve transport planning and operations

The Beijing Statement (2024)

The outcome document from the Second Global Sustainable Transport Conference specifically highlighted:

• The need for a global transport data framework to monitor progress
• Calls for a Global Transport Atlas or equivalent knowledge platform
• The importance of intermodal and cross-border data sharing
• Recognition that data gaps in developing countries are a critical barrier to evidence-based transport policy

Earlier Foundational Resolutions

• A/RES/72/212 (2017): Strengthening links between transport modes for SDG achievement
• A/RES/76/294 (2022): Calling for sustainable transport integration into national development strategies
• The New Urban Agenda (2016): Emphasising data-driven urban transport planning

3. Goals and Thematic Priorities of the Decade

The Decade is structured around several interconnected thematic pillars:

Pillar 1: Climate Action and Decarbonisation

• Transport accounts for approximately 24% of direct CO2 emissions from fuel combustion globally
• The sector is the fastest-growing source of emissions in many developing countries
• Goal: Align transport systems with the Paris Agreement's 1.5 degrees C pathway

Pillar 2: Safety and Health

• 1.35 million road traffic deaths annually (WHO Global Status Report)
• Target aligned with the UN Decade of Action for Road Safety and the Stockholm Declaration
• Reducing air pollution from transport, which causes an estimated 385,000 premature deaths annually

Pillar 3: Access and Equity

• Over 1 billion people lack adequate access to transport
• Rural accessibility, gender-responsive transport, and disability inclusion
• Freight connectivity for landlocked developing countries

Pillar 4: Resilience and Adaptation

• Transport infrastructure vulnerable to climate change impacts
• Supply chain resilience demonstrated as critical during COVID-19
• Disaster preparedness and recovery

Pillar 5: Innovation and Digitalisation

• This is where a Global Intelligence System fits most directly
• Harnessing AI, big data, IoT, and digital twins for transport
• Open data standards and interoperability frameworks
• Intelligent Transport Systems (ITS) deployment

Pillar 6: Governance and Finance

• Mobilising investment for sustainable transport infrastructure
• Policy coherence across transport modes and government levels
• Monitoring frameworks and indicators

4. Institutional Architecture

Lead Agencies

• UNDESA (UN Department of Economic and Social Affairs): Overall coordination of the Decade
• UNECE (UN Economic Commission for Europe): Technical standards, inland transport regulations, and the WP.29/WP.1 framework
• Regional Commissions: UNECA, UNESCAP, UNECLAC, UNESCWA each have regional transport mandates

Key Partners

• World Bank: Transport Global Practice, major funder of transport infrastructure
• World Resources Institute (WRI): Urban transport data, NUMO partnership
• GIZ (Deutsche Gesellschaft fuer Internationale Zusammenarbeit): Bilateral transport programmes, TUMI initiative
• TRL (Transport Research Laboratory): Research and evidence base, particularly for LMICs
• ITF (International Transport Forum): Transport statistics and policy analysis at OECD
• FCDO (UK Foreign, Commonwealth & Development Office): Bilateral funding for transport in developing countries

The Role of Academia

• University of Oxford: Transport Studies Unit, contributions to evidence synthesis
• Imperial College London, MIT, Tsinghua: Key research nodes
• Demand for better data to support research and evidence-based policy

5. The Data and Evidence Dimension

What the Decade Needs from Data

The Decade's monitoring framework requires:

1. Baseline data: Current state of transport systems in all countries
2. Progress indicators: Tracking decarbonisation, safety improvements, access expansion
3. Comparable metrics: Standardised definitions across countries and modes
4. Timeliness: Annual or more frequent updates, not decade-lag census data
5. Disaggregation: By mode, geography, gender, income group

Current Data Architecture (Gaps)

The existing global transport data landscape is fragmented:

The Case for a Global Intelligence System

The Decade cannot achieve its goals without a step change in transport data. A Global Intelligence System would:

• Serve as the data backbone for monitoring Decade progress
• Harmonise disparate data sources into a coherent, queryable platform
• Democratise access so that LDC policymakers have the same evidence base as OECD countries
• Enable AI-powered analysis to surface insights from complex, multimodal datasets
• Provide the "Transport Atlas" called for in UN resolutions and conference outcomes

6. Political Momentum and Windows of Opportunity

2026: The Launch Year

• The Decade formally begins, creating maximum political attention
• Opportunity to position the Global Intelligence System as a founding initiative
• UN Secretary-General's report on the Decade will likely highlight data gaps

Key Milestones Ahead

• 2026: Decade launch events, first Action Plan expected
• 2027: Expected mid-term review of SDG transport indicators
• 2028: COP33, continued pressure on transport decarbonisation data
• 2030: SDG target year -- demand for evidence on transport-related goals
• 2035: Decade conclusion -- need for comprehensive impact assessment

Alignment with Other Frameworks

• Paris Agreement: Enhanced NDCs increasingly include transport targets needing monitoring
• Sendai Framework: Disaster-resilient transport data
• Addis Ababa Action Agenda: Financing sustainable transport infrastructure
• Global Biodiversity Framework: Transport infrastructure impact on ecosystems

7. Implications for the Global Intelligence System

The UN Decade creates a unique, time-bound opportunity:

1. Institutional mandate: UN resolutions specifically call for better transport data systems
2. Political cover: A decade-long framework provides sustained attention, not a one-off event
3. Coalition of the willing: Key organisations (WRI, GIZ, TRL, UNECE, UNDESA, Oxford) are already engaged
4. Funding catalysis: Decade status unlocks funding from multilateral, bilateral, and philanthropic sources
5. Urgency: The 2030 SDG deadline is approaching with major data gaps still unfilled

The question is not whether a global transport intelligence system is needed -- the Decade's existence confirms that it is. The question is whether the proposed coalition can build it effectively, sustainably, and inclusively.

Document prepared: February 2026 Sources: UN General Assembly resolutions, Beijing Conference outcomes, UNDESA transport programme documentation, UNECE transport statistics framework

AI and LLM Applications in Transport: Use Cases and Opportunities

1. The Emerging Landscape

The convergence of large language models (LLMs), geospatial AI, and open transport data is creating new possibilities for how transport data is accessed, analysed, and acted upon. This document surveys current and emerging applications, with a focus on how they inform the design of a "Transport Angel" chatbot and broader Global Intelligence System.

2. Natural Language Querying of Transport Data

The Core Opportunity

Traditional transport data analysis requires GIS expertise, SQL knowledge, or proprietary software skills. LLMs enable natural language interfaces to complex datasets, dramatically lowering the barrier to evidence-based decision-making.

Current Examples and Approaches

Text-to-SQL and Text-to-API

• How it works: A user asks a question in plain English (e.g., "What is the average commute time in Nairobi?"). An LLM translates this into a structured query against a database or API, executes it, and returns results in natural language.
• Precedents:
  • Google BigQuery now supports natural language querying via Gemini integration
  • Microsoft Fabric's Copilot enables conversational queries of analytics datasets
  • Numerous open-source projects (e.g., LangChain SQL agents, LlamaIndex) provide frameworks for building text-to-SQL pipelines
  • Databricks has integrated LLM-based querying into their lakehouse platform

Retrieval-Augmented Generation (RAG) for Transport Documents

• How it works: Thousands of transport policy documents, reports, and standards are indexed. When a user asks a question, the system retrieves relevant passages and uses an LLM to synthesise an answer with citations.
• Transport applications:
  • Querying across the corpus of National Urban Mobility Policies
  • Searching UNECE transport regulations and conventions
  • Synthesising findings from transport safety research databases
  • Cross-referencing Nationally Determined Contributions (NDCs) for transport commitments

Conversational Data Exploration

• Moving beyond single queries to multi-turn conversations where users can refine, drill down, compare, and visualise data iteratively
• Example flow: "Show me road fatality rates in East Africa" -> "Compare that to Southeast Asia" -> "What are the main contributing factors?" -> "Which countries have seen the biggest improvements?"

Technical Architecture Patterns

User Query (Natural Language)
    |
    v
Intent Classification + Entity Extraction
    |
    v
[Route to appropriate backend]
    |---> Structured Data (SQL/API query)
    |---> Document Corpus (RAG retrieval)
    |---> Geospatial Data (Map query)
    |---> Real-time Feeds (Streaming data)
    |
    v
LLM Synthesis + Citation
    |
    v
Response (Text + Visualisation + Sources)

3. Geospatial AI for Transport

Satellite and Remote Sensing Analysis

• Road network extraction: AI models (e.g., from Meta/Facebook AI, Microsoft) can now extract road networks from satellite imagery with high accuracy
• Traffic estimation from space: Emerging techniques use satellite imagery to estimate traffic volumes, particularly valuable where no sensor infrastructure exists
• Infrastructure condition assessment: ML models detect road surface quality, bridge conditions, and other infrastructure states from imagery
• Relevance: These techniques could fill massive data gaps in countries where ground-based monitoring is absent

Key Projects and Platforms

• Google Open Buildings / Roads: AI-derived building footprints and road networks across Africa, South Asia, and Southeast Asia
• Meta/Facebook Map With AI (RapiD editor): AI-assisted road mapping contributed to OpenStreetMap
• Microsoft Road Detections: ML-derived road data for underserved regions
• Mapillary: Street-level imagery platform enabling AI analysis of road infrastructure
• Development Seed: AI-powered geospatial tools used by development organisations
• Allen AI / Satlas: Foundation models for satellite imagery analysis

Urban Mobility Pattern Analysis

• Mobile phone data analytics: Companies like Carto, Unacast, and Veraset analyse anonymised mobile location data to understand travel patterns
• Transit data platforms: Platforms like Remix (by Via), Conveyal, and TransitMatters use data analytics for transit planning
• Shared mobility data: MDS (Mobility Data Specification) and GBFS (General Bikeshare Feed Specification) standards enable aggregation of micromobility data

4. AI for Transport Safety

Predictive Safety Analysis

• Crash prediction models: ML models identifying high-risk road segments based on geometric, traffic, and environmental features
• iRAP (International Road Assessment Programme): Uses AI and street-level imagery to rate road safety and estimate crash risk globally
• Video-based conflict analysis: AI systems analysing traffic camera footage to detect near-misses and predict crash hotspots before incidents occur

Natural Language Processing for Safety

• Crash report analysis: NLP models extracting structured data from unstructured police crash reports
• News monitoring: AI systems scanning local news for transport safety incidents to supplement official statistics
• Social media analysis: Detecting transport disruptions and safety concerns from social media posts

5. AI for Transport Planning and Operations

Demand Forecasting and Modelling

• Traditional transport models (four-step, activity-based) are being augmented with ML techniques
• Deep learning for travel demand: Neural networks predicting origin-destination flows, mode choice, and route selection
• Generative AI for scenario planning: LLMs helping planners explore "what if" scenarios in natural language

Optimisation

• Network design: AI-optimised public transit routes and schedules (e.g., Optibus, Spare Labs)
• Traffic signal control: Reinforcement learning for adaptive signal timing (e.g., Google's Project Green Light)
• Freight logistics: Route optimisation, load planning, and supply chain resilience
• EV charging infrastructure: AI-optimised placement of charging stations based on demand patterns

Real-time Operations

• Incident detection and response: AI systems detecting incidents from sensor and camera data
• Predictive maintenance: ML models for transport infrastructure and fleet maintenance scheduling
• Dynamic pricing: AI-driven congestion pricing and demand management

6. The "Transport Angel" Concept

What It Could Be

A "Transport Angel" chatbot for the Global Intelligence System would serve as the primary interface for diverse users to interact with the world's transport data. The concept draws on several emerging patterns:

User Personas and Use Cases

Technical Design Considerations

Multi-modal Reasoning

The chatbot must integrate across:

• Structured data: Statistics, indicators, time series
• Unstructured data: Reports, policies, regulations
• Geospatial data: Maps, routes, spatial analysis
• Real-time data: Where available, live feeds and alerts

Multilingual Capability

• UN working languages minimum (English, French, Spanish, Arabic, Chinese, Russian)
• Ideally extending to languages of key user communities
• Current LLMs have strong multilingual capabilities but variable quality

Citation and Transparency

• Every response must be traceable to source data
• Confidence levels should be communicated
• Users must be able to drill down from summary to underlying data
• Hallucination mitigation is critical for a system used in policy decisions

Guardrails and Accuracy

• Transport data has real-world consequences (safety, investment, policy)
• The system must clearly distinguish between data-backed answers and analytical inferences
• Human-in-the-loop validation for high-stakes queries
• Regular benchmarking against expert-validated answers

7. Precedents for AI-Powered Data Chatbots

In the Development Sector

• World Bank DataBot: Experimental chatbot for querying World Bank Open Data
• UNICEF Magicbox: Data science platform with conversational elements for child welfare data
• UN Global Pulse: Using AI and big data for development insights
• HDX (Humanitarian Data Exchange): OCHA's platform, increasingly adding AI-assisted search

In the Commercial Transport Sector

• Google Maps Platform: Natural language search for transport options, increasingly AI-powered
• Moovit (Intel): AI-driven transit information across 3,400+ cities
• Citymapper: AI-powered journey planning with natural language interaction
• Waze: Crowdsourced traffic intelligence with increasingly sophisticated AI

In Adjacent Sectors

• Climate TRACE: AI-powered emissions tracking, including transport (see comparable-systems.md)
• Resource Watch (WRI): Data platform with increasingly AI-assisted exploration
• Google Earth Engine: Geospatial analysis platform adding AI capabilities

8. Technical Maturity Assessment

9. Risks Specific to AI in Transport

Hallucination and Misinformation

• An AI system that gives incorrect transport safety statistics could lead to misallocated resources
• Fabricated citations to non-existent reports could undermine trust
• Mitigation: Strict RAG architecture, source citation requirements, confidence scoring

Bias and Equity

• LLMs may perform better for data-rich countries than data-poor ones
• English-language bias in training data could disadvantage non-English speaking users
• Urban bias in available data could marginalise rural transport needs
• Mitigation: Deliberate inclusion of LMIC data, multilingual testing, rural data partnerships

Data Privacy

• Some transport data (e.g., mobile phone-derived mobility data) raises privacy concerns
• Different jurisdictions have different data protection requirements
• Mitigation: Privacy-by-design, aggregation thresholds, compliance with GDPR and local equivalents

Dependency and Sustainability

• Reliance on commercial LLM APIs (OpenAI, Anthropic, Google) creates vendor dependency
• API costs could become unsustainable at scale
• Mitigation: Open-source model options, hybrid architecture, cost-sharing model

10. Recommendations

1. Start with RAG over documents: The most mature and lowest-risk application; index the existing corpus of transport reports and enable natural language querying
2. Add structured data querying: Build text-to-SQL capabilities against well-curated, standardised datasets
3. Layer in geospatial intelligence: Partner with satellite/AI organisations for gap-filling in data-poor regions
4. Design for trust: Every response cited, confidence levels shown, expert validation loops built in
5. Open architecture: Use open standards, open-source components where possible, avoid lock-in
6. Iterative deployment: Start with a limited user group (e.g., UN analysts), gather feedback, expand

Document prepared: February 2026 Sources: Industry reports, academic literature on AI in transport, platform documentation, development sector AI initiatives

Comparable Global Intelligence Systems: Precedents from Other Sectors

1. Introduction

The proposed Global Intelligence System for Transport is not without precedent. Several other sectors have built -- or are building -- global data platforms that aggregate disparate sources, create interoperability, and provide analytical tools for policymakers. This document examines the most relevant precedents, extracting lessons for the transport initiative.

2. Health: WHO Global Health Observatory (GHO)

Overview

The WHO Global Health Observatory is the most established example of a UN-led global data platform serving as the definitive source for a sector's statistics.

Architecture

• Data sources: 194 member states submit data through standardised reporting mechanisms
• Indicators: 1,000+ health indicators covering mortality, morbidity, health systems, risk factors
• Data pipeline: National reporting -> WHO verification -> Harmonisation -> Publication
• User interface: Web portal with visualisations, data downloads, API access
• Update cycle: Annual for most indicators, with some near-real-time (e.g., disease surveillance)

What Works

• Institutional mandate: WHO's constitution gives it authority to request health data from member states
• Standardised classifications: ICD (International Classification of Diseases) provides a universal taxonomy
• Capacity building: WHO invests in national statistical capacity alongside the platform
• Political accountability: Health ministers are answerable for their data submissions
• Long-term funding: Core WHO budget, supplemented by donor contributions

What Struggles

• Data quality: Many LMICs lack vital registration systems; data is modelled/estimated
• Timeliness: Typical 2-3 year lag between data collection and publication
• Interoperability: Despite standards, significant harmonisation effort required
• User experience: The portal is functional but not intuitive for non-specialists
• AI integration: Limited; the platform remains largely a traditional data warehouse

Lessons for Transport

• A mandated reporting mechanism is critical but takes decades to build
• Taxonomies and classifications (equivalent of ICD for transport) are foundational
• Estimation and modelling will be necessary where data is absent
• Consider building AI/chatbot capabilities from the start, not as an afterthought
• The WHO model shows that a global data platform can achieve authoritative status, but requires sustained institutional commitment

3. Climate: Copernicus Climate Data Store (CDS)

Overview

The Copernicus Climate Data Store, operated by ECMWF (European Centre for Medium-Range Weather Forecasts) for the EU, is perhaps the most technically sophisticated global environmental data platform.

Architecture

• Data sources: Satellite observations, in-situ measurements, reanalysis datasets, climate model outputs
• Volume: Petabytes of climate and environmental data
• Toolbox: Users can process data using a cloud-based Python toolbox without downloading
• API: Comprehensive programmatic access
• Update cycle: Near-real-time for some datasets, monthly/annual for others

What Works

• Compute brought to the data: Users process data in the cloud rather than downloading massive files
• Consistent reanalysis: ERA5 reanalysis provides gridded global data back to 1940
• Free and open access: All data freely available under open licences
• High technical quality: ECMWF's scientific credibility underpins the platform
• Strong funding: Copernicus programme has multi-billion euro EU commitment

What Struggles

• Complexity: Steep learning curve for non-climate scientists
• European governance: Global data, but EU-governed -- tensions around global ownership
• Developing country access: Bandwidth and technical capacity barriers
• Limited AI integration: The Toolbox is powerful but requires programming skills

Lessons for Transport

• The "bring compute to data" model is relevant for large transport datasets
• Reanalysis-type products could be valuable (e.g., reconstructing historical transport patterns from partial data)
• The platform shows that open access + high quality can coexist
• Governance must be genuinely global, not dominated by one region
• A "Transport Data Store" could follow the CDS architectural model but with a more accessible interface (the Transport Angel chatbot)

4. Climate: Climate TRACE

Overview

Climate TRACE (Tracking Real-time Atmospheric Carbon Emissions) is a coalition using AI and satellite data to independently monitor global greenhouse gas emissions at the facility level.

Architecture

• Data sources: Satellite imagery, remote sensing, public records, AI inference
• Coverage: 352 million+ assets tracked globally across all sectors, including transport
• AI models: Machine learning models estimate emissions from observable indicators (e.g., inferring factory output from heat signatures)
• Update cycle: Near-real-time, independent of government reporting

What Works

• Independence from government reporting: Can verify or fill gaps in national inventories
• AI-native design: Built from the ground up around ML/AI capabilities
• Granular spatial resolution: Facility-level and road-segment-level data
• Coalition model: Diverse partners including Al Gore's organisation, WattTime, Carbon Mapper, academic institutions

What Struggles

• Accuracy debates: Model-derived estimates are contested by some national governments
• Sustainability: Heavy reliance on philanthropic funding
• Ground truth validation: Limited ability to verify AI estimates in data-poor regions
• Political sensitivity: Countries may resist independent monitoring

Lessons for Transport

• AI-native architecture is viable and can leapfrog traditional data collection
• Satellite-derived transport data (traffic volumes, fleet composition, infrastructure) is a real possibility
• The coalition model (multiple organisations contributing AI models) could work for transport
• Must navigate political sensitivities around independent vs. government-reported data
• Transport Angel could integrate Climate TRACE transport data as a source

5. Agriculture: CGIAR Platform for Big Data in Agriculture

Overview

CGIAR (formerly the Consultative Group on International Agricultural Research) has built a platform for agricultural data sharing across its 15 research centres and beyond.

Architecture

• GARDIAN (Global Agricultural Research Data Innovation and Acceleration Network): Metadata search engine across 800,000+ agricultural datasets
• AgriLAC Data Space: Regional data space for Latin America and the Caribbean
• Crop modelling platforms: Shared computational tools for agricultural research
• Ontologies: AGROVOC and other standardised vocabularies

What Works

• Federated model: Data stays with the source; the platform provides discovery and access
• Ontologies and metadata standards: AGROVOC provides a multilingual agricultural thesaurus
• Open access policy: CGIAR's Open Access/Open Data policy mandates sharing
• Research orientation: Strong link between data platform and research questions

What Struggles

• Fragmentation: Despite the platform, many datasets remain siloed
• Data quality variability: Wide variation in quality across sources
• User adoption: Researchers often prefer familiar tools over the platform
• Sustainability: Dependent on CGIAR reform and funding cycles

Lessons for Transport

• A federated model (metadata and access layer over distributed data) may be more realistic than centralisation
• Transport ontologies and vocabularies are an essential prerequisite
• Open data policies among partners must be agreed upfront
• The platform must be useful enough that users prefer it to alternatives
• GARDIAN's approach of being a discovery layer could inform the Transport Intelligence System architecture

6. Humanitarian: OCHA Humanitarian Data Exchange (HDX)

Overview

The Humanitarian Data Exchange (HDX), run by OCHA's Centre for Humanitarian Data, is a platform for sharing humanitarian data across organisations.

Architecture

• Open platform: CKAN-based data catalogue
• Coverage: 20,000+ datasets from 250+ organisations across 250+ locations
• Data quality framework: Tiered quality assurance process
• HDX HAPI (Humanitarian API): Standardised API for key humanitarian indicators
• Data stories: Analytical products built on platform data

What Works

• Community-driven: Organisations voluntarily contribute data
• Quality tiers: Clear framework for data quality expectations
• API-first design: HDX HAPI provides standardised programmatic access
• Rapid adoption: Has become the go-to platform for humanitarian data
• Low barrier: Easy to contribute and discover data

What Struggles

• Voluntary contributions: Completeness depends on willingness of organisations
• Data sensitivity: Humanitarian data can put vulnerable populations at risk
• Sustainability: OCHA budget constraints affect the platform
• Limited analytics: More of a data catalogue than an analytical platform

Lessons for Transport

• A community-driven, low-barrier approach can achieve rapid adoption
• Data quality frameworks are essential and should be built in from the start
• An API-first design enables others to build on the platform
• The HDX model could work for transport as an initial data-sharing layer
• The Transport Angel could be the analytical layer on top of an HDX-like data sharing platform

7. Statistics: UN Global Platform for Official Statistics

Overview

The UN Global Platform, hosted by UNSD, provides a cloud-based environment where National Statistical Offices (NSOs) can collaborate on big data for official statistics.

Architecture

• Cloud infrastructure: Shared computational environment
• Collaborative workspace: NSOs can share code, methods, and experimental data
• Pilot projects: Testing use of big data sources (satellite, mobile phone, scanner data) for official statistics
• Capacity building: Training for NSOs in data science methods

What Works

• NSO engagement: Builds trust with official statistical community
• Shared infrastructure: Addresses capacity constraints in poorer NSOs
• Experimentation: Safe space to test new methods without affecting official statistics
• UN branding: Gives legitimacy and neutrality

What Struggles

• Adoption: Uneven uptake across countries
• Official statistics conservatism: NSOs are risk-averse, slow to adopt new methods
• Governance: Complex multi-stakeholder governance
• Limited public-facing products: More of a collaboration tool than a public data platform

Lessons for Transport

• NSO engagement is important if transport data is to gain official statistics status
• Shared cloud infrastructure can level the playing field for developing country partners
• The platform shows the value of collaborative method development
• Transport data could be a pilot theme within the UN Global Platform

8. Oceans: Global Fishing Watch

Overview

Global Fishing Watch uses satellite data and AI to track fishing activity worldwide, providing transparency on ocean resource use.

Architecture

• AIS data: Automatic Identification System signals from vessels
• Satellite imagery: Synthetic aperture radar and optical imagery for detecting vessels
• AI models: Classification of vessel behaviour (fishing, transiting, anchoring)
• Open platform: Free access to global maps and data downloads

What Works

• Radical transparency: Making previously opaque data publicly available
• AI-native: ML at the core of data production
• Impact: Used by governments to detect illegal fishing, cited in policy decisions
• Partnership model: Google, Oceana, SkyTruth as founding partners

What Struggles

• AIS limitations: Not all vessels transmit AIS; deliberate disabling common
• Verification: Satellite-derived classifications have error rates
• Political pushback: Some nations resist external monitoring of their waters
• Scope creep: Pressure to expand beyond fishing to other maritime activities

Lessons for Transport

• Satellite + AI + open access can transform transparency in a sector
• AIS for shipping is analogous to ADS-B for aviation or GTFS for transit -- existing data signals that can be aggregated
• The model of using AI to classify and interpret raw signals could apply to transport
• Expect political pushback when making data transparent that was previously opaque
• Demonstrates that a focused coalition can build a globally significant platform

9. Biodiversity: Global Biodiversity Information Facility (GBIF)

Overview

GBIF is an international network and data infrastructure providing free, open access to biodiversity data.

Architecture

• Federated network: 1,800+ data publishing institutions, 100+ participating countries
• Standardised format: Darwin Core standard for biodiversity data
• Data volume: 2.7 billion+ species occurrence records
• Governance: Intergovernmental memorandum of understanding
• Participant funding: Countries fund participation through national nodes

What Works

• Standardisation: Darwin Core provides a universal data exchange format
• Massive scale: Decades of consistent effort have built enormous data volume
• Institutional commitment: Formal government participation
• Incentive alignment: Researchers incentivised to contribute (data DOIs enable citation credit)

What Struggles

• Taxonomic and geographic bias: Better data for well-studied groups and regions
• Data quality: Citizen science data varies widely in quality
• Funding: Core infrastructure funding perpetually under pressure
• Timeliness: Much data is historical occurrence records, not real-time

Lessons for Transport

• Data standards (equivalent of Darwin Core) are the foundation of interoperability
• Incentive structures must reward data contribution
• National nodes could be a model for transport data governance
• Intergovernmental MoU provides legitimacy and commitment
• Demonstrates that federated, open data can achieve global scale over time

10. Synthesis: What Makes These Systems Succeed or Fail

Success Factors

Critical Success Factors for a Transport System

Based on the evidence from comparable systems:

1. Data standards are non-negotiable: Every successful system has a foundational data standard. Transport needs its equivalent (building on GTFS, NeTEx, DATEX II, CDS, SIRI, etc.)

2. Federated beats centralised: Most successful systems are federated (data stays distributed, the platform provides discovery and access) rather than centralised

3. Institutional mandate matters: The stronger the mandate (ideally from an intergovernmental body), the more sustainable the platform

4. Open access drives adoption: Platforms that are freely accessible achieve greater scale and impact

5. AI integration is the frontier: None of the existing systems have deeply integrated AI/LLM capabilities -- this is an opportunity for the transport system to leapfrog

6. Funding is the Achilles heel: Most platforms struggle with sustainable core funding. The transport system needs a clear business model from the start

7. Start useful, scale ambitious: Platforms that launched with immediate utility (even if limited) and scaled over time are more successful than those that tried to boil the ocean

8. Community and incentives: Users must be incentivised to both contribute and use the platform

11. Recommended Architecture Drawing on Precedents

Based on the analysis, the Global Intelligence System for Transport should consider a layered architecture drawing on the best elements of each precedent:

Document prepared: February 2026 Sources: Platform documentation, governance frameworks, published evaluations, and academic analyses of each system