UNFCCC Indigenous Climate Resource Hub Updates: Resilience Action Cards Linking Community Observation, Official Water Data and Cultural Permissions

A useful climate-technology service for an Indigenous township does not necessarily begin with another map. It can begin by putting official risk data, long-term community observations, cultural permissions, service contacts and the next practical action into one record that can be updated and traced.

The UNFCCC Local Communities and Indigenous Peoples Platform (LCIPP) continues to organize resources on Indigenous knowledge and climate action. The LCIPP Resource Hub showed recently uploaded resources on August 31 and, in late August, grouped Pacific materials on culture, education, nature conservation and resilient development. These resources do not prescribe one operating model for every community. They instead offer a useful principle for local digital services: climate adaptation can connect scientific information, local and Indigenous knowledge, cultural practice, governance and community-led action without collapsing them into one undifferentiated data set.

That approach was also visible in the 2026 Pacific Regional Gathering, held in Nadi, Fiji, from August 18 to 21. Public UNFCCC material for the LCIPP Regional Gathering for the Pacific brought together knowledge holders, capacity-building, climate action, living cultural practices and non-economic loss. For Taiwan's Indigenous townships, the most transferable lesson is not a particular institutional label. It is the idea that community knowledge should participate in climate governance through agreed processes, rather than being extracted into a database simply because it can be digitized.

A practical local version can be a “resilience action card.” One card handles one place, hazard or service scenario—for example heavy-rain risk on an access road, heat affecting older adults and crops, rising river levels, short-duration mountain rainfall, a cold-chain interruption or a delayed harvest. Each card can record location or service area, hazard type, official source, community observation, last update, current status, potential impact, next action, backup contact, cultural-sensitivity level, public-access permission, human reviewer and an error-report channel.

Taiwan already has public data that can form the official layer. The Central Weather Administration Open Data Platform provides forecasts, observations, rainfall and warning-related data and supports machine-readable access. A township can normalize township forecasts, rainfall, warning information and update times into stable fields rather than build a new monitoring system. Water information can be linked through the Water Resources Agency IoT platform, which provides water-related observations and sensor services. A chatbot should present source, timestamp and status; it should not independently declare that flooding will occur, that a road is safe, or that an evacuation is required.

The second layer is community observation. Long experience may identify patterns that are not represented by one sensor: a slope that often becomes unstable after sustained rain, a change in river color or sound, a shift in plant phenology, or a seasonal change affecting farming, fishing or gathering. These observations do not need to compete with official data for a single “truth” field. They can be stored separately with observer, time, place, purpose and community review status. The service can then show official measurements and local observations side by side, preserving their different roles.

The third layer is cultural permission. Climate adaptation can intersect with ceremonial timing, gathering places, ancestral routes, medicinal-plant locations, fishing grounds, burial places or family knowledge. Not all such material belongs in a public retrieval system. A township or community can classify data as public service, internal service, designated-role access, re-consent required, or no-AI. This operationalizes principles found in the Global Indigenous Data Alliance CARE Principles and Local Contexts: collective benefit, authority to control, responsibility, ethics, provenance and community-defined use conditions.

RAG and chatbots should therefore come after permissions. Questions about public weather, water information, service contacts or open agricultural adaptation material can retrieve from a public layer. Questions involving sensitive traditional knowledge should pass through a permission policy before retrieval. Depending on the rule, the system may provide only an open summary, show a contact point, require an authorized role, or generate no answer at all. Every response should ideally show the official-data time, community-observation time, sources, human-review status and a correction link. This makes information age and responsibility visible to both residents and township staff.

The Pacific region also offers a broader planning reference. SPREP's description of the Framework for Resilient Development in the Pacific 2017–2030 integrates climate-change adaptation and disaster-risk management within a regional resilience approach. For local deployment, the useful lesson is to treat preparation, risk information, continuity of service and recovery as one workflow. The Pacific Islands Framework for Nature Conservation and Protected Areas similarly emphasizes community participation, governance and natural-resource management, which makes the same action-card structure relevant to forests, rivers, coasts, fisheries, gathering and agricultural landscapes.

A three-month MVP can remain small: one township, two or three communities and three frequent hazards, such as heavy rain/access roads, heat/agriculture and river/water levels. Build 30 to 50 action cards and 10 to 20 human-reviewed community observations that are public or explicitly authorized. Official sources can initially be imported manually or through APIs. The essential requirement is that every record has a date, source, owner, permission and expiration condition. At first, the chatbot only needs to answer what information is currently available, where it came from, when it was updated and which service contact is next. It should not replace official alerts or disclose restricted knowledge.

Seven fields the 55 Indigenous townships can use immediately

1. Where the risk is: community, road, river, farmland, fishing ground or public-service area.

2. What the official source says: weather, water, road, disaster or agricultural-service source and timestamp.

3. What the community observes: long-term local signals, seasonal change and human reviewer.

4. What happens next: check, prepare, contact, activate backup, continue observing or transfer to the responsible authority.

5. Who may see it: public, internal, designated role, re-consent required or no-AI.

6. When it expires: latest update, validity period, next review and withdrawal path.

7. What AI may do: organize, compare, retrieve and remind; formal warnings, evacuation and cultural judgments remain human responsibilities.

Fixing these fields first makes it possible to connect international Indigenous climate experience, Taiwan's official open data and locally governed community observations into a service that can be maintained over time. Models, platforms, GIS layers and IoT devices can change later without rebuilding the underlying governance structure.

Sources

• UNFCCC LCIPP | Resource Hub
• UNFCCC LCIPP | Regional Gathering for the Pacific
• SPREP | Framework for Resilient Development in the Pacific
• Pacific Islands Roundtable for Nature Conservation | Pacific Islands Framework for Nature Conservation and Protected Areas
• Central Weather Administration | Open Data Platform
• Water Resources Agency | Water Resources IoT
• Global Indigenous Data Alliance | CARE Principles
• Local Contexts