top of page

CitizenClimate — Data Integrity & Methodology Protocol

For Verification & Validation Bodies (VVBs) · Document Version: October 2026 (app version 1.25)

​

What's new since the April 2026 version

  • Project site boundaries: upload a KML, KMZ or GeoJSON file, or draw the boundary on a map (Section 12)

  • Daily AI verification of new submissions; faulty entries are archived with a reason and their rewards reversed (Section 13)

  • Plan Vivo (PV Nature) biodiversity module: enrolment, Target Groups and automatic GBIF taxonomy (Section 14)

  • Plan Vivo Pillar Metrics 1–5, including satellite pillars from Sentinel-2 imagery, with a permanent run log and downloadable audit bundles (Section 15)

  • SDG dashboard: side-by-side community comparison for projects tracking one or two SDGs (Section 9)

  • New screenshots throughout, taken from app version 1.25

​

Screenshots come from the Usambara Forest Corridor Restoration demonstration project in Tanzania. Its field records are simulated for demonstration; its satellite metrics are computed from real Sentinel-2 imagery.

Summary for Auditors

Data Type
Deletable?
Timestamped?
Georeferenced?
Photo Evidence?
AI Verified?
Survey submissions
No
Yes
Per survey (required or optional)
Where configured
Species fields only
Faulty submissions
Archived only (with reason)
Yes
Preserved
Preserved
Preserved
Species observations
No
Yes
Yes
Yes
Yes
Grievances
No
Yes
No
Optional
No
SDG survey responses
No
Yes
Yes
Where configured
No
Biodiversity records
No
Yes
Yes
Yes
Yes
Media files
No
Yes (upload metadata)
Via parent submission
N/A
N/A

Key assurances for verification bodies:

  1. Every record in the system has a creation timestamp that cannot be modified

  2. No data is ever deleted — only archived with an auditable reason

  3. GPS coordinates are captured at the device level at the time of field entry, not retrospectively

  4. All AI species identifications are performed by independent third-party services and stored verbatim

  5. Grievances are permanently recorded and cannot be suppressed by Project Developers

  6. VVBs have direct read access to all project data — they do not rely on PD-generated reports

  7. All data exports (CSV/JSON) include raw records, not just summaries

  8. Every new submission with photos, recordings or free text is screened daily by an automated AI reviewer; faulty entries are archived with a reason, never deleted

  9. Plan Vivo Pillar Metrics are calculated by the platform's servers, never typed in; every run is logged permanently and can be re-run independently from its audit bundle

  10. Each project's site boundary is stored with its area in hectares and its source (uploaded file or drawn on the map)

​

CitizenClimate is built on Google Firebase infrastructure (Firestore, Firebase Storage, Firebase Authentication, Cloud Functions). Data residency and Google Cloud audit logging apply in addition to the application-level protections described above.

01_signup.png

1. What CitizenClimate Is

CitizenClimate is a digital Monitoring, Reporting and Verification (MRV) platform designed for nature-based carbon credit projects. It replaces paper-based field forms and manual data entry with a mobile app that works in remote areas without internet access.

​

The platform supports three user roles:

  • Community Members (CM) — field data collectors on the ground

  • Project Developers (PD) — organisations managing the carbon project

  • Verification & Validation Bodies (VVB) — independent auditors who review data

​

All three roles can see the same underlying records. VVBs have read access to everything: submissions, photos, AI species identifications, SDG scores, grievances, and biodiversity observations.

cc_02_survey_options.jpg

2. How Data Cannot Be Erased

​

All submitted data is stored in Google Cloud Firestore, a cloud database with the following permanent protections written into the system's security rules:

  • Submissions can never be deleted. The database's security rules are enforced on Google's servers, not in the app, and they block deletion of every submission for every user: Community Members, Project Developers and VVBs alike.

  • Survey templates cannot be deleted from the app. A Project Developer can edit a survey, for example to add a question or change a reward, but each submission stores the questions and answers exactly as they were given, so editing a survey never changes data already collected.

  • Grievances can never be deleted. Once a community member raises a concern, it exists permanently in the system.

  • Photos and videos can never be deleted. Once uploaded to Firebase Storage, media files have no delete endpoint exposed in the application or its backend functions.

  • Forum posts can never be deleted.

  • Plan Vivo metric results and their run log can be written only by the platform's servers. No user, including the Project Developer, can enter or change a metric value (see Section 15).

​

The only action a Project Developer can take on a submission they believe to be erroneous is to archive it with a reason — explained in Section 3 below.

cc_03_archive.jpg

3. Faulty Data Is Archived, Not Deleted

​

If a Project Developer identifies a submission as problematic (e.g. a duplicate, a test entry, or a suspected error), they cannot delete it. Instead, they archive it. Archiving adds three new fields to the original, untouched record:

  • archived — marks the record as archived

  • archiveReason — the reason, chosen from a fixed list: Duplicate submission, Incorrect data, Test submission, Incomplete submission, Fraudulent entry or Other

  • archiveNotes — additional context or notes

​

The original submission data, timestamp, location, photos, and answers are fully preserved. An archived submission is visible to VVBs with the archive reason attached. This means the audit trail shows exactly what was collected, why it was set aside, and who made that decision.

​

Community Members cannot archive their own submissions. Only Project Developers have this permission, and they cannot change the underlying data — only flag it.

​

Archiving by the daily AI check. The automated reviewer described in Section 13 uses the same mechanism. A submission it rejects is archived with one of the standard reasons above, and its notes begin with "AI verification:" followed by what the reviewer saw. The original record, photos and answers stay intact and visible to VVBs.

cc_02_list.jpg

4. All Data Is Timestamped

Every record in the system carries a submittedAt timestamp generated at the moment of submission:

  • Format: ISO 8601 (e.g. 2025-04-05T14:32:11.000Z) — a universal standard readable by any audit tool

  • Precision: Milliseconds

  • Source: The device clock at the moment the user taps Submit, captured before it leaves the app

​

Additional timestamps are recorded for:

  • Grievance raised — dateCreated (cannot be changed)

  • Grievance reviewed by the PD — dateResponded (set automatically)

  • Grievance resolved — dateResolved (set automatically)

  • Badge awarded — awardedAt

  • Lesson completed — completedAt

  • Media file uploaded — uploadedAt (stored in the file's metadata)

  • Data exported — exportedAt

  • Species matched to the GBIF taxonomy (Plan Vivo projects) — pvTaggedAt

  • Submission checked by the daily AI verification — checkedAt

  • Plan Vivo metric run started — triggeredAt (kept in the permanent run log)

 

When data is collected offline (without internet), the device timestamp is captured at the moment of field entry. When the device reconnects and syncs, the original timestamp is preserved — the sync date is recorded separately. This means a VVB can see when a field observation was made versus when it reached the cloud.

cc_05_geo_map.jpg

5. All Data Is Georeferenced

Every data submission can carry GPS coordinates collected from the device at the time of submission. The system requests the highest available GPS accuracy from the device (navigation-grade, typically < 5 metres), with automatic fallback to lower accuracy if high precision is unavailable.

​

The following location data is captured and stored:

  • Latitude and longitude — stored as a GeoPoint object in Firestore, linked permanently to the submission

  • GPS accuracy in metres — captured alongside the coordinates so VVBs can assess the precision of each reading

  • Farm boundary polygons — where applicable, field teams trace the boundary of a plot, storing an ordered array of GPS waypoints

  • Route walkovers — transect surveys capture a continuous GPS track of the path walked during data collection

  • Project site boundary — the official project area (Section 12)

Location is required or optional per survey type, as configured by the Project Developer. When required, the system will not allow submission without a valid GPS fix. All coordinates are stored permanently and cannot be modified after submission.

Species observations (birds, plants, animals) also capture latitude and longitude at the point of detection, providing georeferenced evidence for each biodiversity record.

cc_06_photo.jpg

6. Photo and Video Evidence

The platform requires photographic evidence for designated survey fields. The following applies to all media:

Collection:

  • Photos and videos are taken in the field using the device camera or selected from the gallery

  • Image quality is configurable per survey field (low / medium / high resolution)

  • Multiple photos can be attached to a single field entry

Storage and linking:

  • Files are uploaded to Firebase Storage using a fixed path structure: submissions/{projectId}/{surveyId}/{submissionId}/{type}/{fieldName}-{timestamp}.{ext}

  • Every file is stored with metadata that permanently records: which submission it belongs to, which form field it was attached to, the original filename, the upload timestamp, and whether it came from an offline sync

  • The download URL is permanently embedded in the submission document

Immutability:

  • Files cannot be deleted through the application

  • The submission document cannot be updated to remove media URLs

  • Even if a submission is archived, all photos and videos remain fully accessible via their permanent URLs

VVBs reviewing a submission will see the photos inline and can download originals. The metadata attached to each file serves as a chain of custody record.

cc_07_ai_id.jpg

7. AI Species Identification

CitizenClimate uses AI services to identify species from photos and audio recordings. Identification is done through secure Cloud Function proxies — the AI services are never called directly from the app, and API keys are never exposed to end users.

​

Supported identification types:

  • Plants — photo — Plant.id

  • Birds — audio recording — BirdNET

  • Animals — photo — AI identification service

  • Insects — photo — AI identification service

  • Mushrooms — photo — AI identification service

  • Crop health — photo — AI identification service

 

What the AI returns:

  • Scientific name

  • Common name

  • Confidence score (0–100%)

  • IUCN conservation status (Least Concern through Extinct)

  • For birds: start and end time of the detected call within the audio, minimum confidence threshold of 25%

  • For plants: health assessment data

​

How results are stored: Each species observation is stored as a permanent record containing the scientific name, common name, category, observation count, IUCN status, the submission it came from, the timestamp of observation, GPS coordinates, observer name, and URLs to the supporting photos or audio.

​

Offline identification: If a field worker has no internet at the time of submission, the observation is saved locally with a status of ai_pending. When connectivity is restored, the AI identification is completed and the result is written permanently into the submission. The original observation timestamp — when the species was actually seen in the field — is preserved regardless of when identification occurs.

​

Rate limiting: To prevent abuse, each user is limited to 20 AI identification requests per minute. Usage is tracked per project with monthly credit allocations.

​

GBIF taxonomy: in Plan Vivo projects, every identified species name is also matched on the server to the GBIF Backbone Taxonomy, so all records use one internationally recognised naming system (Section 14).

​

Daily AI verification: separately from species identification, every new submission with photos, recordings or free text is screened once a day for irrelevant, fake or duplicated evidence (Section 13).

cc_08_grievance.jpg

8. Anonymous Grievance Mechanism

Community members have a dedicated, protected channel to raise concerns about a project. The grievance mechanism works as follows:

Raising a grievance:

  • Any community member enrolled in a project can submit a grievance at any time

  • They provide a title and a description of their concern

  • A timestamp is automatically recorded at the moment of submission

​

What is stored permanently:

  • Title — cannot be changed after creation

  • Description — cannot be changed after creation

  • Date created — cannot be changed

  • Community — cannot be changed after creation

  • Status (Pending / Under Review / Resolved) — updated by the PD

  • Resolution narrative — added by the PD

  • Date responded — set automatically when the PD reviews

  • Date resolved — set automatically when resolved

​

Anonymity in practice: The grievance form does not display the submitter's name or email address to Project Developers or other community members. The community of origin is recorded (for project context) but individual identity is not shown on the grievance card. Community members can also upvote or downvote grievances — votes are aggregated so individual voting behaviour is not visible.

​

VVB access: VVBs can read all grievances across a project, including the full history of status changes and resolution responses. This provides an independent view of whether community concerns are being addressed in good faith and within a reasonable timeframe.

cc_09_sdg.jpg

9. SDG Monitoring

The platform supports structured data collection aligned to the UN Sustainable Development Goals. Each survey can be tagged to a specific SDG (1–17), with fields and scoring weights configured by the Project Developer.

​

How scoring works:

  1. Each survey field has a set of response options, each assigned a score (0–100)

  2. Fields have relative weights (e.g. a field with weight 2 contributes twice as much as a field with weight 1)

  3. The final submission score is the weighted average of all field scores

This produces a single numeric SDG score per submission, per survey

​

What is tracked over time:

  • Scores are tracked by calendar year, allowing year-over-year comparison

  • The system calculates whether a community's average score has improved or declined compared to the previous year

  • Community-level averages are calculated separately, so a VVB can see performance broken down by community within a project

  • All raw submission data (individual question responses, not just totals) is stored and exportable

Community comparison: the SDG dashboard compares communities side by side. With three or more SDGs it uses a radar chart; for projects tracking one or two SDGs it shows grouped bars for each community, so differences between villages are easy to see.

​

Export formats: SDG data can be exported as CSV or JSON, including:

  • Summary metrics

  • Yearly trend data

  • Community-by-community breakdown

  • Full raw submission detail (every question and answer)

 

All SDG submissions follow the same immutability rules as standard submissions. Once a score is recorded, it cannot be altered.

cc_10_biodiversity.jpg

10. Biodiversity Monitoring

The platform includes a full biodiversity monitoring module that goes beyond simple species lists.

What is tracked:

  • Birds, plants, animals, insects, and mushrooms

  • Scientific name, common name, and IUCN conservation status for each species

  • Observation counts per species (multiple sightings tracked individually)

  • GPS coordinates per observation

  • Photos and/or audio recordings as evidence

  • The name of the field observer who made the identification

  • The submission it originated from

​

Diversity indices calculated: The system automatically computes recognised scientific metrics from the raw observation data:

  • Species Richness — total number of unique species

  • Simpson's Index — probability that two random individuals belong to different species

  • Shannon-Wiener Index — species diversity accounting for evenness

  • Pielou's Evenness — how evenly individuals are distributed across species

  • BII Proxy — Biodiversity Intactness Index estimate

  • Alpha Diversity — mean species richness per submission point

  • Beta Diversity — similarity between communities (Sørensen index)

  • Gamma Diversity — total unique species across the entire project

  • Endangered species: Any species with an IUCN status of Vulnerable (VU), Endangered (EN), Critically Endangered (CR), Extinct in the Wild (EW), or Extinct (EX) is automatically flagged and counted in a dedicated endangered species total.

Immutability: Every species observation is permanently linked to its parent submission. Observations cannot be deleted. If a submission is archived, its species data remains in the biodiversity record. This means the biodiversity dataset grows monotonically — records are never removed, only added.

​

Plan Vivo projects: the Biodiversity screen also opens the Plan Vivo dashboard, where the five PV Nature Pillar Metrics are calculated from the same records, using GBIF taxonomy and satellite data (Sections 14 and 15).

cc_11_sync.jpg

11. Offline-First Data Collection

A key design principle is that CitizenClimate works in remote areas with no mobile signal. When offline:

  • Surveys are pre-loaded to the device and available without internet

  • Submissions are saved to a local encrypted database on the device, with the field timestamp recorded at the moment of entry

  • Photos and media are stored locally pending upload

  • All pending submissions are visible to the field worker on a Sync Status screen showing which records are awaiting upload

​

When the device reconnects:

  • All pending submissions are uploaded in order

  • Media files are uploaded first; their URLs are then embedded in the submission

  • Badges and points are awarded retroactively

  • The sync event is logged with timestamp, submission ID, survey name, and success or failure status

​

The sync history is retained on the device (last 50 events). Failed syncs are shown to the user and can be retried.

cc_12_boundary.jpg

12. Project Site Boundary

Every project can store its official site boundary: the area the project covers and is monitored against.

​

Adding a boundary:

  • Upload a boundary file exported from GIS software or Google Earth: KML, KMZ or GeoJSON (WGS84 latitude/longitude)

  • Or draw it on a satellite map in the app: tap to place corner points, drag points to adjust, and add more parts for sites made of separate areas

​

What is stored:

  • The boundary shape, including multi-part sites

  • The total area in hectares, calculated on the curved surface of the Earth rather than on a flat map

  • The number of parts, the source (the uploaded file's name, or "drawn on map") and whether it was edited afterwards

​

Checks when a boundary is added:

  • Files with no polygon, with coordinates outside the valid latitude/longitude range, or in a projected coordinate system are refused with a clear message

  • Very detailed boundaries are simplified to fit the storage limit, and the user is told

  • Sites larger than 10,000 ha are flagged, because Plan Vivo applies a separate sampling protocol to them

​

Only the Project Developer who owns the project can add, replace or remove the boundary. The boundary is part of the project record that VVBs can read, and it defines where the Plan Vivo satellite pillars are measured (Section 15).

cc_04_detail.jpg

13. Daily AI Verification of Submissions

Besides species identification, every new submission is checked once a day by an automated AI reviewer. The check runs after submission, so field data collection stays fast and keeps working offline.

​

What is checked: submissions from the last 48 hours that contain photos, videos, recordings or free-text answers.

​

A submission is rejected only when the reviewer is sure it falls into one of these categories:

  • Irrelevant media — a photo unrelated to the survey, such as a selfie, vehicle, food, document or app screen

  • No subject — an identification photo with no organism in it, such as a blank frame or a finger over the lens

  • Wrong subject — a clear mismatch at kingdom or class level, such as a bird identification with a photo of a cow

  • Not original — a photo of a screen, poster or book page, or a stock image

  • Spam text — gibberish or filler answers with nothing genuine in the submission

  • Duplicate media — the same photo or recording reused from another submission, detected by an exact digital fingerprint of the file rather than by AI judgement

​

Safeguards against wrongful rejection:

  • Disagreeing with a species-level identification is never a reason to reject; that is left to expert quality control

  • Poor-quality photos still pass if the subject is plausible

  • A rejection needs at least 85% confidence; anything less goes to a person for review

  • Each member's submissions are reviewed separately, so one member's content cannot influence verdicts on anyone else's

  • Bird recordings are re-analysed by BirdNET, because audio cannot be judged from a picture

  • If an unusually large share of a day's verdicts would be rejections, all of them are held for human review

  • The reviewer can only read the submission; the outcome is applied by the platform's servers

​

When a submission is rejected:

  • It is archived, never deleted, with a standard archive reason (Section 3) and a note describing what the reviewer saw

  • The points and payment earned for it are reversed

  • The member receives a notification in their own language explaining why

​

Every checked submission permanently records the verdict (passed, review or rejected), the category, the confidence, the reason and the time of the check. Community Members cannot set, change or remove these fields, so a VVB can see which records were screened and with what result.

cc_14_enrol.jpg

14. Plan Vivo (PV Nature) Biodiversity Module

CitizenClimate supports projects certified under Plan Vivo's PV Nature biodiversity standard. Plan Vivo is added on top of any project, independent of its subscription plan.

​

Enrolment — the Project Developer enrols the project when creating or editing it, and records:

  • Ecosystem: terrestrial, terrestrial + freshwater, mixed habitat or marine (freshwater-only is coming soon)

  • Certificate type: restoration or conservation

  • Baseline year: the year that change is measured from

  • Target Groups: the groups of species to be monitored, taken from the PV Nature Methodology & Data Protocol. Terrestrial projects must include the two mandatory groups (herbaceous & woody plants under 2 m, and birds) and at least four groups in total; marine projects need at least three. Each group shows its monitoring method, such as field quadrat, acoustic BirdNET or camera trap. Groups that need eDNA are listed but cannot be selected yet.

  • Satellite settings (terrestrial sites): the three-month NDVI season, the cloud mask, and which land-cover classes count as low-intensity human-use habitat

​

Automatic taxonomy — every species recorded in an enrolled project is matched automatically to the GBIF Backbone Taxonomy (Global Biodiversity Information Facility), the reference taxonomy Plan Vivo requires. Each record stores the GBIF species key, how well the name matched, the full classification from kingdom to species, and its Target Group. This happens on the server straight after submission; field teams do nothing extra.

​

Quality control — every species record in an enrolled project is marked for expert quality control (pending, verified or rejected). Archived submissions, QC-rejected records, and species that cannot be matched to GBIF or to a Target Group are left out of the metric calculations.

cc_15_dashboard.jpg

15. Plan Vivo Pillar Metrics & Auditor Dashboard

The five PV Nature Pillar Metrics:

  • Pillar 1 — Species Richness: the estimated number of species, adjusted for sampling effort. Field records, every year.

  • Pillar 2 — Species Diversity: how many species there are and how evenly individuals are spread among them (Shannon diversity). Field records, every year.

  • Pillar 3 — Taxonomic Dissimilarity: how distinct the recorded species are from each other in the tree of life. Field records and GBIF taxonomy, every year.

  • Pillar 4 — Habitat Health: vegetation condition across the site, from Sentinel-2 satellite greenness (NDVI) inside the site boundary. Satellite, every year.

  • Pillar 5 — Habitat Spatial Structure: the share of the site that is core low-intensity-use habitat, from annual 10 m land-cover maps. Satellite, at baseline and every 5 years.

​

Multimetric — the year-on-year changes of the pillars are combined into one cumulative figure measured from the baseline year, following Plan Vivo's equations for Data Analytic Providers. In the baseline year the dashboard shows that change is measured from the following year.

​

How metrics are produced:

  • Calculations run on a dedicated analysis server using the open-source R packages specified in Plan Vivo's requirements

  • Only the project's owner can start a run, and nobody can type in or edit a result

  • Every run is recorded permanently in an append-only log: who started it, when, which year and pillars, its status, the software versions used, and a digital fingerprint of the exact input data

  • If a run is repeated on identical input, the stored result is returned and logged as "cached", so results cannot be fished for by re-running

  • Satellite runs also record the imagery window, the number of scenes used, the share of clear (cloud-free) pixels and the land-cover year

  • The NDVI season is fixed after the first run so every year is compared over the same months; changing it requires approval from Plan Vivo's Technical Review Panel

​

Audit bundle — each run produces a downloadable .zip with the input data, the exact code, the outputs and provenance records (software and code versions). An auditor can use it to re-run the calculation independently and confirm the same numbers. It is downloaded from the dashboard through a secure link valid for 24 hours.

​

The dashboard — VVBs see the Multimetric, a tile for each pillar with its value and change, the latest runs for the species pillars and the satellite pillars, and the audit bundle download. The project's developer reaches the same dashboard from the Biodiversity screen.

Summary for Auditors

Data Type
Deletable?
Timestamped?
Georeferenced?
Photo Evidence?
AI Verified?
Survey submissions
No
Yes
Per survey (required or optional)
Where configured
Species fields only
Faulty submissions
Archived only (with reason)
Yes
Preserved
Preserved
Preserved
Species observations
No
Yes
Yes
Yes
Yes
Grievances
No
Yes
No
Optional
No
SDG survey responses
No
Yes
Yes
Where configured
No
Biodiversity records
No
Yes
Yes
Yes
Yes
Media files
No
Yes (upload metadata)
Via parent submission
N/A
N/A

Key assurances for verification bodies:

  1. Every record in the system has a creation timestamp that cannot be modified

  2. No data is ever deleted — only archived with an auditable reason

  3. GPS coordinates are captured at the device level at the time of field entry, not retrospectively

  4. All AI species identifications are performed by independent third-party services and stored verbatim

  5. Grievances are permanently recorded and cannot be suppressed by Project Developers

  6. VVBs have direct read access to all project data — they do not rely on PD-generated reports

  7. All data exports (CSV/JSON) include raw records, not just summaries

  8. Every new submission with photos, recordings or free text is screened daily by an automated AI reviewer; faulty entries are archived with a reason, never deleted

  9. Plan Vivo Pillar Metrics are calculated by the platform's servers, never typed in; every run is logged permanently and can be re-run independently from its audit bundle

  10. Each project's site boundary is stored with its area in hectares and its source (uploaded file or drawn on the map)

​

CitizenClimate is built on Google Firebase infrastructure (Firestore, Firebase Storage, Firebase Authentication, Cloud Functions). Data residency and Google Cloud audit logging apply in addition to the application-level protections described above.

One App,
Many Uses

CitizenClimate is a mobile app for collecting environmental data in the field. It's used by communities, project developers, and independent auditors to monitor nature-based carbon credit projects — tracking things like biodiversity, land use, community wellbeing, and SDG (UN Sustainable Development Goal) progress. Think of it as a structured field notebook that syncs to the cloud, works offline, and rewards participation.

Download on the App Store
Get it on Google Play
Citizen Science App Track SDG community led dMRV biodiversity monitoring
bottom of page