Biodiversity Monitoring for Nature-Based Carbon Projects
Biodiversity Verification
Utilizing AI species recognition to log species sightings by local communities. Integrated dashboards correlate wildlife density with habitat restoration metrics.
Rigorous Biodiversity Assessment, Built for Community MRV
Nature-based carbon projects are under growing pressure to demonstrate co-benefits that go beyond sequestration estimates. Biodiversity verification is now a core requirement under Verra's CCB Standards, Gold Standard's NBS framework, and emerging TNFD disclosure guidelines — yet most project communities lack the trained ecologists or equipment to meet that bar.
CitizenClimate closes that gap. Community members collect standardised biodiversity observations in the field using a mobile survey. The platform automatically computes the full suite of ecological diversity indices from accumulated records and presents them in a dashboard designed for VVB auditors and project developers.
Five Taxa, One Survey
Each field visit captures observations across five faunal and floristic groups:
Birds are identified from audio recordings using BirdNET by CornellLab of Ornithology and Chemnitz University of Technology - a convolutional neural network trained on millions of labelled vocalisation recordings. Community members record ambient audio; the model returns scientific name, common name, confidence score, and timestamps for each detected call within the recording. BirdNET is the world's leading bird sound recognition model, trained on over 6,000 species. It is used by researchers, conservationists, and citizen science programmes globally.
Plants are identified from photographs using Kindwise — a professional-grade plant and crop identification API. The engine returns scientific name, taxonomic hierarchy, common names, and a health assessment flag — distinguishing healthy specimens from those showing signs of disease or stress. Kindwise — a professional plant, crop, insect, and mushroom identification API used by agronomists, ecologists, and conservation projects worldwide. Kindwise powers plant health assessment, insect taxonomy, and mushroom edibility flagging within the CitizenClimate platform.
Insects and Fungi are each identified from a single photograph, also powered by Kindwise. Insect identifications include taxonomic classification to species level where confidence permits. Mushroom identifications include edibility status, which is surfaced directly to community members as a safety flag.
Animals are identified from photographs using a vision model prompted with expert wildlife biology criteria, returning scientific name, common name, and behavioural observations where visible in the image.
All identifications are linked to IUCN Red List status at the point of capture. Threatened species (Vulnerable, Endangered, Critically Endangered, Extinct in the Wild, Extinct) are flagged automatically and counted separately in the project record.
Observation Records
Every observation stored in the project record includes:
-
Scientific name and common name (as returned by the identification model)
-
Taxon category
-
GPS coordinates of the observation point
-
Date and time of the field submission
-
Observer identity, linked to the project's community roster
-
Media evidence — audio file for birds, photographs for all other taxa
-
IUCN status
This constitutes a full chain of custody from field observer to audit report, with each observation tied to an immutable submission record.
Diversity Indices
CitizenClimate computes the following indices from all observations recorded against a project, updated in real time as new submissions arrive.
Shannon–Wiener Index (H') quantifies the information entropy of the species community — the degree of uncertainty about the identity of a randomly selected individual. It is sensitive to both species richness and evenness, making it a reliable general-purpose diversity measure across heterogeneous field sampling.
Simpson's Diversity Index (D) expresses the probability that two individuals drawn at random belong to different species. Its dominance-corrected formulation (D = 1 − Σnᵢ(nᵢ−1) / N(N−1)) makes it comparatively robust to sampling effort variation across a project area and to rare species underrepresentation.
Pielou's Evenness (J) normalises H' against ln(S), producing a unitless 0–1 score. A score near 1 indicates that individuals are distributed proportionally across all recorded species; a score near 0 indicates one or a small number of species dominate the count. Evenness is particularly informative for projects in recovery — early-stage restoration sites often show improving ric
hness before evenness catches up.
Alpha diversity (α) is calculated as the mean species richness per individual field submission, representing local plot-scale diversity. It is the most directly comparable metric across sites and time periods within a project.
Beta diversity (β) is calculated as the mean pairwise Sørensen dissimilarity between community areas within the project (β = 1 − 2|A∩B| / (|A|+|B|)). High beta diversity indicates that different community zones are supporting distinct species assemblages — an indicator of habitat heterogeneity that is ecologically valuable but often invisible to simpler richness counts alone.
Gamma diversity (γ) is the total count of unique species recorded across the entire project landscape. This is the primary headline figure for registry reporting and year-on-year trend monitoring.
Threatened Species Tracking
The platform maintains a running count of observations carrying an IUCN threat category (VU, EN, CR, EW, EX). These are surfaced prominently in the dashboard and included in all data exports. For projects applying for CCB Gold or Climate+ label, the presence of documented threatened species sightings with GPS coordinates and photographic or acoustic evidence provides directly usable biodiversity additionality evidence.
Export Formats for Audit and Registry Submission
The biodiversity dashboard supports one-click export in two formats:
JSON — a structured export containing all summary metrics, per-category species counts, and the full detection record for every species observed: scientific name, common name, IUCN status, GPS coordinates, observer, timestamp, and media file URLs. Suitable for direct programmatic ingestion by registry platforms.
CSV — a three-section file covering summary indices, category-level species counts, and a flat per-detection species table. Designed for direct import into R, Excel, or audit review tools.
Both exports carry the project identifier and export timestamp.
Designed for the Field
The survey runs fully offline. Observations made in areas without mobile connectivity are queued locally and synchronised to the project record when a connection is restored. The app is available in six languages — English, Hindi, Indonesian, Spanish, French, and Ukrainian — covering the majority of community contexts where nature-based projects operate.
Supported Standards
The biodiversity module produces data designed for compatibility with:
-
Verra VCS + CCB Standards (Climate, Community & Biodiversity)
-
Gold Standard for the Global Goals — Nature-Based Solutions
-
TNFD (Taskforce on Nature-related Financial Disclosures) pilot biodiversity indicators
-
GBIF data contribution format
API and Dashboard Integration
CitizenClimate's API links questionnaire data straight to dashboards, making Biodiversity tracking more effective:
-
Real-time updates – Questionnaire responses flow through the API to cloud-based dashboards where you can analyse them instantly
-
Visuals you can customise – See Biodiversity scores as bar charts, trend lines, or maps—whatever suits your project best
-
Works with existing systems – Connects to global Biodiversity reporting platforms (like the IUCN database) and conservation tools such as GIS
-
Secure and open – Data gets encrypted, timestamped, and can be shared publicly (anonymised) so it's both credible and transparent
Get Started with Biodiversity Monitoring
Join the CitizenClimate movement to track and push forward Biodiversity progress in your community:
-
Download the app – Get started collecting data using questionnaires built for your needs
-
Access the dashboards – Use our API to visualise and compare Biodiversity indexes
-
Share your impact – Show everyone what your community's doing for Biodiversity
Frequently Asked Questions About Biodiveristy Monitoring
How accurate are the AI identifications?
Accuracy varies by taxon and environmental conditions. BirdNET (Cornell Lab of Ornithology) reports a confidence score for each detected call; identifications below the confidence threshold are flagged rather than discarded, so auditors can review borderline cases. Kindwise model (plants, crop health, insects, mushrooms) similarly return ranked suggestions with confidence values, and the top suggestion is accepted only when it clears a minimum threshold. Animal identifications via Gemini return confidence alongside behavioural notes that can help a reviewer assess plausibility.
For audit purposes, the raw AI output — including confidence score and any alternative suggestions — is preserved in the submission record alongside the media file. A qualified ecologist reviewing an export can inspect and override any identification; the platform does not treat AI output as ground truth.
What is the sampling unit? How should communities structure their field surveys?
Each survey submission is treated as a single sampling event. Alpha diversity (mean species richness per submission) is most meaningful when submissions correspond to standardised plot visits — a fixed area, a fixed duration, or a fixed transect — so that variation in effort does not confound comparisons over time or between communities.
The platform does not enforce a specific field protocol, as appropriate methods differ substantially between forest, grassland, wetland, and agricultural contexts. Project developers should define their sampling protocol in the project design document; CitizenClimate records what is submitted, not how the survey was structured. We recommend aligning with whichever standard monitoring protocol your methodology requires (e.g. point counts for birds, fixed-area quadrats for plants).
Where does IUCN status come from, and how current is it?
IUCN status is returned by the identification API at the time of the observation and stored with the detection record. It reflects the status held in the API provider's database at that point. The Red List is updated periodically; status stored against older observations is not automatically retroactively updated when the Red List changes, which is standard practice in ecological monitoring databases. Projects with high-value threatened species sightings should note the observation date when citing status in audit documents.
How is beta diversity calculated when a project has only one community?
Beta diversity requires at least two distinct community areas to compute. For single-community projects, or projects where all submissions are recorded under a single community identifier, beta diversity returns zero and is displayed as — in the dashboard. This is mathematically correct rather than a data gap — it reflects that no between-community comparison is possible. Gamma and alpha diversity are unaffected.
Can the platform detect changes in diversity over time?
The current dashboard shows cumulative metrics across all observations for the project lifetime. It does not yet provide time-series views or year-on-year index comparisons, which are needed for additionality and permanence monitoring under most methodologies. Timestamped observation records are stored in full, so longitudinal analysis can be performed by exporting the CSV and filtering by date in R or a spreadsheet tool. Time-series dashboards are on the development roadmap.
What happens to observations made offline in areas with no GPS signal?
Submissions made offline are queued locally and synced when connectivity is restored. GPS coordinates are recorded at the moment of the field submission; if the device has no GPS fix at that time, the coordinate fields are left null rather than filled with an erroneous value. Observations without coordinates are included in all diversity index calculations but will appear without a location in the detection record export. In dense canopy or deep valley environments where GPS accuracy is poor, we recommend noting positional uncertainty in your field protocol documentation.
Is the Biodiversity Intactness Proxy comparable to the NHM/GBIF BII?
No, and it is not intended to be. The Natural History Museum's Biodiversity Intactness Index requires baseline land-use pressure models and reference site data that are not available at the project level. CitizenClimate's composite score — Simpson's D × Pielou's J, bounded 0–1 — captures the combined effect of species richness and community evenness in a single intuitive figure suitable for within-project trend monitoring. It should be described in project documentation as a community diversity composite, not as BII. Projects that require a formal BII estimate should use PREDICTS or similar modelled baselines.
How many observations are needed before the indices are meaningful?
Shannon–Wiener and Simpson's indices are sensitive to sample size at low N, and both will underestimate true diversity when sampling effort is insufficient. As a rough guide, indices computed from fewer than 30–50 total individual observations should be treated as provisional. Pielou's evenness requires at least two species and becomes more stable above S = 5. For beta diversity, a minimum of two community areas with at least 10 observations each is advisable before drawing conclusions. The dashboard displays all indices regardless of sample size; interpreting them in the context of observation count (shown in the summary row) is the ecologist's responsibility.
Can we add taxa beyond the current five?
Yes. Birds, vascular plants, animals, insects, and fungi represent the five groups for which sufficiently accurate automated identification models are available at scale. Extending to, for example, amphibians, freshwater macroinvertebrates, or lichen would require either additional AI identification partners or a manual expert review workflow. Both are under consideration for future releases. Alternatively Animal Identification tool can be used for less specific purposes.