From Photo to Species: The Five-Step Identification Flow

​

The plant identification process is built as a guided five-step survey flow, visible in the screenshot as Step 3 of 5. The stepped structure is deliberate — it breaks what could feel like a complex or technical task into manageable stages, and it ensures that users engage meaningfully with each part of the process rather than rushing through to submission.

​

The flow takes a community member from field observation to verified GPS-tagged species record in a single coherent experience. The photograph is taken. The AI analyses it. The suggestions are presented. The user makes an informed selection. The record is submitted and logged. Each step is explicit, clearly labelled, and designed for users who may be doing this for the first time — or for the hundredth time, in which case the familiarity of the flow makes the process efficient.

​

The step indicator — Step 3 of 5 with a progress bar — gives users a clear sense of where they are and how much remains. This is a small design detail that has a meaningful effect on completion rates. Users who can see they are three-fifths through a process are significantly more likely to complete it than users who have no sense of how much further they have to go.

​

Probability Scores and Multiple Suggestions: Handling Uncertainty Honestly

​

One of the most important design decisions in the plant identification feature is the explicit display of probability scores alongside each suggestion.

 

The screenshot shows two suggestions returned for a photographed plant in northern Norway. The first — Cornus suecica, the dwarf cornel or bunchberry, a small flowering plant in the dogwood family native to subarctic and cool temperate regions of Europe and Asia — carries a probability of 51.38%. The second — feathery false lily of the valley, also known as Solomon's plume or false Solomon's seal — carries 35.17%.

​

A probability of 51.38% is not certainty. It's the AI's best assessment based on the image, but it leaves room for the user to disagree, to apply their own local knowledge, or to choose the second suggestion if it better matches what they're seeing. This is the right design. An AI that presents single definitive identifications without confidence scores creates a false sense of certainty and removes the user from the identification process. An AI that shows its working — here are the two most likely candidates, here is how confident I am in each, here is the reference information you need to make your own judgement — invites the user to be an active participant in the identification rather than a passive recipient of a machine answer.

​

This matters both for data quality and for education. A user who has compared the reference photograph of Cornus suecica against the plant they photographed, read the species description, noted the characteristic red berry clusters, and actively selected the identification has learned something. A user who was simply told "this is dwarf cornel" by an app has not.

​

Taxonomy, Species Descriptions, and Edible Parts: What Communities Learn

​

The information returned alongside each species suggestion is designed to be genuinely educational, not just functionally useful for identification purposes.

​

The taxonomy data — in the case of Cornus suecica, class Magnoliopsida, family Cornaceae, genus Cornus, kingdom Plantae, order Cornales, phylum Tracheophyta — connects a single plant observation to the scientific classification system that organises all plant life. For a community member who has never encountered these categories before, seeing them displayed consistently alongside every identification they make gradually builds familiarity with the structure of botanical science. After fifty or a hundred identifications, the concepts of genus, family, and order become intuitive rather than abstract.

​

The species description — in this case, information on Cornus suecica as a species of flowering plant in the dogwood family Cornaceae, native to cool temperate and subarctic regions of Europe and Asia and locally present in extreme northeastern and northwestern North America — places the plant in its geographic and ecological context. Community members learn not just what a plant is called, but where it lives, what its distribution is, and what kind of environment it belongs to.

​

The edible parts information — for Cornus suecica, the fruit is edible — is practically valuable and adds a dimension of relevance that purely scientific information sometimes lacks. In many of the communities where CitizenClimate is used, local knowledge of which plants are edible is already present. The edible parts field creates a connection between that traditional knowledge and the scientific identification system, reinforcing both.

​

GPS-Tagged Records: Each Identification as a Biodiversity Data Point

​

Every plant identification submitted through the app is automatically accompanied by the precise GPS coordinates of the observation. The screenshot demonstrates this with a coordinate precision that speaks to the reliability of the location data: 68.169649275664°N, 14.693419301429874°E — a location in the Lofoten archipelago of northern Norway, above the Arctic Circle.

​

This level of coordinate precision is what distinguishes citizen science data collected through a purpose-built platform from informal observations or manually transcribed records. The GPS coordinates are captured automatically at submission, tied to the photograph and the species identification, and stored as a permanent, auditable record. The observation can be mapped, analysed spatially, compared to previous observations from the same location, and used as part of a longitudinal biodiversity record for the survey area.

​

For conservation NGOs and project developers, this GPS-tagged species data is genuinely useful. Biodiversity baseline surveys that would traditionally require expensive specialist field teams can be supplemented — or in some contexts, partially replaced — by community monitoring data collected through the platform. CCB Standards require evidence of biodiversity co-benefits, and GPS-tagged community species records, accumulated over time, provide exactly that evidence in an auditable, location-specific format.

​

The coordinate precision shown here — to fifteen decimal places of latitude and longitude — is a function of modern smartphone GPS hardware. It places each observation with greater spatial precision than most traditional field survey methods can achieve.

​​