WildGap

Problem

Biodiversity records are uneven. Observations often cluster where people already look, leaving spatial and taxonomic gaps that make environmental monitoring less useful. A low record count, however, is not evidence that wildlife disappeared; it may simply mean that nobody looked recently.

School eco-clubs and citizen scientists can help, but raw occurrence databases do not tell them where another hour of observation would be most useful or how to turn a data gap into a safe, repeatable outing.

Solution

WildGap lets a user search a city, paste exact coordinates, use browser location, or click the map to pin a survey area. It covers a 2, 5, or 10 kilometre area with seven comparable H3 cells and queries coordinate-valid GBIF occurrence coverage. It compares the latest 12 completed months with the preceding 36-month baseline, identifies recently under-observed target groups, and explains every score component and confidence level.

Open-Meteo adds recent climate context and ranks upcoming field windows. A user can select a candidate cell and create a printable, shareable 60-minute mission with a target group, weather guidance, safety checks, a timed protocol, and an optional public evidence link.

WildGap calls the result survey priority. It never labels observation gaps as habitat health, species abundance, population decline, confirmed absence, or ecological causation.

Product Tour

Why This Matters

Environmental decisions are only as good as the observations behind them. WildGap turns an abstract monitoring gap into a concrete action that a student or community naturalist can take this weekend. It supports the challenge theme by combining real environmental data with an end-to-end conservation-monitoring workflow, while making sampling bias and uncertainty visible instead of hiding them.

The product reports only defensible action metrics: missions planned, missions completed, priority cells visited, and evidence links added. This submission claims deployed functionality and technical verification. It does not claim a completed field outing or independent human usability study.

How We Used AI

WildGap does not use generative AI or a hosted model at runtime. The product's technical core is deterministic, inspectable geospatial computation over GBIF and Open-Meteo data. This was a deliberate choice: a survey-priority score that affects real field activity should be reproducible and explainable from its source data.

AI-assisted development was used through Codex to research implementation options, challenge unsupported ecological claims, implement and refactor the application, generate tests, diagnose live API behavior, review accessibility and security, and prepare deployment and submission materials. No model-generated ecological conclusions are shown to users.

How We Used Codex

Codex worked as the coding and review partner across the build. It helped translate the product plan into a Next.js application; implement H3 coverage, GBIF/Open-Meteo integrations, ranking suppression, caching, portable mission validation, and accessible map alternatives; and test the complete location-to-mission flow.

During hardening, Codex measured uncached production behavior, found upstream concurrency and retry risks, added bounded request scheduling and cache verification, ran TypeScript and unit checks, reviewed the public repository for exposed secrets, and removed the unused sign-in stack when the product moved to a fully account-free design. The final deployment and this write-up were also prepared with Codex, with field and tester results deliberately left unclaimed.

Key Features

City search, exact coordinate paste, browser location, and map-pin selection with 2, 5, and 10 kilometre radii Focused seven-cell H3 coverage for fast, readable local comparisons Live GBIF comparison of the latest 12 months with the preceding 36 months Transparent 55% density gap + 30% coverage change + 15% target gap score High, Medium, and Low confidence levels with low-data ranking suppression Target-group guidance for plants, fungi, birds, and insects Previous-month climate context and seven-day field-condition ranking MapLibre/OpenFreeMap map plus a keyboard-accessible ranked-list equivalent Printable 60-minute field mission with weather, safety, readiness, and evidence fields No accounts: mission history stays on-device and complete missions travel in validated portable links Clearly labeled Winnipeg demo snapshot for third-party outage resilience

Architecture

WildGap uses Next.js App Router, React, TypeScript, Tailwind CSS, MapLibre, OpenFreeMap, and h3-js. Its server-side analysis route builds H3 geometry, queries GBIF occurrence counts with paced two-request concurrency and bounded retries, retrieves Open-Meteo climate and forecast data independently, calculates the transparent score, and returns a serializable habitat analysis. If one source is partially unavailable, WildGap labels the exact completeness and omits missing cells instead of estimating them. If the hosting worker cannot reach the forecast service, the browser retries the public climate and forecast endpoints and still preserves manual date entry if those retries fail. Biodiversity metrics are cached for 24 hours and weather data for one hour. A complete labeled Winnipeg snapshot remains available for demo continuity and is never silently substituted for a global live result.

Missions require no authentication or central mission database. The browser validates the complete mission payload at the trust boundary, saves it in local storage, and embeds it in the portable share URL. Anyone holding the complete URL can read its field-card data, so the interface asks users to share intentionally and avoid sensitive evidence.

Data and technology attribution

GBIF occurrence API; individual occurrence records retain their publisher licenses and citations Open-Meteo geocoding, historical weather, and forecast APIs; weather data is CC BY 4.0 with attribution OpenFreeMap map style, OpenMapTiles, and OpenStreetMap contributors H3 spatial indexing library Next.js, React, TypeScript, Tailwind CSS, MapLibre GL JS, Lucide, Vitest, vinext, and Cloudflare/Sites deployment tooling under their respective licenses Original WildGap code is published under the MIT License What Comes Next

Community users can run WildGap missions in permitted public locations and, if they choose, add observation evidence. Future work can validate target-group protocols with local naturalists, add optional eBird/iNaturalist integrations, and evaluate whether repeat use improves spatiotemporal coverage. No Winnipeg field mission is part of this submission.

Built With

• cloudflare
• gbif
• h3
• maplibre-gl-js
• next.js
• open-meteo
• openfreemap
• react
• tailwind-css
• typescript
• vitest