The Problem
Species distribution models are widely used in conservation planning, but they often lack systematic ground-truth validation. Models built from occurrence records may overpredict suitability or fail to capture habitat nuances that experienced field biologists recognize immediately.
For the Maui Nui project, expert input was essential. However, traditional email-based review processes do not scale effectively. They make it difficult to track which experts reviewed which species, aggregate spatial corrections, or preserve structured metadata for downstream analysis.
A system was needed that could:
- Assign species or planning criteria to specific experts
- Capture spatial feedback in a structured format
- Allow simultaneous multi-user participation
- Maintain clear data provenance
The Solution
I developed a Shiny-based web application centered on an interactive Leaflet mapping interface. The platform is organized into two primary modules.
- Experts review habitat suitability maps for assigned taxonomic groups
- Interactive layers display model predictions, known occurrences, and planning unit boundaries
- Users select planning units to flag as unsuitable or highly suitable
- Confidence ratings and comments capture qualitative nuance
- Feedback is stored per expert, per species, enabling cross-review comparison
- Experts evaluate spatial planning criteria such as habitat accessibility
- Map overlays include trails, roads, landing zones, and accessibility scores
- Structured ratings allow refinement of spatial planning assumptions
The result is a structured, auditable feedback system that integrates expert knowledge directly into spatial datasets.
Technical Architecture
The application was built using R Shiny with the golem framework to support modular structure and maintainability. The UI leverages bs4Dash and Leaflet for interactive spatial visualization.
Spatial data layers (species models, planning units, base maps) are stored in Amazon S3 and loaded dynamically. User authentication assigns species or planning criteria to specific experts.
Expert feedback is stored in AWS DynamoDB, enabling:
- Concurrent multi-user participation
- Structured storage of per-user, per-species responses
- Aggregation of expert consensus while preserving individual input
- Clear tracking of review progress and data provenance
This architecture separates front-end interaction from persistent storage, ensuring scalability and reliability.
Outcome
The platform was deployed for the Maui Nui Landscape Conservation Planning project. Conservation experts used the system to review and refine species distribution models and planning criteria across the landscape.
Expert feedback identified model inaccuracies, corrected habitat assumptions, and improved spatial data inputs used in subsequent optimization workflows (including optimTFE).
The project demonstrates how domain-specific web applications can bridge the gap between quantitative modeling and field-based knowledge — strengthening conservation decisions through structured expert input.