# Building a Reliable Citizen-Science Bird Survey

Citizen-science bird surveys can reveal migration patterns and population changes across areas that professional field teams cannot monitor alone. Our spring survey recruited 184 volunteers to record birds at fixed observation points, producing 12,640 checklists over eight weeks. Each checklist included the site identifier, observation time, survey duration, weather conditions, species, count, and detection method.

## Standardizing Field Observations

To reduce variation between observers, volunteers followed a ten-minute stationary-count protocol and completed an online identification exercise before collecting data. Surveys began within three hours of sunrise and were postponed during heavy rain or strong winds. Participants recorded every bird seen or heard, including explicit zero counts when no birds were detected.

```yaml
survey:
  duration_minutes: 10
  max_wind_beaufort: 3
  required_fields: [site_id, started_at, species_code, count]
```

## Validating Incoming Data

Submissions passed through automated checks before entering the analysis dataset. The pipeline rejected missing site identifiers, negative counts, observations outside the survey window, and coordinates more than 100 metres from the assigned point. Unusually high counts and locally rare species were flagged for review rather than deleted; reviewers could confirm them using notes, photographs, or audio recordings.

## Accounting for Uneven Detection

Raw totals are useful summaries, but they do not directly represent abundance. Detection probability changes with observer experience, habitat, weather, time of day, and whether a species is primarily identified by sight or sound. We therefore modelled detections using survey duration, canopy cover, wind, and observer experience as covariates, while treating site and observer as random effects.

## Sharing Reproducible Results

After validation, records were published with sensitive nesting locations spatially generalized to protect vulnerable species. Each release included the cleaned dataset, field protocol, species taxonomy version, validation rules, and analysis scripts. Publishing these materials makes the survey auditable and allows future teams to distinguish ecological change from changes in sampling effort or data processing.