Praven Pro 2.2 - Automated Biological Validation

What is Praven Pro?

Praven Pro is an open-source Python toolkit that bridges the gap between BirdNET automated bird sound recognition and Raven Pro bioacoustic analysis software. It streamlines the verification workflow for acoustic monitoring studies by:

Converting BirdNET detections to Raven Pro selection tables for expert review
Generating spectrograms with publication-quality rendering
Enhancing audio with Wiener filtering and HPSS noise reduction
Batch processing thousands of detections automatically
🆕 Automatic habitat detection from GPS using OpenStreetMap (v2.2)
🆕 Automatic weather fetching from historical weather APIs (v2.2)
🆕 Smart review selection - Top 3 per species for manual review (v2.2)
🆕 Biological validation to auto-reject impossible detections (v2.0+)

GitHub Repository

https://github.com/Ziforge/praven-pro

Non-Commercial License • Python 3.9+ • Cross-platform (Windows/Mac/Linux)

🌐 Live Demo: https://ziforge.github.io/praven-pro

🆕 New in Version 2.2: Full Automation

Completely Automated Workflow

Version 2.2 introduces complete automation requiring only CSV upload and GPS coordinates. Everything else is detected automatically:

Automatic date extraction: Reads from CSV columns (recording_date, absolute_timestamp) or filenames (YYYYMMDD pattern)
Automatic habitat detection: OpenStreetMap Overpass API queries 1000m radius for land cover types and hybrid habitats
Automatic weather fetching: Open-Meteo Historical Weather API retrieves temperature, precipitation, visibility, wind, and clouds
Smart review selection: Top 3 highest-confidence detections per species (Gaulossen dataset: 6,202 → 192 detections for manual review)
Biological validation: Activity patterns, geographic constraints, seasonal timing, habitat requirements, vagrant flagging

Gaulossen Dataset Results

✅ 580 Accepted

Expected species in correct time/place

❌ 23 Rejected

Biological impossibilities auto-rejected

⚠️ 6,202 Review

Smart review selected 192 detections for manual verification

Three-Way Classification System

✅ Auto-Accept

Expected species in correct time/place

Still requires audio quality verification

⚠️ Flag for Review

Rare/vagrant species or low confidence

Requires expert manual verification

❌ Auto-Reject

Biological impossibilities

Nocturnal woodpeckers, wrong season, etc.

Real-World Performance: Gaulosen Study

We tested the biological validator on the complete Gaulosen dataset (82 species, 4,108 detections). The validator was run after manual expert review to evaluate its effectiveness.

Errors Found by Biological Validator

Black Woodpecker (Dryocopus martius)

❌ Rejected: 2 nocturnal detections (01:35, 02:18)

📚 Literature: No evidence of nocturnal calling behavior

🔍 Missed by: Manual expert review

Common Grasshopper-Warbler (Locustella naevia)

❌ Rejected: All 59 detections (October 13-15)

📚 Literature: Summer visitor, departs by late September

🔍 Missed by: Manual expert review

💡 Additional evidence: Spectrograms showed rain noise, not bird vocalizations

Corn Crake (Crex crex)

❌ Rejected: 1 detection (October 14, 18:47)

📚 Literature: Long-distance migrant, should be in Africa by mid-October

🔍 Missed by: Manual expert review

Validation Summary

Result: Biological validator successfully identified all 8 biologically implausible species (5 from manual review + 3 newly discovered) based on temporal, habitat, and geographic constraints.

Impact: Species count corrected from 77 → 74 verified species, improving scientific accuracy of the study.

Quick Start Example

Install and validate your BirdNET detections in under 2 minutes - everything is automatic:

CLI Usage (Recommended)

# Install Praven Pro
pip install -r requirements.txt

# Validate - just provide CSV and GPS coordinates
# Everything else (date, weather, habitat) is detected automatically!
python validate.py gaulossen_all_detections.csv --lat 63.341 --lon 10.215

# Output:
# ✅ CSV date extraction: 2025-10-13 from recording_date column
# ✅ Habitat auto-detection: grassland (39%), oceanic (23%), wetland (16%)
# ✅ Weather auto-fetching: -18.9°C, rain 0.00, fog 0.00
# ✅ Processed: 6,805 detections
# ✅ Validated: 580 accepted, 23 rejected, 6,202 review
# ✅ Smart review: 192 detections selected for manual verification

Web Interface

# Start web server
python web_app.py

# Open http://localhost:5001
# Upload CSV, enter GPS coordinates
# Download results with automatic habitat and weather data included!

Automatic Detection Features

Version 2.2 eliminates manual data entry through intelligent automation:

Date extraction: Reads recording_date, absolute_timestamp columns or parses YYYYMMDD from filenames
Habitat detection: 1000m radius OpenStreetMap query identifies land cover (permanent caching)
Weather fetching: Historical weather API retrieves conditions normalized to 0-1 scale (30-day TTL caching)
Taxonomic validation: 40 bird families with species-specific rules for activity patterns and habitats
Smart review: Selects top 3 highest-confidence detections per species for manual review

Impact on This Study

                Statistics Before vs. After Biological Validation
                
                    
                            Metric
                            Before
                            After
                            Change
                        

                    
                            Verified Species
                            77
                            74
                            -3
                        

                            Verified Detections
                            4,085
                            4,023
                            -62
                        

                            Rejected Species
                            5
                            8
                            +3
                        

                            Species Pass Rate
                            85.6%
                            82.2%
                            -3.4%
                        

                            Stage 2 Pass Rate
                            93.9%
                            90.2%
                            -3.7%
                        

                
            

Metric	Before	After	Change
Verified Species	77	74	-3
Verified Detections	4,085	4,023	-62
Rejected Species	5	8	+3
Species Pass Rate	85.6%	82.2%	-3.4%
Stage 2 Pass Rate	93.9%	90.2%	-3.7%

While the pass rate decreased slightly, this represents increased scientific accuracy. The three species removed were biological impossibilities that should never have been in the final dataset.

Future Development Roadmap

⚠️ Current Limitations

Praven Pro 2.2 is a proof-of-concept system validated on a single wetland study (Gaulossen Nature Reserve, October 2025). The system demonstrated practical utility for this specific dataset but has significant limitations:

Limited species coverage: Only 40 bird families (primarily European species)
Incomplete weather data: Temperature, precipitation, visibility, wind, clouds only - missing humidity, barometric pressure, snow/ice conditions
Single habitat type: Tested only on wetlands, not validated for forests, grasslands, urban areas, or other ecosystems
Limited geographic scope: Validated only in Norway (63°N), untested across different latitudes, continents, and climate zones
No call type validation: Cannot distinguish between song, flight calls, and alarm calls
No seasonal migration models: Lacks comprehensive phenology databases for regional migration timing variations
Peer review required: Validation methodology and biological rules need formal scientific peer review

Extensive development required for broader real-world deployment beyond proof-of-concept status.

Planned Features for v2.3+

Expanded species coverage: North American, Asian, and African species rules
Call type validation: Distinguish song vs. flight calls vs. alarm calls
Enhanced ML integration: Improved rare/vagrant probability prediction
Regional phenology: Migration timing varies by latitude (eBird data integration)
Confidence thresholds: Species-specific confidence requirements
Advanced habitat modeling: Integration with eBird occurrence data and elevation constraints
Multi-species validation: Check for biologically impossible co-occurrences

Community Contributions Welcome

We're actively seeking contributors to help expand the species rules database and add regional validation rules. Visit the GitHub Issues page to participate in development discussions.

How to Cite

Redpath, G. (2025). Praven Pro: Automated Biological Validation Toolkit
for BirdNET Detections (Version 2.2). GitHub repository.
https://github.com/Ziforge/praven-pro
https://ziforge.github.io/praven-pro