87% of all detections came from social/flock species, revealing a highly social avian community.
8,778 geese-crow co-occurrences suggest pattern consistent with sentinel mutualism hypothesis where crows act as early warning sentinels for feeding geese.
Extreme flock behavior documented in Graylag Goose: largest event contained 620 calls over 91 minutes.
37 migratory species detected, including active nocturnal migration with 47 flight calls peaking at 03:00-04:00.
Great Snipe migration stopover calling shows dusk timing (20:00-21:00 peak) during autumn migration to Africa.
8,778 co-occurrence events detected between geese (Graylag, Pink-footed) and crows (Hooded, Carrion) within 10-minute windows.
Pattern consistent with sentinel mutualism hypothesis: Co-occurrence pattern resembles documented sentinel relationships in mixed-species flocks where sentinel species provide early predator warnings. However, acoustic data alone cannot prove functional benefit.
Mechanism (hypothesis): Potential heterospecific eavesdropping - geese may recognize crow alarm calls. Behavioral observations needed to test this hypothesis.
2,877 Graylag-Pinkfoot co-occurrences suggest regular formation of mixed-species goose flocks.
Benefits: Increased vigilance, collective predator detection, and shared knowledge of feeding sites.
47 nocturnal flight calls detected between 01:00-06:00, peak at 03:00-04:00.
Species involved: Pink-footed Goose (18 calls), Common Crane (11 calls), Greater White-fronted Goose (18 calls).
Significance: Documents active nocturnal migration through Gaulosen, likely along established flyway routes.
⚠️ Detection time does not necessarily indicate species ecology - may reflect sampling period
Detections: 189 total, 61% during crepuscular periods
Peak timing: 20:00-21:00 (82 calls at 20:00)
Behavior: Dusk calling during autumn migration stopover. October = migration season to Africa, NOT breeding season (breeds May-June).
Conservation significance: Declining species across Europe. Stopover site documentation important for migration monitoring and habitat protection.
Important: These are migration calls, not lek displays. Lek behavior only occurs during May-June breeding season.
Original detections: 59 (now rejected)
Rejection reason: Spectrogram analysis revealed rain noise, not bird vocalizations
Evidence:
See: GRASSHOPPER_WARBLER_REJECTION.md for detailed analysis with spectrogram images
Detections: 57 total, 75% crepuscular
Peak timing: 07:00-08:00 (31 calls at 08:00)
Behavior: Dawn/dusk calling during autumn migration. October = migration season, NOT breeding (roding displays occur March-July).
Important: These are migration calls, not "roding" breeding displays. Roding is a territorial breeding behavior only observed March-July.
86.6% of all detections from wetland-associated species
Wetland species: 19 species, 3,559 detections
Top wetland species: Graylag Goose (2,871), Pink-footed Goose (189), Great Snipe (189)
Conclusion: Gaulosen functions primarily as wetland habitat with extensive waterfowl and wader activity.
All behavioral interpretations are supported by 26 peer-reviewed scientific sources covering:
Species: All 74 species verified through manual review of spectrograms and biological verification
Behavioral patterns: Flock clustering, temporal patterns, co-occurrences documented through systematic analysis
False positive removal: 8 species rejected due to biological impossibility (nocturnal woodpeckers, oceanic seabirds inland, seasonal impossibilities)
Verification rate: 82.2% (74/90) overall pass rate through two-stage verification
Detection counts: Only best example per species verified, not all 4,023 detections individually reviewed
Rare species: Species with <10 detections have limited validation
Confusable pairs: Some closely related species may require additional acoustic analysis
IMPORTANT: 80%+ of recording occurred during rain/fog conditions
Impact: Cannot determine if species presence correlates with weather - I only recorded during poor weather
Conclusion: This study documents species detectable during rain, not general weather preferences
Great Bittern lesson: Rain noise caused 129 false positives at up to 91% confidence - highlights challenge of wet-weather acoustic monitoring
Detection: BirdNET v2.4 (Kahl et al., 2021) deep learning classifier
Audio enhancement: Wiener filtering + HPSS (Harmonic-Percussive Source Separation)
Spectrograms: Raven Pro-style (2048 FFT, 512 hop length, 0-12kHz)
Verification: Human expert review of spectrograms and audio for all species
Analysis: Temporal clustering (5-min windows), co-occurrence (10-min windows)