BirdWeather

BirdWeather makes unattended listening visible. Fixed Raspberry Pi stations, mobile devices, uploaded recordings and dedicated sensors send BirdNET classifications into a live map. Each detection can carry time, location, species, confidence, probability, a playable soundscape and spectrogram. Station pages reveal daily patterns, while the Data Explorer filters across species, places and recording modes; authenticated owners can export CSV data, and documented APIs support other applications. This lets a birder inspect dawn activity far beyond one morning walk or operate a continuing backyard monitor.

A detection count is not a bird count. One loud individual may trigger repeatedly, microphones and thresholds differ, and stations cluster where volunteers can install hardware. Classifier errors remain possible even when a confidence ring looks persuasive. Location privacy can obscure a station, and audio publication requires thought about household speech and neighbouring property. BirdWeather is best used to find recordings and temporal patterns for review, not to rank sites by population from raw totals. Its openness becomes useful when sound, settings and deployment context travel with the number.