Two kinds of events

We track two types of train activity from our 24/7 audio monitoring:

While monitored 24/7, it takes 24-48 hours to fully process and analyze the data, so what's available here is a day or two behind.

🔊 Horn events (red on the calendar) are train horns detected by our audio pipeline. These are the loud, unmistakable blasts — sometimes a few short toots as required by law, sometimes a lot more. Each horn clip is a short snippet (usually under a minute) around the detected horn sound.

🚆 Idle events (amber on the calendar) are periods where a train is present but not moving. The engine is still running, air compressors are cycling, and hydraulics are releasing pressure — all of which are loud enough to be picked up by our microphones. Idle clips are longer samples captured at intervals throughout a recording session.

What "idle" actually means

An idle train isn't a quiet train. The locomotive engines keep running to power onboard systems, maintain air brake pressure, and keep the train ready to move. The hissing, chugging, and periodic pressure releases you hear are entirely normal — but when a train sits parked in a residential neighborhood for hours, that noise adds up. It also means diesel exhaust is being pumped into the air the entire time.

How we merge idle periods

A single recording session often captures multiple idle clips spread across several hours. Rather than showing them as a bunch of overlapping one-hour blocks on the calendar, we merge consecutive idle clips from the same recording session into a single idle period. For example, seven clips captured between 10:00 AM and 4:00 PM become one event spanning that full window.

This also means that when you click a merged idle event and download the audio, you're getting just the first clip from that period — a representative sample, not the entire multi-hour recording.

A note on accuracy

Everything on this site is a best effort. Our audio pipeline uses silence detection and basic classification to separate horn blasts from idle locomotive noise. It gets things right most of the time, but:

• Background noise (wind, traffic, neighbors) can occasionally produce false positives.
• Very quiet trains or distant activity may not trigger detection at all.
• Merged idle periods assume consecutive clips from the same session represent one continuous idle train, which is almost always true but not guaranteed.

We're continually improving the pipeline. If you notice something that looks wrong, drop by the forums and let us know.