When do they sing,
and why.
Every detection has a timestamp, a species and a probability. But also a weather moment, a seasonal indicator and an hour-of-day. By laying these side by side, you can see when birds are active and when they stay silent. Here are the patterns so far.
The weather decides whether the garden sounds.
Activity measured across Bladerdak and Struikgewas combined, grouped by hour and weather condition from the Davis station, then compared with the overall average of the last sixty days.
Data not reachable yet.
The vertical line marks the overall average (100%). Scale runs up to 150%.
Temperature versus activity.
Detections per hour, paired with the average temperature of that hour, totalled per 2-degree bucket over the past thirty days.
Do they start earlier on clear mornings?
Detection rate per minute, plotted from one hour before to three hours after sunrise. One line per weather type, averaged across all mornings with that weather. The sunrise peak shifts with the season, so anchoring to sunrise is fairer than anchoring to clock time.
Data not yet available.
The busiest cells.
A grid of time-of-day against weather type. Each row is scaled to its busiest cell. Darker means more active. So you can see which weather brings a part of the day alive, and which weather keeps it quiet.
Data not yet available.
The temperature where it lives shifts along.
Each cell counts how many detections fell in that month and that temperature bin. The colour is scaled to the busiest cell in the whole grid. Grey means that temperature simply didn’t occur in that month.
Data not yet available.
Activity turns with the compass.
Eight wind sectors around the compass. The bar points to where the wind came from; its length is activity relative to the average (100 = normal, 150 = busier than usual, 50 = quieter). The grey ring behind it shows how often that wind direction actually occurred.
Data not yet available.
Birds read the barometer.
For each hour we look at the pressure change over the last six hours. Falling fast = front approaching, rising fast = front just passed. The bar shows how much busier or quieter the microphones got in each of those conditions.
Data not yet available.
Data not reachable yet.
Each detection is stamped at write-time with the most recent Davis reading: temperature, humidity, wind, pressure and rainfall. After that a nightly script bundles the detections per hour, per day and per season, and calculates the averages. Nothing fancier than a handful of SQL queries and some Python aggregation. It's still time-consuming to get honest averages, though, because some hours have a single reading where others have thousands.
The raw data is open under CC BY-NC 4.0. An Open Data page with downloadable CSV/JSON will follow.