Machine learning:

  • Entered the STAC Overflow (get it?) contest. I managed to go through three iterations of my model, and improve a bit over the baseline model they had as a tutorial. I finished in 71st place out of 664 – not bad! However, my final score (0.5314) was nowhere near the winners; the top four were all over 0.8. Still, this was a good exercise.

Mapping/GIS:

Hardware hacking:

  • More fan runtime experiments. It’s interesting to see the different battery behaviours.

  • Got the anemometer hooked up to the weather station and working at last! 🎉 The one thing it’s not is calibrated – so I’ve got RPM, but I don’t know what that translates to in wind speed. Yes, you can calculate the circumference of one rotation & figure it out from there, but…well, it’s complicated.

  • This also required rejigging the cable on the tipping bucket rain meter to use the same cat6 cable I used for the anemometer. Twisted pair cabling, people, it’s the bomb.

  • Big refactoring of the Arduino code for the weather station; it’s a lot more readable now. And I’m reasonably confident that my floating point math is probably okay.

  • Begin plans for a bird feeder camera. I’ve got some Coral dev boards, and it turns out the example code for it includes a bird species recognition model based on iNaturalist data. Bought a cheap pair of binoculars to try using as a telephoto lens for a webcam.

Nature/science:

  • Continuing phenology measurements for Nature’s Notebook.

  • Lots of observations for iNaturalist and eBird, including submitting some historical bird counts from the past year.

  • After a lot of looking around, I bought two big hardcover sketch books to use as phenology/nature journals: one page per day, and observations from each year on that page.

Climate emergency: