A recent discussion of migration on the LA County Birding Listserv got me thinking about how I could come up with some data to throw at the questions that have been batted about recently.
Here's what I did:
I created a daily index of migrant numbers and diversity in the coastal plain of LA and Orange counties based on data from eBird, focusing on the 5 species of passerine migrants with pronounced spring migration peaks in LA County: Willow Flycatcher, Hammond's Flycatcher, Warbling Vireo, Swainson's Thrush and Wilson's Warbler. I would have included Nashville Warbler but couldn't because the AKN is apparently still discombobulated by the change in it's scientific name. I tested which weather variables best explained daily variations in this index in the period from April 20 to May 20 in each of the last 10 years. Data was sparse in earlier years, so many days were omitted from the analysis due to lack of data.
I repeated this process for the Antelope Valley using Lancaster weather data.
Here's what I found (with the caveat that these are preliminary results and could contain errors):
- the LAX weather variables that do the best job of predicting migration peaks on the coastal plain are those related to humidity, dew point and visibility. Temperature, barometric pressure and wind direction are also predictive but to a lesser extent. The best single variable is Mean Dew Point. The best days are often associated with lower dew points in the range of ~35-40, lower visibility, lower max temperature.
- the Lancaster weather variables that do the best job of predicting migration peaks in the Antelope Valley (AV) are 1-day Wind Direction change, Dew Point and Visibility.
- Oddly, wind speed seems to be among the least predictive variables for migrant levels, both on the coast and in the Antelope Valley (AV). However, wind direction and changes in wind direction are predictive in the AV and to a lesser extent on the coast.
- I found it interesting that several variables that are "predictive" of migrant levels relate to weather the next day. It would appear that migrants can anticipate tomorrow's weather and use that information to make decisions about when to move on. Perhaps Kevin Larson is on to something with his idea about the beginnings of weather patterns.
[If you are still with me at this point, you clearly enjoy self-inflicted brain damage and should see a psychologist.]
The net result of all of this is that I constructed a model that predicted good/bad migration days correctly roughly 60-65% of the time. That isn't very good, and the model often gets it totally wrong. I assume that it would be possible to do somewhat better using weather variables from farther down the coast and also by focusing on migration in specific locations rather than over a broad area.
Coast: Saturday: average; Sunday better-than-average
Antelope Valley: Saturday: average; Sunday below average
The models I came up with are:
coastal score (tomorrow) = 23.48 - 0.0098 * MeanDewPtLAX(tomorrow's forecast) + 0.0577*MeanVisibilityLAX(today) - 0.7466 * MeanSeaLevelPressureInLancaster(yesterday)
Scores above 1.2 indicate a ~70% likelihood of a good migration day tomorrow. I calculate 1.14 for Saturday and 1.25 for Sunday
Antelope Valley score (tomorrow) = -1 + 0.78 * SinWindDirectionLancaster(today) + 0.0263 * MinTempF(tomorrow's forecast) + 0.0065 * WindDirDegrees(today)
AV scores over about 1.2 indicate a ~75% chance of a good migration. I calculate 1.18 for Saturday and 0.94 for Sunday.
Still, it is good enough to warrant a bit of further digging ... and it indicates that eBird data is useful for measuring and modeling day-to-day changes in bird species diversity.
Hmmm....
