We collected weather data hourly during each focal run (see Telemetry). A computerized Weather Monitor II weather station (Davis Instruments Corporation, Hayward, California, USA) was installed at 0.7-m height in the dominant habitat type on TWMA and GP. Temperature, relative humidity, dewpoint, rainfall, wind speed, wind direction, barometric pressure, and barometric trend (rising, falling, steady; change of 0.02 within 1 hr considered to indicate trend) were recorded hourly.
We used multiple linear regression with stepwise variable selection to examine relationships between weather conditions and movement rates across seasons. To examine the influence of weather conditions on activity (active or inactive) we used logistic regression to build a predictive model for activity by season. Inactive locations were coded as 0 and active locations were coded as 1. As active and inactive locations were never equal in occurrence, we determined prior probability of group membership for each model.