National Quail Symposium Proceedings


A newly developed technique for estimating fall northern bobwhite (Colinus virginianus) density is currently being employed in parts of the United States. One aspect of this technique involves predicting morning covey calling rates (i.e., the proportion of coveys that call on a given morning). We monitored 60 radiomarked coveys, a total of 229 covey observations, to determine whether or not each covey called. Calling rates were evaluated in relation to date, year, area, temperature, relative humidity, barometric pressure, barometric status, cloud coverage, and wind speed. We used logistic regression to test 9 a priori models as predictive models of bobwhite covey calling behavior. Models were compared using Akaike information criteria (AICc) values to determine the relative importance of 6 different variables (wind speed, date, temperature, cloud coverage, barometric pressure, and relative humidity). An exploratory analysis was then conducted to find the best predictive model using the best subsets model selection procedure. Standard errors of the coefficients in the best models were calculated using a traditional bootstrapping technique. We found an overall calling rate of 78%. Wind speed and date were the most influential of the 6 variables used in a priori model tests. Nine of the 19 exploratory models fit the data reasonably well. The best model included area and wind speed as independent variables, and was a better model than the best a priori model. There was a difference in calling rates between areas, and as a consequence, we recommend caution in application of our models to new areas.