A Dynamic State Variable Model of Mate Desertion in Cooper's Hawks

In a 4—yr study of the reproductive strategies of Cooper's Hawks (Accipiter cooperii) nesting in north—central New Mexico, >50% of the females deserted during the fledgling—dependency period and did not renest. A dynamic state variable model was developed to study the females' brood—rearing strategies. In this model a strategy consisted of combinations of staying at the nest, hunting, and deserting. The modeling assumptions were: a female's strategy during brood rearing maximizes her reproductive fitness, defined as the weighted average of the expected probability of survival of ether current offspring and her expected future reproduction; and the reproductive fitness function depends on the physical condition of the female and nestlings, the risk to the nestlings associated with each strategy, and the male's foraging capabilities. The model predictions were compared to the observations of female strategies in this population of Cooper's Hawks. To insure a valid comparison, the model parameters were estimated from sources other than the observed population. The best match between observations and predictions (84—96%) was obtained when the nestlings' survival and the female's future reproductive potential were equally weighted during the nestlings stage, but weighted in favor of the female's productive potential during the fledgling stage. A sensitivity analysis showed that the model predictions corresponded well with the observations of staying and hunting at all parameters bounds. However, those combinations of parameter values that reflected conditions with the least pressure to desert missed 70—85% of the desertions. The sensitivity analysis also indicated that a key factor influencing the female's choice of strategy was the interaction between the threat to her future reproduction due to her poor physical condition and the nestlings's risk of death from predation and exposure. The agreement of model predictions and observed strategies supported the modeling assumptions. These results combined with the sensitivity analysis indicated that dynamic state variable modeling is an excellent tool for studying mate desertion.