Reanalysis of Historical Mourning Dove Nest Data by Using a Bayesian Approach

Abstract

Abstract We reconstructed a historical mourning dove Zenaida macroura nesting dataset to estimate nest survival and investigate the effect of covariates by using a Bayesian hierarchical model. During 1979–1980, 106 study areas, across 27 states, were established to conduct weekly nest searches during February–October. We used roughly 11,000 data sheets to reconstruct the dataset containing 7,139 nests compared to 6,950 nests in the original study. Original and reconstructed nest survival estimates showed little difference by using the original analysis methodology, that is, the Mayfield method. Thus, we assumed we closely replicated the original dataset; distributions of nests found, birds hatched, and birds fledged also showed similar trends. After confirming the validity of the reconstructed dataset, we evaluated 10 different models by using a Bayesian hierarchical modeling approach; the final model contained variables for nest age or stage, nest height, region, but not habitat. The year 1980 had a higher probability of nest survival compared to 1979, and nest survival increased with nest height. The nest encounter probability increased at days 4 and 11 of the nesting cycle, providing some insight into the convenience sampling used in the original study. Our reanalysis with the use of covariates confirms previous hypotheses that mourning doves are habitat generalists, but it adds new information showing lower nest survival during nest initiation and egg laying and a decline when fledglings would be 4 or 5 d old. Regional differences in mourning dove nest survival confirm existing hypotheses about northern states demonstrating greater nest success compared to southern states where differences may reflect trade-offs associated with northern latitudes, weather differences, or food availability.