The occupancy model (Fig 7; Table 4) shows that at low HSI values, cattle presence significantly increases the probability of occupancy of cowbirds. However, at higher HSI values, the presence of cattle no longer increases the probability that sites will be occupied by cowbirds. Areas with high a HSI value for cowbirds would be expected to be occupied regardless of the presence of cattle.
Fig 7. Probability of cowbird occupancy predicted from HSI only (red), and probability of cowbird occupancy in the presence of cattle (the interaction between cattle presence and HSI (blue).
Table 4. Model structure and output using the 'unmarked' package in R. Time of day (hr) was used as a detection covariat, and cattle presence/absence (cattle_pa), habitat suitability (HSI), and the interaction between habitat suitability and cattle presence (int) were used as site covariates. The interaction term was defined using dummy variable coding (i.e., cattle_pa * HSI).
Cowbird Generalized Linear Model
The GLM (Fig 8; Table 5) shows that both HSI and cattle presence or absence are significant predictors of cowbird presence or absence. The results indicate cattle presence increased the probability of cowbird presence regardless of the HSI value.
Fig 8. A generalized linear model predicting the interaction between cattle presence and HSI and the probability of cowbird presence.
Table 5. Model structure and output of a generalized linear model using presence/absence of cowbirds (pa) as the response variable, and habitat suitability (HSI) and cattle presence/absence (cow) as predictor variables. The model was fitted using a binomial distribution and a logit link function.
Conclusion
Cattle presence significantly influences the probability that a site will be occupied by cowbirds in areas with otherwise low habitat suitability for cowbirds. The effect diminishes as habitat suitability increases, as would be expected since these areas are suitable for cowbirds regardless of the presence or absence of cattle. This has implications for range management in the context of bird conservation in areas where at risk bird species are naive to cowbird parasitism.
The difference in results between the occupancy model and the GLM at higher HSI values could be due to the different assumptions made by the models. The occupancy model assumes that the probability of occupancy is influenced by both detection probability and actual presence or absence of the species, while the GLM only models the probability of presence or absence based on the predictor variables. Therefore, the GLM may not have the ability to capture nuances in the relationship between the predictor variables and the response variable that are related to detection probability, as it does not explicitly model this parameter. Additionally, the GLM assumes that the errors follow a binomial distribution, which may not be appropriate for modeling occupancy data. This may have contributed to differences in results between the two models.
This study did not assess potential consequences to naive bird species that may be subjected to nest parasitism by cowbirds. This would be a logical next step for further research and would require evaluation of nest success in areas where cowbird occurrence is influenced by the presence of cattle and may be causing reduced nest success for bird species of conservation concern.