GEOGRAPHIC VARIATION IN BROWN-HEADED COWBIRD (MOLOTHRUS ATER) PARASITISM ON DICKCISSELS (SPIZA AMERICANA) IN GREAT PLAINS TALLGRASS PRAIRIE

Auk, The, Apr 2005 by Jensen, William E, Cully, Jack F Jr

Response variables of interest. -Response variables were (1) cowbird parasitism frequency, (2) parasitism intensity on Dickcissel nests, and (3) measures of Dickcissel nest success. Female cowbird density was explored as both an explanatory variable in relation to parasitism levels and a response variable in relation to local and landscape habitat variables, host community variables, and Dickcissel nest attributes. Parasitism frequencies were measured as the proportion of nests containing ≥1 cowbird egg or nestling when found, whereas parasitism intensity was measured as the number of cowbird eggs or nestlings per parasitized nest.

The main cost of cowbird parasitism expected to affect Dickcissel reproductive success was removal of host eggs by female cowbirds (Zimmerman 1983), which obviously results in fewer Dickcissels fledged per nest. However, there is also evidence that parasitism intensity influences prédation rates during the nestling stage (Jensen and Cully 2005). Therefore, daily nest mortality rates were measured separately for incubation and nestling stages (Mayfield 1975). Final clutch size of Dickcissel nests (after all apparent egg removal by cowbirds) was recorded, in addition to the number of Dickcissel young present in nests on visits prior to fledging (assumed fledged), for all nests and for successful (undepredated) nests only. These Dickcissel nest-success parameters were analyzed at two scales. First, they were compared with levels of parasitism intensity (0 to ≥6 cowbird eggs per nest) across all nests to examine the effect of parasitism intensity at the level of individual nests. Then those parameters were calculated for all nests per study site and compared to geographic variation in cowbird parasitism frequency and intensity.

Statistical analyses.-The many explanatory variables of interest in the current study were too numerous for inclusion in a single, global categorical regression (e.g. logistic regression) for selection of "best-fit" models including several explanatory variables, or to examine interactions among the explanatory variables. Therefore, we separately compared the measures of parasitism with each explanatory variable using a nonparametric correlation analysis. Spearman rank correlation was used to examine categorical cowbird parasitism frequencies and parasitism intensities on Dickcissel nests in relation to the many habitat and host community attributes, Dickcissel nest attributes, and female cowbird density per site. Female cowbird density was compared with the same habitat and host community attributes using Pearson correlation, because cowbird density estimates were continuous and not proportional data. As noted below, the results should be interpreted with caution, because multiple comparisons increase the probability of Type I error.

The effect of cowbird parasitism on the various nest productivity parameters of Dickcissels across all nests was assessed using categorical regression analyses (logistic and Poisson rate regression), where the independent variable was parasitism intensity (i.e. maximum number of cowbird eggs found per nest: 0 to ≥6). Logistic regression was used to compare Dickcissel egg hatchability (number of eggs hatched divided by eggs incubated in nests surviving until the nestling stage) and daily nest mortality during incubation and nestling stages (number of nest losses per total exposure days; Mayfield 1975, Hazier 2004) in relation to parasitism intensity. Poisson rate regression was used to compare Dickcissel final clutch size (number of host eggs incubated after egg removal by cowbirds) and number of Dickcissels fledged per nest ("apparent fledge rate") and per successful nest (i.e. undepredated nests) in relation to parasitism intensity. Lastly, associations of Dickcissel nest productivity parameters with local cowbird parasitism rate and intensity were assessed using Spearman rank correlation.