Linking patterns in elk aggregation and brucellosis to variation in group size, land use, climate and wolves

Abstract

Recent data suggest that elk were the source of brucellosis outbreaks in cattle and that the disease is on the rise in elk across the Greater Yellowstone Area. As the distribution of elk group sizes is often right-skewed and spans several orders of magnitude, the largest elk groups may be driving brucellosis dynamics. To investigate whether the increases in brucellosis could be explained by elk grouping patterns, we examined 21 years of serologic data and three years of aerial surveys to record group sizes across 10 wintering elk populations, and we evaluated the relationships between rates of increase in brucellosis and seven measures of elk aggregation. We also examined the relationships between large elk groups and land use, habitat conditions, levels of predation risk, and snow accumulation. To do this, we used quantile regression to focus on group sizes in the tail of the distribution. We found that brucellosis increased over time in eight of the 10 populations, and that these increases were positively related to all measures of aggregation. We also found that group sizes were larger on irrigated land and as the habitat got more open. Because we were interested in the effects of snow on elk grouping behavior, we also examined the consequences of using snow model predictions in place of direct measurements. We found substantial model prediction uncertainty could directly impact our inferences. Lastly, we tested a noninvasive tool for conducting wolf surveys and found the method was a poor substitute for other techniques currently used during the winter, particularly in areas where pack territories overlap. Overall, our findings suggest that (i) most measures of elk aggregation had similar utility to predict changes in brucellosis over time, (ii) there may be more than one way to be dense and spread disease in populations that are structured by grouping (iii) it may be important to focus on more than one metric of the group size distribution to inform management, and (iv) it is important to consider the implications of prediction uncertainty on inferences when using model predictions as explanatory variables.