Climate change presents challenges to global biodiversity that require data covering large spatial and temporal scales. Bird population trends have been monitored with several datasets with different protocols that may influence analyses of abundance. eBird and the Breeding Bird Survey (BBS) are two of largest and most commonly used avian datasets in the US that differ in spatial sampling biases and observer expertise and consistency. We examined the differences in species counts reported to eBird and BBS and determined whether the quantity of eBird data reduces these differences. We further hypothesized that differences between the two might be related to species traits. We compare eBird and BBS records for 300 species by aggregating eBird checklists within a spatial and temporal buffer around BBS routes. We fit the bird count data with environmental covariates and data source (eBird vs BBS) as a factor using a Generalized Joint Attribute Model (GJAM). The coefficients for data source are here termed the “observer effect”. We also examined the mean squared error (MSE) between counts from eBird records and BBS records at each site and determined the relationship between MSE and total minutes of surveying (“effort”).
Results/Conclusions
We found that overall, most counts per observation time were lower in eBird than BBS, but increased observation time in eBird reduces these differences. For 75.6% of species, the estimates for the eBird effect were negative. The eBird observer effect size was negatively correlated with commonness (R = -0.5), indicating that common species are under-reported to a greater degree than rare species. Conversely, there is a tendency for rare species to be overreported in eBird. Body size is weakly correlated with the eBird effect (R = 0.17), meaning that small species may be underreported in eBird more than large species. Color contrast is not correlated with the eBird effect (R = 0.07). MSE of BBS and eBird observations are negatively correlated with effort for most species (90.1%), meaning that the divergence between BBS and eBird declines with higher observation effort. Overall, these results indicate that reporting to eBird and BBS show the widest differences for short observation times in eBird, and there are additional differences by species commonness and size. Understanding these reporting differences can help to properly integrate datasets and improve monitoring of species populations.