Growing evidence supports that some bee populations are declining, raising concerns about consequences for pollination services. One major obstacle to understanding bee declines is a scarcity of mechanistic models that link environmental drivers to population dynamics and have been validated with independent data. Previous time series analyses of bumble bee forager abundances in coastal California showed significant declines from 1999-2014, linked with loss of floral resources during drought and increased competition with Apis mellifera. I collected four years of new data from 2015-2018 to further evaluate the mechanistic hypotheses developed through my previous analyses: 1) Higher spring rainfall promotes increased Scrophularia californica flowers, benefitting Bombus; 2) Niche overlap between A. mellifera and Bombus goes down as A. mellifera abundances increase; 3) This shift is primarily driven by reduced Bombus foraging on the common aster Eriophyllum staechadifolium as A. mellifera visitation rises; and 4) Higher A. mellifera abundance leads to lower Bombus numbers in the following year. I also assessed whether the previously developed statistical model was effective at qualitatively predicting Bombus abundances for novel data, from either flower patches not used in the original model fitting, the new 2015-2018 time period, or both.
Results/Conclusions
Mean A. mellifera densities changed little in 2015, fell sharply by 58% in 2016, then returned to higher levels in 2017-18. Bombus densities remained low in 2015-2016, then rebounded by 2.8 times in 2017 before dropping again in 2018. Scrophularia californica flowering in 2015-2018 continued to correlate strongly with spring rain. Yet this species was less abundant than in 2009-2014, potentially because of persistent effects from previous drought. Both Bombus preference for E. staechadifolium and niche overlap went down significantly with higher densities of A. mellifera. As predicted, niche overlap with A. mellifera strongly, positively correlated with Bombus density in the following year. Predictions from the previously developed Bombus model were unbiased and showed strong rank correlation with data that were either spatially or temporally novel. However, the model consistently under predicted Bombus abundances when both spatial location and time period were novel. Overall, four new years of data further strengthen evidence that loss of floral resources and competition with feral Apis mellifera are key drivers of Bombus declines at this site. This work also reinforces the value of even simple models that are mechanistically framed, not only in understanding past patterns of change but for qualitatively predicting new data.