Regional distribution models with lack of proximate predictors: Africanized honeybees expanding north

ABSTRACTAim Species distribution models have often been hampered by poor local spe-cies data, reliance on coarse-scale climate predictors and the assumption thatspecies–environment relationships, even with non-proximate predictors, areconsistent across geographical space. Yet locally accurate maps of invasive spe-cies, such as the Africanized honeybee (AHB) in North America, are needed tosupport conservation efforts. Current AHB range maps are relatively coarse andare inconsistent with observed data. Our aim was to improve distribution mapsusing more proximate predictors (phenology) and using regional models ratherthan one across the entire range of interest to explore potential differences indrivers.Location United States of America.Methods We provide a generalized framework for regional and local speciesdistribution modelling with our more nuanced and spatially detailed forecast ofpotential AHB spread using multiple habitat modelling techniques and newlyderived remotely sensed phenology layers.Results Variable importance did differ between the two regions for which wemodelled AHB. Phenology metrics were important, especially in the south-east.Main conclusions Results demonstrate that incorporating a combination ofboth climate drivers and vegetation phenology information into models can beimportant for predicting the suitable habitat range of these pollinators. Regio-nal models may provide evidence of differing drivers of distributions geograph-ically. This framework may improve many local and regional speciesdistribution modelling efforts.KeywordsAfricanized honeybee, Apis mellifera, habitat suitability, species distributionmodelling, vegetation phenology.INTRODUCTIONSpecies distribution modelling (SDM) has become a com-mon tool over the last few years with applications to diversedisciplines and biological taxa including conservation biology(e.g. Urbina-Cardona & Flores-Villela, 2010), biological inva-sions (Measey et al., 2012), risk assessments (Bolliger et al.,2007), restoration (Fei et al., 2012) and climate changeimpacts (Thomas et al., 2004). While these models are oftencorrelative in nature, physiological information about aspecies should inform environmental factors included indistribution models (Austin, 2002). However, it can be diffi-cult to obtain spatially continuous information for relevantfactors. Indirect predictors such as elevation are often used assurrogates for those thought to be causal due to their highcorrelation with direct predictors such as temperature (Guisan& Zimmermann, 2000). For plant species, direct predictorsare often environmental or abiotic factors that are measuredsuch as climate or soil data. For fauna species, however, directpredictors may be different, including factors such as foodavailability and competition. Creating spatially explicit contin-uous surfaces describing these factors may be difficult.