Arctic shrub cover has been increasing along with rapid warming over the recent decades. Shrub expansion remains challenging to predict yet is crucial for future carbon budgets, ecosystem vulnerability, and climate feedback. We explore factors controlling the spatial pattern of shrub expansion across the NASA Arctic-Boreal Vulnerability Experiment (ABoVE) core domain covering Alaska and western Canada. Here, shrub expansion is defined as the dominance of shrubs in 2014 in areas that were originally dominated by non-woody species in 1984 based on an annual land-cover product derived from Landsat. We identify bioclimatic and topographic conditions most informative in estimating environmental suitability of shrubs. While suitability explains shrub cover in 1984, it poorly explains newly established shrublands by 2014, suggesting areas becoming more suitable is not sufficient for shrub expansion to occur. By contrast, seed-arrival probability, calculated based on spatial proximity to existing shrub area using convolution of seed dispersal kernels, significantly improves shrub expansion estimation. Wildfire is also found to enhance shrub expansion during 1984-2014. These results indicate seed dispersal and fire have been a stronger limiting factor on shrub expansion than suitability over the recent decades. The inferred relationships predict that ~18% of current non-shrub tundra will be dominated by shrubs in 2100, with seed dispersal being the most important control. Our findings highlight the need to better integrate seed dispersal and wildfire into land surface model projections.