Land use and land cover change (LULCC) plays an important role in the interactions between human and earth systems. To project gridded LULCC under future scenarios, recent research has developed different spatial downscaling methods that disaggregate regional LULCC projections from integrated assessment models, such as the Global Change Analysis Model (GCAM). However, the uncertainty induced by using different downscaling methods and its impacts on the subsequent carbon cycle simulations remains unclear. This study used two different spatial downscaling methods (i.e., the Future Land Use Simulation model (FLUS) and the Demeter model) to generate 0.25-degree gridded LULCC data based the regional projections from GCAM over the Arctic-Boreal Vulnerability Experiment (ABoVE) domain in the high-latitude North America from 2015 to 2100 under SSP126 and SSP585. The gridded LULCC datasets (i.e., LULCC_FLUS and LULCC_Demeter) were then used as input for the Community Land Model version 5 (CLM5) to prognostically simulate the terrestrial carbon cycle dynamics over the ABoVE domain. The results show that there are large differences between LULCC_FLUS and LULCC_Demeter, especially for the spatial distributions of the needleleaf evergreen boreal tree, broadleaf deciduous boreal tree, broadleaf deciduous boreal shrub, and C3 arctic grass. The LULCC differences further lead to large discrepancies in the spatial and temporal patterns of projected gross primary productivity, net ecosystem exchange, ecosystem respiration and land use emission under both low and high emission scenarios. This study stresses the importance of accurate LULCC spatial downscaling methods and provides insights for understanding the impacts of LULCC uncertainties caused by the spatial downscaling methods on projecting future carbon cycle dynamics under climate change.