The Arctic Coordinated Regional Downscaling Experiment (Arctic-CORDEX) uses regional climate models (RCMs) to downscale selected Fifth Coupled Model Intercomparison Project (CMIP5) simulations, allowing trend validation and projection on subregional scales. For 1986-2015, the CORDEX seasonal-average near-surface temperature (tas), wind speed (sfcWind), precipitation (pr) and snowfall (prsn) trends are consistent with the ERA5 analysis for the Arctic Ocean regions considered. The projected Representative Concentration Pathway 8.5 (RCP8.5) 2016-2100 subregional annual tas trends range from 0.03 to 0.18 K/year. Projected annual pr and prsn trends have a large inter-model spread centered around approximately 5.0x10−8 mm/s/year and -5.0x10−8 mm/s/year, respectively, while projected sfcWind summer and winter trends range between 0.0 and 0.4 m/s/year. For all variables except prsn, and sometimes total precipitation, the driving general circulation model (GCM) dominates the trends, however there is a tendency for the GCMs to underestimate the sfcWind trends compared to the downscaled simulations. Subtracting the Arctic-Ocean mean from subregional trends reveals a consistent, qualitative anomaly pattern in several variables and seasons characterized by greater-than or average trends in the central and Siberian Arctic Ocean and lesser or average trends in the Atlantic Sector and the Bering Sea, related to summer sea-ice trends. In particular, a strong proportional relationship exists between the summer sea-ice concentration and fall tas and sfcWind trend anomalies. The RCP4.5 annual, multi-model mean trends are 35-55% of the corresponding RCP8.5 trends for most variables and subregions.