This dataset contains North American Regional Reanalysis (NARR) precipitation, downscaled/bias-corrected for the Kaskawulsh River headwaters catchment in Yukon, Canada. There is 1 file per year, with dimensions (t, x, y) = 2920 x 230 x 329 (3-hourly timestep, 230 gridcells x 329 grid cells). Gridcells are 200m x 200m. Downscaled precip: NARR precipitation is downscaled following Young et al. (2021), using a regression-based approach from Guan et al. (2009) that relates NARR surface precipitation to the Easting, Northing and elevation of NARR gridcells. NARR gridcells used for downscaling are a maximum distance of 90 km from the catchment and exclude gridcells on the maritime side of the ice divide. This choice captures the southeast-northwest precipitation gradient in the raw NARR data without including coastal gridcells that represent a different precipitation regime (see Robinson, 2024). Site-specific corrected precip: Downscaled precipitation is partitioned into rain and snow using a prescribed temperature threshold of 1ºC. Snow accumulation is bias corrected by multiplying downscaled accumulation by an elevation-dependent correction factor based on in-situ measurements from within the Kaskawulsh River headwaters (see Robinson et al. (2024)). Rain is not corrected in this dataset.