Does applying quantile mapping to subsamples improve the bias correction of daily precipitation?

ABSTRACT Quantile mapping (QM) is routinely applied in many climate change impact studies for the bias correction (BC) of daily precipitation data. It corrects the complete distribution, but does not correct for errors in the annual cycle. Therefore, QM is often applied separately to temporal subsamples of the data (e.g. each calendar month), which reduces the calibration sample size. The question arises whether this sample size reduction negates the benefit from applying QM to temporal subsamples. We applied four QM methods in a cross‐validation approach to 40 years of daily precipitation data from 10 regional climate model (RCM) hindcast runs, without and with (semi‐annual, seasonal, and monthly) subsampling. QM subsampling improved the BC of daily RCM precipitation; less distinct for independent data but considerably for the calibration data. The optimal subsampling timescale for the correction of independent data depended on the chosen QM method and ranged between semi‐annual and monthly. Overall, a sub‐annual QM improves the forcing for climate change impact studies and thus their reliability. Sub‐annual quantile mapping (subsampling) results in a clear benefit and improves the forcing for climate change impact studies and thus their reliability. For corrections of the calibration data, the monthly subsampling timescale was optimal for all quantile mapping methods. The optimal subsampling timescale for the correction of independent data was found to depend on the quantile mapping method.