Bias and variance correction of sea surface temperatures used for dynamical downscaling

Bias correction is a widely used method to deal with the deficiencies of climate models in representing the current climate. While it is mainly applied to prepare the output of global or regional climate models (GCMs/RCMs) for climate impact assessment, it has also been used recently to correct GCM output before it is downscaled by RCMs. For most RCMs, 3-D atmospheric fields as well as sea surface temperatures (SSTs) should be corrected in order to create forcing fields. The global stretched grid, conformal-cubic atmospheric model (CCAM), is able to run regional simulations with SST-only forcing. Therefore, only the monthly SSTs obtained from the GCM need to be corrected. In previous studies, the climatological bias was removed, while the bias in the temporal variability was still present. In this study, a simple method for correction of the mean and variance is proposed. The impact of the bias correction is tested using global even-grid CCAM simulations forced with raw and corrected SSTs from ACCESS1.0. Results indicate an improved precipitation pattern in the tropics for all seasons using corrected SSTs. There is also a slight improvement in the precipitation pattern in December–February and March–May and in the response to the El Nino–Southern Oscillation due to the additional variance correction.