Code for variance-constrained bias correction of three-hourly global CMIP6 near-surface climate variables

Code Structure and Script Types The code package contains two subdirectories with NCL and bash shell scripts for processing ERA5 and CMIP6 data, respectively. ERA5 Directory In the ERA5 directory, there are files named by three steps for detrending, calculating variance of detrended data (anomalies) and calculating reference climatology. CMIP6 Directory In the CMIP6 directory, there are files named by six steps. Step 1 includes detrending of CMIP6 data for the histrocal reference period (1980-2009) and backward running detrending of CMIP6 data (Eq. 2) for the future period (2010 to 2099) using CDO operators. Step 2 calculates the variance of the detrended data (anomalies) and the scaling factor for variance correction (Eq. 1). Step 3 conducts variance correction of CMIP6 for the historical and future periods (Eqs. 3-4). Step 4 calculates historical climatology of variance-corrected CMIP6 data. Step 5 calculates the difference (bias) between ERA5 and CMIP6 climatologies using NCL (Eq. 5). Step 6 conducts the final mean bias correction of CMIP6 for the historical and future periods (Eqs. 6-7). The following information applies to scripts: Script Suffix Explanation In the ERA5 directory, scripts with the suffix _1 are used for processing general climate variables, while scripts with the suffix _2 are for radiation variables in a different data path. In the CMIP6 directory, scripts with suffix _AWI and _IITM are for processing two special variables 6-hourly "hurs" and daily "ps" of the model AWI-CM-1-1-MR and IITM-ESM, while others with suffix _3hr or without any special suffix are for general processing of most 3-hourly climate variables for most models. Script Header Settings All scripts require user customization of the model/variable names to process and input/output data paths, that can be specified at the beginning of the script. Please update these settings to match your own data directories as needed; subsequent code will then directly use your customized paths and model/variable selections. Generally, splitting the models and variables to subsets can speed up processing. There is also a file named "job.slurm" in the root directory which shows how to submit the scripts to the computing nodes depending on the configuration of your cluster. Special Note on "hurs" Variable in AWI-CM-1-1-MR Most variables were variance- and bias-corrected, except for one variable “hurs” of the model “AWI-CM-1-1-MR”. This model didn’t provide 3-hourly “huss” but only provided 6-hourly “hurs” for the SSP5-8.5 scenario. We derived 3-hourly hurs of the historical period from 3-hourly dew point temperature “tdps”. The resulting hurs of 1980-2014 does not match perfectly with 6-hourly hurs after 2015. There is a sudden change in the mean and trend of the two records, which causes overcorrection of the 6-hourly hurs after 2015 if variance correction is applied based on the variance correction factor of the 3-hourly hurs. Therefore, it is better to skip variance correction for this variable of AWI-CM-1-1-MR. The variances of both 3-hourly and 6-hourly hurs are very similar to the ERA5 reference data and thus is sufficient to perform mean bias correction for this variable. Data Access The resulting variance- and bias-corrected dataset is provided in the ScienceDB repository (DOI: 10.57760/sciencedb.28321) with free access. The original dataset of CMIP6 models can be accessed through the Earth System Grid Federation (ESGF) nodes (https:// esgf-node.llnl.gov/search/cmip6/). Contact If you have any questions about the use of this code, please email to: [email protected]