Search

We investigate the influence of bias correction of Global Climate Models (GCMs) prior to dynamical downscaling using regional climate models (RCMs), on the change in climate projected. We use 4 GCMs which are bias corrected against ERA-Interim re-analysis as a surrogate truth, and carry out bias corrected and non-bias corrected simulations over the CORDEX Australasia domain using the Weather Research and Forecasting model. Our results show that when considering the effect of bias correction on current and future climate separately, bias correction has a large influence on precipitation and temperature, especially for models which are known to have large biases. However, when considering the change in climate, i.e the Δ change (future minus current), we found that while differences between bias-corrected and non-corrected RCM simulations can be substantial (e.g. more than 1 ∘ C for temperatures) these differences are generally smaller than the models' inter-annual variability. Overall, averaged across all variables, bias corrected boundary conditions produce an overall reduction in the range, standard deviation and mean absolute deviation of the change in climate projected by the 4 models tested, over 61.5%, 62% and 58% of land area, with a larger reduction for precipitation as compared to temperature indices. In addition, we show that changes in the Δ change for DJF tasmax are broadly linked to precipitation changes and consequently soil moisture and surface sensible heat flux and changes in the Δ changefor JJA tasmin are linked to downward longwave heat flux. This study shows that bias correction of GCMs against re-analysis prior to dynamical downscaling can increase our confidence in projected future changes produced by downscaled ensembles. [ABSTRACT FROM AUTHOR]

NARCliM2.0 comprises two Weather Research and Forecasting (WRF) regional climate models (RCMs) downscaling five CMIP6 global climate models contributing to the Coordinated Regional Downscaling Experiment over Australasia at 20 km resolution, and south-east Australia at 4 km convection-permitting resolution. We first describe NARCliM2.0's design, including selecting two, definitive RCMs via testing seventy-eight RCMs using different parameterisations for planetary boundary layer, microphysics, cumulus, radiation, and land surface model (LSM). We then assess NARCliM2.0's skill in simulating the historical climate versus CMIP3-forced NARCliM1.0 and CMIP5-forced NARCliM1.5 RCMs and compare differences in future climate projections. RCMs using the new Noah-MP LSM in WRF with default settings confer substantial improvements in simulating temperature variables versus RCMs using Noah-Unified. Noah-MP confers smaller improvements in simulating precipitation, except for large improvements over Australia's southeast coast. Activating Noah-MP's dynamic vegetation cover and/or runoff options primarily improve simulation of minimum temperature. NARCliM2.0 confers large reductions in maximum temperature bias versus NARCliM1.0 and 1.5 (1.x), with small absolute biases of ~0.5 K over many regions versus over ~2 K for NARCliM1.x. NARCliM2.0 reduces wet biases versus NARCliM1.x by as much as 50 %, but retains dry biases over Australia's north. NARCliM2.0 is biased warmer for minimum temperature versus NARCliM1.5 which is partly inherited from stronger warm biases in CMIP6 versus CMIP5 GCMs. Under shared socioeconomic pathway (SSP)3-7.0, NARCliM2.0 projects ~3 K warming by 2060–79 over inland regions versus ~2.5 K over coastal regions. NARCliM2.0-SSP3-7.0 projects dry futures over most of Australia, except for wet futures over Australia's north and parts of western Australia which are largest in summer. NARCliM2.0-SSP1-2.6 projects dry changes over Australia with only few exceptions. NARCliM2.0 is a valuable resource for assessing climate change impacts on societies and natural systems and informing resilience planning by reducing model biases versus earlier NARCliM generations and providing more up-to-date future climate projections utilising CMIP6. [ABSTRACT FROM AUTHOR]