Drift in CMIP5 decadal precipitation at catchment level

摘要

Abstract Over the last few years, decadal prediction has been paid much attention for its potential applications in agriculture, hydrology and for assessing the climate impact on the various aspects of human life. Though the fidelity of decadal prediction specifically the hindcasts experiments through Coupled Model Inter-comparison Project Phase 5 (CMIP5) has been examined for many climate variables and at different temporal and spatial scales, the drift in CMIP5 decadal precipitation at a local scale remains unknown. Drift is the long-term time varying systematic bias generated by GCMs while they revert to their equilibrium state from the forced initialized state. This study used seven general circulation models (GCMs) from five different modelling groups to examine the drift in monthly and seasonal mean precipitation from the CMIP5 decadal hindcasts for Brisbane river catchment, Australia. Drifts of individual model’s ensemble mean and multi-model ensembles’ mean (MMEM) at monthly and seasonal time scales were quantified and examined using four different skill tests. Results revealed that the magnitude of drifts are higher in monthly precipitation than the seasonal mean precipitation. Next, the drift in hindcast precipitation was corrected using mean drift correction method and found that the mean drift correction method is not sufficient to alleviate the drift in CMIP5 decadal precipitation. This suggests further research for an appropriate drift correction method for decadal precipitation. Comparing the drift and skill test results over the entire catchment, this study finds, MMEM showing the lowest drifts and outperformed in all models in skill tests.