Bias correction of surface air temperature and precipitation in CORDEX East Asia simulation: What should we do when applying bias correction?

• Published in: Atmospheric research Vol. 280; p. 106439
• Main Authors: Chen, Jueying, Yang, Yi, Tang, Jianping
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 15.12.2022
• Subjects: air temperature; Bias correction; climate; climate change; CORDEX East Asia; East Asia; Extreme events; Quantile mapping; seasonal variation; East Asia; Bias correction; Quantile mapping; CORDEX East Asia; Extreme events
• ISSN: 0169-8095; 1873-2895
• DOI: 10.1016/j.atmosres.2022.106439

Three groups of bias correction (BC) methods, which are linear scaling, distributional-based quantile mapping (QM) and empirical-based QM method, have been applied to correct surface air temperature and precipitation from Weather Research and Forecasting model (WRF) simulation within the second phase of Coordinated Regional Downscaling Experiment East Asia (CORDEX-EA-II) framework. WRF simulation and bias-corrected results are evaluated with gridded observations from CN05.1 and APHRODITE. The evaluation is conducted in terms of climate mean, seasonal cycle and extreme indices. Results show that WRF exhibits large biases in simulating surface air temperature and precipitation, which can be significantly reduced by bias correction. Among the three groups of BC approaches, empirical-based QM is the most comprehensive method in adjusting WRF simulation. However, when considering different time scales and regions, distributional-based QM can better correct precipitation higher than 25 mm/day than empirical-based QM, and a simple linear scaling can also show comparable skills in correcting seasonal cycles. Furthermore, based solely on statistical relationships between simulations and observations, bias correction cannot adjust circulation controlled and consecutive events. These results emphasize the importance of BC validation before climate change application. In addition, BC methods should be used carefully, considering regional climate and research subjects. •Model biases can be largely corrected by bias correction in terms of climatological mean statistics and daily indices.•Compared to linear scaling and distributional-based QM methods, empirical-based QM generally gives better corrections.•Bias correction can well-correct intensity-related extreme indices, but is less effective for duration-related indices.