Bias correcting climate model multi-member ensembles to access climate change impacts on hydrology

Authors: CHEN, JIE - Wuhan University; BRISSETTE, FRANCOIS - University Of Quebec; Zhang, Xunchang; CHEN, HUA - Wuhan University; GUO, SHENGLIAN - Wuhan University; ZHANG, YAN - Wuhan University

Submitted to: Climatic Change
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 2/14/2019
Publication Date: 2/25/2019
Citation: Chen, J., Brissette, F.P., Zhang, X.J., Chen, H., Guo, S., Zhang, Y. 2019. Bias correcting climate model multi-member ensembles to access climate change impacts on hydrology. Climatic Change. 153(3):361-377. https://doi.org/10.1007/s10584-019-02393-x.
DOI: https://doi.org/10.1007/s10584-019-02393-x

Interpretive Summary: The coarse resolution and inherent biases of climate models prevent the direct use of their simulations as the inputs to hydrological models for impact studies. Downscaling/bias correction is usually applied before driving hydrological models. However, the most commonly used bias correction methods (also referred to as individual correction methods) for climate change impact studies treat each climate model simulation individually, and thus may not be suitable for correcting multi-member ensembles. This individual correction approach could bring all of the members close to the observed data, thereby removing the ensemble spread. This study proposes three variants of a bias correction method and compares their performance with the individual correction method in terms of their ability for hydrological simulations. Two multi-member ensembles (5-member CanESM2 and 10-member CSIRO) were used for a subtropical monsoon watershed in China. The results showed that all bias correction variants are capable of reducing biases for simulated precipitation and temperature, as well as the individual correction method. As expected, individual correction reduced the spread of multi-member ensembles for the calibration period, while their spread can be overestimated for the validation period. All three bias correction variants could preserve the multi-member ensemble spread of precipitation and temperature. However, the aggregate correction and median methods also performed slightly better than the one-member correction method at reproducing the ensemble mean. Similar results were also observed in hydrological modeling. Taking into account the easiness of operation, aggregating all members in a bias correction method is recommended when using the multi-member ensemble of climate models for climate change impact studies. This work will be useful to hydrologists and meteorologists for studying the impacts of climate change and variation on water reaources.

Technical Abstract: The coarse resolution and inherent biases of climate models prevent the direct use of their simulations as the inputs to hydrological models for impact studies. Downscaling/bias correction is usually applied before driving hydrological models. However, the most commonly used bias correction methods (also referred to as individual correction methods) for climate change impact studies treat each climate model simulation individually, and thus may not be suitable for correcting multi-member ensembles. This individual correction approach could bring all of the members close to the observed data, thereby removing the ensemble spread. This study proposes three variants of a bias correction method and compares their performance with the individual correction method in terms of their ability for hydrological simulations. Two multi-member ensembles (5-member CanESM2 and 10-member CSIRO) were used for a subtropical monsoon watershed in China. The results showed that all bias correction variants are capable of reducing biases for simulated precipitation and temperature, as well as the individual correction method. As expected, individual correction reduced the spread of multi-member ensembles for the calibration period, while their spread can be overestimated for the validation period. All three bias correction variants could preserve the multi-member ensemble spread of precipitation and temperature. However, the aggregate correction and median methods also performed slightly better than the one-member correction method at reproducing the ensemble mean. Similar results were also observed in hydrological modeling. Taking into account the easiness of operation, aggregating all members in a bias correction method is recommended when using the multi-member ensemble of climate models for climate change impact studies.