Bias correction of reanalysis data for assessment of climate change effects on river systems in a data scarce area : a case study for precipitation and temperature
Wednesday, 16 December 2015
Poster Hall (Moscone South)
Outputs from Global Climate Models (GCM) are increasingly used to assess the impacts of climate change on water resources, however it is often necessary to correct biases in GCM predictions relative to available observations. Bias correction should be based on over 20 years of observational data in order to reproduce long-term trends (e.g. yearly and monthly trend and less frequent extreme events). In previous studies, reanalysis data has been used instead of observation data when bias correction method is applied for outputs from GCM in the world and a data scarce area like observation data is under 20 years. Reanalysis data has biases between the observation data in local scale since it has been reproduced for global scale. However, bias correction method for reanalysis data has not been clarified for local scale. Therefore, this study aims to develop the bias correction method for re-analysis data based on observation data. Our application compared 6 hourly precipitation and monthly temperature data from an Australian site with the ERA-Interim (ERA) provided by ECMRWF reanalysis data. Initially, patterns in the 6 hourly precipitation data from the ERA were poorly predicted (r = 0.34), however, monthly precipitation and temperature were in good agreement (r: 0.79 and 0.996, respectively). Moreover, monthly precipitation events, average precipitation and standard deviation of precipitation in observation are associated with monthly precipitation in ERA (CC: 0.7011, 0.8168 and 0.7523, respectively). Therefore, we applied a bias correction method to correct the bias within ERA. Firstly, we corrected monthly precipitation and temperature and estimated monthly precipitation events, average and standard deviation of precipitation based on relationship observation and reanalysis data. Then, we corrected average and standard deviation of precipitation using them that was estimated. Finally, we corrected precipitation applying cdf method to reproduce low frequency large precipitation. As a result, it is revealed that precipitation and temperature pattern can be corrected using relationship monthly precipitation and temperature in re-analysis data and monthly precipitation events, average and standard deviation of precipitation and monthly temperature in observation.