An advanced bias correction technique for improved Seasonal Forecasts: a focus on extreme events in Southern Africa

In the context of climate change adaptation, the use of climate predictions is steadily gaining importance. One critical challenge to consider when dealing with climate simulation outputs is the systematic bias affecting the modelled data. While bias correction methods are commonly employed in impact models to assess the effect of climate events on human activities, their effectiveness is often reduced in the case of extreme events, due to the scarcity of data for these low-probability and high-impact phenomena. 

This study, conducted as part of the European project FOCUS-Africa, is dedicated to advancing innovative climate services in the southern regions of Africa. Our primary objective is to respond to the needs of risk assessment studies, focusing on the impact of extreme events and their implications for climate change adaptation. To this end, we designed a novel bias correction method to consistently correct extreme events of temperature and precipitation, but is adaptable to other climate variables, such as wind speed. Our approach conceptually extends one of the classic Quantile Mapping (QM) methods by improving the description of the tail ends of the distribution through a generalised extreme value distribution (GEV) fitting. Our methodology also incorporates a downscaling component. QM is indeed frequently both as a bias correction method and for downscaling simulations to finer observed scales. Therefore, our method not only corrects the climate data but also enhances the raw resolution of the model outputs (typically around 100 km) to match the 9 km grid of the observational reference. 

In this study, we applied our technique to daily mean temperature and total precipitation data from three seasonal forecasting systems: SEAS5, System7, and GCFS2.1, developed respectively by ECMWF, Météo-France, and DWD. The bias correction efficiency was tested over the Southern African Development Community (SADC) region, which includes 15 Southern African countries. The performance was verified by comparing each of the three models with a reference dataset, the ECMWF reanalysis ERA5-Land. The results reveal that this novel technique significantly reduces the systematic biases in the forecasting models, yielding further improvements over the classic QM. For both the mean temperature and total precipitation, the bias correction produces a decrease in the Root Mean Squared Error (RMSE) and in the bias between the simulated and the reference data. After bias correcting the data, the ensemble forecasts members that correctly predict the temperature extreme increases. On the other hand, the number of members identifying precipitation extremes decreases after the bias correction, highlighting the challenge of obtaining robust statistics due to the lack of information about extreme events.