Hotspots of sensitivity to GCM biases in global modelling of mean and extreme runoff

Climate model outputs feature systematic errors and biases that render them unsuitable for direct use by the impact models, especially when hydrological parameters are studied. To deal with this issue many bias correction techniques have been developed to adjust the modelled variables against observations. For the most common applications, adjustment concerns only precipitation and temperature whilst for others more driving parameters (including radiation, wind speed, humidity, air pressure) are bias adjusted. Bias adjusting only a part of the variables required as biophysical model input could affect the physical consistency among input variables and is poorly studied. In this work we quantify the individual effect of bias correction of each climate variable on global scale hydrological simulations of the recent past. To this end, a partial correction bias assessment experiment is conducted. Six climate parameters (precipitation, temperature, radiation, humidity, surface pressure and wind speed) from a set of three Global Climate Models are tested. The examined hydrological indicators are mean and extreme (low and high) runoff production. A methodology for the classification of the bias correction effects is developed and applied. Global hotspots of hydrological sensitivity to GCM biases at the global scale are derived, for both mean and extreme runoff. Our results show that runoff is mostly affected by the biases in precipitation, temperature, specific humidity and radiation (in this order) and suggest that bias correction should be applied in priority to these parameters. Surface pressure and wind speed had a minor effect on runoff simulations for the majority of the land surface. Low runoff has an increased sensitivity to the GCM biases compared to mean and high runoff, underlying the importance of bias correction for the study of low flow conditions and relevant hydrological extremes, such as droughts.