Reducing biases in low cloud cover over the tropical Atlantic in the Norwegian Earth System Model

Biases in the representation of low cloud cover in climate models has been identified as one of the leading causes of uncertainty in equilibrium climate sensitivity. It is therfore crucial to reduce current climate model biases in low cloud cover in order to reduce uncertainty in projected climate change. We have conducted perturbed parameter simulations to assess the sensitivity of the simulated low cloud cover in the Norwegian Earth System Model to parameters within the CLUBB scheme. The CLUBB scheme unifies the atmospheric boundary layer turbulence scheme with the clouds schemes and so has the potential advantage of reducing inconsistencies between these components of the atmosphere. However, there are many parameters within the CLUBB scheme that are not well constrained and have unknown effects on simulated climate. We demonstrate that of the 12 parameters in the CLUBB scheme selected for perturbed-parameter experiments, there are just 2 which control the low cloud cover in the model. We used a combination of multi-linear regression models and offline data assimilation with parameter estimation to identify the optimum values for these two parameters to eliminate the bias in low cloud cover, and confirmed this through a second iteration of perturbed-parameter experiments.