Biases in Global Climate Models due to PBL Parameterization

Global climate models (GCMs) use a variety of parameterization schemes to describe the Planetary Boundary Layer (PBL) and this results in large differences between models in their description of the PBL depth. These differences are greatest under stably-stratified conditions, related to shallow boundary layers. The planetary boundary layer depth modulates the magnitude of the surface air temperature (SAT) response to forcing: the strength of the temperature change on a given timescale is reciprocally proportional to the boundary-layer depth. Given that it is these shallow boundary layers that most strongly modulate the air temperature, we expect the greatest uncertainties in model temperature response to occur in stably-stratified conditions.

Here we have investigated GCM performance with regard to some key metrics affected by the boundary-layer modulation of forcing: the SAT mean, trend and variability. We assessed the GCMs individually by determining each model's departure from observations in the historical simulations of CMIP5, and as an ensemble, by assessing the inter-model spread in each metric. These assessments are performed both geographically, and as a function of the PBL climatology. We show that the greatest model error in these metrics occurs under conditions where the PBL depth is most uncertain i.e. in shallow/stably-stratified boundary layers, both for GCMs individually, and as an ensemble. Furthermore, we show that the bias in the models towards over-estimating the PBL depth in shallow layers corresponds with an under-estimation of the SAT variability. This work highlights the importance of improving the treatment of the stably-stratified PBL in GCMs.