Surface energy balance complexity in GCM land surface models. Part II: Coupled simulations

Global coupled simulations with the Bureau of Meteorology Research Centre climate model and the CHAmeleon Surface Model (CHASM) are used to examine how four general extensions to the representation of the basic land surface energy balance affect simulated land-atmosphere interface variables: evaporation, precipitation, skin temperature and air temperature. The impacts of including separate surface energy balance calculations for: Vegetated and non-vegetated portions of the land surface; an explicit parametrisation of canopy resistance; explicit bare ground evaporation; and explicit canopy interception are isolated and quantified. The hypothesis that these aspects of surface energy balance parametrisation do not contain substantial information at the monthly time scale (and are therefore not important to consider in a land surface model) is shown to be false. Considerable sensitivity to each of the four general surface energy balance extensions is identified in average pointwise monthly changes for important land-atmosphere interface variables. Average pointwise changes in monthly precipitation and land evaporation are equal to about 40 and 31-37% of the global-average precipitation and land evaporation respectively. Average pointwise changes for land surface skin temperature and lowest model layer air temperature are about 2 and 0.9 K respectively. The average pointwise change and average pointwise biases are statistically significant at 95% in all cases. Substantial changes to zonally average variables are also identified. We demonstrate how the globally averaged surface resistance parameter can vary from 150 to 25 s/m depending on which aspects of the surface energy balance are treated implicitly. We also show that if interception is treated implicitly, the effective surface resistance must vary geographically in order to capture the behaviour of a model which treats this process explicitly. The implication of these results for the design of land surface models is discussed.