This study explores whether regional climate model scenarios are sensitive to uncertainty in the representation of surface energy balance (SEB) complexity. Simulations with the Bureau of Meteorology Research Centre climate model and the CHAmeleon Surface Model (CHASM) were used to explore the sensitivity of changes in air temperature (T), evaporation (E), precipitation (P) and soil moisture (W) over Australia resulting from a doubling of atmospheric carbon dioxide (ΔCO₂). The 1 × CO₂ and the 2 × CO₂ simulations of T, E, P and W were sensitive to the complexity of the SEB, even though the grand mean of these quantities was almost always insensitive to SEB complexity. Seasonal variations in T, E, P and W at 1 × CO₂ and 2 × CO₂ were sensitive in terms of the point-by-point temporal mean and temporal variance. The overall spatial and temporal variances of T and P were insensitive to SEB complexity, but E and W were sensitive during periods of drying. The simulated seasonal change in T, E, P and W was insensitive to the SEB, and uncertainty in SEB parameterisation does not limit the reliability of existing climate change scenarios for Australia. However, the temporal variance of E, P and W was sensitive to the SEB complexity during periods of drying. Use of temporal variances of these quantities in future impact assessments are therefore likely to be very limited until uncertainty in the representation of SEB in climate models is reduced. To simulate the climate over Australia at either 1 × CO₂ or 2 × CO₂, a reasonably complex representation of the SEB, including a temporally and spatially variable surface resistance and an explicit representation of canopy interception, is required. Finally, our results say nothing about the importance of the land surface in general, since our analysis is restricted to a consideration of the SEB alone.