The biosphere-atmosphere transfer scheme (BATS) land surface parameterization was developed to provide a realistic representation of hydrometeorology at the atmosphere-land interface for coupling with general circulation models. This paper presents the results of a comprehensive sensitivity analysis of BATS for the conditions of midlatitude land surface and climatology. The analysis employs visual ''feature-curve'' representations to study the sensitivity of key model outputs to variations over a range of model parameters, atmospheric forcings, and model initialization conditions. The analysis provides insights into the behavior of the model and indicates that (1) the dominant energy flux shifts from being the soil sensible heat flux in arid/semiarid regions to evapotranspiration in rain forest regions; (2) in off-line studies the absence of important feedback mechanisms may result in unrealistic results at extreme climatic conditions; (3) the model functions which determine the atmosphere-controlled and soil moisture limited latent heat fluxes from soil and vegetation may need to be refined in order to obtain a more realistic modeling of the energy fluxes from the land surface; and (4) certain components of the model can possibly be simplified. The results also show that the influence of errors in the initial state estimates can persist for several years, indicating that caution should be exercised when attempting to validate or calibrate BATS using incomplete observation data and short-term simulation runs. ¿ American Geophysical Union 1996