Over the last decade, improved understanding of plant physiological processes has generated a significant change in the way stomatal functioning is described in advanced land surface schemes. New versions of two advanced and widely used land surface schemes, the Biosphere-Atmosphere Transfer Scheme (BATS) and the Simple Biosphere Model (SIB), reflect this change in understanding, although these two models make different assumptions regarding the response of stomata to atmospheric humidity deficit. The goal of this study was to evaluate the new, second version of BATS, here called BATS2, using: Amazon field data from the Anglo-Brazilian Amazonian Climate Observational Study (ABRACOS) project, with an emphasis on comparison with the original version of BATS and the new, second version of SIE (SiB2). Evaluation of SiB2 using a 3-yr time series of ABRACOS data revealed that there is an unrealistic simulation of the yearly cycle in soil moisture status, with a resulting poor simulation of evaporation. Improved long-term simulation by SiB2 requires specification of a deeper rooting depth, and this requirement is general for all three models. In general, the original version of BATS with a revised root distribution and rooting depth gave good agreement with observations of the surface energy balance but occasionally showed excessive sensitivity to large atmospheric vapor pressure deficit. Evaluation of BATS2 revealed that changes are required in the parameters that determine stomatal behavior in the model for realistic simulation of transpiration, time-averaged respiration, and net carbon dioxide (CO2) uptake. When initiated with default values for carbon stores, BATS2 takes several hundred years to reach an equilibrium carbon balance. Aspects of the model's representation of instantaneous carbon allocation and respiration processes indicate that BATS2 cannot be expected to provide a realistic simulation of hourly variations in CO2 exchanges. In general, all three models have weaknesses when describing the field data with default values of model parameters. If a few model parameters are modified in a plausible way, however, all three models can be made to give a good time-averaged simulation of measured exchanges. There is little evidence of sensitivity to the different forms assumed for the stomatal response to atmospheric humidity deficit, although this study suggests that assuming that leaf stress is related linearly to relative humidity is marginally preferred.