Post-processed data from the coupled land-atmosphere CPTEC-BAM model simulations used in the study by Sampaio et al. (2021 Biogeosciences; doi: 10.5194/bg-18-2511-2021) This dataset is organized into three data types as follows: monthly_mean: last 10 years of each simulation, month-by-month (120 data slices per model simulation per variable) annual_mean: long-term monthly mean for the last 10 years of each simulation (12 data slices per model simulation per variable) ensemble: average between the three runs (each with a different initial conditionl, see below) for the controls, deforestation and physiology simulations. Control and experimental runs were composed each of three model runs, each of them with different initial conditions derived from three different days of NCEP analysis (2003.Jan.01; 2007.Oct.10; 2017.Dec.17). [Except for the physiology scenario in which the simulations starting with the 2003.Jan.01 initial condition was not carried out] Data files labels are as follows: …controle: control run for the deforestation scenario. Deforestation: no. Vegetation: static. CO2 concentration affecting the atmosphere (atmospheric radiative fluxes): 388. CO2 concentration affecting plant physiology: 388. …controle2: control run for the physiolgy scenario. Deforestation: no. Vegetation: dynamic. CO2 concentration affecting the atmosphere (atmospheric radiative fluxes): 388. CO2 concentration affecting plant physiology: 388. …deforestation: experiment run for the deforestation scenario. Deforestation: yes (100%). Vegetation: static. CO2 concentration affecting the atmosphere (atmospheric radiative fluxes): 388. CO2 concentration affecting plant physiology: 388. …physiology: experiment run for the physiology scenario. Deforestation: no. Vegetation: dynamic. CO2 concentration affecting the atmosphere (atmospheric radiative fluxes): 388. CO2 concentration affecting plant physiology: 588. Spatial resolution: 192 longitude x 97 latitude x 18 vertical levels Files are in NetCDF format, with a header specifying the contained variables and units, reproduced in a separate README file for sake of convenience. This dataset and related publication are part of the AmazonFACE∫ME (AmazonFACE/ME) research project (https://labterra.cpa.unicamp.br/amazonface-me/).