We estimated the seasonal patterns of evapotranspiration (ET) using measured field data in a typical headwater basin and four sub-basins, in subtropical climate region of Serra da Mantiqueira/South-East Brazil, with a multi-instrument hydrometeorological set (standard weather station variables, discharge, soil moisture, and surface-atmosphere turbulent fluxes), towards providing regional data, to our knowledge, hitherto non-existent, of independent ET from micrometeorological methods (eddy covariance EC and Bowen ratio BR), energy balance closure corrections, and water budget based modelling, to discuss the magnitude and seasonality of the most accurate ET possible to be obtained. We had an unsatisfactory closure of the energy balance for the EC method, and the BR method likely also did not meet satisfactory estimates, as both suffered from the non-ideal conditions of the experimental site imposed by secondary atmospheric circulations. The comparison of the mean annual ET among 3 correction methods with the water budget showed an agreement in the range from 2.6 to 2.9 mm d −1, that was particularly met by the approaches of Mauder et al. (2013) and Twine et al. (2001), of 2.8 mm d −1. These corrections respected upper limits of potential evapotranspiration. Nighttime corrections were not applied, that possibly explains how daily totals of the H + LE sum overestimated about 9% of the available energy, that led to an overestimated ET of roughly less than 5% by those corrections. The seasonal variability of ET ranged from the minimum of 1.3 mm d-1 in July, and high fluxes in the wet season of about 3.5 mm d-1. The maximum H flux in September indicated how the onset of rainfall and the quick response of soil moisture recovery prevented H to increase onwards on the rhythm of radiation. Soil moisture appeared to play a key role in not only ET but in controlling the energy partition.