This paper presents a method of subtracting the effect of atmospheric conditions from thermal response test (TRT) estimates by using data on the ambient air temperature. The method assesses effective ground thermal conductivity within 10% of the mean value from the test, depending on the time interval chosen for the analysis, whereas the estimated value can vary by a third if energy losses outside the borehole are neglected. Evaluating the same test data using the finite line-source (FLS) model gives lower values for the ground thermal conductivity than for the infinite line-source (ILS) model, whether or not heat dissipation to ambient air is assumed. (C) 2011 Elsevier Ltd. All rights reserved.