An approach using an experimentally built low-order model is proposed for the estimation of time-varying heat sources. In a first step, a low order dynamical system of equations, linking up temperatures at a set of specific points to heat sources strengths, is identified from experimental data using the Modal Identification Method. In a second step, the low order model is used to efficiently solve the transient inverse problem for the estimation of heat sources intensities from temperature measurements. The proposed approach is illustrated with an experimental set-up involving thermal diffusion with convective and radiative boundary conditions. (C) 2009 Elsevier Ltd. All rights reserved.