A predictor-corrector method for solving inverse convection problems has been developed and tested against both numerical and experimental data. The method was applied to the simple convection problem of a two-dimensional plume in a crossflow. Crossflow velocities up to 1.0 m/s. The plume was generated by electrically heating a copper plate to a temperature up to 425 K. The method attempts to predict both the source strength, and the source location, with a self imposed requirement on sampling and simulation data points. The samples and simulations required are found to be 5 and 2 respectively. Tests based on simulation alone indicate the methodology has a source strength prediction error of less than 1%, and less than 6% for source location. Experimental tests bring the overall error up to 5% for source strength and 10% for source location. This study indicates the potential of the methodology and demonstrates some of its limitations. The approach can be extended to applied areas such as environmental flows, room fires, and thermal management systems. (C) 2013 Elsevier Ltd. All rights reserved.