2D distribution measurement on temperature and concentration of combustion flow field can be obtained by tunable diode laser absorption tomography technique. A new method using multiple absorption lines to reconstruct 2D temperature distribution was developed. In the method, different lines were combined based on the property of temperature sensitivity. And the area interested were reconstructed by the line pairs which were optimal for the temperature. In this paper, four H2O lines were selected to reconstruct the temperature distribution using Gauss and random models in a range of 300-1 500 K. The simulation results show that the new method improves the reconstruction quality with reconstructed error of 0.039 6 comparing with two-line thermometry of 0.095 2. In engineering application, the multiple absorption lines combination optimization method can effectively improve the reconstruction quality when the temperature range of the area of interest can be predicted.