The paper describes the development and validation of an efficient and cost effective method for the prediction of the NOx emissions of turbulent gas turbine burners in the early burner design phases, which are usually focused on the optimization of the swirler aerodynamics and the fuel-air mixing. Since the method solely relies on nonreacting tests of burner models in the water channel, it can be applied before any test equipment for combustion experiments exists. In order to achieve optimum similarity of fuel-air mixing in the water channel tests with engine operation the model is operated at the engine momentum ratio. During the laser induced fluorescence (LIF) measurements the water flow representing the fuel is doped with fluorescent dye, a plane perpendicular to the length axis near the burner exit plane is illuminated with a 5W Ar-ion laser, and the fluorescence is recorded with a video camera from downstream. From the video sequence,s the local probability density functions (PDF) of the dye concentration fluctuations are calculated from the data. Furthermore, the time mean velocity fields are measured with particle image velocimetry (PIV). The PDFs of the local equivalence ratio are derived from the LIF data. Assuming flamelets, the NOx generation in the entire equivalence ratio range observed in the water channel tests is computed using the unstrained freely propagating one-dimensional flame model in Cantera and the GRI3.0 reaction scheme. Although neither flame stretch nor post flame NOx generation were considered, the computed NOx values were in excellent agreement with the experimental data from perfectly premixed combustion experiments. The local time averaged NOx mole fraction is obtained by integrating the flamelet NOx over the mixture PDF. Finally the global NOx emission of the burner at the considered operating point is obtained by spatial integration, considering the measured velocity field. The method was validated using a conical swirl burner with two fuel injection stages, allowing the degree of premixedness to be adjusted over a wide range, depending on the specific fuel injection scenario. For the case with fuel injection along the air inlet slots NOx values slightly above the minimum NOx limit for perfectly premixed combustion were computed. This is consistent with the emission measurements and indicates the finite mixing quality of this injection method. In the partially premixed regime the configurations with potential for low NOx emissions were reliably identified with the LIF and PIV based water channel method. The method also shows the steep increase of the NOx emissions with the decreasing degree of premixing observed in the experiments, however, quantitative predictions would have required a postprocessing of the data from the LIF mixing study with a higher spatial resolution than available.