Data missing is inevitable in long-term eddy covariance system measurement. Several gap filling methods are chosen and applied in the data sets of Miyun station in 2007. These methods include mean diurnal variation (MDV), look-up tables (LUT), nonlinear regressions (NLR), Dynamic linear regression (DLR), Harmonic analysis of time series (HANTS) as well as the method of Penman-Monteith recommend by Food and Agricultural Organization (FAO-PM). The impacts of different gap filling methods on the evapotranspiration are investigated and the results are tested. The result shows: LUT method is stable in every data missing cases (RMSD less than 8 W/m2); MDV and NLR methods are more suitable for short data missing; The gap filling results of DLR and FAO-PM methods are not good if the observed data have large change continuously. The Annual ET obtained by LUT, DLR, NLR, HANTS, FAO-PM methods are 395.8mm, 409.9mm, 393.5mm,390.7mm,399.4mm. The difference between annual ET filled by different methods resulted in a range of 2.3~19.2mm per year. The annual ET obtained by LUT method is compared to net radiation, precipitation and latent heat flux measured by LAS; it is shows that the results are reasonable.