Network of eddy covariance observation is measuring long-term carbon and water fluxes in contrasting ecosystems and climates. As one important reference of independently evaluating scalar flux estimates from eddy covariance, energy balance closure is used widely in study of carbon and water fluxes. Energy balance closure in ChinaFLUX was evaluated by statistical regression of turbulent energy fluxes (sensible and latent heat) against available energy (net radiation, soil heat flux, canopy heat storage) and the energy balance ratio (EBR) and the frequency distribution of relative errors of energy balance (δ). The trends of diurnal and seasonal variation of energy balance in ChinaFLUX were analyzed. The results indicated that the imbalance was prevalent in all observation sites, but there were little differences among sites because of the properties variation of sites. The imbalance was greater during nocturnal periods than daytime and closure was improved with friction velocity intensifying. Generally the results suggested that estimates of the scalar turbulent fluxes of sensible and latent heat were underestimated and/or that available energy was overestimated. Finally, we discussed certain factors that are contributed to the imbalance of energy, such as systematic errors associated with the sampling mismatch, systematic instrument bias, neglected energy sinks, low and high frequency loss of turbulent fluxes and advection of heat and water vapor.