We have examined the random error of eddy covariance (EC) measurements on the basis of two-tower approach during daytime. Two EC towers were placed on the grassland with different vegetation density near Gumi-weir. We calculated the random error using three different methods. The first method (M1) is two-tower method suggested by Hollinger and Richardson (2005) where random error is based on differences between simultaneous flux measurements from two towers in very similar environmental conditions. The second one (M2) is suggested by Kessomkiat et al. (2013), which is extended procedure to estimate random error of EC data for two towers in more heterogeneous environmental conditions. They removed systematic flux difference due to the energy balance deficit and evaporative fraction difference between two sites before determining the random error of fluxes using M1 method. Here, we introduce the third method (M3) where we additionally removed systematic flux difference due to available energy difference between two sites. Compared to M1 and M2 methods, application of M3 method results in more symmetric random error distribution. The magnitude of estimated random error is smallest when using M3 method because application of M3 method results in the least systematic flux difference between two sites among three methods. An empirical formula of random error is developed as a function of flux magnitude, wind speed and measurement height for use in single tower sites near Nakdong River. This study suggests that correcting available energy difference between two sites is also required for calculating the random error of EC data from two towers at heterogeneous site where vegetation density is low.