The eddy covariance method is widely used for measuring the vertical flux densities of heat and mass near the surface. In some studies, the sum of latent and sensible heat fluxes estimated by eddy covariance has been too small to close the surface energy balance, relative to the sum of the net radiation, soil heat flux, and heat storage terms measured independently. In this work, the covariances sensed at a height of 4 m above short grass were found by 30-min block averaging, as well as by high-pass filtering with a recursive filter using effective time constants of 50-800 s. After coordinate rotations to set the mean vertical and lateral wind components to zero, the sum of the two heat fluxes found with 30-min block averaging was sufficient to close the energy balance, and the fluxes were consistently larger than the fluxes computed with the recursive filters. Usually the differences during the daytime were largest for water vapor, intermediate for carbon dioxide, and smallest for heat. These differences were typically 10-18% for the recursive filter with a 50-s time constant and 5-10% with an 800-s time constant. One interpretation is that relatively low-frequency turbulent fluctuations contributed to the fluxes during the daytime. Also, because the effects of the 800-s mean removal computation should be roughly equivalent to block averaging over periods of 30-40 min, another mechanism seems to have contributed to the computation of smaller values with the recursive filters. Over the course of several days, the magnitude of the net carbon dioxide exchange was significantly larger when found by using the block averaging than by applying the recursive filters.