Surface energy imbalance not only imposes constraints on the application of eddy covariance observations in research on land surface model development and evaluation, but also creates uncertainty regarding measurements of long-term net ecosystem CO₂ exchange. Lakes are a main component of the climate system and their surface energy balance is the dominant driver of biophysical and biogeochemical processes in lake ecosystems. In this study, observations of energy fluxes, radiation components, micrometeorological conditions, and water temperature profile were used to investigate the energy balance closure of Lake Taihu on different temporal scales (e.g., hourly, daily, and monthly). Energy balance closure was evaluated by linear regression of turbulent energy fluxes (sensible heat flux plus latent heat flux) against available energy (net radiation minus heat storage in water volume) and by calculating the energy balance ratio, the ratio of turbulent heat fluxes to available energy. Furthermore, the effects of three mechanisms-the stability of boundary layer atmosphere, friction velocity, and lake-breeze-on energy imbalance were analyzed quantitatively. The results showed that the heat storage in lake water volume had a diurnal variation similar to that of net radiation with comparable magnitudes, reaching a maximum at noontime. Both the sensible and latent heat flux showed much smaller diurnal variations and peaked in the morning and afternoon, respectively. Energy balance closure was observed at only 0.59 for the smooth Lake Taihu with half-hour averages, but increased to 0.73 using daily averages. Compared to land observations, there was less obvious diurnal variation in energy balance closure at Lake Taihu due to its aerodynamically smoother surface. Throughout the year, energy balance showed an obvious deficit during the warming months but perfect closure during the winter months. At an annual scale, the energy imbalance was 27%, which is comparable to values reported from eddy covariance observations on land and a few field surveys of lakes. The energy balance closure significantly improved with friction velocity during both daytime and nighttime, which indicates that mechanical turbulence is the main constraint on energy balance at Lake Taihu. The energy balance closure was approximately 0.7 for less unstable conditions (atmospheric stability parameter— -0.1), but was only 0.4 for very unstable conditions (atmospheric stability parameter— -1.5). Our results indicated that the poorer closure for very unstable conditions compared with less unstable conditions was due to reduced friction velocity (from 0.25 to 0.1). Although lake-breeze reduced the energy balance closure by 0.1 on an hourly scale, this was indiscernible on a daily scale. In addition, large-scale atmospheric motion, the stratification of lake water, and the source area mismatch between turbulent heat fluxes and available energy also contributed to the observed energy imbalance at Lake Taihu. Cospectra analysis showed that large-scale atmospheric motion was obvious at Lake Taihu, particularly during stable conditions, which may be filtered by a half-hour block average. Good energy balance closure was achieved when water thermal convection occurred with a 100 cm depth water temperature higher than that at 20 cm depth. The footprint of turbulent heat fluxes was much larger than that for available energy, and also varied with atmospheric stability and surface roughness. However, it is difficult to qualify the energy imbalance resulting from footprint mismatch.