Aquatic systems are sources of greenhouse gases on different scales, however the uncertainty of gas fluxes estimated using popular methods are not well defined. Here we show that greenhouse gas fluxes across the air-water interface of seas and inland waters are significantly underestimated by the currently used static floating chamber (SFC) method. We found that the SFC CH4 flux calculated with the popular linear regression (LR) on changes of gas concentration over time only accounts for 54.75% and 35.77% of the corresponding real gas flux when the monitoring periods are 30 and 60 minutes respectively based on the theoretical model and experimental measurements. Our results do manifest that nonlinear regression models can improve gas flux estimations, while the exponential regression (ER) model can give the best estimations which are close to true values when compared to LR. However, the quadratic regression model is proved to be inappropriate for long time measurements and those aquatic systems with high gas emission rate. The greenhouse gases effluxes emitted from aquatic systems may be much more than those reported previously, and models on future scenarios of global climate changes should be adjusted accordingly.