Most earth system models that simulate wetland methane emissions ignore spatial variability of water table depth. We compared methane emissions estimated using two commonly-used water table approximations (wet-dry and uniform) to a scheme that accounts for spatial variability, for several combinations of methane model parameter values and a range of water table depth spatial variability (σzwt). The approximations' biases were consistent across methane model parameter sets. The wet-dry scheme's biases were smallest in absolute value (+/−5–10%) when σzwt was large (38 cm) but increased to between −85% and −95% as σzwt approached 0. The uniform scheme's biases were largest (+/−100% or more) when σzwt was large and vanished as σzwt approached 0. For σzwt values typical of boreal wetlands, these schemes' previous predictions that global wetland methane emissions will double by 2100 are in error by factors ranging from 0.7 to 1.3. Similar biases may affect paleoclimate studies.