Based on the analysis on the difficulty of wetland boundary identification and its causes, and having reviewed the wetland boundary indicators and their criteria selected at present, the paper suggested that, under the condition that the mechanism of wetland formation process has not been fully understood, all of the wetland boundaries based on hydrology and vegetation criteria are to some extent arbitrary boundaries. Wetland boundaries based on arbitrary criteria are not wetland's natural boundaries, because the natural boundaries of any complex can only be cognized, not be "man- prescribed". Even if there is no controversy on the criteria, it is usually impractical to identify wetland boundaries in the field accurately, because it takes repeated visits over a lengthy period of time (several years), and both seasonal conditions and then recent weather conditions should be considered when these indicators are applied. Wetland boundaries based on instantaneity indicators are unrepeatable and controversial. Aiming to resolve the problems on wetland boundary delineation, the paper analyzed the principles of the System Theory and physical geography related to wetland boundary delineation, and put forward a new method for wetland upper boundary identification. The method suggested that, under the condition that the mechanism of wetland formation processes has not been fully understood, "black box"method of System Theory can be used to identify wetland boundaries scientifically; when it is difficult to identify wetland boundary based on the characteristics of wetlands itself, if the boundaries of aquatic and terrestrial systems can be identified, they can be regarded as the wetland' s lower and upper boundaries; the lower limit of rooted plants is chosen as the indicator for delineating wetland lower boundary, and the thinning- out point of the diagnostic horizon of zonal soil is chosen as the indicator for wetland's upper boundary, because the thinning- out point is a kind of critical point, and it can factually reflect the long- term interaction result of all environmental elements. The scientificity and feasibility of the method was also analyzed. The characteristics of the diagnostic horizon, which is less responsive to short - term change than wetland vegetation and hydrology, make sure that the delineating result is repeatable, and that the controversy can be avoided.