Between 96 to 98% of the total annual input of 16 billion m³ into the Okavango Delta is estimated to return to the atmosphere by evapotranspiration, and only about 2% is leaving the Delta by surface outflow through the Thamalakane and Boteti Rivers. This outflow, however, is of great importance for the water supply of Maun and downstream areas along the Boteti, where groundwater is scarce and usually saline. A severe problem is the extreme variability of the outflow. Predictions have been based on two numerical models, outflow from the Delta (being considered as a function of inflow, precipitation), and evapotranspiration. Both models have shown remarkable fluctuations - socalled regime shifts - in the Delta's outflow. Attempts were made to interpret these in terms of flow changes inside the Delta, for example, channel diversion resulting from blockages by vegetation and sedimentation, or simply in terms of systematic errors in rainfall measurements. This paper shows that substantial model improvement is possible when both long and short term antecedent climatic conditions are taken into account. The time span of the long term conditions which is in the order of 10 years, seems to be influenced by shallow groundwatersurface water interactions in the lower Delta, while the short term conditions, with a memory of one year or less, are linked to a combination of peak rain and flood events. The modelling therefore suggests that the outflow variations observed during the period of record are neither caused by physical changes in the Delta nor result as an artefact of errors in the rainfall data. They are indicated to be entirely the result of climatic variations.