The study examines the ability of a set of atmospheric variables to predict major floods in the Negev Desert in the southern part of Israel. The two dominant synoptic types, which contributed over 70% of the major floods during 1965-94 in that region, were examined. These are: the active Red Sea trough (ARST), a surface trough along the Red Sea, combined with a pronounced upper trough; and the Syrian low (SL), an intense Mediterranean cyclone centred over Syria. For each synoptic type, a set of atmospheric variables (predictors) was chosen to reflect its unique atmospheric features, and a prediction score was calculated as the ratio between the number of flood-producing storms and the total number of events in which all the variables exceeded their threshold values. The prediction score for the ARST type is 86%. Moreover, the predictors identify the major flood dates without any 'false' date (100%) for 4 of the 5 months in which major floods of this type had occurred. Most of the predictors are found at the 500 hPa level, and the most powerful is the v/u ratio, which represents the southerly wind component and is responsible for the transport of moist tropical air masses (essential for convection) toward the Middle East. The prediction score for the SL type is 73%. The intensity and structure of the surface cyclone are found to be most powerful predictors, although the importance of geopotential height at 500 hPa indicates that these typqs of flood depend on the combined effect of several factors. Transforming these remarkably high scores into a high-skill operational forecast of major floods in the Negev requires reliable forecast models to supply the desired variables with reasonable accuracy. It seems that the current operational models, together with our derived predictors, have the potential to yield a successful forecast of major floods 2 days in advance.