The paper presents a method for predicting flood hazards based on the integration of predictive data from the NCEP GFS (Global Forecast System) global atmosphere model into a statistical scheme. The relevance of the study is driven by the increasing frequency and destructiveness of floods affecting infrastructure, as well as the scarcity of the ground-based hydrometeorological network in the mountainous regions of the Northern Caucasus. It is proposed to compensate for the information deficit of spatially distributed data by utilizing hydrometeorological fields from global numerical modeling. The aim of the work is to develop a predictive model that links the thermodynamic parameters of the atmosphere, precipitation amounts, and the excess water levels in rivers above the datum line (average longterm level). The research materials included archival data from hydrological field observations of floods and the associated synoptic conditions. To identify predictors of hazardous convective phenomena, the vertical profiles of temperature, humidity, and wind velocity vectors on standard isobaric surfaces were analyzed. The application of multiple regression analysis methods in the SPSS software environment yielded equations describing the relationship between flood intensity, total precipitation over the preceding three days, and the expected amount of precipitation on the day of runoff formation. Extreme precipitation forecasting for the current day is based on regression relationships of atmospheric thermodynamic indices extracted from the GFS Global Model. The implementation results of the developed approach will enable warnings of adverse weather conditions 24 hours prior to their onset. This time interval optimizes the decision-making process for government authorities and civil protection units when planning preventive measures in areas prone to catastrophic floods.