Extremes of weather conditions, be it high or low, may have a devastating impact on the agricultural and industrial production of a country. We develop a statistical framework for prediction of areal impact of heat waves. The methodology applies to cold waves as well. The approach adopted is the quantification of the area in US under profound heat wave activity at given time point. The Pickands-Balkema-de Haan theorem as well as extensive model diagnostic plots reveal that the generalized Pareto distribution (GPD) serves as an efficient tool in modeling the peaks over threshold for the so-obtained time series. Several other factors like intensity level of the heat wave, grid network of the US, season of the year and duration of heat wave events have been explored in connection to the analysis of heat wave distribution. As a main contribution, we obtain estimates for the out-of-sample return levels for a variety of heat wave events as a function of the season, El-Nino Southern Oscillation (ENSO) index and location in US. These estimates are based on the analysis of daily temperature records for a period of 100 years for 424 stations spread across the continental US.