Climate studies based on global climate models (GCMs) are projecting an overall steady increase in annual average temperature as well as hotter heat extremes for central North America through mid-century and beyond. Such a temperature trend is predicted, for example, for Austin, but for certain localities in the Midwest such as Des Moines, annual extreme temperatures have not increased over the past three decades, likely due to agricultural factors, thus deviating markedly from GCM predictions. This study analyzes regionally-downscaled GCM and observational data for these two cities with different extreme temperature trends and formulates a statistical model to define for each city a 95% probability interval of heat extreme values by 2040. The statistical model uses a linear combination of downscaled GCM annual extreme temperature forecasts. Regression coefficients for individual forecasts are determined by optimizing model results against observations over a training period (1950-2002). The model is validated against observations during a testing period (2002-2017).