These data were compiled for/to provide an example and assess methods and results of pre-fire estimation of predicted differenced normalized burn ration (dNBR) for predicting post-fire debris flow hazard classification. Objective(s) of our study were to develop predictive models for burn severity, using variables of pre-fire conditions, for two large wildfires from 2020 in Colorado, USA. These data represent pre-fire predictions of post-fire differenced normalized burn ratio (dNBR) as a proxy of burn severity and further understand pre-fire modeling of burn severity. These data were collected/created in the fire perimeters the East Troublesome Fire (10/14/2020 – 11/30/2020) and the Grizzly Creek Fire (8/10/2020 – 12/18/2020), Colorado, USA. These data were collected/created by use of random forest modeling of variables representing pre-fire conditions (satellite spectral data, landscape biophysical data, GIS topographic data, and meteorological/climate data) against observed estimates of post-fire difference in normalized burn ratio (dNBR). These data can be used to provide estimates of burn severity for post-fire hazard analysis.