Hurricane Matthew Power Outage Forecasts

Colors on the map below show the fraction of the population predictd to lose power at the census tract level. The total number of people predicted to lose power is shown in the figure legend. Note that we are predicting the number of people without power while most utilities report the number of customer meters without power. Our numbers should be higher than their number in most cases. The ratio of population to meters is approximately 3 people per meter on average in those areas where we have found the data needed to estimate this. This means our estimates should be approximately three times higher than utility-reported customer (meter) outage totals.

Models can change quickly as the storm progresses. Visit for details on the storm and currently active storm warnings and advisories.

This model is current as of October 8, 2016 at 18 UTC (2PM EST).

Developed by Seth Guikema (University of Michigan), Steven Quiring (Ohio State University) and Brent McRoberts (Texas A & M University), this predictive model factors in a variety of data including:
  • Hurricane track and intensity forecasts and wind speed estimates from the National Hurricane Center
  • Population density data from the United States Census Bureau
  • Tree data from the United States Department of Agriculture
  • Drought indices from the National Drought Mitigation Center
  • Soil moisture levels gathered by the University of Washington

  • The estimated number of outages has been stabilizing at 8-10 million in the last few model runs as the track and intensity forecast have begun to stabilize.

    For a full-resolution version without the university logos for media use click on the figure.

    Further Information and Disclaimers

    This model is a statistical outage forecasting model based on wind speed estimates, population density, soil moisture levels, drought indices, and information about trees in each census tract. These outage estimates are provided for informational purposes only and are the product of a research project at the collaborating institutions. The information is provided "as is" without warranty of any kind. The investigators and their universities do not accept any responsibility or liability for the accuracy, content, completeness, legality, or reliability of the information contained on this website. For more information on an earlier version of this model, you can find the journal paper from IEEE Access here. The model used for the predictions above builds from this previous model. The paper describing the new model has been accepted for publication in the journal Risk Analysis and will appear online soon. For more information you can also contact Seth Guikema via email. This research was funded by the U.S. Department of Energy, the U.S. National Science Foundation, and a private utility.