A Look Back on Forecasting Hurricane Harvey

By : Emmett Soldati |September 05, 2017 |weather-report |0 Comment

Hurricane Harvey made its mark on history as one of the most devastating hurricanes to hit a coastal region of the United States. It stands as a powerful reminder to insurers about the value of early – and accurate – warnings of landfall. While hurricane season remains open, a fresh take on hurricane forecasting emerges to help us assess the past and plan better for the future.


Throughout the life-cycle of Hurricane Harvey, Weather Analytics published its 10-day ensemble hurricane forecast model, Beacon, to anticipate storm movements. This forecast, built with observational and forecast data for over 200 tropical cyclones combined with machine-learning algorithms, exhibited impressive accuracy in forecasting the track of Hurricane Harvey prior to making landfall.

This figure shows the overall recorded track of Hurricane Harvey as of August 30th.

This figure from Weather Analytics’ Beacon Hurricane platform shows a forecast of Harvey two days prior to making landfall.


The tool, Beacon Hurricane, is the country’s leading hurricane forecast built on an artificial intelligence platform that pulls together all major global hurricane forecasts – including the Canadian Model, the European Centre, and the U.S.’ National Hurricane Center model.  Originally developed for a U.S. Government agency, the model uses pattern recognition to bias-correct forecasts based on the historic performance of a given forecast from the last 30 years.

“When such an Agency asks for a better hurricane forecast – an agency tasked with managing, among other things, some of our Nation’s most critical infrastructure – it’s all hands on deck – our meteorologists, data scientists and engineers.” says Bill Pardue, Chairman and CEO of Weather Analytics. “The novelty here was combining a team of skilled meteorologists who had served in the deepest levels of global forecasting outfits, with a team of machine-learning data-scientists who are capable of building highly astute pattern-recognition software.  What we didn’t know until the development started is that even something as dynamic as hurricane forecasts are subject to patterns and biases that we can control for.”

This figure shows the forecasted ‘cone of uncertainty’ of Harvey by The National Hurricane Center when Harvey was 36 hours away from making landfall.

This figure shows the forecasted tracks of Harvey by the Weather Analytics Beacon Hurricane forecast model when Harvey was 36 hours away from making landfall.

As senior atmospheric scientist at Weather Analytics, Dr. Stefan Cecelski puts it, “this approach brings together the best of both worlds: state-of-the-industry hurricane ensemble forecasts with the latest in machine-learning data science techniques.”

As insurers look to Irma, the tropical storm currently forming in the south Atlantic, this new predictive analytics technology will provide a timelier opportunity to understand, with confidence, the direction the impending hurricane will take. This will help customers better forecast downstream effects and potential damage before it occurs.

Weather Analytics urges you to contribute in whatever way you can to the relief effort in the aftermath of Hurricane Harvey.

1. National Hurricane Center Harvey Graphics Archive. –