Hurricane season is expected to be above average, as is forecast
During the record-breaking Atlantic hurricane season of 2020, something happened that many people overlooked: Hurricane forecasts have improved. And it saved lives.
During the record-breaking Atlantic hurricane season 2020, something happened that many people overlooked: Hurricane forecasts have improved. And it saved lives.
With another active hurricane season expected, this is great news. The National Hurricane Center (NHC) believes it can build on last year’s success with focused initiative, new strategies and long-standing IT model updates.
It is essential to have a good forecast as evacuation costs can reach tens of millions.
The actual cost is a difficult calculation because it depends on several factors, according to a 2003 study by Dr. John Whitehead. This includes the size and speed of the approaching storm and the population density along the affected coastline.
When the NHC plots the familiar “cone of uncertainty”, people on the way base their evacuation decisions on where that prediction hits their part of the shoreline.
As Hurricane Laura swept across the Gulf of Mexico in mid-August last year, Houstonians feared the worst. However, three days later forecasters focused on a landing in Louisiana. Houston emergency officials breathed a sigh of relief and avoided a multi-million evacuation cost.
This level of precision is part of an expected trend, according to NHC hurricane specialist John Cangialosi.
“Advances in technology and science have improved dramatically in recent years,” he said during a recent NOAA hurricane webinar.
Cangialosi considers hurricane track prediction to be a relatively straightforward theoretical problem. He says it’s like predicting where a cork moving in a stream will go. Usually, the cork follows the flow of the stream current, but it may be impossible to represent every jog that the cork takes.
Basics of the model
Most atmospheric systems that direct storms are large, easy to measure and understand, so most models predict the track well. When a hurricane forms, NHC targets an arsenal of computer models to predict where the storm will go and how strong it will be.
Some of these models are no more sophisticated than your car’s GPS. If you’re driving north at 60 mph, you should go 120 miles in two hours. Other models determine the course based on knowledge of the behavior of past storms forming at similar locations.
Then there are models based on more robust calculations. They are designed to mimic the behavior of low pressure systems in the heart of hurricanes, as well as the atmosphere around them. Over time, these models improved with improvements in model resolution, mathematical calculations used to simulate physical properties, and the speed of computers to perform the calculations.
NHC forecasters manually evaluate model data about every six hours to develop the familiar cone forecast. The envelope surrounding the solid line of the track (think of Cangialosi’s “cork”) represents the uncertainty based on the results of the preferred models.
The error in the NHC tracking forecast – the difference between what is expected and what is happening – is the baseline metric that shows improvement. The average error in the 24-hour forecast fell from 100 miles in 1990 to less than 50 in 2020. When the five-day forecast debuted in 2000, the average error was 400 miles. He fell below 200 miles last year.
Here’s a visual representation of the improved 3-day forecast:
While NHC forecasters are encouraged about the improvement in the predicted path, they are also starting to see improvement in an area they have long struggled with – forecasting hurricane intensity.
Better intensity forecast
Predicting the intensity of a storm is a “second-rate problem,” Cangialosi said. This means that it depends on the track. “If you take the wrong track you can get the wrong intensity just because you thought the storm was going to be in a different place and in a different environment,” he said.
In addition, knowing the intensity depends on the patterns of wind, humidity and temperature on the core and the environment of the storm. It also depends on internal processes, such as the strengthening that occurs when the eye of the storm suddenly reforms. This action cannot be predicted with much precision.
Yet something interesting has happened in recent years. The three-day average intensity error from 1970 to 2009 was approximately 20 knots (23 mph). Since 2010, this has also been almost halved.
The improvement is mostly due to an intentional effort called the Hurricane Forecast Improvement Project (HFIP), according to NHC researcher Jason Sippel.
The project incorporates a unique NHC model called HWRF, which uses the input of a real-time dataset, including Doppler radar data on board reconnaissance planes flying in the storm – the Hurricane Hunters. A special radar in the nose and tail of these planes detects the speed and direction of winds inside a storm as well as the intensity of precipitation, the essence of what Doppler radar does.
Sippel says Doppler radar data is essential for improving both trajectory and intensity predictions. He notes 2013 as a turning point for the HFIP, when researchers generated two forecasts for Tropical Storm Karen in the Gulf of Mexico.
One used the Doppler radar input and the other did not. According to Sippel, “The forecast that didn’t use Doppler data was terrible, heralding a full-blown hurricane. However, the first forecasts incorporating Doppler data were not perfect but significantly improved.
Since then, the HFIP continues to add more and more operational inputs to the HWRF and other models. One of these entrances are the cylindrical tubes dropped in storms during reconnaissance missions, called dropsondes. Sippel asked to change the deployment strategy to take different actions along the flight path. Adding this new data to the models resulted in a further reduction in path errors.
A final reason to be optimistic about improving hurricane forecasts is updates to the long-standing Global Forecast System (GFS) model, also known as the American model.
Earlier this spring, researchers overloaded the model to better represent the vertical complexities of the atmosphere and ran it on a faster computer, speeding up the time required for calculations.
The 2021 season is set to be another above average. Hopefully, these forecasting improvements will increase confidence when coastal residents and emergency managers face tough evacuation and safety decisions.