As waves rolled in on Galveston Island before the deadly 1900 Storm, residents, visitors and even the chief local forecaster for the U.S. Weather Service at the time had little idea of what would come next.
Climate scientists now know the storm made landfall with winds in excess of 130 mph and with a 15-foot storm surge that flooded Galveston. It killed more than 8,000 and is said to be the deadliest natural disaster in the United States still.
At that time, weather forecasters had few tools and little way to track and predict where such a deadly storm would strike. Isaac Cline, chief local forecaster for the U.S. Weather Service, said in his autobiography that while he stood on the Galveston beach measuring waves as they rolled in, he tried to warn people of the impending storm.
Now, 115 years later, forecasting storms still can be tricky, but the margin of error has dropped dramatically, and the advanced warning for residents along the coast has improved.
Forecasting, of both a storm’s path and its intensity, has improved tremendously thanks to modern tools such as computers, satellites and forecasting models.
There is a difference between track forecasting and predicting the intensity of a hurricane, said James Franklin, branch chief of the Hurricane Specialist Unit at National Oceanic and Atmospheric Administration’s National Hurricane Center.
While there have been improvements to both, there has been steady improvement since the 1970s in the ability to forecast the track a hurricane will take, Franklin said.
That improvement is due in large part to increased computer power that allows for increasingly sophisticated models.
“They can represent smaller and smaller features in the atmosphere over time,” Franklin said. “The more detail that a model can represent, generally speaking, leads to a more accurate forecast.”
Along with the increasingly powerful models and computers are the increase in observations of storms thanks to satellites. Better data from satellites allows scientists to feed their models with better starting points to work from, Franklin said.
“We basically cut the average track error in half over the last 10 or 15 years,” Franklin said.
Forecasting intensity of hurricanes has been a bit more difficult and improvements have been slower, he said.
“For a long time, the intensity errors very stubbornly sat where they were,” Franklin said. “There was very little or no improvement.”
But over the last four or five seasons, the intensity errors have come down dramatically when it came to storms in the Atlantic, he said.
Part of the reason for the improvement could be that there have been few strong or rapidly intensifying storms — which typically give forecasters the most problems — in the past few years, he said.
But improvements also could be because of the significant investment in the 10-year Hurricane Forecast Improvement Project, he said.
“We are now finally starting to see some intensity forecast improvements, and I hope they can continue,” Franklin said.
And of course those forecasts can make the difference between staying put or evacuating.
When the 1900 Storm hit, residents in Galveston and the surrounding areas had little to no warning. The weather service of the time did issue a storm warning — stretching from Pensacola, Fla., to Galveston — on Sept. 7, 1900. The hurricane hit Galveston the next day.
Now the National Hurricane Center can issue two-, three- and five-day forecasts and has even begun experimenting with a seven-day track forecast, Franklin said.
And those forecasts have all improved dramatically since the turn of the century.
The average three-day track error was a little over 200 nautical miles, Franklin said. That has improved and now the average error is down to just less than 100 miles, he said.
The five-day forecast has improved from being accurate within 300 miles to now less than 200 miles, he said.
While the seven-day forecast is not ready to be published and released, Franklin said the agency has been working on it, and someday there will be forecasts of where a storm is headed as far as a week a head of time.
“Because of the improvements that have been ongoing, we are now at least looking at trying to extend that forecast horizon,” he said.