Grading the forecasts for Irene; Katia organizing; threat of a Gulf of Mexico storm
Recovery from the destruction left behind by Hurricane Irene continues in the mid-Atlantic and New England states today. Irene's storm surge, winds, and record rains likely did $3 - $6 billion in insured damage to the U.S., according to AIR-Worldwide. Since actual damages are typically double insured losses, Irene's total price tag will likely be $6 - $12 billion, making it one of the top 20 most expensive hurricanes to hit the U.S. Irene will be one of the most expensive Category 1 hurricanes ever; the record is held by 1972's Hurricane Agnes, which did $11.8 billion in damage (2010 dollars.) As AIR Worldwide notes in their press release, part of this damage is due to the costs of evacuation for the 2 million people that were evacuated. It costs approximately $1 million to evacuate each mile of U.S. coast warned (Aberson et al., 2006). This number will be higher for more densely populated areas of the coast, such as Miami, and may be a factor of six lower for the North Carolina coast (Whitehead, 2003). So were we over-warned during Irene? Could the costs of the storm been lower due to better forecasts and fewer evacuations?
Figure 1. The National Hurricane Center forecast for Hurricane Irene issued five days before it hit Long Island, NY, compared with the actual track of Irene. The landfall locations along the coasts of North Carolina, New Jersey, and New York were pretty much spot-on, though the time of arrival was off by a few hours. The NOAA Environmental Visualization Laboratory has a nice satellite animation of the storm's track superimposed on the NHC's cone of uncertainty forecast.
Well, the official NHC track forecast for Irene was remarkably good; the 5-day forecast was pretty much spot-on for landfall locations, though the timing of when the storm would arrive at the coast was off by a few hours (Figure 1.) This remarkably accurate forecast undoubtedly reduced the costs of unnecessary preparations, and probably saved many lives. NHC track forecasts have improved by over 50% since 1990. The average error in a 24-hour forecast was about 105 miles in 1990, and has averaged near 50 miles the past few years. NHC director Bill Read stated in a interview this week that had Hurricane Irene come along before the recent improvements in track forecasting, hurricane warnings would have been issued for the entire Florida, Georgia, and South Carolina coasts. At an average cost of $1 million per mile of coast over-warned, this would have cost over $700 million. We can credit the investments made in hurricane research, improved satellites, and better computer models for the majority of this improvement. When we consider that government funding for hurricane research has averaged $20 million per year during much of the past two decades, the roughly $200 million spent on hurricane research over the past 20 years was paid back by over a factor of three during just one storm. According to a 2007 presentation at the 61st Interdepartmental Hurricane Conference, the improved hurricane forecasts between 2000 - 2006 resulted in savings of $3 billion compared to what the forecasts of the 1990s would have cost.
What about intensity forecasting?
Progress in making better intensity forecasts of hurricanes, though, has lagged. Over the past twenty years, there has been virtually no improvement in forecasting how strong or weak a hurricane will grow. NHC predicted Irene would hit North Carolina as a Category 3 storm, but it hit at Category 1 strength. Had the intensity forecast been better, many evacuations that were done for Irene could have been avoided. The failure of the intensity forecast led to many accusations that the storm was over-hyped, and an unnecessary amount of expensive preparations and evacuations were done. While I did see some over-hype by the media, I did not think it was more excessive than what has been the case for previous hurricanes. Nate Silver of the New York Times makes some interesting comparisons of the media attention given to Irene versus previous storms, and finds that Irene had about the same amount of media attention as hurricanes Ike and Gustav of 2008. Given in inexperience of the mid-Atlantic and New England coasts with hurricanes, our lack of skill in making intensity forecasts, and the potential for high storm surge damage due to the size of Irene and its landfall during the highest tides of the month, I thought that the overly-cautious approach to evacuations along the coast was warranted.
Better intensity forecasts threatened by budget cuts
Better intensity forecasts of hurricane are possible, but it will take a large investment in hurricane research over an extended time to do that. Such an effort is underway; we are currently in year three of a ten-year program called the Hurricane Forecast Improvement Project (HFIP), funded at just over $1 million per year. The goals of the HFIP are to reduce the average errors of hurricane track and intensity forecasts by 20% within five years and 50% in ten years with a forecast period out to 7 days. In an interview I did last fall with the leader of the project, Dr. Frank Marks of NOAA's Hurricane Research Division, he expressed to me optimism that the program could meet its objectives, provided it remains fully funded. Some of the experimental computer models developed by HFIP have done very well so far during the 2011 hurricane season, so I see reason for optimism, too. However, this project is in serious danger of failure, due to the current budget-cutting emphasis in Washington D.C. A key tool we need to make better intensity forecasts is to have detailed measurements inside the core of the hurricane from instrumented aircraft. Without detailed observations, there is no hope of making a good intensity forecast, no matter how good your model is. During Hurricane Irene, the two P-3 hurricane hunter aircraft and G-IV jet operated by NOAA's Aircraft Operations Center flew continuously into the storm, taking detailed measurements via dropsonde and Doppler radar that were fed in real time to the experimental HFIP computer models. In theory, these measurements by the Hurricane Hunters should be able to significantly improve our intensity forecasts over the coming years. However, the current proposed budget from the House of Representatives mandates a $400 million cut for NOAA, and the NOAA Hurricane Hunters are slated to have their budget cut by 40%, from $29 million to $17 million per year. If these cuts materialize, the ability of the NOAA Hurricane Hunters to continue to aid improvements in hurricane forecasting will be seriously impacted. Many of the critical technologies used operationally now by the Air Force Hurricane Hunters and NOAA jet to improve hurricane forecasts--dropsondes, real-time high-density observations, and the SFMR surface wind measuring instrument--were developed on the NOAA P-3s as research projects, then were migrated to operational use once they proved their worth. The cost of hurricane damages in the U.S. has been doubling every ten years since the 1960s, and is expected to continue to double every ten years, even without the likely coming increase in storm surge damages due to accelerating sea level rise. A Category 1 hurricane doing $10 billion in damage should be a wake-up call that we need to continue our investments in hurricane research to reduce the costs of the inevitable coming storms. Slashing funding by 40% for a research group that was instrumental in saving $700 million in costs from just one storm makes no sense, and I hope Congress will reconsider the proposed cuts for NOAA's Aircraft Operations Center.
Whitehead, J.C., 2003: "One million dollars per mile? The opportunity costs of Hurricane evacuation", Ocean and Coastal Management 46, 1069.
Tropical Storm Katia
Tropical Storm Katia continues its long trek across the Atlantic Ocean today, and is expected to arrive at a position several hundred miles north of the Northern Lesser Antilles Islands by Monday. At this time, it appears unlikely that the islands will receive tropical storm-force winds from Katia. Satellite images show that Katia is a well-organized storm with plenty of heavy thunderstorms. The storm has good upper-level outflow channels to the north and south, is under light wind shear, and is traversing warm waters, so it should be able to overcome any dry air problems by Thursday and intensify into a hurricane. It is looking less likely that Katia will affect land. Dr. Bob Hart's Historical Tropical Cyclone Probability web page suggests shows that tropical storms in Katia's current position have an 11% chance of hitting North Carolina, a 12% chance of hitting Canada, a 5% chance of hitting Florida, and a 62% chance of never hitting land. It will be two more days before our computer models will be able to assess the threat to land, though, as Katia is currently still very far out at sea.
Figure 2. The morning run of the GFS Ensemble prediction. The ensemble prediction was done by taking a lower-resolution version of the GFS model and changing the initial distributions of temperature, pressure, and humidity randomly by a few percent to generate an ensemble of 20 different computer projections of where Katia might go. The operational (highest-resolution) version of the GFS model (white line) is usually more accurate, but the ensemble runs give one an idea of the uncertainty in the forecast. Very few of the ensemble members are currently showing a threat to the U.S. Canada is more at risk than the U.S., according to this model.
Gulf of Mexico disturbance a threat to develop
Surface winds over the Gulf of Mexico are rising today in advance of the approach of a tropical wave currently over the Western Caribbean, western tip of Cuba, and Mexico's Yucatan Peninsula. This wave is headed west-northwest at 10 - 15 mph, and is under a high 20 - 30 knots of wind shear. The wave is slowly beginning to build an increased amount of heavy thunderstorms, and this process will accelerate on Thursday when the wave enters the Gulf of Mexico. By Friday, when the wave will be near the Louisiana or Texas coast, wind shear is expected to drop to low to moderate levels, and the wave may be able to organize into a tropical depression. This process will likely take several days, and formation of a tropical depression is more likely Saturday or Sunday. NHC is giving the wave just a 10% chance of developing into a tropical depression by Friday morning. Regardless, this system will spread heavy rains to portions of the Gulf Coast by Friday, with the Upper Texas coast and the coast of Louisiana the most likely recipients of heavy rain. Strong onshore winds raising tides to 1 - 2 feet above normal are likely over Louisiana beginning on Friday, and coastal flood statements have been issued for the region. Three of our four top models for predicting tropical cyclone development forecast that a tropical depression will form this weekend or early next week, and I think it is at least 50% likely we will have Tropical Depression 13 on our hands by Monday. However, steering currents will be weak in the Gulf, and it is difficult to predict where the storm might go.The GFS model has a possible tropical depression forming by Sunday off the coast of Mississippi, then moving east-northeast over the Florida Panhandle on Monday. The ECMWF model forms the storm on Monday off the coast of Texas, and leaves the storm stalled out there through Wednesday. The UKMET model forms the storm Saturday off the coast of Louisiana, and leaves it stalled out there through Monday. If the storm did remain in the Gulf of Mexico for three days as some of the recent model runs have been predicting, it would be a threat to intensify into a hurricane.
Big money for hurricane research? My October 27, 2006 post.
Jason Samenow at the Washington Post has an excellent post, Hurricane Irene hype: over the top media coverage or justified?
Andrew Freedman at the Washington Post talked earlier this month how lack of funding to replace an aging weather satellite may degrade weather forecasts beginning in 2016. Michael Conathan at climateprogress.org had a more detailed analysis of the issue in a February blog post.
Andy Revkin at the New York Times discussed in his Dot Earth blog yesterday how cuts in the USGS stream gauge network will hamper flood forecasting.