Hurricane Intensity Forecast Models Are Improving

Forecast accuracy in predicting where a hurricane will go has improved dramatically over the past 20 years, with official NHC track errors for 1 - 5 day Atlantic forecasts improving by more than a factor of two (Figure 1). Improving hurricane intensity forecasts, though, has proved to be very difficult--there has been very little improvement in official NHC intensity forecasts over the past 20 years (Figure 2). However, the models used to predict hurricane intensity have steadily improved over the past six years, and this improvement may herald the arrival of significantly improved hurricane intensity forecasts in the coming years. A good portion of this credit goes to the Hurricane Forecast Improvement Project (HFIP), a ten-year project that began in 2009 with the objective of reducing hurricane track and intensity errors by 20% over five years (by 2014) and by 50% over ten years (by 2019.)


Figure 1. Verification of official NHC hurricane track forecasts for the Atlantic, 1990 - 2015. Over the past 15 years, 1 - 5 day track forecast errors have been reduced by approximately 50%. Image credit: 2015 National Hurricane Center Forecast Verification Report.


Figure 2. Verification of official NHC hurricane intensity forecasts for the Atlantic, 1990 - 2015. Intensity forecasts have shown much slower improvement that track forecasts. There is some support for the idea that 1-day and 2-day intensity forecasts since 2010 (red and green lines) have shown a modest increase in improvement. Image credit: 2015 National Hurricane Center Forecast Verification Report.

Improvements in NOAA's HWRF model
The main focus of development efforts in HFIP have been to improve NOAA's HWRF model, which was the top-performing hurricane intensity model in 2015 and 2012 - 2015. If we consider the HWRF model alone, its improvement over the five-year period ending in 2014 has been 20% (Figure 3), meeting the HFIP goal of a 20% improvement in hurricane intensity models in a five-year period. (Note that HFIP cannot take full credit for the improvement of the HWRF model during this period, since the National Weather Service made independent substantial improvements to NOAA's GFS model, which supplies the initial conditions needed to run the HWRF model.) Official NHC intensity forecasts have also improved since 2009, though the numbers in 2015 for 3 - 5 day forecasts did not follow this trend. The 2015 numbers may be skewed because of the relatively few number of forecasts made last year, as NHC made forecasts as far out as five days for only four storms, two of which proved difficult to forecast--Danny and Joaquin. This allowed a few ugly forecasts to have an unrepresentative influence on the yearly stats. Perhaps most encouraging, the HWRF model showed significant progress in 2015 in making the most important intensity forecasts there are--ones of rapid intensification (RI). The model's probability of detection of an RI event increased, and the false alarm rate went down, compared to forecasts from previous years.


Figure 3. Intensity forecasts from NOAA's HWRF model for Atlantic tropical cyclones for the 2015 version of the model have shown a 20% improvement, meeting the 5-year improvement goal for the Hurricane Forecast Improvement Project (HFIP), a ten-year project that began in 2009 with the objective of reducing hurricane track and intensity errors by 20% over five years (by 2014). Image credit: Vijay Tallapragada, NOAA/EMC.

Improvements coming in 2016 for NOAA's HWRF model
The HWRF model should be even better this year. In July of 2016, the HWRF model is scheduled to receive a major upgrade to its code to improve the boundary layer and surface physics and vertical wind structure. The model will also increase the size of the "zoomed-in" region where its highest-resolution calculations are performed immediately surrounding a hurricane, and HWRF will be connected to a separate hurricane wave model which will allow the two models to interact and improve the forecasts of both models. I look forward to seeing if the HFIP program can meet its goal of a 50% improvement of hurricane intensity forecasts by 2019; I think 2016 will be a crucial test. Even if this goal is not met, HFIP has shown its worth in training a new generation of hurricane scientists. Many of these researchers will be moving into operational weather forecasting in the next few years in support of NOAA's latest modeling effort, the Next Generation Global Prediction System (NGGPS).


Figure 4. Skill of computer model intensity forecasts of Atlantic named storms in 2015, compared to a "no skill" model called "Decay-SHIFOR5" that uses just climatology and persistence to make a hurricane intensity forecast (persistence means that a storm will tend to maintain its current behavior.) The official NHC intensity forecasts were close in skill to three of their four main intensity models. These four models were the dynamical Hurricane Weather Research Forecasting (HWRF) and Geophysical Fluid Dynamic Laboratory (GFDL) models, which subdivide the atmosphere into a 3-D grid around the storm and solve the atmospheric equations of fluid flow at each point on the grid, and the statistics-based Logistic Growth Equation Model (LGEM) and Decay Statistical Hurricane Intensity Prediction Scheme (DSHIPS, the SHIPS model with inland decay of a storm factored in.) The GFDL model did poorly in 2015. Note also that NOAA's Global Forecast System (GFS) and the European Center for Medium Range Weather Forecasting model (ECMWF) made lousy intensity forecasts relative to climatology and persistence except at 5 days; these models are generally disregarded by NHC when making intensity forecasts. Image credit: 2015 National Hurricane Center Forecast Verification Report.

It’s PWS Owner Appreciation Week--and your chance to win a personal weather station!
Our thousands of members with personal weather stations (PWSs) are the heart and soul of Weather Underground; it’s the data they provide that helps give WU forecast apps their hyper-local edge. Our first-ever PWS Owner Appreciation Week, running through next Wednesday, May 11, pays tribute to our backyard observer-members. You’ll find profiles of PWS owners and their stations posted on the WU PWS blog. On Thursday, WU’s Madeline Rae introduced us to GoodGreen House, located in central New York between Binghamton and Ithaca, where Jeff White and Liz Smith use a PWS to keep track of conditions both outside and inside the working greenhouse that also serves as their home. Check out the profile to learn more. Earlier this year, Madeline spotlighted Koe Kellen, a firefighter in northern Illinois whose PWS helps Koe and his fellow volunteers assess fire risk.

There’s still time for you to participate in our PWS giveaway. By taking a quick quiz, you’ll have a shot at winning one of 10 personal weather stations or 500 WU T-shirts. Winners will be announced on Wednesday, May 11, so don’t delay!


Have a great weekend, everyone!

Jeff Masters and Bob Henson