Seasonal Hurricane Forecast Skill

Seasonal hurricane forecasts, with varying lead times, have been produced in the Atlantic basin since 1984 by the late Dr. Gray from Colorado State University (CSU).  Partly as a result of the early success of those forecasts, seasonal named storm forecasts are now a dime a dozen with many different research and operational groups making seasonal hurricane forecasts for tropical basins around the world.  Although there are some very brilliant minds making these forecasts, some might argue there has been little skill improvement over the last several years. The insurance industry is accustomed to using climatology to understand the risk of a U.S. landfalling hurricane.  This climatology also suggests, on average, the more active the overall Atlantic basin hurricane season is, the greater the probability of U.S. hurricane landfall which can be useful guidance to the insurance industry heading into the hurricane season. However, when it comes to seasonal landfall forecasts, really no skill has been demonstrated; this is ultimately what matters for the insurance industry.  In fact, there isn’t a forecast group that has skillfully predicted the current drought in major hurricane strikes over the last 11 years.

CSU Atlantic Basin wide hurricane forecasted and observed hurricanes with missed forecasted number shown.

CSU provides easy access to historical forecasts which allows for simple validation of there seasonal forecasts.  In a simple analysis, the CSU forecast has done fairly well with their April predictions when compared to using historic climatology (6.5 hurricanes per year) .  Fourteen CSU forecasts since 1995 have come within three hurricanes of the seasonal hurricane total in the Atlantic Basin.  On seven occasions, they have only missed the cumulative number by one.  They hit the mark in 2008.  Of course, CSU have posted a few wayward forecasts as well: 2005 (undershot the total by eight hurricanes) and 2012 (fell short by six hurricanes).

CSU Atlantic Basin wide major hurricane forecasted and observed hurricanes with missed forecasted number shown.

Regarding the major hurricane forecasts, CSU has made seven forecasts that failed to fall within two storms of the final tally. This should not obscure the fact that they’ve also posted four perfect forecasts of major hurricane occurrence and 13 others that approximated the seasonal total within one storm.
It is very easy for one to sit back and grade objective season hurricane forecasts like CSU, and there is no doubt a lot of work goes into the refinement of these forecasts each year. I am a firm believer that seasonal forecasts and even landfall forecasts will improve over time. Currently, many of these seasonal forecasts are statistical schemes which will inherently fail some years. There is some promise that a hybrid of statistical and dynamical forecast could be the future of seasonal hurricane forecasting.

https://www.gfdl.noaa.gov/hyhufs/

The Real Value – Landfall Forecasts

As mentioned, the real value to the insurance industry is in getting an accurate picture of potential landfalling storm activity. CSU issues probabilities for landfalling hurricanes across various regions of the U.S. This year, they are forecasting a 75% probability of a hurricane landfall. The probability of landfall for any one location along the coast is very low and reflects the fact that, in any one season, most U.S. coastal areas will not feel the effects of a hurricane no matter how active the individual season is. And, given the variation of exposure along the coast, providing accurate estimated insured loss ahead of season is very difficult if not impossible.
However, given the current state of the science of seasonal forecasting it is possible to provide guidance to which area of the coastline will see a higher than normal probability of landfall. This year, landfall forecasts may prove to be one of the more challenging years. This is due to the large amount of uncertainty of many of the climate forcers used to predict Atlantic named storm activity. One of the most common climate forcers used is El Niño Southern Oscillation (ENSO) and, this year, many forecasts are uncertain if an El Niño event will take hold during the summer, which historically suppresses hurricane activity. CSU thinks an El Niño will be weak or moderate by the peak of the Atlantic hurricane season based on their April forecast. However, as mentioned a few weeks ago, spring El Niño forecasts can be notoriously misleading, and result in difficult forecasts. (This phenomenon is so familiar that it has its own name – the “spring predictability barrier”.) Over the last few months many of the monthly climate models have backed off of the idea of a strong El Niño during the peak of the Atlantic hurricane season and now have a weak to neutral ENSO condition. If the El Niño is weak, it could result in increased landfall chances, and El Niño historically has little to do with any activity in the Western Caribbean / Gulf of Mexico and any activity in this region increases the chances of U.S. landfall.

Official Climate Prediction Center ENSO probability forecast for Mid-April (Top) and Early May (Bottom) Showing in just a half of month the probability of a El Niño during the hurricane season (Gray Box) went form 70% to below 50%. There is an increasing signal that ENSO will be neutral during the hurricane seasons.

The other challenge this year is that earlier this spring, water in the far North Atlantic and water off the coast of Africa had a relatively cold signal which is potentially indicative of a negative phase of the Atlantic Multi-Decadal Oscillation (AMO).  However, over the last month or so, the Atlantic water has warmed substantially.  This continues the debate as to whether the AMO has switched to a negative phase.  The U.S. government suggests the Atlantic is still in a positive state, whereas the CSU data suggests the AMO has trended negatively over the past few years.  If water continues to warm over the main development region, like it has over the last month, it could allow more storm formation in the heart of the hurricane season.

NOAA SST anomaly (degree C) for 4/10 (left) and 5/11 (right) showing a warming of SST off the west coast of Africa off the East Coast of the US. The cold SST in the North Atlantic have expanded and SST after months of above normal in the Gulf Of Mexico have recently turned colder than normal which should be short lived.

Another factor that has had a large influence on named storm activity over the last few years has been the Saharan Air Layer (SAL).  Named storms interact with the SAL in several ways.  Some named storms get embedded in the SAL their entire life-cycle and often struggle to intensify beyond strong tropical storm strength.  Other named storms can be overrun by the faster moving SAL and are quickly weakened.  The SAL in general can hamper convection, making it hard for named storms to develop.  This year could be similar to the last couple of years which would hamper development of named storm activity in the main development region but, as tropical waves get closer to the U.S. and Caribbean, the SAL decreases, and conditions become more favorable for development outside of the SAL.

SST anomaly for 5/11 showing warmer than normal SST around Hawaii.

Not to forget about Hawaii, which is also a target for named storm activity and, like the U.S. mainland, has gone a long time without a major hurricane impact.   This year, the insurance industry should also keep an eye on this 50th state as, just like the warm sea surface temperature (SST) off the East Coast of the U.S., the SSTs are above normal near Hawaii, and any tropical activity that might come close to the islands could be enhanced by these warmer than normal SSTs.

Summary
Forecasting of Atlantic named storm activity is not easy, and CSU has demonstrated variability in skill year over year, with no real improvement in the April seasonal forecasts over the last several years, but overall there is skill improvement over using basic climatology. Although some forecasts are currently calling for below normal activity in the Atlantic basin, climate forcers are pointing toward the ideas that these forecasts will be adjusted upward as we approach June 1 to a more active or normal season. The major climate forcers suggest there could be less development in the main development region of the Atlantic but, as tropical waves move closer to the U.S., they could have a better chance at developing, due to lack of wind shear and lower SAL environment. SSTs near the U.S. coastline are also expected to be warmer than normal which would add fuel to any disturbances that develop. Therefore, parts of the U.S. coastline have a higher probability of a landfalling named storm this season, with the Mid-Atlantic and Northeast seeing the highest chance.
Resources:
Barcelona Supercomputing Center has a nice summary of the various Atlantic Hurricane Forecasts
http://www.bsc.es/ESS/seasonalhurricanepredictions/
NOAA Climatology Products – Avg Start Dates, Return Periods, Develop Origin by Date, Strike Density
http://www.nhc.noaa.gov/climo/
CSU U.S. Landfalling Hurricane Probability Project
http://www.e-transit.org/hurricane/welcome.html