BMS News

When Hurricane Michael made landfall on October 11th 2018, it had a central pressure of 919mb, which is the third lowest pressure of any U.S. landfalling hurricane. Yet, at the time, the storm was classified on the Saffir Simpson Hurricane Scale as a Category 4 hurricane with winds of 155 mph – only one mph below the Category 5 criteria. On April 19, the National Hurricane Center (NHC) announced what some meteorologists were already thinking, that, at the time of landfall, Hurricane Michael was actually a Category 5 storm with a wind speed intensity of 160 mph. The NHC bumped up the maximum wind speed based on a detailed analysis of aircraft, Doppler radar velocities, and new surface wind speed observations. It is worth noting the caveat from the NHC report that the maximum winds were in a very small area and future revisions are possible. Reconstruction of Stepped Frequency Microwave Radiometer (SFMR) winds showed possible higher values, but the reliability is still unknown, and more research is being done on this data.

There is also some other useful information to the insurance industry in the NHC report. Along a wealth of observational data the report also details some important loss information. The National Centers for Environmental Information (NCEI) currently estimates total economic damages from Michael in the United States at $25B. There have been several media articles written about the continuing insurance loss development. With a current insurance industry loss at $10B, the overall figure is coming close to the generally-held rule that the economic loss is about half the insured loss for U.S. landfalling named storms. Another interesting aspect of the report is the overall damage figures. For example, $3B of loss was on Tyndall Air Force Base near the landfall location. The damage survey suggested that every building at the Base was damaged to some degree. In Mexico Beach, 1,584 out of the 1,692 buildings were damaged with 48% being completely destroyed. The report also goes into damage figure details for Bay County and Gulf County. For example, Marianna, Florida, which is over 50 miles inland, had 1,000 buildings destroyed or with major damage. Seminole County in the Southwest part of Georgia reported 99% of homes were damaged from wind gusts in excess of 100 mph.

Not to be overshadowed, the storm surge section of the report is just as impressive. The storm surge surveys and analysis revealed maximum inundation of 14.7 feet above what is normally dry ground in Mexico Beach, which would also explain much of the damage found in that area.

The formality of upgrading Michael from a Category 4 to a Category 5 is a big deal. Michael is now tied with the San Felipe Hurricane of 1928 or also known as Okeechobee hurricane as the fourth strongest hurricane to strike the United States (including Puerto Rico) since 1900, behind the Labor Day Hurricane (1935), Camille (1969), and Andrew (1992).
If you believe empirical landfall probabilities, the addition of Michael in the U.S. record since 1900 has increased the probability of a Category 5 storm making landfall from 2.52% to 3.36% or, in terms of a return period, 40 years to 30 years. As respects Florida only, the same calculations went from 1.68% to 2.52% and 60 years to 40 years.

Constantly Changing View of Landfalling Hurricane Risk
So, has the overall risk for the insurance industry increased? How will this new data change the insurance industry’s view of landfalling frequency and severity?

Hurricane risk across the U.S. is understood through the use of catastrophe risk models that employ stochastic catalogs containing thousands of events, each one with a small impact on model results. There is an ongoing process of recalibration of the stochastic catalogs. This constant reevaluation of current and past storms contributes materially to the development of a rich stochastic set of events informed by the historical record that is adjusted based on history or future forecasts.

Due to advanced observation techniques and improved understanding of tropical cyclones, the Hurricane Research Division (HRD) has implemented a data set (HURDAT2) that addresses errors and biases identified in the historical record. A working group of scientists makes corrections to location and intensity information in the six-hour track points for select tropical cyclones. Additionally, this research, supported by new evidence and data, adds previously undocumented cyclones to the record. Currently, the scientists are looking at storms after 1960 with further changes likely as a result of new findings.

One does not need to look very hard for proof that catastrophe modeling companies use changes in HURDAT2 to help adjust landfall frequencies. Recently, HURDAT2 updated Hurricane King (1950) from a Category 2 to a Category 3 storm resulting in a long-term rate increase in Miami Dade County in one of the stochastic catalogs. Another notable example of a tropical cyclone reanalysis is the Labor Day Hurricane of 1935. Upon reanalysis, the maximum wind speed at the time of its first landfall over the Florida Keys was increased (23 mph), increasing its intensity to strong Category 5 status. This reanalysis revised the Labor Day Hurricane to be the strongest U.S. landfalling hurricane in the historical record. Additional reanalysis upgraded its second landfall, in northern Florida, from Category 1 to Category 2 intensity and shifted its track closer to land. This change also resulted in changes in loss results from the Florida region in particularly using this event as a historical what if scenario.

As Hurricane Michael continues to be examined by researchers, recent activity will also adjust landfall rates. In fact, the most recent updates of HURDAT2 reevaluated landfalling events during the 2015 – 2016 season, including Hurricane Hermine (2016) and Hurricane Matthew (2016), yielding localized loss increases in the landfall areas. Similarly, losses increased in Canada due to changes in landfall categories for some storms during the 1956 – 1960 period as part of the HURDAT2 updates. In areas that didn’t see any new landfalling storms, small decreases in loss across mainland U.S. resulted.

With the knowledge that Hurricane Michael is, in fact, the fourth Category 5 hurricane to make landfall along the U.S. coastline, it will surely change the historical view of risk, particularly in the Eastern Florida Panhandle, which already has a limited historical record of major hurricanes making landfall. Michael is now the strongest hurricane landfall of record in the Florida Panhandle and only the second known Category 5 landfall on the northern Gulf Coast. Additionally, Michael marks the latest in the season a Category 5 hurricane has made landfall in the United States. All of this should be factored into new landfall rate adjustments.

It is too early to determine potential changes to future stochastic catalogs, but these revisions tend to be in the single percentages based previous Atlantic Hurricane model updates. It should also be a reminder, however, that just because the landfall frequency and severity might change for major hurricanes, the insurance industry should not forget storms of lesser intensity can also cause billions of dollars in loss. Any hurricane can cause great devastation.

The Season Has Begun
There has undoubtedly been a lot of attention given to severe weather across much of the U.S. since the end of April, but in the meantime, the Atlantic hurricane season has arrived. In fact, most people may not have noticed that the Atlantic Basin already had its first named storm (Andrea) of the season. Andrea was a short-lived (less than 24 hours) storm that formed on May 20, about 300 miles south of Bermuda. Andrea became the fifth tropical storm to form before the official start of the Atlantic hurricane season, which did not officially begin until June 1.

Dates of first named storm formation over the past 50 years. The official start of hurricane season is marked by the horizontal green line, and the median date of the first formation is marked by the horizontal dashed black line. The trend over the past 50 years has been for the first named storm to form earlier, with a large spread. There is no correlation between an early start of the season and the season’s overall activity, as discussed here. Source: (Brian McNoldy)

According to the National Hurricane Center, there is a 60% chance of named storm development in the near term in the southern Gulf of Mexico. It’s typical for us to see early season development in this area. At this time, most of the forecast models are bringing tropical weather up into Texas, which will also bring more moisture to the Central Plains where it is surely not needed at this time.

Invest 91 is a tropical disturbance in the southern Gulf of Mexico, which could develop into a named storm as it moves northwest over the coming days. Source: Weathermodels.com

Seasonal Forecasts

I was recently asked: “Does the insurance industry place any weight on seasonal hurricane forecasts?” My quick answer was Yes and No, as it depends.

To elaborate on the Yes part: The insurance industry has always been supportive of seasonal hurricane forecasts by subscribing to private forecasting companies and by funding academic research in these areas. Most companies don’t like surprises and, naturally, being prepared is the logical thing to do for both the reinsurance and insurance industries. However, insurance companies may be more limited in what they can do than reinsurance companies.

The seasonal forecasts can influence the strategic planning of an insurance company by making sure they have adequate claims adjusters in place going into the hurricane season to better serve their customers. Of course, an active hurricane season might require an insurance company to consider service level agreements or loss adjustment expenses and the effects of demand surge that might need to be paid in an active hurricane season. However, since the insurance industry is heavily regulated, companies have little ability to adjust rates for an active or inactive season ahead of time. The reinsurance industry, on the other hand, can be a bit more opportunistic when dealing with seasonal forecasts in terms of planning ahead to provide reinsurance in the secondary market.

To the No side of the answer: Capital requirements for insurance companies are regulated by rating agencies and the insurance commissioners of each state, and do not allow for rate adjustments on a seasonal basis. As part of its strategic planning, an insurance company may want to stress test a portfolio to a certain loss level, as it has the ability to buy more reinsurance during the season if it is uncomfortable with the potential losses that may come from a particular seasonal forecast. The extent of just how much seasonal forecasting plays into each insurance company’s strategic planning is unknown, but generally, insurance companies spend considerable time with strategic planning to understand the potential of losses in any given season, regardless of the Atlantic Basin seasonal activity.

From a business perspective, it’s hard to rely heavily on forecasting when the accuracy of seasonal forecasting is low. Historically, in June seasonal forecasts, the forecasting of activity has not correlated well with actual insurance industry losses. How many times have you heard about 1992’s Hurricane Andrew and how that season was a below normal year? Yet, Andrew happened and was a very big loss for the industry. “It only takes one” will haunt the industry until forecasting gets better.

The readers of my past Atlantic hurricane seasonal forecasts may know that I am all for our industry moving away from simply looking at activity in terms of the number of named storms and hurricanes the Atlantic basin might produce, and rather focus more on what the pattern is suggesting in terms of landfall impacts. This is where we will find the most value, as I describe in more detail below.

Seasonal Forecast and Landfall Threats
The seasonal forecasts we see each year from various sources that provide a number range of expected named storms and hurricanes in the Atlantic Basin are a dime a dozen. There is usually a considerable spread in these forecasts. Currently, there are 19 different forecast groups that have submitted Atlantic seasonal hurricane forecasts to http://seasonalhurricanepredictions.bsc.es/ . The average of these forecasts calls for 6 hurricanes, which is closer to an above normal season.

What do El Niño and La Niña have to do with tropical cyclones? During El Niño, wind shear increases in the Atlantic and we see cooler sea surface temperatures. In La Niña, wind shear decreases and the sea surface temperatures become warmer in the Atlantic, fueling more tropical cyclone activity. Source: NOAACliamte.gov

One of the major factors being considered in many of these forecasts is what the state of El Niño Southern Oscillation (ENSO) may be during the hurricane season. Currently, there is a weak El Niño occurring in the tropical Pacific, and this is forecasted to maintain into the heart of hurricane season. El Niño historically limits named storm development by increasing wind shear across the Caribbean and the Main Development Region. However, not all El Niños are the same, and this El Niño is a Modoki El Niño. This means that the warmest water is found in the central Pacific rather than off the coast of South America. We may see less of an overall impact on Atlantic hurricane season activity with a Modoki El Niño.

General difference between La Niña and  El Niño season.  However, the current El Niño is more of a Modoki El Niño which has less overall impact on the Atlantic hurricane season. Source: NOAA

Another factor to consider, which I think some of the seasonal forecasts may be missing, is that the depth of warm water is shallow and there is actually cooler than normal water beginning to show up at depth, which actually suggests more of a La Niña signature. This could mean that we see very little El Niño impact this hurricane season and instead begin more of a transition to a La Niña weather pattern for 2020.

The image to the left is the  vertical profile of the water temperatures at depth across the Pacific Ocean from South America to Australia as represented by the black line in the image to the right. Note the cold water below the warmer sea surface temperature closer to South American coastline with much of the warmer than normal water in the Central Pacific, which is more of a classic Modoki El Niño signature. Source. NOAA CPC

Sea Surface Temperatures across the Atlantic are currently warmer than usual, which would suggest a more active season. However, we may also want to consider one of the indexes that the insurance industry has looked at for over two decades now. According to Colorado State University, the Atlantic Multidecadal Oscillation (AMO) is currently in a cold phase, which historically would limit named storm activity in a given season, however, SST will be plenty warm enough for named storm development.

The other two factors to watch this Atlantic hurricane season will be Saharan dust and the Madden Julian Oscillation (MJO). Both of these are difficult to predict at seasonal time scales, but understanding the phases of the MJO can help determine when named storm development will occur. The MJO is the major fluctuation in tropical weather on weekly to monthly timescales that comes from pulses of tropical convection over the subcontinent of Asia. The MJO can be characterized as an eastward moving ‘pulse’ of cloud and rainfall near the equator that typically recurs every 30 to 60 day. In fact, the Atlantic basin is currently going into a phase (The MJO is currently in phase 3 and about to enter phase 4) that is not conducive to tropical development. It would appear unlikely that anything of substance develops in the Atlantic Ocean through early July as El Niño and the phase of the MJO limit convection development, which would also limit named storm development.

Overall, it looks like we have the potential for another late blooming season for 2019 with some subtropical development between now and the middle of July, along with a chance for a weak named storm in the Gulf of Mexico this week.

Saharan dust can be an inhibitor of Atlantic Hurricane activity, but it often moves off Africa in waves. In between these breaks of dust, and when combined with the right phase of the MJO, you can find that named storm development has a higher probability of occurring.

Above are the phases of the MJO and the tropical storm tracks that have occurred across the world. Atlantic Ocean is to the far right in the images above.   The MJO can also considerably influence hurricanes in the Gulf of Mexico, Caribbean Sea, and tropical Atlantic. More hurricanes tend to occur in MJO phases 2 and 3 than in phases 6 and 7. Differences in major hurricane numbers and hurricane days in the main development region are a factor of 3. Source: https://journals.ametsoc.org/doi/pdf/10.1175/BAMS-D-12-00026.1

As we’ve discussed, there are multiple factors to weigh when trying to predict the timing for named storm development. However, what are the seasonal climate models suggesting in terms of the possible tracks for named storms? As seasonal climate models continue to get better, we can begin to pick up on the overall pressure and precipitation patterns to help determine where storms will track.

The ECMWF Climate model is suggesting higher wind shear across the Caribbean which would limit tropical storm development in this area.  However, a window of lower shear is forecasted for the Atlantic hurricane season from the Bahamas to south of Bermuda.  Source: Ben Noll Weather

The CanSIPS  climate forecast model provides an view of where the Bermuda Azores high might be positioned for the Atlantic hurricane season. The lower than normal pressure coming from the model could suggest area of more named storms. Pressure below average near U.S. coast, above average in Main Development Region. Source: TropicalTidbits.com

Seasonal climate models are suggesting that above normal precipitation will be occurring this summer off the eastern coast of the U.S. This suggests that this is where the storm track may set up for the season, as storms bring above average rainfall to the eastern coast of the U.S. and the island of Bermuda. Also notice less overall rainfall in the central Atlantic suggesting storms might form closer to the East Coast if they develop. Source: Ben Noll Weather

Summary

I am expecting a more active than normal (Named Storm and Hurricane Storm Counts) Atlantic hurricane season as I think the Modoki El Niño will have less of an overall impact. When one combines this with the warmer than normal SST and weaker than normal wind shear near the East Coast, the conditions should allow for more than normal named storm development. The climate models are suggesting above normal precipitation on the periphery of the west side of the Bermuda high, which could come in the form of named storms during the summer months.  The key to the overall season will be how the MJO and Saharan dust enhance convection, with the next best possible window after this week coming in early July.  How all of this insight will impact the insurance industry is a big question at this point in time, but risks along the East Coast need to be watched.

Catastrophic weather events in 2017 left the re/insurance business with a conundrum. Record-breaking insured catastrophe losses exceeded any previous year. However, the impact of those unprecedented losses on re/insurance pricing has been less than many had hoped for or expected. Claims of nearly $140 billion have not delivered a traditional hard market.

To solve the conundrum and learn the year’s lessons, we must view the storms, fires, and floods of 2017 in a long-term historical perspective. We should not use loss data alone to define weather and climate trends; we need to also understand the impact of the growth of global wealth, as well as patterns of weather extremes.

When we do so, we find that the record-breaking losses were not so enormous, and could have been worse.  About $92 billion of 2017’s insured catastrophe losses arose from three major hurricanes according to the latest Swiss Re Sigma Report. When indexed to 2017 values, Hurricanes Katrina, Rita, and Wilma in 2005 caused insured losses of $112 billion.

We also need to consider that extreme weather memories are typically short. Catastrophe models that simulate the impact of historic storms in today’s built-up environments show that  events such as 1926’s Miami Hurricane or the 1928 Okeechobee Hurricane would cause insured losses that surpass those of 2005 and 2017 hurricane losses. When other named storms and natural catastrophes that occurred worldwide during those years are included, the insured impact on our modern built environment would easily exceeds $140 billion.

Bay Shore Drive in Miami then 1926 after Hurricane and Now. Source NOAA and Google Street View

Meanwhile there is a perception that extreme weather patterns may be changing.  In some cases the data supports this perception.  It is likely that, since 1951, the number of statistically significant regional increases in heavy precipitation events is greater than the number of declines. Strong regional and sub-regional variations modify the trend, but in short, when it rains it often rains more. However, despite the increase in these extreme precipitation events, little evidence suggests the rainwater has led to an increase in floods, with lots of regional variations.

The U.S. Climate Extreme Index quantifies the observed changes in one-day precipitation extremes across the U.S. Since 1990, there has been an increase in the area of the U.S. that has experienced a one-day extreme rainfall.

Tropical cyclones seem not to be changing. After accounting for past shifts in observational capabilities over time, the best records that started in the 1970’s suggest, globally, there are currently no significant observed trends in the number of named tropical cyclones, but the data is suggesting that, when named storms form, they have become stronger in recent years.

2017 had 18 total landfalling named storms. 14 of these had a category 1 or 2 and four of the three major categories were in North America.

As for severe weather, poor data quality makes conclusions about long-term trends problematic. Observed trends in small spatial-scale phenomena such as tornadoes and hail can be made with only low confidence levels, but more recently the data is suggesting clearly a downward trend in the number of occurrences of major tornadoes.

A lot of talk today about the anniversary of the April 2011 Super outbreak. With Joplin later next month many said after 2011 we would see more frequent major tornadoes. Here is the trend which shows a noticeable downward trend since 2011. #MajorTornadoDrought pic.twitter.com/HD3idKeAYI

— Andrew Siffert (@AndrewSiffert) April 27, 2018

With all this in mind, we should learn lessons from the 2017 experience.

• Whether it was the major wildfires in California, flooding of Harvey, the extreme winds of Irma over south Florida, or one of the many damaging hail storms; exposure management is very important and there needs to be a refocus on this simple task to limit losses.
• Valuable data will be collected from the 2017 events, especially from the high wind speeds from HIM. This data will fill critical gaps in the historical record and will be valuable both meteorologically and financially, as it will help insurers and customers by providing better loss models, and lessons for loss mitigation and resilience.
• Building codes matter. When wind-speed data is combined with claims data, we get critical knowledge of how high winds damage different types of construction. We now have validation for some wind-tunnel testing which has been lacking due to the lack of hurricane landfalls over the last decade.
• Hurricane forecast track and intensity still have a great deal of uncertainty. The 2017 landfalls raised questions about the value of early landfall loss projections, and how they could be used.
• Effectively, the people who need flood insurance do not have it.  How many more flood events like Harvey and Katrina do we need before more action is taken?
• Companies with the best business interruption (BI) insurance came out on top after an event, but few small businesses don’t have enough BI.
• Demand for certain post-event services seriously tested supply. The lack of adjusters, for example, created a major bottleneck that led to very high loss adjustment expenses.
• Given the long 10 years without major hurricane landfalling activity, 2017 was a great systems test for InsurTech.  It would appear no major lapses in these new technologies have occurred which should further expedite system enhancements in InsurTech.

You might have noticed that the first named storm (Arlene) of the 2017 Atlantic Hurricane season has formed in the middle of the North Atlantic Ocean – 815 miles west of the Azores Islands.   The biggest impact from this storm will be the discussion in the meteorology community as to whether this system should even be monitored by the National Hurricane Center.  Currently Arlene is over relatively cold ocean water (66 F (19 C)) for a tropical system, and according to satellite data, it is questionable if the system has a warm core.

IR Satellite image show the system is over 19C water.

Typically a tropical cyclone is characterized by lack of warm/cold fronts attached, a “warm core” (air is warmer in center of the cyclone than elsewhere), and persistent deep convection wrapped close to the center; these attributes are commonly referred to as “tropical characteristics” of a cyclone.

It would appear that the decision to track Arlene is very subjective: there have likely been dozens of similar systems over the last 100 years (systems that develop in water warmer than 66 F and with a marginally warm core) that have gone unclassified, including systems that have impacted the U.S. (Sept 2008 SC coast, Sept 2009 NJ).

All this matters to the insurance industry because our hurricane catalogs in the catastrophe models are tuned to the historical data, and after decades there is still no objective guidance as to what type of system gets tracked by the NHC. In some cases like Sandy, questionably categorized storms can have large impacts on landfall definitions as well.

Regardless, Arlene is now in our record books as a storm in the North Atlantic, but it will not be around for long as a strong mid-latitude trough will merge with this system later this weekend and make it unrecognizable. These early-season tropical systems should be no surprise (this will be the sixth early-season storm in the past six years). In fact, history suggests that tropical/subtropical systems in April are uncommon, but not necessarily rare. There have been a few, but they are typically short-lived and innocuous.

Historical Storms that have occurred in April

I am currently tracking the various early season forecasts for the 2017 Atlantic basin season.  I should have my views formulated in a couple of weeks.  What I can tell you right now there is no correlation for April storms and the rest of the season, partly because of a small sample size.

Three weeks ago, James Kossin published a major study in the scientific journal Nature: “Hurricane intensification along United States coast suppressed during active hurricane periods” (http://www.nature.com/nature/journal/v541/n7637/full/nature20783.html). This paper has gotten a bit of press in various insurer publications as over the last decade near term hurricane risk has dramatically influenced the insurance industry in hurricane-prone states. In short, Kossin’s paper provides valuable insight into rapid hurricane intensification during active and inactive periods of hurricane activity in the Atlantic basin (known as the warm and cold phase of the Atlantic Multi-Decadal Oscillation [AMO]). The findings suggest that approaching hurricanes are more likely to intensify before landfall during colder phases of the AMO and are more likely to weaken before landfall during warmer phases. Suggesting that during times when the sea surface temperatures are anomalously warm in the main development region, regions along the U.S. coast experience higher then normal vertical wind shear. Furthermore, the local sea surface temperatures are substantially cooler than the main development region which tend to inhibit intensification and form a “protective barrier” along the United States coastline during this period.

My very first thought as I read about this “protective barrier” was, finally, someone has scientifically proven the existence of HAARP and electromagnetic capabilities of the ionosphere that conspiracy theories suggest can also be used in weather modification and more. (i.e., a “protective barrier” created to control the weather (hurricane landfall))! (If you’re unfamiliar with the HAARP conspiracy theories that abound, then you won’t get this reference, but it’s one of the top questions meteorologists get asked, and you can learn more at http://www.nbcnews.com/science/weird-science/conspiracy-theories-abound-u-s-military-closes-haarp-n112576.)

How the Science has Changed

In all seriousness, to truly understand the impact of Kossin’s paper, it’s important to review how our understanding of active and inactive phases of the Atlantic Ocean has evolved and influenced the insurance industry. Since the 1960’s, the late Dr. Gray showed distinctly differing environmental conditions for tropical cyclone development. But not until the marked increase beginning in 1995 did insurers begin to notice what drives this activity. After the costly 2004 and 2005 hurricane seasons, a dozen and half papers highlighted a link to increased hurricane activity due to a warming world and/or the current phase of the AMO. About the same time, catastrophe modeling companies were being pressured to adjust for this increased frequency and severity that had impacted the insurance industry since 1995. Modeling companies incorporated a range of statistical analyses, and in some cases expert elicitation, to make assessment on future hurricane landfalling risk.

Mother Nature, however, does not always follow weather trends; and since 2005, the U.S. has entered what some have called a “U.S. landfall hurricane drought.” While overall named storm activity remains elevated across the Atlantic basin with some very active years since 2005 (2010, for example), many hurricanes have not made landfall as expected, which is actually influencing the long term landfall rates across the U.S. A few recent research papers suggest that scientists never settled on the exact impact the warmer phase of the AMO might have had on U.S. landfalls. Some of these papers suggest that in a warmer world, African dust and/or North Atlantic sea surface temperatures could actually encourage hurricanes to form further east or recurve before impacting North America, resulting in a lower U.S. hurricane landfall rate.

But Kossin’s paper provides new evidence that the science is not settled, and there is a lot to learn about hurricanes and the rate at which they make landfall. In general, the correlation between the number of hurricanes that develop in the Atlantic basin and number of hurricanes that make landfall is weak. This new study in Nature could explain part of this relationship as it highlights that elevated wind shear (changing wind speeds with height) and cooler sea-surface temperatures along the U.S. coast during the warm active phase of the AMO create conditions that tend to weaken hurricanes as they approach the U.S. coast. Conversely, the cool phase of the AMO makes approaching hurricanes more likely to intensify. This might conflict with ideas that some scientists suggested prior to 2004 and 2005, that the warmer phase of the AMO was correlated to more landfalls and possibly stronger landfalling storms (and fewer landfalls during the cooler phase of the AMO).

The Hurricane “Drought” can be Explained

Kossin suggests in his paper’s closing that this new research could possibly explain the recent “drought” of major hurricane landfalls. However, in late 2015 another research paper funded by Risk Prediction Initiative (“The Arbitrary Definition of the current Atlantic Major Hurricane Landfall Drought”) suggested the major hurricane drought is more a function of the definition of a major hurricane (defined as a Category 3 or higher on the Saffir-Simpson Scale) and the uncertainty in wind estimates. As an example, Ike (2008) made landfall as a high-end Category 2, but could just as easily have been classified as a low-end Category 3 at landfall. For contrast, consider a storm like Sandy (2012), which was not even classified as a hurricane at landfall, but still resulted in large wind and storm surge insurance losses.

Image created by Roger Pielke Jr. who has been keeping track of the days between major hurricane landfalls.

It’s important to remember that only 78 major hurricanes have made landfall since 1900 (117 years), which results in a particularly small sample size, and that there are likely many other factors that affect intensification before landfall. Hurricanes can experience higher levels of disruptive wind shear while turning north, which could be due to a large scale trough of low pressure or drawing in dry air from landmasses. Hurricanes can also send large waves and surge ahead of the storm causing warm water to mix with cooler water and deplete important energy from the storm as it approaches land. These vacillating factors highlight the complex interactions hurricanes may experience during a life cycle. These factors could also explain why many hurricanes have not intensified prior to U.S. landfall (with the exception of Charley [2004] and Humberto [2007]).

Future Hurricane Landfall Rate Models

As always, new research will continue to become available and catastrophe modeling companies will continue to judge how the science best fits into catastrophe models. This new research should not discredit near term view modeling work. Hindsight is 20/20, and as Hall, T.M., and K. Hereid, 2015: The frequency and duration of U.S. hurricane droughts paper suggests several observations point to the current drought being more a case of good luck than any shift in hurricane climate. Therefore this luck could have swayed higher then normal insured losses in a completely different direction over the last 10 years. These facts highlight why it is important to continuously evaluate new science. Modeling companies need to ensure they avoid knee-jerk reactions to industry demands or speculative science. Seasonal hurricane forecasts are improving, and with time, multidecadal sea surface patterns and other variables could help clarify the cause of reduced landfall rates, not just in the Atlantic basin, but in other basins near high exposure. El Niño/Southern Oscillation (“ENSO”) and other variables could also be worked into the landfall rate catalogs produced by model vendors.

As pointed out in this blog, the science might not be settled on hurricane landfall rates and what drives these rates during particular climatic conditions. Although it is short by climate standards, the U.S. landfall hurricane record is one of the best natural catastrophe databases insurers have to understand frequency and severity. There is a reason why catastrophe model companies calibrate their models to the longest period of record available.

There is plenty of room for debate about near term hurricane landfall rate models, the drivers of this risk, and what position an insurer should take regarding rate models. Are they a good idea or a bad idea? Should a blended approach be used? Those are interesting questions and a worthy topic, which is why I will be moderating the session, “Is Near Term vs. Long Term View of Hurricane Risk Over?” at the February RAA Cat Management Conference. We hope this subject generates lively discussion and hope to see you there!

Now that Matthew’s story is complete, immediate attention will turn to Nicole, a tropical storm currently 450 miles south of Bermuda. The models generally agree that Nicole will slowly strengthen back into a hurricane and that there is a good chance that Matthew will become a strong category 1 or weak category 2 hurricane as it tracks close to Bermuda this Thursday.
So with no tropical troubles threatening the U.S. coastline in the immediate future, this is the time of year that the insurance industry often wonders if there is any other tropical trouble forecasted for the remainder of the year.

Climatologically over the next two weeks we tend to see stable named storm activity, but after October 18th, the activity in the Atlantic Basin drastically trails off.

This is also roughly the time of year that the West African – Cape Verde type hurricane season trails off due to the equator-ward shift of the African Monsoon. This past weekend the overall tropical rainfall seemed to shift below 10 degrees north latitude, which is the benchmark for tropical waves coming off Africa to obtain enough spin to become named storms.

Here is a look at the tropical rainfall totals since mid July. It shows the height of the African Waves during middle of August with a trailing off over since this period. The axis to the left is latitude showing a trend towards the equator.

This is also the time of year that colder Canadian troughs of low pressure air start to invade the U.S. Often these weather systems leave trailing cold fronts that sometimes extend into the southern Gulf of Mexico or Caribbean. Low pressure can form at the tail end of these fronts, and if the water is warm enough in the Gulf or Caribbean, it can encourage tropical convection which could then become organized. These typical developments are consistent with some of the models’ long-range weather forecasts for later next week.

Forecasts for Thursday morning show an area of low pressure moving across the upper Great Lakes, which could provide the first freeze of the year for the upper Midwest. This forecast also shows Nicole moving toward Bermuda.

This low pressure moves across the north Atlantic late next week, and it leaves a trailing cold front with an area of low pressure off the east coast of Mexico in the western Caribbean. This is the area that needs to be watched for tropical trouble later next week as the water in this area is plenty warm enough to support tropical convection.

The long range forecast from Oct 17  – Oct 24th  total precipitation plot below show a decent amount of rain which could be  from  convection thunderstorms.

The graphic below illustrates the climatological pattern that is typical of tropical storms that originate in October so the Western Caribbean is the place to watch over the next three week.

Mathew End Game

Matthew is now classified as a post tropical storm as the majority of its energy has merged with a mid-latitude frontal system.   As a result Matthew will once again defy the long range forecast models that had suggested Matthew’s energy would travel southward towards the Bahamas.  Now Matthew will go out to sea just like the majority of tropical systems and Matthew’s long story will end.

Matthew Forecast issued at 8 am EDT Saturday Morning. Showed Matthew likely taking a southern track in the long term

Almost a complete 180 degree change in the direction Matthew is expected to take in today’s 8 AM NHC official forecast. Now Matthew is expected to track north of Bermuda.

Flooding

However, the damage is done and it is becoming apparent that water related damages could add up to more than the wind related damages.   In fact’ the water damages will continue to add up as many major rivers in Carolinas are expected to hit record flood levels.  In some cases there have been dam breaks which is reminiscent of the incredible rain event just over a year ago from tropical moisture from Hurricane Joaquin.  So with major rivers still rising, the flood situation is still unfolding, particularly in North Carolina which will lead to more losses in the next few days. However, it is assumed the majority of the inland flooding / river/ storm surge losses are either uninsured or are being assigned to flood policies by the NFIP.   As shown yesterday, the National Flood Insurance Program take-up rate by counties which are heavily coastal in nature across northern Florida through the Carolinas decreases for inland counties.  Therefore the majority of insured losses should be driven by wind related claims.

Record flooding forecasted at Goldsboro, NC

Record Flooding forecated for Kinston, NC

This record river flooding is a result of heavy rainfall which in some places has totaled more than 12”

Past 7 Days of Rainfall observed. It should be noted BMS iVision also has a rainfall layer users can use to understand risk impacts to flooding.

As highlighted in yesterday’s update storm surge has also done a lot of damage along the entire coast from Flroida up to North Carolina.

• Mayport, FL 3.28 feet:  Previous record of 2.47’, during Hurricane Jeanne on September 27, 2004 (records since 1928.)
• Fort Pulaski, Georgia 5.06 feet:  Previous record of 3.40’ during the October 15, 1947 hurricane (records since 1935.)
• Wilmington, NC 8.21 feet:  Previous record of 8.15 feet during Hurricane Hazel in October 1954.

Vilano Beach homes, and what’s left of the dunes they were built on @FCN2go#FCNStormpic.twitter.com/MXhUjIF6ss

— anne schindler (@schindy) October 8, 2016

Notice in the image above however that one of the areas hardest hit by winds in Florida right along the coastline. Most of the homes are relatively undamaged from this view.   Other damage from University Florida Damage Assessment group find minor roof and exterior damage along parts of the Florida coastline with surge leading to most of the damage observed.

Wind damage

As pointed out several days ago, treefall would be enhanced due to the saturated soil conditions in the Carolinas. With this combined with wind gusts and the lack of any significant large scale wind event over the region in several years, wind related tree damage could increase insured losses.

Since wind related damages will likely drive the insured loss, users of BMS iVision can now get a full picture of Matthew’s wind swath by looking at the 1 minute wind speed and 3 sec gust wind speed.   In the coming days   users will find wind duration information.

BMS iVision Verisk Climate 3 sec wind gust swath for Matthew

Industry Loss

The only public report of insurance industry loss numbers was in a report issued Saturday afternoon by, CoreLogic, which estimates Matthew’s insured losses to range between $4B to $8B with 90% of loss being driven by wind claims, 10% by storm surge.  These estimates are well below the industry loss estimates projected by the major catastrophe modeling firms late last week.  However, they are somewhat inline with the other industry loss numbers from other modeling firms (not yet released publicly).

While Matthew was within 20 miles of bringing much larger insured property losses, Phil Klotzbach of Colorado State has compiled the historic statistics on Hurricane Matthew.

Hurricane Matthew Records/Notable Facts Recap (through October 8)

Intensity

• 80 mph intensification in 24 hours – the 3rd strongest rapid intensification in the Atlantic on record (trailing Wilma-2005 & Felix-2007).
• 31st Atlantic Category 5 hurricane on record and the 1st since Hurricane Felix (2007)
• Lowest latitude Atlantic Category 5 hurricane on record
• 6th lowest MSLP for any Atlantic October on record at 934 mb (trailing Joan, Opal, Mitch, Wilma & Joaquin). Consistent MLSP records date back to 1979.

Longevity

• Longest-lived Category 4-5 hurricane in the eastern Caribbean (<=20°N, 90- 60°W) on record.
• Generated the most Accumulated Cyclone Energy on record for any hurricane in the eastern Caribbean
• Maintained Category 4-5 hurricane strength for 102 hours in October – the longest that a hurricane has maintained Category 4-5 strength on record during October in the Atlantic
• Maintained major hurricane strength for 7.25 days – the longest-lived major hurricane forming after September 25 on record and longest lasting at any time of year since Ivan (2004). Tied with Fabian (2003) for 5th longest major hurricane in satellite era (since 1966)
• Currently ranked 8th for Accumulated Cyclone Energy by an Atlantic hurricane in the satellite era

Landfall

• 1st Category 4 hurricane to make landfall in Haiti since Cleo (1964) – 1st Category 4 hurricane to make landfall in Cuba since Ike (2008)
• 1st time on record that a major hurricane has made landfall in Haiti, Cuba and the Bahamas
• 2nd time that a Category 4 hurricane has made landfall in the Bahamas since 1866 (Joaquin-2015 was the other)
• 1st hurricane to make landfall in South Carolina since Gaston (2004)
• 1st hurricane to make landfall north of Georgia in October since Hazel (1954)

Note: While Atlantic hurricane records go back to 1851, there are likely underestimates in storm intensity prior to the satellite era (since 1966) and especially prior to aircraft reconnaissance (since 1944).

Matthew Weakening

Matthew has weakened from a category 3 hurricane yesterday afternoon down to a category 1 hurricane overnight.  Matthew has officially made landfall near Mcclellanville, SC as a category 1 hurricane.   Matthew has weakened due to a shallower continental shelf and lower ocean heat content. Also as Matthew has tracked further north it has been able to wrap in some cooler, drier air into the center of its circulation allowing some weakening.

Over the next day Matthew will continue to weaken as it moves along the South Carolina coastline before it moves out into the Atlantic near Wilmington, NC where it will start to track southeastward. The biggest impacts over the next 24 hours will be inland flooding as the rain shield from Matthew extends across the Carolinas and into the southern  Mid-Atlantic states as the warm moist air from Matthew rides up and over a frontal boundary across the eastern United States.  Just one year after hurricane Joaquin, enhanced rainfall parts of the Carolina’s could see 5 – 14 inches of rain again.

Total radar observed rainfall over the last 7 days

Percent of homes that have flood policy thru National Flood Insurance Program (NFIP). Very low take-up rate inland.

Overnight Impacts

Storm surge has been the biggest impact overnight. Savannah, GA  has seen record flooding breaking the previous record which was created during hurricane David in 1979.  Other hard hit areas by high storm surge were Tybee Island, GA, Hilton Head Island, SC and Charleston, SC.   However, it would appear the overall height of storm surge was 2 feet below what was forecasted for many of these areas.

Tide Gage at Fort Pulaski near Savannah, GA

It has been a long time since a hurricane of this strength has impacted the Georgia, South Carolina coastline.  In fact a few years ago I took a trip to Hilton Head Island, SC and I could not believe the over growth of foliage on the island, which was likely due to the fact this area has not recently been hit by a hurricane.  Hurricanes are nature’s way of cleaning out some of that foliage.  I expect this could lead to some higher than normal insured loss from tree fall right along the coastline.   The maximum wind gust record at Hilton Head, SC was 87 mph and Tybee Island, GA was 96 mph.

Check out this huge tree become uprooted after strong winds from #HurricaneMatthew! pic.twitter.com/MrNNceeyAB

— News4JAX (@wjxt4) October 7, 2016

BMS iVision shows the historical 3 sec Max Wind Gust and now the forecasted wind gust. This allows clients to get a complete picture of possible wind damage.

Matthew Extended Forecast    

Later next week Matthew will be a tropical storm or depression off the southeastern U.S.   It would now appear that Matthew’s energy will reach the Bahamas and even track south into the southern Gulf of Mexico later next week, however, not as a named system.   Hurricane Ivan in 2004 made a similar large loop and reformed in the Gulf of Mexico.  No models at this time forecast Matthew reforming in the southern Gulf of Mexico, but this is something to watch out for. In the next three weeks this would be the area to watch for new tropical development.     There is still a chance that Matthew will get caught up with hurricane Nicole in the Atlantic and move out into the open ocean later next week as well.

ECMWF ensemble forecast showing many of the 51 members tracking the remnants of Matthew south of the Florida keys next week.

Summary

• 4,001days: that is how long it has been since a major hurricane (Wilma) made landfall in south Florida. And today, Matthew, the violent end to the U.S major hurricane drought, rages into day two.
• Many of the hurricane models that forecasted a central Florida landfall were clearly wrong, as Matthew is currently 35 miles off the coast of Cape Canaveral as a 120-mph category 3 hurricane. Even the trusted ECMWF model predicted a central Florida impact yesterday. But there are some model winners. The BMS iVision Verisk Climate wind model never predicted a landfalling outcome. And yet again, we see that there is continued room for improving hurricane track and intensity forecasting.
• As I suggested on October 5, hurricane track will drive outcomes. I also mentioned that the storm size would make a difference, and Matthew’s hurricane force winds never really expanded.
• Matthew is now expected to parallel the Florida coastline and slowly weaken, so insured loss should be confined to near-coastal areas close to its current location and up to the southern North Carolina coastline.
• Matthew is forecasted to stall or loop near the southeast coast of the U.S.

Landfall and Wind Impacts

“Landfall” is defined as the center of low pressure that intersects with the coast. This has not occurred with Matthew. For several days, weather models suggested that Matthew’s ultimate landfall (or lack thereof) would be determined by just a matter of miles. What might be more interesting than whether or not Matthew makes landfall is the fact that the last two largest U.S. insured losses from a hurricane will ultimately be caused by storms that failed to make landfall. Sandy (2012) was never officially classified as a hurricane and thus didn’t make landfall. Matthew may never make landfall, but will still likely cause significant insured loss; most likely the highest insured loss since Sandy.

Matthew never grew in size as expected, and for the most part, always maintained a very small area of hurricane force winds. This area grew only nominally as it moved north as illustrated in the NHC wind history plot below.

Past extent of the hurricane force winds and tropical storm force winds

Matthew has maintained its track parallel to the Florida coastline overnight, as it never got pulled landward, and the hurricane-force wind around its small center of circulation have primarily stayed off shore.
This will likely mean that insured losses will be limited, so in some cases insurers will ultimately retain a greater portion of their losses (versus collecting on higher attaching reinsurance programs designed to respond to major natural catastrophes). These retentions could add up, especially since it has been 35 days since hurricane Hermine impacted the state and caused minimal insured loss primarily retained by insurers. Still, the full extent of insured losses will continue to develop as Matthew continues to track up the north-Florida, Georgia, and South Carolina coast.
Wind Reports

I have yet to identify a hurricane-force wind speed recorded at an official weather station. The highest wind gust I have seen so far is 74 mph, recorded at Vero Beach, Florida. This would not translate into a hurricane-force wind, which must be measured at 1 minute resolution. There have been other reports of high wind, especially on the high tower near the Cape Canaveral coastline. Keep in mind, however, that many of these towers are hundreds of feet tall, and much higher than the official 30-foot measuring height.  Other higher windspeed gusts could occur yet today as Matthew tracks very close to the Florida coastline.

Strongest wind guest from observations last nights

Although structures will likely sustain damage given the observed wind speeds, ultimate damage should be minimal. To illustrate, see the damage plot below for a pre-2002 home that experienced category2 wind speeds. The outcome shows 35% probability of at least roof sheathing damage. Probability drops to a 5% for a category 1, and even less for a tropical storm force wind.

General  probability of at least roof sheathing damage vs wind speed

Rainfall
As one would expect, and as my blog previously suggested, flooding can be expected in Georgia and the Carolinas as Matthew tracks northward.

NOAA rainfall forecast showing 13 inches rainfall for areas of the coastline

Surge
Storm surge has not been as bad as forecasted. Areas with 9-foot surge forecasts ultimately experienced only 6-foot surges. However, high storm surge values are still forecasted for coastal Georgia and South Carolina as Matthew works it way northward. Currently some of the highest surge forecast values are found along the North Georgia.

Water value observed near Trident Pier, FL showing the build up storm surge at this station in this area and the fall of that water as the storm passed and the water gets pulled away from the shoreline.

Matthew Just Won’t Go Away
The burning questions seem to be, “Is Matthew really going to do a loop off the southeast coastline? Is this possible?” Yes, it is very possible, but also unusual.
Gordan (1994), Felix (1995), Ivan (2004) and Jeanne (2004) are all examples of looping storms. However, as indicated in the plot below, south or even southeast hurricane movement is rare, as most trajectories for Atlantic hurricanes track northeast.

Hurricane Gordon 1994 did a nice loop off the southeast coast

Another burning question is, “Will Matthew loop back and impact the Bahamas and Florida again?” It is possible and still a possible outcome on many of the long-range ensemble forecasts. My bet is that Matthew will loop northeast of the Bahamas, but then be kicked back out to sea toward Bermuda around Thursday next week.

ECMWF track model showing the chances of the Matthew getting back to Florida or Bahamas

Summary

• Matthew is a weak category-4 hurricane and will likely slowly strengthen today as passes over Nassau, Bahamas.
• Matthew will likely make landfall between Jupiter, Florida, and West Palm Beach, Florida, as a category-4 hurricane. Landfall should weaken Matthew to a category 2 by the time it reaches Daytona Beach, Florida, and then travels up the coastline. I don’t anticipate that the center will track more than 10-20 miles inland.
• There is a very small chance that Matthew will become a category 5 hurricane before it reaches the Florida coastline, as it will likely strengthen until its inner core starts to interact with the Florida coastline.
• Estimating the maximum storm surge is a bit complicated as these forecasts depend on Matthew’s track and strength as it travels up the Florida coast. If Matthew weakens due to land interaction, surge will be less severe; but if Matthew stays off shore, more water may move onto the coast, and this surge could be further complicated by multiple tide cycles that are expected during Matthew’s move up the coastline.
• Given the large amount of exposure along the coast and expected wind gusts of 144 mph, this will no doubt be a significant insured loss event. For specific impacts, each National Weather Service (NWS) office is offering detailed forecasts for Matthew.
• In the longer-range forecast, Matthew will likely loop over or the stall off the southeast coast this weekend or early next week. This will limit insured losses in locations that are north of South Carolina.

Forecasted Landfall Location and Specifics
As I have mentioned, Matthew is tracking over very warm water in the Bahamas. This warm water will continue to fuel Matthew until it starts to interact with the Florida coastline tomorrow morning. Keep in mind that hurricane-intensity forecasting is typically more predictable than hurricane-track forecasting. Matthew already surprised us this past weekend with rapid intensification to a category-5 hurricane in the southern Caribbean. It should be noted that, currently, no model guidance is forecasting further rapid intensification.  The intensity guidance below show most models are in the camp that Matthew will be category 4 hurricane at landfall or along the Florida coastline.

With Florida land interaction expected in the 36 hours. The intensity of Matthew should weaken considerable especially if it tracks inland a bit vs staying over open water.

For the last week, meteorologists have used the long-range hurricane models to predict Matthew’s track. We are now within what is considered the short-term forecast window where higher-resolution weather models become a bit more useful for providing specific hurricane forecasting.
HRRR (High Resolution Rapid Refresh) is a high-resolution (3K grid resolution) model which shows the details of a hurricane and its impacts. This model is updated hourly and forecasts the next 18 hours. Here is the latest hourly forecasts.

My general expectation is that Matthew will make landfall and track inland 10-20 miles along interstate highway 95. This counters some high-resolution models that suggest Matthew will track along the Florida coastline. Three of these models are the HWRF, GFDL and Verisk Climate model (which is the basis for BMS iVision).

HWRF Forecasted Wind Swath. This is not surface wind speeds These are winds at 900 mb which close to the surface.

GFDL Forecasted Wind Swath. This is not surface wind speeds These are winds at 900 mb which close to the surface.

This is the National Weather Service (NWS) Forecated wind swath

BMS iVision Verisk Climate wind swath which is a version of the WRF model. Users of iVision can use this product to get an ideas of risks impacted by wind speed.

One important point I want to make is no numerical weather prediction model is at a high enough resolution to resolve the details of surface roughness. It is this surface roughness that would likely weaken the winds inland. So many of the modeled forecast plots above are likely overestimating the wind speed. This is particularly apparent in the National Weather Service forecasted wind swath.

At this time, much can be gleaned from the various NWS forecast products because they offer great detail on potential impacts. These details include likely power outages, specific areas that are prone to flooding, tornado threats, etc. Many of these NWS offices offer public video briefings as well. In some cases the wording is quite strong, and for good reason, as Matthew is a dangerous, life-threating storm.

Estimating the maximum storm surge is a bit complicated as these forecasts depend on Matthew’s track and strength as it travels up the Florida coast. If Matthew weakens due to land interaction, surge will be less severe; but if Matthew stays off shore, more water may move onto the coast, and this surge could be further complicated by multiple tide cycles that are expected during Matthew’s move up the coastline.  Please utilize the new NHC storm surge forecast guidance. This guidance will change with the NHC track forecast. Right now the worse storm surge will occur in north Florida where the continental shelf extends out into the Atlantic and water will have more time to pile up.  Surge values in this general area could be greater than 9 feet in isolated locations.

Storm Surge Forecast For Jupiter, FL

Storm Surge Forecast For St. Augustine

Another issue that is not being talked about much is the possible flooding of Lake Okeechobee. The lake level is very high right now at 15.78 feet.  A  risk of flooding starts to occur if this lake reaches 17″. If the lake level goes up to 20″ this could be a major concern as this is a known dike that is dire need of upgrades and if given to much stress it could break if water levels get too high. Matthew will no doubt raise this lake level.   The Weather Channel has a good summary  of the risk.

Long Range Forecast

As we have seen for over the last week, Matthew’s long-range forecast remains uncertain. But the chance of significant insured loss at locations north of South Carolina is declining. Many global models suggest Matthew will loop off the southeast coast. Some of these model solutions even take Matthew back into the Bahamas or Florida, and in one extreme case, back into the Gulf of Mexico. Most likely, Matthew will sit and stall near the southeast coast and weaken while doing so. Matthew will eventually be kicked out into the Atlantic Ocean. However, as with any long-range hurricane track forecast, a lot of variables are at play, so Matthew’s long-range tracking still requires watching.

The newest ECMWF forecast which has shown the best skill in the long range 3+ day range