Article

Forecasts for the 2019 Atlantic Hurricane Season

Accurate hurricane forecasting is a bit like the holy grail for reinsurers and ILS managers. If you know whether a hurricane season will be active or inactive, or whether to expect no landfalls or many landfalls, you may be able to adjust your portfolio, for example by taking on more risk during an “inactive” season or buying more hedges during an “active” season. But how reliable are hurricane season forecasts? And what is the forecast for the 2019 hurricane season, which has already begun?

August 20, 2019

Berit Gehring

Head ILS Product Specialist, Credit Suisse Insurance Linked Strategies Ltd

A little background information on seasonal hurricane forecasting

The “pre-season” discussions already started last December when the Department of Atmospheric Science of Colorado State University (CSU) published a first qualitative discussion of the 2019 Atlantic seasonal hurricane activity. CSU is regarded as one of the expert institutions on hurricane forecasting, together with Tropical Storm Risk (TSR) of University College London and the National Hurricane Center of the National Oceanic and Atmospheric Administration (NOAA). The first quantitative seasonal hurricane forecasts, which include figures on how many storms and hurricanes are forecasted, followed in early April. Updates are then typically published every month, with the latest update published in early August. Not surprisingly, forecast accuracy is highest for forecasts issued in July or August, close to the start of the “peak” of the hurricane season. The US hurricane season officially begins on June 1 every year, and ends on November 30. Historically, August and September have been the “peak” months of the hurricane season with the highest storm and hurricane activity.

For more than 100 years, hurricane forecasting has included scientific elements used to predict storm and hurricane activity. In the beginning, it simply focused on detecting approaching storms early on in order to warn and prepare citizens. Seasonal forecast models have only been in operation for around 20 years, but have since greatly improved with enhanced data availability and computing power. Thanks to today’s highly sophisticated statistical forecast models, late season forecasts typically have significant “forecast skill”, which means they are better at predicting storm and hurricane activity compared to when only relying on climate statistics.

The main focus of hurricane forecasting is still to monitor developing storms and accurately forecast their short-term track and strengthening in order to warn and prepare citizens along the US coast or on affected islands. So why do research institutes bother at all with long-term seasonal forecasts? According to CSU, this is done because “it is possible” and because “people are curious to know.” However, there are also more practical and business-relevant applications of seasonal hurricane forecasts: as early as 2003, Niklaus Hilti, the current CEO of Credit Suisse Insurance Linked Strategies Ltd, together with Mark Saunders from TSR and David Simmons from Benfield Group, co-authored a study¹ that showed that reinsurance buying based on late-season hurricane forecasts delivered superior results compared to a static or climatology-focused buying strategy. Since then, and with the rapid development of the ILS market, seasonal hurricane forecasts have become more important for market participants and can influence buying and hedging strategies. It should be noted, however, that an ILS portfolio will enter into most of its US hurricane transactions during the January and May/June renewals when only early-season forecasts with lower forecast skill are available. Late season forecasts, therefore, are most useful in order to make adjustments to existing US hurricane exposures.

The 2019 seasonal hurricane forecast

The first quantitative forecasts for the 2019 hurricane season, issued by TSR and CSU in early April, predicted a season with activity slightly below the long-term average. The updated forecasts have increased the predicted storm activity but it is still only “near-average” or “average”.

	Number of tropical storms	Number of hurricanes	Number of major hurricanes	ACE index²
“Active” season average (1995-2018)	16	8	7	132
Long-term average (1950-2018)	11	6	3	105
CSU forecast, 5 August 2019	14	7	2	105
TSR forecast, 6 August 2019	13	6	2	100
NOAA forecast, (mid-point) 8 August 2019	14	7	3	n/a
2019 season YTD, 12 August 2019	2	1	-	n/a
2018 season	15	8	2	129
2017 season	17	10	6	225

Source: Colorado State University, Tropical Storm Risk, National Oceanic and Atmospheric Administration, Credit Suisse
² The Accumulated Cyclone Energy (ACE) Index is a measure of the power of a hurricane multiplied by the length of time it exists. Therefore, hurricanes with a long duration and particularly strong hurricanes produce high ACE figures.

The main factors driving the current forecasts are the near-average sea surface temperatures (SST) in the tropical northern Atlantic and Caribbean Sea, along with enhanced trade winds due to neutral or weak El Niño – both of which are inhibiting to hurricane activity. The current SSTs in the Atlantic are illustrated below (deviation from long-term climatology). As can be seen, a large area off the east coast of Africa shows significantly lower temperatures compared to the longer-term average. The Caribbean Sea and Gulf of Mexico show slightly higher-than-average temperatures but still significantly below temperatures observed, for example, in 2017. Hurricanes derive their strength from moving over warm sea surface waters. Colder surface sea waters will likely inhibit storms from strengthening. Off the northern US east coast, there is an area with SSTs above the long-term average but storms moving over these waters typically track away from the coast, which means they pose a limited threat of making landfall.

The image show the deviation of the current SST’s (Sea Surface Temperatures) from long-term climatology. It can be seen that a large area off the east coast of Africa shows significantly lower temperatures compared to the longer-term average. The Caribbean Sea and Gulf of Mexico show slightly higher-than-average temperatures but still significantly below temperatures observed, for example, in 2017.

The El Niño Southern Oscillation is another climate factor that is closely observed. Neutral to weak El Niño, which were forecast for this summer, generally indicated the appearance of enhanced trade winds in the Atlantic basin. These trade winds or “vertical wind shear” make it more difficult for hurricanes to form. In contrast, during the second half of 2017, weak La Niña conditions – the opposite of El Niño – prevailed, creating conditions which were more conducive to hurricane formation. And, indeed, August through October 2017 saw the formation of one hurricane after another, the three most notorious of which – Harvey, Irma and Maria – caused billions in economic and insured losses in the US and across the Caribbean.

The latest trend of near-average or even cooler-than-normal SSTs in the Atlantic somewhat contradicts many people’s perception of ongoing climate change and global warming, which are typically believed to cause an increasing number of ever more destructive weather events such as hurricanes. It also highlights the intrinsic difficulty in monitoring and forecasting such complex weather phenomena. Many climate factors interact and can change throughout the season, thereby altering the conditions for hurricane activity.

What does this hurricane forecast mean for the reinsurance and ILS market?

Given the very active 2017 hurricane season, and repeated hurricane losses during 2018 due to Florence and Michael, an “average” hurricane season with limited insurance losses would certainly be welcome. However, a significant level of uncertainty remains given that climate factors such as SSTs or El Niño can still change even during the peak of the hurricane season. And it only takes one strong hurricane to make landfall in an otherwise inactive season to cause high insurance losses – as happened with hurricane Andrew in 1992. On the other hand, a very active hurricane season such as 2010 can result in very low damages if none of the hurricanes make landfall in the US. As such, it appears that overall market conditions, insurers’ normal reinsurance appetite, and catastrophe losses from 2017 and 2018, as opposed to seasonal hurricane forecasts, have been driving both pricing and demand for reinsurance capacity during the 2019 mid-year reinsurance renewals for US hurricane risk.

Latest seasonal update

During July and August, the weak El Niño signal has faded and has now reached “neutral” El Niño status. To date, we have observed above-average trade wind or wind shear in the main hurricane development area in the Atlantic. We believe this trend will continue for the next few days. SSTs are near normal levels and, on average, about 0.3 degrees Celsius warmer than the long-term average which is a moderate level. The forecast for the next few days is for high levels of Sahara Dust over the Atlantic which is cooling the SSTs and stabilizing the air and, hence, is suppressing hurricane formation. Consequently, there are currently no active storm systems in the Atlantic. We believe, however, that conditions may change in the last week of August and the first two weeks of September. At that time, we believe the weather will be more conductive to hurricane formation. As of today, we have roughly 82% of the hurricane season left in terms of statistical activity.

As of the day of publication of this blog, the 2019 hurricane season had produced two named storms – Andrea and Barry – with Barry reaching hurricane strength and making landfall in Louisiana in mid-July. However, insured losses are expected to remain limited.

¹ Mark Saunders, Benfield Hazard Research Centre; David Simmons, Benfield Risk Consultancy; Niklaus Hilti, Helvetia Patria Group (2003). Hurricane Forecasting. The Move Towards Business Relevance. B / Issue Four / Spring 2003 / Benfield Industry Analysis & Research Team.