Today, we feature a post from guest blogger Dr Richard Dixon, a catastrophe risk modelling expert with over 20 years of experience in meteorological research, catastrophe model building and model evaluation. Richard is a consultant in the insurance industry on model evaluation and is also currently a part-time visiting Research Fellow in the Department of Meteorology at the University of Reading assessing how we can better use climate model output in the insurance industry.
This weekend saw the landfall of the second typhoon of the season in Japan, Typhoon Hagibis. Japanese typhoon risk is a complicated affair for those trying to understand risk impacts for a few reasons. Let’s take a look at the struggles for those of us trying to make head or tail of typhoon losses, both from the real-time and the wider risk perspective.
Jet stream frequency
The real problem for risk assessment in Japan a lot of the time boils down to the Jet stream: how it accelerates hurricanes and changes their structure — and the resultant impact on risk from this. The benchmark for this that many risk modellers will be familiar is the 1938 Hurricane in the US, where an intense tropical cyclone off the coastline of Florida was rapidly accelerated northwards by the Jet stream into Long Island.
As the hurricane and typhoon seasons overlap the extratropical windstorm season, the Jet stream can interact with typhoons more frequently, especially when they move further north into the basin. The chart below shows the mean October windspeed at the Jet stream level: 300 millibars in meteorological speak, or approximately 8km, across the Pacific and Atlantic basins from 1979-2018.
What is fairly obvious is how much more intense the Jet stream is close to Japan when compared to the NE US. Storms that approach Japan towards the end of the season are typically fairly likely to encounter a strong Jet stream that can disrupt, weaken and eventually potentially reintensify dying hurricanes as extratropical storms. We are still in the learning phase about how typhoons interact with jet streams and their resultant behaviour, so this does not naturally help us in risk assessment.

Extratropical Transition
Part of the problem around the frequent interaction of the Jet stream with hurricanes is how they behave during the interaction. Reask’s Thomas Loridan was lead author on a ground-breaking paper with fellow RMS co-authors that shed light on what happens post- extratropical transition.
We’re often used to thinking about extratropical transition as a transition from the circular, symmetrical windfield of a tropical system to an asymmetrical windfield similar to that of an extra-tropical windstorm like Lothar or Anatol where the strongest winds are on the right-hand side of the storm. They showed this to be the case in some typhoons like Halong (2002) on the left of the chart below. However, a significant proportion, like Rammasun (2002) showed strong winds on the left-hand side of the track as well as the right:

Given how close hurricanes can get to Japan, recurve, or make glancing blows — as well as make landfall — you can see how the importance in getting more understanding around the wind footprint of tropical systems interacting with the Jet stream.
Tropical cyclone strength assessment
The major issue in terms of landfalling cyclones — even before we consider the interaction with the Jet stream — is understanding their strength. US landfalling cyclones are blessed with now multiple daily flights to understand their structure and strength and also feed atmospheric information into models to help understand how they will be steered in future.
We have a fairly good idea from these research flights what the likely strength is as and when they make landfall. The best we can manage for typhoons are surface buoy observations and satellite-derived intensity estimates which, pretty good as a guide, don’t tend to nail the intensity on the head. Knowing intensity is all-important for loss impacts given how sensitive damage impacts are to small changes in intensity. Add into the equation here the tendency for typhoons to fall apart rapidly (take Typhoon Lan (2017), for example that went from 915 mb to 980 mb at landfall), and estimating the landfall intensity can be fraught with problems.
At landfall for Hagibis, the Japan Meteorological Agency was suggesting an 80 knot 10-minute sustained wind, or around 90 kt (103 mph) 1-minute sustained wind (Category 2) in US terms. The strongest winds were observed were at Haneda Airport on the outskirts of Tokyo close to the water where 77 mph mean winds – marginally at Category 1 storm – were reported. This suggests that the storm was probably weaker than imagined at landfall (albeit with the impact of surface roughness), but without aircraft observations in a rapidly changing situation, it’s very hard to estimate this.
The Rainfall Problem
It looks as though rainfall will end up getting the most attention for this event. Interaction with the mid-latitude jet stream in extra-tropical transition brings about the additional problem of widespread rainfall. The broadening of the cyclone and interaction with the widespread rising and falling air surrounding the Jet stream can lead to significant rainfall on a broad scale.
Hagibis is nothing new in that sense; across in the Atlantic Basin, think Hurricane Hazel in 1954 whose extratropical transition led to floods that killed more than 80 people some nearly 700 miles inland, or Hurricane Floyd in 1999 whose speed of movement led to widespread flooding and deaths from North Carolina at landfall as far north as Vermont. In both these cases, the two hurricanes were accelerated inland by upper level winds.
Japan is not aided in terms of potential rainfall over land by its west-east orientation, often parallel to the Jet stream, which can stretch the precipitation out along the length of the country. This is evidenced from the map below of a forecast of precipitation (in mm) before landfall. Japan’s rugged terrain is also responsible for orographically enhancing the precipitation, which only exacerbates the rainfall issue.

Urban wind hazard
Tokyo is huge in terms of urban sprawl. The chart below shows the population density of Tokyo alongside that of New York and by way of another loss-driving location in the US, Miami. The densely populated area in the darkest greys (equivalent to 20,000 people per km) stretches over a larger area that New York, as does the lesser, but still highly populated regions in grey. Miami pales in comparison to Tokyo’s population density.


The degradation of wind hazard in major urban areas is a relatively unknown quantity because of lack of official wind observations in these areas: often wind observations are preferred to be in areas with little obstruction within the nearest 50-100 metres. How the wind drops off across high exposure areas is clearly key to understanding the loss impacts.
Interestingly, two stations are fairly closely located to one another in Japan. Haneda Airport (H) on the map below is on the outskirts of the city and central Tokyo (T) in a park, about 15-20 km to the north-west of Haneda Airport.
The data shown below highlights two aspects of the storm at these two locations. The lines show the central pressure as Hagibis passed over the city. Both locations saw a similar minimum pressure of around 967mb. The two sets of dots at each location point to the mean wind and the gusts at each location.

What is marked is how the mean wind speeds at the more exposed Haneda location are significantly higher than inland at the central Tokyo site, something we expect: cities degrade the mean wind. What is remarkable however in this example is how the gustiness of the winds over and above the mean at central Tokyo in the high density urban area is considerably higher. This means that the peak gusts at the two locations are fairly similar, where one might expect gusts to be lower in dense urban areas: and is certainly food for thought. With Tokyo experiencing both Hagibis and Faxai this year, is clearly one of the more highly urbanised areas with a fairly regular typhoon risk from which we could benefit from further understanding as to how tropical cyclones affect large cities to help us better assess risk in urban areas.
Finally, anecdotally, central Tokyo seemed fairly unscathed the following day in spite of these strong winds passing over the area. We might even ask whether the strictness of building codes given the ever-present earthquake risk is making buildings more resilient here?
There are a lot of unknowns at play when typhoons make landfall in Japan, so it’s little wonder that losses take a little longer to be crystallised in this part of the world.