by Judith Curry
“Impending massive hurricanes bring the best out of weather twitter and the worst out of climate twitter” – Joseph Maykut
Every time we have a tropical cyclone landfall in the U.S., there is an explosion of public statements regarding the role (or not) of human caused global warming on the impacts suffered upon landfall.
This blog post sorts through the various claims, and also provides a way forward for objectively assessing them.
Note, I am almost finished with my big report on sea level rise. After that, I will begin a comprehensive assessment of the hurricanes and global warming topic. This post is a preliminary taste, before I really dig in and focus on this.
Meteorological recap
Florence originated from a strong African easterly wave that emerged off the coast on August 30. Heading on a west-northwest trajectory, the system became a tropical storm on September 1. An unexpected bout of rapid intensification ensued on September 4–5, Florence emerging as a Cat 4.
Wind shear then tore the storm apart and Florence degraded to a tropical storm by September 7. The system regained hurricane strength on Sept 9 and major hurricane status by the following day, becoming a Cat 4 on Sept 10. Afterwards, Florence weakened slightly as it underwent an eyewall replacement cycle. After some restrengthening wind shear reduced the intensity although Florence continued to increase in horizontal size. On Sept 14, Florence made landfall on the southern coast of North Carolina as a Cat 1.
The main meteorological impact has been record breaking precipitation, that is continuing as the post-tropical storm slowly moves across the mid Atlantic states, expecting to exit back into the Atlantic later today.
The forecast for Florence has to be regarded as a success story. 5-7 days lead time was provided for a landfall (or pseudo landfall) near Wilmington, SC. Florence presented some challenges for intensity forecasting, owing to its very large size and evolving complex structure.
Pre landfall forecasts of storm surge, wind speed and precipitation were qualitatively correct, sufficiently accurate to support emergency management planning 3-5 days in advance.
A full verification report will be forthcoming, comparing track forecasts from the different models (including CFAN’s forecast), plus CFAN’s landfall wind forecasts and precipitation forecasts.
The controversy
In an unusual move, on Sept 12 a ‘pre-event attribution’ analysis was published online [link]:
We find that rainfall will be significantly increased by over 50% in the heaviest precipitating parts of the storm. This increase is substantially larger than expected from thermodynamic considerations alone. We further find that the storm will remain at a high category on the SaffirSimpson scale for a longer duration and that the storm is approximately 80 km in diameter larger at landfall because of the human interference in the climate system.
This was followed by a USA Today op-ed by Roy Spencer, entitled Hurricane Florence is not climate change or global warming. Its just the weather. The title pretty much summarizes the article.
Seth Borenstein of AP published interviewed 17 meteorologists climate scientists in an article entitled A warmer world makes hurricanes more intense. This list included a number of climate scientists/activists who apparently know nothing about hurricanes, who got most of the space in the article.
Climate Feedback responded to Roy Spencer’s op-ed with a fact check entitled USA Today op-ed ignores evidence to claim climate change had no role in Hurricane Florence. It includes extensive commentary by Kerry Emanuel that is worth reading.
And finally Jim Hansen has posted an article entitled Global Warming and East Coast Hurricanes. “So, does global warming have a hand in the magnitude of the Hurricane Florence disaster on the U.S. East Coast? Yes, we can say with confidence, it contributes in several ways.“
Another report from Seth Borenstein: Hurricane rating system fails to account for deadly rains which makes some very good points.
Attribution methods
The key issue is how different scientists mentally frame the problem of attribution of the causes of extreme weather events. Here are 4 different frames that I’ve identified:
- It’s just weather (e.g. Roy Spencer). This is the null hypothesis, and looks to identify more extreme events in the past.
- Conditional approach (e.g. Kevin Trenberth, Jim Hansen). Examines basic thermodynamic impacts, and changes in circulation, focused on understanding AGW impacts on basic mechanisms.
- Climate model based attribution (pre-event attribution analysis, Fredericke Otto, Myles Allen). Compares climate simulations without anthropogenic forcing, with simulations including anthropogenic forcing.
- Detailed observational analysis that includes not just trends but also mechanistic covariances (e.g. Jim Kossin).
Re #1. This is indeed the null hypothesis, but we cannot rule out AGW effects based on this argument. The detection component of this approach is very useful in identifying previous extremes in historical and paleo records.
Re #2. The conditional approach is useful in pointing to #4 type observational analyses, but is insufficient on its own for attribution
Re #3. The climate models simply are not fit for this task. Such analyses implicitly assume the a) climate model sensitivity is correct; b) climate models correctly simulate internal variability; c) climate models correctly simulate extreme events; d) climate models correctly simulate thermodynamic feedbacks; e) climate models correctly simulate large-scale weather events and blocking patterns.
Re #4. This is the way to go, relatively few studies taking this approach. Results need to be carefully interpreted in terms of the impacts of natural modes of variability.
Intensity
The debate on hurricane intensity has been summarized in these previous blog posts:
- Hurricanes and global warming: 5 years post Katrina
- Hurricanes and global warming: 10 years post Katrina
- JC interview: hurricanes and global warming
The punchline is that while we would expect a signal of increased intensity from global warming, any signal is at present lost in the noise of natural variability.
The point I want to make here with regards to Florence (and this also relates to Harvey, Irma and Maria in 2016) is that these hurricanes (which went to Cat 4 or 5) are over performing their thermodynamic base. My basis for this conclusion is the operational hurricane forecasts provided by my company Climate Forecast Applications Network. We make intensity forecasts based on ensemble wind speeds predicted by global weather models, and also apply a statistical thermodynamics based intensity models to these ensembles. The statistical model assess the thermodynamic intensity potential. The high intensities and the rapid intensification was driven primarily by storm dynamics, and not thermodynamics that could be attributable to AGW.
And finally with regards to Florence, she was only a Cat 1 at landfall, after reaching Cat 4 (twice).
From Phil Klotzbach:
Also of interest in this regard. Part of what did in Florence’s intensification is that it was a victim of its own success. Look at the cold wake following the track, resulting from wind-driven mixing that brings cold water to the surface.
Size matters
Seth Borenstein’s second article (cited above) makes the important point that Cat 1 is woefully inadequate for describing Florence’s landfall impacts. Florence had a large horizontal extent once it started the eyewall replacement process. Large horizontal extent directly relates to landfall impacts: storm surge, large horizontal extent and duration of tropical storm level wind speeds, rainfall, and tornadoes.
5-10 years ago, there was interest in the financial/insurance sector in including size in measures of hurricane intensity (e.g. the Carvill Index). Also the metric Integrated Kinetic Energy (IKE) is a 2-D version of Accumulated Cyclone Energy (ACE)
Is hurricane size relate to global warming? I’m not aware of any studies that really address this, the most relevant one that I know of is the masters thesis by Angela Fritz (of Washington Post fame, yes she was my student) entitled North Atlantic Tropical Cyclones: A Kinetic Energy Perspective
Here are figures that Angela Fritz created:
R34 refers to the radius of tropical storm force winds
It would certainly be interesting to update this analysis for the N. Atlantic and extend to global. What I see is that the big jump in size and IKE occurred in 1995, with the shift to warm phase of the Atlantic Multidecadal Oscillation. Interpreting this as a trend from global warming would be misleading, although there is possible some component there that could be associated with global warming.
For another perspective on Florence’s size:
Florence’s size was much much smaller than Typhoon Manhkhut in the West Pacific, where tropical cyclones tend to have much larger horizontal extent than in the Atlantic. Understanding this difference is a prerequisite before attempting to attribute anything related to the horizontal size of Atlantic tropical cyclones to global warming.
Track and stalling
The track for Florence, with a North Carolina landfall, was unusual for storms on this initial path:
The reason for Florence’s unusual path relates the to presence of a massively large high pressure system across the North Atlantic and Northeastern U.S. The figure below shows the 1018 hPa isobar, with higher pressure to the north. Florence was unable to punch through this region of high pressure, and continued moving to the wast towards the Carolinas.
The horrendous rainfall associated with Florence was associated with its stalling once it reached the coast. Again, this is the presence of high pressure system to blame:
Some of the analyses of Florence cite a paper by Jim Kossin [link] that found that tropical-cyclone translation speed has decreased globally by 10 per cent over the period 1949–2016. This is most likely a robust result, although the interpretation of what has caused this remains open to debate. However, the effect described by Kossin is very different than the out-and-out stall that we saw for Florence and Harvey in 2016. These are blamed on blocking high pressure systems.
So are these blocking systems changing with global warming? Rahmstorf and Mann say yes [link], related to Arctic warming. This is a hotly debated issue, stay tuned for a more detailed analysis of this issue as I work on my new hurricane assessment. But if you consider multi-decadal to interannual internal variability, it is pretty difficult to argue for a global warming signal.
Storm surge
The maximum storm surge from Florence was 10-12 feet, although because of the complex coastal geography, surge was felt inland
Jeff Masters of Weather Underground has a good article on Florence’s storm surge [link]. The title of the article is Florence’s 1-in-100-year storm surge breaks all records.
The record for all-time highest water level at Wilmington, NC, where records extend back to 1935, fell.
“Note that sea level in Beaufort has risen by about 0.7 feet since the time of Hazel, largely due to human-caused climate change, and Florence would not have been able to break Hazel’s record without it.”
According to NOAA, sea level has risen at these locations by the following amounts:
- Wilmington: average rate of SLR of 2.3 mm/yr, or 9 inches per century.
- Beaufort: average rate of SLR of 3.04 mm/yr, or 12 inches per century
So, how much of this sea level rise can we blame on warming (human caused, or natural)? Karegar et al. (2016) provides a GPS-based analysis of vertical land motion on the U.S. Atlantic coast:
- Wilmington: -1.41 mm/year, for a value of absolute sea level rise of 0.9 mm/yr
- Beaufort: -1.13 mm/year, for a value of absolute sea level rise of 1.9 mm/yr
It is somewhat surprising to see such different values of absolute sea level rise for two locations that are relatively close, but this probably depends partly on how close the GPS station is to the actual tide gauge. In any event, the maximum amount of sea level rise that could be blamed on human caused global warming (assuming all warming in last century is human caused) is substantially less than the sea level rise measured at tide gauges.
According to Masters, that little boost of sea level rise made the difference in breaking the record Beaufort. However the amount of sea level rise that you can potentially blame on human caused global warming is substantially smaller than what is measured at tide gauges, owing to vertical land motion.
Rainfall
Florence set tropical rainfall records for North Carolina and South Carolina, there was also considerable flooding inland as far north as Massachusetts.
To what extent can we blame this substantial rainfall on human caused global warming. Well, the major cause of the large rainfall (same as for Hurricane Harvey) was the stalling and slow motion of the storm.
What of the claims that the warm sea surface temperatures in the Atlantic contributed to the heavy rainfall? Well the Atlantic surface temperatures weren’t particularly warm in the main development region of the North Atlantic where Florence spun up its intensity. The warm Atlantic temperatures were further north and didn’t particularly influence Florence’s intensity.
The basic thermodynamic argument is that warm surface temperatures result in a larger equilibrium vapor pressure (then wave hands) and therefore more rainfall in tropical cyclones. It is not implausible to imagine that warmer sea surface temperatures could contribute to increased rainfall in tropical cyclones (#2 style of attribution argument).
Well, for this argument to be convincing, the following analyses need to be done (a #4 approach):
- Look at the liquid water path and precipitation data from microwave satellite back to ~1979 for all Atlantic tropical cyclones (best to extend globally as well). I used to be very active in this research [link] I need to catch up on this research but I suspect that there has not been much done re tropical cyclone water climatology.
- Calculate the “characteristic life time” (CLT) of the condensed water (sum of liquid and ice) and water vapor may be determined by dividing their amounts by rainfall rate. Also the precipitation efficiency is the precipitation divided by the condensed water
- Relate above parameters on a storm by storm basis to the local sea surface temperatures.
- Assess the influence of SST on liquid water path, precipitation, characteristic life time and precipitation efficiency.
- If no significant relationship with SST, then give up. If significant relationship with SST, then assess the relationships in context of the Atlantic Multidecadal Oscillation, etc.
I’m not signing up to do these things, but I’m laying out what needs to be done.
Summary
I’ve scratched the surface of the complex issues surrounding the weather and climate dynamics of Florence, but the take home point is that convincingly attributing any of this to human caused global warming is very challenging, and the strategies used by the mainstream climate community to do this (#2, #3) are woefully inadequate and misleading to scientists, the public and policy makers.