The American Meteorological Society (AMS) is the flagship organization for most atmospheric and climate scientists around the world, but especially in the U.S. Every other year, for as long as I can remember, AMS has organized a Conference on Tropical Meteorology and Hurricanes. The 33rd one was held recently in Ponte Vedra, Florida, and no—I don’t remember the first!
Several of us from the Meteorology Team at AIR took the week-long opportunity in April to present some of our latest research, hear about the great work others have been doing, and escape the never-ending Boston winter (it snowed back home while we were in Ponte Vedra). There were hundreds of talks and hundreds more posters at the conference, which made it difficult to catch them all. We took as many opportunities as we could to chat with everyone about the presentations we attended and those we could not get to. In this blog we’ll outline some of the issues discussed.
Machine learning was brought up in several talks. It was discussed as an approach to improve genesis prediction (sorting through tens of thousands of predictors) and as a methodology within numerical models to predict the rapid intensification of tropical cyclones and second eyewall formations. One talk by Keqin Wu, a scientist at the National Oceanic and Atmospheric Administration (NOAA), described how machine learning algorithms can test different processes that lead to the formation of a secondary eyewall. This methodology will help achieve a consensus on the causes that lead to the formation of a secondary eyewall, a process that is not fully understood.
In an update to an ongoing project, Chris Hennon talked about crowdsourcing as an approach for determining tropical cyclone intensities through straightforward visual inspection of satellite images by humans—lots of them—leading to more accurate results. You can participate too by visiting the Cyclone Center to inspect past tropical cyclones and current storms in real time.
Although it is not AIR’s mission to forecast hurricanes or other weather phenomena, we did take great interest in attending talks that addressed various aspects of forecast dissemination. With numerous graphical and data products available to clients, it’s helpful for us to understand potential differences between how information is intended to be interpreted, and how it is interpreted. Some of the confusion is due to evolution from deterministic to probabilistic analysis and forecast products, with some criticism directed toward the National Hurricane Center itself. Two examples were in talks by Derrek Ortt (StormGeo) and Josh Wurman (Center for Severe Weather Research).
Derrek Ortt’s talk was a critique of the forecast track error (the “cone” of uncertainty); he showed cases of the cone being a very poor representation of true forecast uncertainty. This was especially evident in bifurcating cases when predictability is lost. He suggested multi-model ensembles are a much better representation of short-term (one- to five-day) uncertainty.
Josh Wurman presented mobile Doppler radar data from Harvey and Irma, but the fireworks were set off in the final 10 minutes when he was highly critical of the hurricane intensity estimate (Vmax, 1-minute maximum sustained wind) issued by the National Hurricane Center. According to Wurman, this single number is “fake news” because 1) it in no way represents actual wind hazard and 2) it becomes manipulated by media outlets that in turn misinform the public. Many generally agreed with his point, as we tend to be heavily focused on wind footprints matching this value near the landfall location.
Not to be overshadowed, storm surge talks were a noteworthy part of the conference. While many focused on improvements to real-time forecasting, others such as the one from Kathryn Fossel discussed sensitivities to hurricane characteristics, and showed that storm size is second only to storm track in terms of storm surge importance. Importantly, the talks and posters from others at the conference supported the notion that AIR is on the technological cutting-edge of storm surge modeling.
Too Much of a Good Thing
Conferences like this are all about the exchange of ideas, and AIR made a strong contribution to the conference by delivering six presentations outlining a range of new research that you’ll hear about in coming blogs. We gave as good as we got; we found the talks we attended truly inspirational. Jason Dunion, for example, highlighted the importance of the diurnal cycle for tropical cyclone intensity and precipitation and gave us some ideas for even more improvements to our next-generation hurricane wind field and precipitation and flood models! The Bill Gray Symposium and talks by Dr. Phil Klotzbach, Jim Kossin, and Dr. Greg Tripoli that we found particularly interesting have been summarized briefly here.
There were far too many presentations, however, for us to do even half of them justice in a single blog.
The good news is that most of the conference presentations were recorded and are on line for anyone to see and hear. Thanks to AMS you can catch details of the talks noted in this blog, others we could not mention, and all of those from AIR.
Editor’s note: Eric Uhlhorn, Anna Treviño, Michal Clavner, Suz Tolwinski-Ward, Rich Yablonsky, and Sylvie Lorsolo contributed to this blog.