Final commentary on Joaquin, models and operations
It was definitely a busy week as we were all monitoring Hurricane Joaquin and its threats to the US east coast. Fortunately, it has turned out to sea and also appears to be tracking well west of Bermuda which may spare them a direct hit. It rapidly intensified once again yesterday to just under Category 5 intensity, but has since weakened to a Category 2 storm. It should continue northeast and lose its tropical characteristics early this week well to the southeast of Newfoundland.
We must not forget the fact that, partially due to Joaquin and also due to a stalled front, devastating flooding is currently ongoing in South Carolina. Some areas have received 15 to 25 inches of rain with more forecast. That was well forecast and remarkably accurate, and many houses have been inundated, bridges washed away and people have had to be rescued. It remains to be seen what will happen there in the next 24 to 48 hours as the much larger rivers begin to rise. Catastrophic main stem flooding is certainly possible. This will almost certainly be that state’s worst natural disaster since Hurricane Hugo in 1989.
Recommend Twitter connections:
NWS has reported 22.47" of rain near Charleston, SC over the past 60 hours. Map shows rainfall over the past 7 days. pic.twitter.com/q9kSWE9hoO
— AerisWeather (@AerisWeather) October 4, 2015
The “model wars”
I was quite confident, but not over-confident, most of the week thinking that Joaquin would make landfall, most likely in the Southeast or Mid-Atlantic coast. From Monday to Wednesday, it was supported by nearly even available model – except for the ECMWF. That has reopened the so-called “model wars”, also understanding of the fact that the European model successfully forecast Hurricane Sandy as far out as 8-9 days before landfall, which the GFS and most other models had the storm going out to sea until about 3-4 days before landfall when additional data was available from aircraft dropsondes, special weather balloon launches and short-term trends. That was a moment that realized the pan-European group running ECMWF were the kings of medium-term weather forecasting, while the American groups (along with the Canadian CMC, the British UKMET and the Japanese JMA, among others) were falling behind.
It is well known that the European model has an excellent reliability record, but has also had several critical misses over time. For example, the ECMWF predicted that New York City would experience a major blizzard last January, and streets and subways were cleared as a result for one night. However, the blizzard did not materialize for them (it did for eastern Long Island and most of New England) and many people became skeptical of meteorologists as a result. In addition, the GFS was significantly upgraded in early 2015 and incorporated much more data than the previous version.
It was believed by most meteorologists (and myself, with caveats) that the upgraded GFS would win out when it was joined by many of the parallel runs and the less accurate models (GFDL, HWRF, CMC, UKMET, JMA, NAVGEM, etc.). During that time, only a small number of non-ECMWF ensembles predicted an out to sea track. At one point on Wednesday, about a third of the ECMWF ensembles joined in with the other models predicting landfall, while the majority continued to stick to the operational model. That was the reasoning that I went with an 85% chance of landfall, most likely in the Southeast or Mid-Atlantic (in retrospect, the probability might have been a bit high, but certainly the odds favored such). As we know now, that is not what happened. I kept open the possibility that it would stay out to sea, knowing the ECMWF has succeeded before against all odds.
Remember that these computer models are highly advanced. They incorporate many functions, such as derivatives, integrals, matrices and other challenging statistical analyses which are usually learned in advanced level college math courses. Every wind pattern, every temperature, every dew point, moisture levels and energy levels must be incorporated in every model run, typically with all that data received every 6 or 12 hours. The NOAA Gulfstream-IV missions were extremely valuable in studying the data over the open ocean where there is very little in the way of weather observations (only islands, ship data and buoys).
That is why calculus is extremely important when studying meteorology, which I made note of a few weeks ago. It is not as simple as just looking at maps and studying the lines and wind patterns. Most of these functions are extremely complicated and the average person does not understand them. Even some meteorologists have trouble with the calculations – it is normal. I know I am not ready to do them yet!! If it were that simple, the trough would make it look obvious that Joaquin would track due north and make landfall, likely in the Mid-Atlantic or Northeast.
It has become clear that the European model knew something the rest of us didn’t. That could be why it is very expensive for companies to acquire most of the data from the ECMWF and they generally only sell the data at very high cost to the public, or intermittent at best. Meanwhile, the GFS, GFDL, CMC, UKMET, HWRF and NAVGEM models provide most of the data to the public online at little or no cost in comparison. Running these computer models 24 hours a day, 7 days a week is an extremely expensive proposition, and in order to provide the best data possible, it requires even more investment. There needs to be a return on investment as well.
The National Hurricane Center: A job very well done
Some may criticize the National Hurricane Center for a missed track, forecasting that Joaquin would come ashore in the Mid-Atlantic or Northeast when it really didn’t. However, they really didn’t have a choice. They did many things right in their communication and operations during Joaquin, as I will explain below.
- The forecast was an extremely difficult one with two very divergent solutions. The majority of solutions supported landfall farther south and west of the forecast locations (which were around the mouth of Chesapeake Bay). However, the NHC had to respect the fact that the ECMWF called for an out to sea track and that it was a very reliable model.
- There were additional balloon launches every 6 hours instead of every 12 hours by many NWS offices from Tuesday to late Friday. The additional data would then be placed into the computer models in an attempt to obtain a consistent and accurate forecast.
- The NOAA Gulfstream-IV aircraft flew many missions around the western Atlantic and Gulf of Mexico, assessing the synoptic environment of areas that have little or no synoptic observations. The missions increased later in the week and, come Wednesday, the additional data was helpful in slowly turning other models towards the ECMWF solution.
- It can be very costly to launch additional balloons, obtain additional observations and fly supplemental reconnaissance and synoptic aircraft. It likely cost many millions of dollars from the NOAA budget alone.
- However, had that data not been available at the time that it was and the models were still split on Thursday and especially on Friday, major preparations would have had to begin along the east coast (far beyond what was necessary). Major evacuations would have likely been required for hundreds of thousands, if not millions, of people on Friday and Saturday, when a Hurricane Watch would have been issued for many areas. That alone costs hundreds of millions of dollars in preparation and operational costs, not to mention the cost of lost business and commerce which could be in the billions. That is a return of investment of at least 25:1, if not 100:1 or more. I would consider it an excellent investment. Even 10 or 15 years ago, the data would likely not have been available and such preparations would have surely been required.
- Had the evacuations and other preparations been undertaken and Joaquin moved back out to sea (as it did), the general public would have likely been angry as a result, and complacence would have likely set in. That was somewhat of a factor in Hurricane Sandy in 2012 after some coastal areas were largely spared during Hurricane Irene in 2011 (although others were battered in both storms) and is a common problem in areas especially prone to hurricanes. There would be a risk that the next time a major hurricane was threatening the coastline, many people would decide that it was another case of “crying wolf” and not prepare properly. That could easily have catastrophic consequences with an enormous – and preventable – loss of life and property.
- As for the NHC operations, they also added, for the first time, a series of “key messages” to the forecast discussions from Wednesday morning to Friday afternoon. They explained, in a much easier tone for the public to understand, the uncertainty that existed, mentioning that the threat of a major landfalling hurricane did exist but that Joaquin could easily go out to sea as well. They were also very useful in explaining the non-tropical systems which are currently resulting in the Southeast flooding, as well as the reasoning for the difficult forecast in a less technical language. Those were extremely useful and are highly valuable for future storms – as well as for other agencies like the Storm Prediction Center and local offices to use on challenging days.
Hurricane forecasting will always be a work in process. I was as unsure as everyone else all week, and had to mention every possibility that was considered plausible. In storms earlier in the season, I used multi-cone graphics to try to explain the possibilities. However, those were not well received by others and criticized so I avoided using them this time. The probabilities for any single possibility were also kept fairly low (generally no higher than 30%) in order to show the wide variety of solutions. Confidence was not high, so I couldn’t settle on anything or make a single forecast until late Friday, after I published the previous blog post. It is important to remember that forecasts often have a large error range, and that was especially true with Joaquin.
Weather forecast continues to improve, but as we have learned, we still have a long way to go. Regardless, it was a devastating impact for the central Bahamas with catastrophic damage and several fatalities reported in that region. In addition, 33 people are missing right now after a US-based ship El Faro went missing during the storm. We are all praying for good news in the search. Although direct impacts into intense hurricanes from ships are very rare, the unusual track into the Bahamas from the northeast likely caught the ship off guard. My thoughts and prayers are with everyone affected by both Joaquin directly and the Southeast flooding, and also the families of those missing.
Forecaster Craig Ceecee – @Ceecee_Wx