With the winds aloft layer being added to the map, I think upper air data has been on everyone’s mind lately. But where does this data actually come from? Surface data is (relatively) easy; there are ground stations all around the world which record meteorological conditions both for local reporting and to be fed into the super computers that power global models like GFS, ECMWF, and others. Even over the ocean, there are hundreds, if not thousands, of buoys, boats, and island-based weather stations reporting in. While more data would always be useful, and aviation does play a major part in this network with many stations colocated at airports, I don’t think this network requires much explanation. When you start to look high up in the atmosphere, though, that is when things get really interesting.
The first question most of you probably have is, can’t we just multiply or through some other mathematical means, determine wind speeds at altitude from ground-based wind speeds? Unfortunately, it isn’t quite that simple. This could be a whole explainer on its own, but it has to do with friction slowing wind near the surface down more than it is aloft. This, through the Coriolis force, also will have an impact on direction. In the most extreme cases, the speed and direction at the surface can be completely opposite that aloft, for example, a tropical cyclone with cyclonic inflow at the surface and anti-cyclonic outflow aloft.
So, we need upper air data; how do we get that? The oldest method, and still really the only one meteorologists control, is radiosonde balloon launches, weather balloons. These balloons are launched by around 800 sites around the world and are launched at 00z and 12z worldwide. This is called the Global Upper-Air Observation Network (GUAN). The uniform timing is important so that there is a complete set of data to feed into the next model run, though not all sites launch at these times. This is a big part of why you will find just about every model has a 00z and 12z run time. Some also have 06z and 18z (or other intermediate times) run times, but these are known as interpolated model runs because there is not as much new data.
Map of ballon launch sites:
Source: WMO
So 800 sites around the world twice a day is a lot, but really not that much on the scale of the world. Firstly, there are 12-hour gaps with no data; a lot can change in that time. Secondly, there is the issue of geographic coverage. 800 sites gives us decent coverage over most major landmasses, but oceans? There are a couple of island-based sites, but you can easily go thousands of miles without hitting a site that participates in GUAN. While weather over oceans may be less pressing to our daily lives, the weather over the Pacific is tomorrow’s weather over the US, and the weather over the Atlantic is tomorrow’s weather over Europe. Not to mention in our case, aircraft flying over these regions.
Alright, so there are holes in our data, how do we fill those in? More balloons? That cost adds up pretty quickly. Where else could we get upper air data? Well, as you might have guessed, the answer is aircraft. Specifically, commercial aircraft. If you were thinking of dedicated weather reconnaissance aircraft, those do exist, probably most famously NOAA and the USAF’s Hurricane Hunters. These dedicated aircraft are very few, and so even if pushed to their absolute limit of coverage, their coverage would be practically insignificant for global forecasting. This is why they are used in specific cases where more data is needed, for example, improving the accuracy of hurricane forecasting.
Today, the World Meteorological Organization (WMO), which is a component of the United Nations (UN), runs the Aircraft Meterological Data Relay (AMDAR) observational system. As of 2019, 43 airlines around the world participated in AMDAR, and airlines from every continent (minus Antarctica) and most major regions are represented.
Participating airlines as of 2019
Aerolineas Argentinas, Dragon Airlines, Lufthansa Passager, Aer Lingus, easyJet, NAV Canada, Aeromexico, Eurowings Europe, Northwest Airlines, Air France, Eurowings Germany, Qantas Airways, Air New Zealand, Federal Express, Scandinavian Airlines, Air Nippon Airways, Finnair, Shandong Airlines, Alaska Airlines, Germanwings, SkyTraders, American Airlines, Hawaiian Airlines, South African Airways, Asiana Airlines, Japan Airlines, Southwest Airlines, Austrian Airways, Jazz Aviation LP, Thomas Cook Group, British Airways, JetConnect (Qantas), United Airlines, Cathay Pacific, Korean Air, United Parcel Service (UPS), China Eastern Airlines LATAM Airlines, Xiamen Airlines, China Southern Airlines, Lufthansa Cargo, Delta Air Lines, Lufthansa CityLine
(And yes, I did notice that their list includes, for example, Northwest Airlines which merged into Delta in 2010. Their website says this is the list as of 2019, but evidently isn’t entirely correct.)
AMDAR data includes temperature, pressure altitude, wind speed and direction, turbulence (EDR), and more. This data is transmitted from the aircraft through various methods, either to ground stations or to satellites, which communicate with ground stations. Most typically, this is transmitted using existing ACARS systems. Furthermore, some aircraft have additional sensors necessary to transmit humidity data, another important meteorological data point.
This is a huge deal for global weather forecasting. Firstly, observations over water and at high altitudes are increased dramatically. Secondly, this creates hundreds of temperature profiles at these airports nearly continuously throughout the day. Temperature profiles, typically presented as a skew-t for those of you who have heard of these, are a core piece of local forecasting. While the amount of data does fall off at night, this is where the participation of multiple air freight companies becomes even more important.
The WMO themselves says this contributes to a 15-20% increase in computational weather modeling accuracy and lists AMDR as the 4th largest contributor to global forecasting accuracy out of all of their global observation programs.
Source: WMO
While there are many, many more contributors, and they are all important, the fact that a program which uses data that is already collected already by aircraft that are already flying, meaning it has shockingly few direct costs (there are, of course, some costs associated with data collection, quality control, processing, etc.), can be as important, or more important than satellite programs costing tens of billions of dollars is amazing.
Aviation and meteorology have always gone hand in hand. Even the location of the first flight was chosen for winds. Every airport’s design is dictated by prevailing winds. Every flight is bound by the weather of the day. Much of the meteorological observation systems we have today are directly in support of aviation operations, or at least used for that if it isn’t a primary mission. I think AMDAR is a super cool way that aviation gives back to the meteorology world, so to speak.
So if you have ever been looking at upper air data for your flight and wondered where they get this data from, the answer may well be another aircraft. Pretty neat if you ask me!
