14th August 2019
Getting the weather right
The INEOS 1:59 Challenge – the work of the Performance Team
The setting of any new record in athletics needs a number of things to come together; the right people with the right mindset, all in peak condition, at the right venue and with the right weather.
Athletes in the majority of disciplines will have many opportunities for the stars to align. This is not the case with the marathon, where the toll on elite competitors is so great that most only compete twice a year. So, the chances of any one of those events producing perfect conditions for any historical barrier are slim.
It’s this fact that drives much of the technical work behind the INEOS 1:59 Challenge. This is a focused attempt to achieve a true first in human history — a sub-two-hour time for the marathon distance of 26.2miles.
Eliud Kipchoge is not in a position to wait for everything to come together again at another big race (he already holds the world record) — this is his moment. Eliud is in the form of his life, winner of 11 of the 12 marathons he has competed in. He wants to do this now, and at a time that doesn’t disrupt his Olympic preparation ahead of Tokyo 2020.
“Back in 2017 — as part of the Nike Breaking2 Project — I tried my best and ran 2:00:25. I did not expect a second opportunity to come about this year, because such events take a long time to organise, but I am very grateful INEOS have put on this challenge. I believe it has come at the right time. This is a golden chance for me to make history and show the world that no human is limited,” explained Eliud in one of his diary entries.
This is what the INEOS 1:59 Challenge is about; to make it a golden opportunity for Eliud to fulfil his dream. The supporting Performance Team cannot leave anything to chance.
They have to control as many of the contributory factors as possible, encouraging them to work in Eliud’s favour without reducing the credibility of the run. This will ensure that it could one day be — as Performance Manager Peter Vint put it — ‘replicated in the wild’. The moment when serendipity does eventually bring everything together for... someone.
In this series of articles, we look at the work of the INEOS 1:59 Challenge Performance Team as it tries to make the stars align.
"This is a golden chance for me to make history and show the world that no human is limited"
Getting the weather right
One of the biggest factors in running a fast marathon is the weather — temperature, humidity, precipitation, wind speed and direction will all have a significant impact, even the air pressure has some effect — although running through slightly denser, low pressure air is the least of the worries.
The optimums and the consequences of straying outside good conditions are well understood for the runner. Temperature is perhaps the most important; “In long distance running events the ability of bodies to maintain or regulate heat is what ultimately determines how long athletes can last at a given intensity, along with their internal energy state and their fitness,” explained Peter Vint.
The temperature interacts with the humidity, which makes it harder to sweat and makes it harder for the athlete to stay cool. While any rain at all will add water and weight to clothing and footwear, draining energy faster; and running into the wind, or even being buffeted by variable wind adds drag and slows the runner. But it’s one thing to understand the impact of the weather on an athlete, another altogether to ensure that the run takes place in the correct environment.
Ensuring that Eliud Kipchoge’s sub-two-hour marathon attempt takes place in the best possible weather conditions is one of the most challenging parts of the whole endeavour. The weather team must work on three different forecasts, each relevant at different times.
The first will decide when Eliud should travel to Vienna. This decision will require the team to identify a period of the right weather in advance of the event. The second will pick the day for the event and will determine when Eliud enters his final tapering phase. And the final decision will be to identify the precise two-hour window for optimal conditions on the day. These decisions will all be right at the limits of forecasting technology.
Robby Ketchell is the man in the hot seat — a spot he’s been in before when he worked on Eliud’s previous attempt to go sub-two (Breaking2). He will also be familiar to cycling fans as the data scientist behind three of Team Sky – now Team INEOS’s - Tour de France wins. Robby’s first job was a data analytics task; to pick the needle out of the haystack and find a venue that would provide — amongst other things — the best chance of getting the right weather.
“We did an iterative, recursive search for where all of the parameters would be optimal individually. So looking not just for the temperature, the humidity, the pressure, the wind, elevation and precipitation. When you do that, you end up finding some general areas that match where your parameters would likely occur during the time frame [that the Challenge was set for].”
The initial search went back 15 years and was set on a grid of 100km, and then if it found somewhere promising the grid was refined to 20km. “It [the software] was all custom built to search for this, we wrote a bespoke tool that went out and searched the weather data from stations all around the world.
“We were looking between October 10th to October 20th originally and over those ten days you get these areas where the parameters are good. So when you combine them all, you end up with a very limited number of locations that would be optimal. Now you have those locations, that doesn't necessarily mean there's a venue there. So you have to go and actually search to see if there's an actual venue.”
Vienna was eventually chosen as the venue for the challenge because it offered the best chance of getting the right environmental conditions — along with other parameters like time zone and altitude. There were many good reasons for Vienna being chosen, but Robby highlights one that was particularly important for his task of trying to get the right conditions for Eliud’s run.
The INEOS 1:59 Challenge course in Vienna, Austria.
“The main thing around Vienna... was the fact that it had coverage from the trees on the left and the right side.” The reason that this was important was that it reduces the unpredictability of one of the most unpredictable of the six parameters that Robby is working to optimise — the wind.
“If you have coverage [in this case, the trees] on both sides of the course, there may be a certain wind speed that is high, but if it's coming from a certain direction it doesn't necessarily matter because down on the road surface you have less wind.”
The wooded nature of the Vienna course would allow them to run the Challenge on a greater number of days — but only if they could gain a complete understanding of the relationship between the wind up high (above the tree canopy) and the wind that Eliud would face on the road.
“We want to know to a very high precision what the wind is going to be on the course. The weather stations that we're using through the local Met office are located a pretty far distance away from the course and some of them are located in the hills. Others are located on the other side of the city centre or on the other side of the river, the Danube.
“Because of that, you don't have an accurate indication of what the free flow is right above the course. So we are mounting a reference station that is going to sit about three meters above the trees so that we will be able to infer what the conditions are from the [other local] weather stations — or the Met office — to our reference station. Then we can better align what the historical data means for our specific location.
“In order to downscale it to what the conditions are on the course, we then have wind sensors that are mounted at the street-light level that will be permanently fixed there from now until the record attempt.”
These sensors are mounted below the trees and above the centre of the road, but because they must be high enough [4.3m] for local traffic to pass underneath they still tell a different story to the actual conditions that Eliud will face on the road. To solve this problem, the team have also run tests with wind sensors on tripods at head height, to collect data that will allow them to relate the 4.3m measurements to conditions on the road. All of the equipment has had to be built, shipped to Vienna and installed by the team.
The goal of all this effort is simple. “We're basically trying to downscale it so that if we're getting a wind measurement from a weather station that is 20kms away, or 15kms away we know what that means on the course,” explained Robby. In this way, the team can relate all of their historical data from weather stations around Vienna to conditions on the road.
It will allow them a much greater understanding of what macro-scale weather conditions will be acceptable on the race track — and that brings us to the forecasting, and those three big calls.
“Here's the difficult piece; making a good decision that’s within a couple of days from far enough out for Eliud to travel. That's really difficult to do,” commented Robby.
“The second piece is that we need to then know which day is our day because Eliud needs to start modifying his nutrition. He can't extend it very much after he starts up because he will start to put on weight from having more fuel... We may be able to modify the time within a few hours, but we need to know the day.”
And the third call, the decision when to go on the day? “When we did Breaking2 I was responsible for the weather and it rained... everybody was a little uneasy and then the rain went away, roads dried and we got optimal conditions. But that's part of the variability. Generally, you're not going to have a full two-hour window, but you need to find the majority of that two hours and you need to understand that outside of that you're going to have some pretty significant variation. So yeah, it's never perfect, but you try to do the best [you can].”
Fortunately, Robby Ketchell won’t be handling all the heat on his own, he has support from Walter Zwieflhofer, a former ECMWF Director of Operations and a meteorologist with the INEOS TEAM UK sailing team.
Walter explained their approach to each of the three calls. “The last one [the start time] is obvious, because you have these very high-resolution models of Vienna already running. The Austrian Met service (Zentralanstalt für Meteorologie und Geodynamik or ZAMG) is already running a model over the whole area.”
ALARO has been developed for over 20 years by a group of 16 predominantly European national meteorological services, it forecasts up to 72 hours ahead and provides hourly resolution using initial data from the ECMWF as well as data from ZAMG's automatic station network. It’s this model that will be used — with help from local forecasters — to pick the start time.
“There's a human element to it... of forecasters and observers that work in weather analysis locally. We are seeking their involvement in the project, so somebody that understands when fog is coming, for example, just because they've seen it before in this area... it goes a long way beyond just the data itself,” explained Robby Ketchell.
Walter Zwieflhofer continued, “The decision to pick a day is a typical medium range weather application, medium range forecast. This comes from a global model, because you have to know how the pressure systems move and the big picture.
“So the model is actually forming that system, before it exists, moving it into the right place, and you are only out a couple of hours, that is amazing, yes? But for that particular application it's not amazing, because people...” Walter tailed off.
“They want better,” I added, unhelpfully.
“It's about preparing for it, and understanding the requirements best, so that one can get out of the statistical forecast the best guidance that's possible,” Walter concluded.
That leaves the decision for Eliud to get on a plane to Vienna. “This far in advance you don’t know whether the rain front comes through at 2:00 in the morning, or 7:00 in the morning...” explained Walter, “which will make all the difference to it even being possible. So the work involves a historical analysis, to see how we can best employ the state of the art for this particular application.”
It seems that — despite all the work, and even with the very best forecasting technology -- the team will still need a little help from mother nature for the stars to fully align on the day.