Pinkpop is undoubtedly the best known Dutch pop festival. The weekend of Whitsunday, or Pinksteren in Dutch, tens of thousands (up to 60,000 per day) of music lovers gather in Landgraaf, Limburg, for three days of music and fun.
An event of this size takes a lot of planning that almost starts the day after the event itself. When planning in advance you can take care of a lot of things, but at least one thing is out of your control: the weather. Of course you know what you can expect, based on the seasonal boundaries. Below zero temperatures really rare in May/June (Whitsunday occurs 49 days after Easter so it fluctuates), however temperature can fluctuate 25 degrees easily. Only when Easter falls really early in March, Pinkpop is moved to another date.
The so-called ‘festival season’ runs from approximately May to September when the weather is good enough for outdoor activities.
With the climate becoming more extreme and with events like Pukkelpop in 2011 in mind (a storm killed 5 people with dozen more in hospital), more attention is given to the emergency and contingency planning in case of (natural) disasters.
Pinkpop 2014 started off great with famous bands like the Rolling Stones headlining the first day. On Monday, the last day of the festival severe weather occurred in the evening just before Metallica was scheduled to go on stage. Over 1600 strikes of lightning took place in a ten-minute timespan near the venue at Landgraaf combined with torrential rainfall.
A lose-lose situation
What should and can the organization do? It looks like they can never win. If they cancel the show and the weather is not as bad as feared people are mad. If the show goes on and the weather is indeed as feared and people die or get hurt people are mad. In any case they need to do something, do nothing is not an option.
There is another dimension to this problem and that is the smartphone using festivalgoer.
Can Science help?
Science cannot prevent violent storms or bad weather that is a fact. We, as humans, do have an influence on the weather in a general sense like the use of fossil fuels, deforestation and so on. This will, in the end, influence the weather overall. Recently, the Royal Netherlands Meteorological Institute (KNMI) issued 14 new climate scenarios in which extreme rain and thunderstorms will become more prevalent.
In order to develop these climatological scenario’s a lot of calculations need to be done. This is the point science steps in. So what kind of calculations are we talking about?
Let’s say that due to global climate change the ice in Greenland melts and a lot of fresh water enters the Atlantic Ocean. This changes circulation in the water and some place the water will rise and at other places it will fall. But where will the water rise and where fall. In order to calculate that you need to look at the ocean and for instance divide it in a 10 x 10 km grid. For each grid you need the temperature, amount of salt, circulation speed etc. and calculate how much water is transported to the next grid, which also has its own temperature and so on. It is one giant interlinked system of cause and effect that you need to calculate for a period of a minimum of 100 years to have the beginning of a believable model.
When people talk about big data, this is really it! The example above is highly simplified, by the way.
So you just needs some CPU’s and calculate? Wrong, you don’t need so much CPU’s but Graphical Processing Units (GPU’s). These processors, used for advanced computer games in PC’s and game consoles, have 1000’s of cores that can perform simple repetitive calculations as opposed to only 8 cores in a CPU (that are more able to do complex tasks and calculation). They can process up to 200GB of data per second in this way.
In the Dutch research program COMMIT/ scientists are working on the infrastructure that can handle the data and a methodology to optimally deploy clusters of GPU’s to do the calculations. This will optimize the process and reduce the amount of energy and time needed, which is also better for the environment.
Sensing the mood
The advent of smartphones and twitter has created a platform where people express themselves: what they do; what they like and how they feel (to mention but a few communicative functions). The scale and accessibility of these communications is unprecedented; the messages and moods of many people can be monitored at the same time.
The research from COMMIT/ helps identify the mood or emotions that people have when at for instance an event like Pinkpop by analyzing those tweets that concern Pinkpop. There are a number of challenges with that:
- Try and identify those tweets that actually concern Pinkpop;
- Recognize the mood or emotion from the tweet;
- Use only those of people who are actually there if the goal is to monitor an event happening at the event itself.
For challenge one and two, hashtags can play an important role. It is much easier to scan for #PP2014 or #Pinkpop2014 than to try and understand or infer it from the tweet itself. Based on the combination of location (e.g. on stage), a name of a band (e.g. Rolling Stones) and time (using the date timestamp of the tweet) you could say that this about Pinkpop and perhaps from a spectator. Being able to do that requires a lot of knowledge for a system, semantic knowledge, knowledge about the line-up of Pinkpop and the ability to combine it all. Our language is very flexible; the same event can be, and is described in countless different ways.
Identifying emotions is even more complex, but when people use hashtags like #zinin or #geenzinin this problem is again made relatively easy; it is possible to train automatic ‘spam filters’ that recognize tweets that would typically carry these hashtags even when they do not. Another issue to take into account is that when people are afraid or in a stressful situation in which they have to pay attention to what is going on, the use of their smartphone may not be their first priority. Our brain in programmed to be vigilant in those situation. The hypothesis of the researchers is that a drop in the number of tweets together with a potential dangerous situation might actually indicate a sense of danger.
Discontent is easier to identify, using the right hashtags like #fail or #epicfail.
And there is of course the question of actual presence: is the person really at Pinkpop? Luckily, about one per cent of tweets has a GPS location (voluntarily given by the user as metadata) so researchers are able to identify those tweets coming from people at the venue accurately.
What they found was that there was not really a sense of emergency based on the analysis of tweets during the severe weather on Sunday evening. Jokes where being made about Metallica’s hit Ride the Lightning and other jokes indicating a lighter mood, directly after the skies cleared.
Until now the research has focused on listening in on tweets. In the future scientists might study the possibility to send out tweets by for instance a festival organizer to reassure people or otherwise try and influence the crowd or mood. The problem with that is that the reaction to that sort of tweets can vary between people.
To evacuate or not to evacuate, that’s the question
There was a lot of discussion on social media if they should evacuate or not. But can you evacuate a crowd at any time and any situation?
The answer is of course NO, the possibility of an evacuation depends on many things. Let’s assume that the venue can handle the number of people that will attend, so there is enough room, exits and so on.
What you need when you want to monitor and control a crowd is knowing where they are and what they are doing. The latter is not so much about what they are looking at or listening to, but if they are moving, standing still etc.
Smartphones are a good way to measure that sort of data, but require an app to do so. It is possible to combine that app with information about the festival like who is performing, where and when, but for Pinkpop a wearable device in the form of a bracelet would probably be a better solution. The bracelet could combine entry, payment, loyalty and other functions in combination with crowd insight.
The prototype that the COMMIT/ project EWIDS envisions is a small bracelet that has a limited capability to transmit data (both in the amount as well as in the distance it can cover). In order to see how many people are present at a certain place it is not necessary that every node sees all other nodes. It suffices that one node shares its information with a few direct neighbours (so called gossip mode). With algorithms under development the scientists are able to estimate how many neighbours are present in total, where they are and how closely together they are standing, and even a path (gradient) to the nearest, available exit.
The contribution of science to festivals like Pinkpop can be significant. In this article we have described three contributions that will help:
- Prepare (for possible bad weather)
- Be aware (of the mood of festivalgoers)
- Get them out of there
Science cannot control the weather but can for instance help define climate scenarios that should be kept in mind when organizing such a festival. Although such scenarios do not describe the actual weather they describe the fluctuations and frequency of such weather. Forewarned is forearmed. Another issue that science can help identify is the mood of a crowd. Twitter is often used for expression of mood and to report what is happening. By identifying the general mood organizers can sense what is happening in the crowd.
Lastly, science can help control the crowds. With knowledge of the whereabouts of people they can proactively try and disperse people before the situation gets possibly out of hand. When there is a real emergency they can steer crowds in different safe directions hopefully preventing chaos.
Who knew that organizing a festival like Pinkpop could use so much science?
Written by Rob Blaauboer, Innovation Manager & Blogger & Gadgeteer
This article was written based on three interviews:
- Antal van den Bosch (Radboud University Nijmegen) on his work with Ph.D. students Florian Kunneman and Ali Hürriyetoglu, of the work package ADNEXT, part of the COMMIT/project INFINITI.
- Ben van Werkhoven (VU University) on his work as work package leader for work package SALSY from the COMMIT/project IV-E.
- Koen Langendoen (TU Delft) on his work as projectleader for work package Adaptive Data Dissemination from the COMMIT/project EWIDS.