In May 2016, a digger being towed on a trailer struck the Penrose overbridge on Auckland’s Southern Motorway. A few of the Qrious team were caught up in the ensuing traffic snarl. It got us thinking - how could we use data to understand the impact of crashes like this? And how could we use that data to create a predictive model, to help alleviate the worst, next time an accident happens?
The road to congestion
In 1955 the city of Auckland was in the throes of its first proper transport crisis. Despite one of the highest rates of public transport use anywhere in the world (a population of around 400,000 was taking almost 100 million public transport trips per annum, or around 250 trips for every single citizen), there was real concern over roads. And this concern was driven by one thing - congestion.
In 1922 there were around 37,000 private cars in New Zealand. By the late 1930s this had rocketed to over 200,000 (close to one car for every two people), giving us one of the highest rates of per capita private car ownership in the world. At the time our public transport was largely based on buses and trams, meaning many of those 100m trips per annum shared road capacity with all those cars. The pressure on the network grew steadily decade by decade, until by 1955 something had to give.
The Master Transportation Plan for 1955 was supposed to be the answer. Strongly influenced by trends out of the US, it made a series of recommendations that largely set the tone for Auckland’s traffic and transportation management strategies for decades to come.
The road's conquest of Auckland
A network of urban motorways would be built, corresponding to the main existing arterial routes into Auckland. These would be supported by a ring road, allowing for easy interchange between the urban motorways. In fact, right across the plan roads were the winners. Public transport would be boosted - but largely via bus. And though rail was included in the plan, it was as a future project - as the years dragged on initial recommendations were quietly shelved, and although rail would periodically appear in planning visions again over the next few years, it wasn’t until after 2000 that a concerted effort was made to deliver strong rail public transport.
Fast forward sixty years and the Master Transportation Plan looks a little like a wrong turn. Auckland’s roads perform reasonably well outside of peak hours, but the real problems come when there’s an incident on one of the interlinked motorways. And if the incident is close enough to impact peak hour traffic, that’s when the problems really start. Traffic soon clogs, epic delays and four hour trips home become reality, and the city itself grinds to a halt - with massive impact on the economy and citizen well-being.
What happens when it all goes wrong
The Southern Motorway’s Penrose overbridge was built in 1952, and has a clearance height of around 4.5m. That’s one of the lowest on the motorway, meaning strikes are relatively common. Amazingly, there were on average over five strikes a year between 2008 and late 2016.
But even in this context May 9th 2016 was one out of the box. The impact ripped the digger from the trailer and sent it sprawling across the motorway. Miraculously, cars following behind the trailer managed to avoid the wrecked digger, but the crash still closed two of the southbound lanes for well over three hours.
Some of the Qrious team were on the motorway at the time, enroute to visit a client. Coincidentally, a client who wanted us to use data in helping to improve traffic management… the opportunity was just too good. Crawling along at less than ten kilometres an hour we knew we had the perfect test case. Traditionally an understanding of the impact of a crash like this would have sat with traffic engineers - but what if there was a different way to analyse a crash’s impact? Could data help to enhance the view delivered by traditional traffic engineering methodologies?
What the data told us
We started by looking at Google data, accessing anonymised data from a range of their services, including Maps, Navigation, and Traffic. Because of the numbers using Google products and continually pinging data back to Google throughout their journey, these products create remarkably accurate travel times and create a great base from which to measure an incident’s impact.
Within minutes of the crash occurring we could see abnormal build ups of traffic southbound on the Southern Motorway. Detour routes were affected within fifteen minutes of the impact as cars were diverted or tried to find an escape themselves. Thirty minutes after impact the congestion had hit both sides of the motorway, and stretched for several kilometres either side of the crash site.
So far, so predictable. What we saw next was more intriguing. The congestion we noted at thirty minutes didn’t stay limited to motorway and detour routes. Rapidly it started to affect intersections at arterial routes feeding the motorway - by tracking Google’s travel time data we could see how the impact of the crash spread out, eventually overwhelming the road system far beyond the initial impact.
By 4PM, nearly three hours after the crash (and an hour after all lanes were re-opened), sites as far away as Western Motorway’s St Lukes on-ramp were still experiencing far heavier congestion than usual. The St Lukes motorway connection is over twelve kilometres from the scene of the crash.
By 6pm the gridlock was common knowledge. The motorway had been reopened for three hours and was flowing reasonably well. But that didn’t stop drivers from avoiding the expected delay by taking alternative routes. In suburban Orakei and Panmure drivers were now stuck in abnormally long queues as traffic swelled with people who would normally have been on the motorway.
In fact, the impact on traffic as far as 10-15 kilometres away from this single vehicle and non-fatal crash did not begin to noticeably subside until after 7pm. By this time tens of thousands of people had been affected, Auckland’s road transport system had ground to a near halt, and although never formally calculated the likely cost to the economy would have been in the millions.
Better traffic management through data
There’s another thing to consider when looking at all this disruption. The Penrose crash wasn’t particularly unusual. Sure, clipping a bridge is a pretty spectacular event. But in terms of the traffic management issues it created? They are reasonably regular events in Auckland. Five days after it opened, a car broke down in the brand new Waterview Tunnel. Cue delays, frustration, and social media disbelief. But we shouldn’t be surprised - instead we should look to data for solutions.
Over the last year the team at Qrious have focused on traffic management as one of our product offerings. With the ability to integrate data from multiple data sets, we are already playing an active role in minimising the impact of roadworks. Having seen the unexpected impacts from the Penrose crash, predictive understanding of traffic flows is our aim, diverting traffic in ways unlikely to create downstream impact.
That’s where our ability to integrate multiple datasets is proving a success. The data adds layers of understanding to basic travel times, allowing us to predict likely traffic painpoints. The ability to be proactive and predictive is absolutely vital in traffic management - at the next level this means connecting into messaging systems used to alert drivers (everything from motorway signs to news media), to see realtime impacts of routing decisions, and help make those messages more effective.
As imperfect a solution as the Master Transportation Plan may have been, it was the city of Auckland’s first stab at solving the problem. By adding datasets to pre-existing knowledge of traffic management we think we can revolutionise the approach. Less traffic congestion has massive benefits - from economic and environmental, all the way through to less obvious measurements, such as happiness. It’s the start of the journey for us here at Qrious, but we’re confident we can help find better ways for us all to get around.