Any meteorologist will tell you that predicting the weather is far from an exact science. But apparently we’re not all that good at working out what’s it been like either. Obviously we know that yesterday was sunny, but how many of us – for example – think about local weather trends in five or ten year cycles?
In South Africa we thankfully very rarely see a hurricane, but we do have rain – and lots of it. We are blessed with a steady supply of around 464mm each year, and while it is significantly below the world average of 860mm per year, it still has the power to disrupt industrial operations. In its financial results presentation for the first quarter of this year, ArcelorMittal highlighted that its earnings decreased by almost 40% when compared to the previous quarter due to “lower production and higher energy costs caused by the severe weather disruption” thanks to heavy rains at the end of the summer.
Extreme flooding in Queensland, Australia’s mining sector in 2012 was blamed on great changes in precipitation. According to a report by the International Council on Mining and Minerals, “the floods severely limited access to mines, resulting in a rapid decline of export stocks at port and financial losses in excess of $1 billion for the state coal industry”.
According to big data execs at Microsoft, however, there is something that can be done. After working with one company in the area – which preferred not to be named – a team of business analysts actually discovered what they believe is climate change-induced shifts in local weather patterns that are affecting the sector.
The gold mining company in question had been using historical weather data pulled from a weather station in nearby Polokwane to plan its future projected output and create annual budgets and profit expectations.
It used the data which it believed to be accurate to formulate precipitation predictions for when rainfall in the area would be the heaviest, and which months the precipitation would be at its lowest. The data allowed it to plan production cycles, work output and revenue as times of heavy rainfall would result in fewer deposits being brought to the surface.
Where light or no rainfall was predicted by using the historical data, the company could ramp up production to make up for the loss from the wetter months, and also plan for an increase in revenue due to a better mineral yield.
But it didn’t go according to plan…
After a number of seasons, the mining company discovered that while its calculations were correct, the mining output did not match its projected earnings and mineral turnover. Nevertheless, it forged ahead year-on-year, relying on the same weather data to predict the seasonal rains ahead.
It wasn’t until the company turned brought in Microsoft that they realised that the weather data they have been using was completely wrong. Using analysis tools that link Azure servers with programmes such as PowerBI and PowerView, historical weather data from the New Pietersburg weather station was overlaid it with the weather data that the mining company had been using.
The results from the new data showed that rainfall patterns in the area appear to have shifted by two months over the last five years alone, and April 2011 seeing just as much rain as February of 2006, when they weren’t expecting it.
“When we presented our findings back to mine management, we highlighted the change in weather patterns. This is when the proverbial lights came on and they admitted that they were still planning their business on what they believed was accurate data, but in fact wasn’t,” Microsoft’s Solutions Sales Professional Mark Schneeberger told us.
Knowing now that the annual rainfall patterns have shifted every year from 2006 until 2011, the company can save money by adjusting their projected output and better manage staff around the times that higher precipitation is expected – which would in turn lead to less injuries during shifts.
Microsoft’s findings need to be further analysed and built into complex climate models to see if they do indeed show evidence of man made climate change or are part of natural weather patterns, says a spokesperson for the South African meteorological office. Nevertheless, they are intriguing evidence of the kind of macro effects that modern data analysis techniques can uncover using relatively simple tools.
That wasn’t the only interesting piece of productivity-enhancing data that Microsoft uncovered, in fact. It also found that by the company could save money by switching its heavy trucks to a different brand.
Microsoft took diesel prices and overlaid that with all the trucks’ efficiencies. It took the average price of diesel from the AA website and looked at the tonnage that trucks carry, average waiting time in a loading queue and how much fuel they use per ton.
“We could work out the diesel costs incurred by each truck brand, tonnage moved during the selected period and then established a tonnes per litre measure. This highlighted that the investment in Komatsu trucks were in fact the right move as they were proving to be more efficient – according to the calculation we created – than the CAT,” Schneeberger added.
“The organisation in question now has the ability to run its business more proactively by mining the vast amount of data from numerous data sources available both within and external to the organisation. Inefficiencies are quickly exposed allowing corrective action to be taken before costs run out of control.”[Image – CC by 2.0/Michael Becker]