Over the last six weeks we Brits have been starved of our usual pastime of talking about the weather, as the topic of the Olympics and Paralympics has taken over the conversation. Now that our summer of sport is over, we can safely go back to our favourite topic. After the ‘wettest summer for 100 years’ on the back of a prolonged dry spell that led to hosepipe bans in April, Figure It Out took to wondering whether the extremely wet weather we’ve had is any indicator of whether we are in for a wet winter or not.
So how wet was summer 2012?
The Meteorological Office defines summer as June-August and winter as Dec-Feb, and to measure how abnormal our weather is, then we need a definition of what is normal. Now the UK is world-famous for its variable weather, so we will base our observations on variations on the average rainfall per month over the last 20 years. Also we will consider the UK as whole, knowing that regional variation in weather is considerable; well you’ve got to start somewhere.
We see from the chart that the last winter has indeed been dryer than average, and the summer considerable wetter. This is on the back of below average rainfall for 2010 and 2011. Immediately it is apparent that even in this observation there is a big regional variation with England drier than the monthly UK averages in the winter, and perhaps more amazingly Scotland drier than this average in the summer! In fact even within Scotland the picture is varied, with North West Scotland considerably drier even than South East Scotland. Next year’s summer holiday destination perhaps?
Why the UK’s weather is so difficult to forecast
So before we try to get our heads around weather factors, we need to appreciate something about the variability in weather across the UK. We are in a unique position of being between the Atlantic Ocean on one side and continental Europe on the other, and it’s this location that makes our weather so worthy of study, as even subtle changes in the wind direction can bring marked changes in the weather (a local sort of butterfly effect). As well as this, the UK is sandwiched between tropical warm air coming from the south, and Arctic cold air from the north, which collide to create numerous weather systems. Tied to the fact that height from sea level, proximity to the coast, and geographical micro-climates (like big cities) also affect the weather, we have the conditions for a highly variable climate. Despite this inherent difficulty this does not stop us getting relatively accurate forecasts in the main, and improving all the time. For example the Met Office’s own analysis shows that a three-day forecast today is more accurate than a one-day forecast in 1980.
The factors that go into the specifics of our weather are many and varied, from the measurable outcomes that we notice, like temperature, rainfall, and humidity, to the key drivers for the UK, like the jet stream, and sea temperature. But in fact there are bigger global forces at play that are also factors in the weather we get. Such things as sunspot cycles, the El Nino South Pacific weather systems, and the North Atlantic oscillation have an effect, albeit a delayed one. This is clearly a very complex business and one that teams of meteorologists around the world occupy themselves with.
Are there any patterns in the data?
Although a complex system, could there be some simpler correlations between measured data that would help to provide clues for our weather six months ahead?
This situation is a common one in business – the commercial context that businesses work in is a complex mix of cause and effect, where it is impossible to account for every variable driving performance. But often there are key drivers that provide insight even without a full understanding of every feature of the problem. We have a number of problem structuring tools we can apply as well as statistical analysis of large data sets like product sales data, customer spend, and inventory levels. The idea is to find patterns or correlations, and the key is to spotting the difference between data that is connected (ice cream sales and barbecue purchases both rise together, but that is because they are both driven by warm weather), versus identifying the driver and making predictions (warm weather influences these sales but also affects garden furniture, beachwear, and air conditioning unit sales as well).
So is summer rain an indicator for a wet winter?
So back to our weather problem. Let’s start with a simple hypothesis: is there is a link between rainfall in the summer and the following winter? One way to investigate this is to see whether the data is correlated i.e. if one rises then so does the other. Remember we are not saying one causes the other, just that the measurement of today (summer rainfall) is an indicator of a future outcome (winter rainfall). Behind that conjecture is the assumption that in the main our rainfall pattern persists for many months, and once it is in one mode (persistently wet or persistently dry), it remains that way.
Well not surprisingly if you chart the actual summer rainfall against following winter rainfall (Met Office data) there is no obvious direct link, telling us that firstly summer rainfall is no indicator for whether the next winter will be wet or dry, but also that rain patterns in general do not persist for long enough to have any effect six months ahead. The known complexities of our weather suggest that picking any two parameters in a bid to find a pattern is unlikely to yield results.
Do we experience extreme weather?
Despite our relatively benign weather the UK does experience some extremes of weather, including hurricane force winds, tornados and snow in May.
Of course what we call extreme pales into insignificance compared to the worst the rest of the world can provide. Our wettest year since 1910 was the year 2000 with 1337mm of rain. In comparison La Reunion experience 1825mm of rain in 24 hours in January 1966!
Applying knowledge, dedicated tools and analytics delivers value from data
In a business setting we would start with a more detailed understanding of the issue, and apply specific sector or content knowledge to narrow down the possibilities before embarking on the analysis needed. We are able to apply sophisticated analytics to large data sets in order to reveal patterns, but we can also combine together data from different sources, leading to even greater insight. Thanks to developments in software and combining our experience and business analytics capability, we can explore big data for business benefit, demonstrating the value from business data. For example for one client we applied sales data with retail store performance measures, plus product demand forecasting to improve stock control and sales as well as reducing costs.
So in conclusion this analysis provides no clues as to whether to invest in shares in umbrellas, loft insulation or snow boots this winter, and that it’s best to leave weather forecasting to the professionals!